Electric Power Research Trends

ELECTRIC POWER RESEARCH TRENDS
ELECTRIC POWER RESEARCH TRENDS
MICHAEL C. SCHMIDT
EDITOR
Nova Science Publishers, Inc.

New York
Copyright 2007 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical photocopying, recording or otherwise without the written permission of the Publisher. For permission to use material from this book please contact us: Telephone 631-231-7269; Fax 631-231-8175 Web Site: http://www.novapublishers.com NOTICE TO THE READER The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained in this book. The Publisher shall not be liable for any special, consequential, or exemplary damages resulting, in whole or in part, from the readers use of, or reliance upon, this material. Independent verification should be sought for any data, advice or recommendations contained in this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage to persons or property arising from any methods, products, instructions, ideas or otherwise contained in this publication. This publication is designed to provide accurate and authoritative information with regard to the subject matter covered herein. It is sold with the clear understanding that the Publisher is not engaged in rendering legal or any other professional services. If legal or any other expert assistance is required, the services of a competent person should be sought. FROM A DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS. LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA Electric power research trends / Michael C. Schmidt, Editor. p. cm. Includes index. ISBN-13: 978-1-60692-758-8 1. Electric power. I. Schmidt, Michael C. TK1001.E35 2008 621.31--dc22 2007033923
Published by Nova Science Publishers, Inc.
New York
CONTENTS
Preface Chapter 1 Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal in the Electricity Generation Sector Panagiotis Grammelis and Emmanuel Kakaras Micro-CHP Power Generation for Residential and Small Commercial Buildings Louay M. Chamra and Pedro J. Mago Wind Energy in Electricity Markets with High Wind Penetration Julio Usaola and Edgardo D. Castronuovo Leakage Current on High Voltage Contaminated Insulators P. T. Tsarabaris and C. G. Karagiannopoulos Use of Organic Rankine Cycles to Produce Electric Power from Waste Heat Sources Pedro J. Mago and Louay M. Chamra Real Time Congestion Monitoring and Management of Power Systems Jizhong Zhu Power Systems Disturbance Analysis Abhisek Ukil Arcing and Fusion Process of Exploding Wires in Industrial Applications C. S. Psomopoulos and C. G. Karagiannopoulos Wide-Area Monitoring of Power System Oscillations in Thailand Power Network Using Synchronized Phasor Measurement Units via Demand Side Outlets Issarachai Ngamroo and Yasunori Mitani vii 1
Chapter 2
47
Chapter 3
103 155
Chapter 4 Chapter 5
177
Chapter 6
209 227
Chapter 7 Chapter 8
251
Chapter 9
269 295
Index
PREFACE
The world is becoming increasingly electrified. For the foreseeable future, coal will continue to be the dominant fuel used for electric power production. The low cost and abundance of coal is one of the primary reasons for this. Electric power transmission, a process in the delivery of electricity to consumers, is the bulk transfer of electrical power. Typically, power transmission is between the power plant and a substation near a populated area. Electricity distribution is the delivery from the substation to the consumers. Due to the large amount of power involved, transmission normally takes place at high voltage (110 kV or above). Electricity is usually transmitted over long distance through overhead power transmission lines. Underground power transmission is used only in densely populated areas due to its high cost of installation and maintenance, and because the high reactive power gain produces large charging currents and difficulties in voltage management. A power transmission system is sometimes referred to colloquially as a "grid"; however, for reasons of economy, the network is rarely a true grid. Redundant paths and lines are provided so that power can be routed from any power plant to any load center, through a variety of routes, based on the economics of the transmission path and the cost of power. Much analysis is done by transmission companies to determine the maximum reliable capacity of each line, which, due to system stability considerations, may be less than the physical or thermal limit of the line. Deregulation of electricity companies in many countries has led to renewed interest in reliable economic design of transmission networks. This new book presents leading-edge research on electric power and its generation, transmission and efficiency. Chapter 1 - In this chapter, various issues related to the co-firing of secondary fuels coming from biomass or waste, in existing coal-fired power plants are discussed. The main objective is to present the current situation of co-firing technology in Europe and its future developments. Towards that direction, the most applied technological co-firing options in the power industry are described. Co-firing can be carried out either with the direct or indirect approach. The general trend in co-firing is to adopt the former approach due to time saving, i.e. fewer modifications on the existing power plant, shorter shutdown periods etc, and lower cost-investment compared to the latter option. In order to co-utilize biomass in an existing coal fired power plant some form of retrofitting is needed. Although no two plants are the same, the practical considerations for retrofitting and co-firing can be generalized into four areas of interest, namely fuel availability, plant modification, legislative framework on environmental issues and financial evaluation. The technical constraints met in a co-firing
viii
Michael C. Schmidt
project are such as potential increase in slagging and fouling, ash deposition and utilization of solid by product, and pollutant formation, whilst the non-technical barriers mainly concern fuel supply and financial issues. The chapter concludes with a review on co-firing experience at European level, mainly focusing on the utilization of novel solid bio-fuels, such as Solid Recovered Fuels. It is summarized that co-firing with limited biomass shares is technically feasible and has been demonstrated in several European countries. Research efforts are now focused on the deployment of co-firing applications in even more power plants and the increase of solid bio-fuels shares in the mixture. For this reason, solid bio-fuel qualities have to be standardized for their use in the power sector and new fuel markets should be established. Chapter 2 - The traditional structure of the electrical utility market, green engineering issues, and environmental acceptability have resulted in a relatively small number of electric utilities. However, today's technology permits development of smaller, less expensive power systems, bringing in new, independent producers. Competitions from these independent producers along with the re-thinking of existing regulations have affected the conventional structure of electric utilities. The restructuring of the electric utility industry and the development of new "onsite and near-site" power generation technologies have opened up new possibilities for buildings, building complexes, and communities to generate and sell power. Competitive forces have created new challenges as well as opportunities for companies that can anticipate technological needs and emerging market trends. Micro-cooling, heating, and power (micro-CHP) is decentralized electricity generation coupled with thermally activated components for residential and small commercial applications. A micro-CHP system consists of a prime mover, such as a reciprocating engine, which drives a generator, which produces electrical power. The waste heat from the prime mover is recovered and used to drive thermally activated components and to produce hot water or warm air through the use of heat exchangers. Micro-CHP holds some of the answers to the efficiency, pollution, and deregulation issues that the utility industry currently faces. A review of micro-CHP systems, specific types of distributed power generation, and thermally-activated technologies are introduced and discussed in this chapter. Chapter 3 - The great amount of wind power recently connected to electric power systems makes necessary new grid analysis tools and control strategies. Some of these tools are short term wind power prediction programs, which have been recently developed and are already in use in many System Operators and wind power producers. Prediction programs also allow the participation of wind energy in the electricity market by keeping the economic losses due to power imbalances for wind generators within acceptable values, specially if the uncertainty of the predictions is evaluated and used in the bidding process. Such participation intends to make the power system operation easier. Distributed control of power system, and specially optimized control of clusters of wind generators is another measure that is being taken in order to maintain the levels of power system security and to minimize the consequences of possible and temporary wind power curtailments, due to grid security reasons, on the affected wind farms. The contents of this chapter include a detailed description of short term wind power prediction programs, an evaluation of the economic losses due to imbalances incurred by wind generators participating in an electricity market, and an optimization strategy aimed at minimizing the consequences of power curtailments.
Preface
ix
Chapter 4 - The present chapter will present a short review of research work regarding leakage current. In addition, measurement system of the leakage current is introduced using a high sampling frequency analogue/digital converter. The application of this system in 20 kV insulators contaminated from a compound of salt and kaolin are described. Using this system measurement in a time frame of one period (50 Hz) were done and an investigation of the observed phenomena is attempted with the assistance of i-u characteristic curves plotted for one cycle of voltage application. The fact that existence of partial discharges on the surface of the contaminated insulators, beyond threshold field intensity leads to radiation emission, is also be examined. An estimation of the free electrons energy has been done and the corresponding emitted radiation which seems to include acoustic waves, radio waves, microwaves and infrared waves is also examined. The classification of the leakage current values of a typical porcelain insulator of 20 kV, contaminated by salt and kaolin, is presented. The classification is based on the collaboration between the above high precision data acquisition system having high sampling rate and an unsupervised self-organized neural network. In addition a simulation model for contaminated insulators is presented. The proposed model will be provided, together with a mathematical function that simulates the behavior of the dry band resistance as a function of time, even in cases where arcs or partial discharges occur. The models parameters of a typical porcelain insulator of 20 kV, contaminated by salt and kaolin, are presented. Chapter 5 - Several industrial processes have low-temperature waste heat sources that can not be efficiently recovered. Low grade waste heat has generally been discarded by industry and has become an environmental concern because of thermal pollution. A solution for this problem is the use of Organic Rankine Cycles (ORC) which can make use of low temperature waste heat to generate electric power. This chapter presents an ORC analysis using different types of organic fluids and cycle configurations. Some of the working fluids under investigation are R134a, R113, R245ca, R245fa, R123, Isobutane, and Propane, with boiling points between -43C and 48C. These organic working fluids were selected to evaluate the effect of the fluid boiling point temperature on the performance of ORCs. The results are compared with those of water under similar conditions. In order to improve the cycle performance, modified ORC are also investigated. Regenerative ORCs are analyzed and compared with basic ORCs in order to determine the configuration that presents the best thermal efficiency with minimum irreversibility. The evaluation for both configurations is performed using a combined first and second law analysis by varying certain system operating parameters at various reference temperatures and pressures. Some of the results show that ORC using R113 as working fluid has the best thermal efficiency while those using Propane show the worse efficiency. Also, it is shown that the organic-fluid boiling point has a strong influence on the system thermal efficiency and the electric power production. In addition, results from these analyses show that regenerative ORC produces higher efficiency compared to basic ORC. Furthermore, regenerative ORC requires less waste heat to produce the same electric power with a lower irreversibility. ORCs are good candidates to produce electric power from low temperature heat sources. Chapter 6 - Power systems operation data could include a large list of variables. The power system uncertainties as well as restructuring are introducing new class of dynamic phenomena into the power grid and substantially increasing the data requirements for real time congestion monitoring and management (RTCMM). Generating units and customer load respond to market signals that are affected by grid conditions, such as congestion. One of the
Michael C. Schmidt
most visible impacts is the increased volatility in the system, which results in the increased volumes of data to manage. The increasingly severe data overwhelm problem results in many challenges in terms of development of effective analytical and software tools. The information availability is also an issue. Secure operations under open access require the widespread availability of real time information about physical, as distinct from financial, variables. Unfortunately, the competitive environment is not conductive to information sharing. For example, data for state estimation is becoming increasingly difficult to obtain because of some uncertain factors or the ability of that data to reveal sensitive financial information. Consequently, state estimation may be failed or couldnt be solved, which increases the difficult to analyze the behavior of real time system operations. Another challenge in real time congestion monitoring and management is to design real time controls to maintain the security and reliability of the grid under the market condition and the contingency condition as well as uncertain environment. The implementations of real time congestion monitoring and management are discussed in the chapter. Chapter 7 - The analysis of faults and disturbances has always been a fundamental foundation for a secure and reliable electrical power supply. The introduction of digital recording technology opened up a new dimension in the quantity and quality of fault and disturbance data acquisition, resulting in the availability of a huge amount of new information to power systems engineers. Information from the analysis of digital records can provide much-needed insight into the behavior of the power system as well as the performance of protection equipments. Manual analysis of these records, however, is both time-consuming and complex. Today the challenge is to automatically convert data into knowledge, which frees the human resources to implement corrective or preventive action. Therefore, much focus is laid on automatic disturbance analysis. This also goes towards the direction of blackout prevention which is a great threat from utility perspective. This chapter will provide the insight on the recent trends on power systems disturbance and fault analysis. Basics of disturbance analysis, Fourier transform and wavelet transform-based signal processing techniques in the power systems disturbance analysis are discussed in details. Chapter 8 - The fusible wires and electrically exploded conductors are frequently used in fuses that protect the electrical installations of the low and medium voltage distribution network and in a wide variety of power applications. They generally operate either under a minus value of their nominal current or momentarily under excess or sort-circuit currents. Despite a wide variety of experimental work, there remains much that is not understood about the electrical explosion of conductors. Furthermore the surrounding medium has a significant role in the fusion process. Most of the experimental work is focused in air and some in SiO2 due to use as the most common surrounding material in industrial fuses. Recently the fusion process in capillaries presented also interest among researchers. The power supplies used in these experimental works were pulses mostly, produced by typical L-C circuits. This chapter will investigate the fusion process of exploding wires using common industrial power supply directly feed from the network of 50Hz. The measured magnitudes were the voltage drop and the current waveforms during the current interruption process and the total duration of the fusion. Experiments performed for several diameters and lengths and different current densities. The resistance and the I=f(V) waveforms during fusion process were directly calculated from the experimental curves.
Preface
xi
In this chapter, measurements better than before through fast digital oscilloscopes connected with PC are presented and the observed phenomena are investigated. Measurements of the total time duration of the fusion process are presented and analysed. The presented measurements are compared with existing ones (taken with the use of A/D converters) and the differences are evaluated. In the related literature the experiments are carried out usually with pulse voltages or currents, as it is mentioned above. The measurements presented here were performed in random position of the supply voltage curve. These measurements give the possibility to investigate the physicochemical process during the fusion process further, and especially the arcing behavior. The obtained experimental results are investigated and evaluated with the use of analytical and statistical methods, and are compared with other relevant ones in literature. The analysis of the results is focused on the fusion duration and the arcs formation for a variety of parameters (current density, geometric characteristics, etc) because of their importance in the protection of networks during short circuit. The digital measurements with PCs, can provide a more accurate approximation and understanding of the phenomena during the fusion of exploding wire. Another parameter that contributes to the above is the use of common industrial voltage source and random sinus values in the experiments, because provide a more accurate simulation of the actual conditions. Chapter 9 - The longitudinal structure of the tie-line interconnection between central and southern areas of Thailand power system causes the dynamic stability problem due to the inter-area oscillation with poor damping. To detect the dominant inter-area oscillation mode, a sophisticated wide-area dynamic stability monitoring system based on the phasor measurement units (PMUs) is proposed. The main feature of the presented system is the convenient installation of PMU at 220 V level. PMUs are located at two universities representing both areas. The single-phase phasor voltage data measured by PMUs at both sites are synchronized by global positioning system. The phase voltage difference data are used to analyze the dynamic characteristic of the inter-area oscillation based on signal processing techniques. The proposed monitoring system provides vital information of widearea dynamic stability.
In: Electric Power Research Trends Editor: Michael C. Schmidt, pp. 1-45
ISBN: 978-60021-978-8 2007 Nova Science Publishers, Inc.
Chapter 1
CO-FIRING SOLID BIO-FUELS OR WASTE RECOVERED FUELS WITH COAL IN THE ELECTRICITY GENERATION SECTOR
Panagiotis Grammelis*and Emmanuel Kakaras
Institute for Solid Fuels Technology and Applications / Centre for Research and Technology Hellas, Ptolemais, Greece
ABSTRACT
In this chapter, various issues related to the co-firing of secondary fuels coming from biomass or waste, in existing coal-fired power plants are discussed. The main objective is to present the current situation of co-firing technology in Europe and its future developments. Towards that direction, the most applied technological co-firing options in the power industry are described. Co-firing can be carried out either with the direct or indirect approach. The general trend in co-firing is to adopt the former approach due to time saving, i.e. fewer modifications on the existing power plant, shorter shutdown periods etc, and lower cost-investment compared to the latter option. In order to co-utilize biomass in an existing coal fired power plant some form of retrofitting is needed. Although no two plants are the same, the practical considerations for retrofitting and cofiring can be generalized into four areas of interest, namely fuel availability, plant modification, legislative framework on environmental issues and financial evaluation. The technical constraints met in a co-firing project are such as potential increase in slagging and fouling, ash deposition and utilization of solid by product, and pollutant formation, whilst the non-technical barriers mainly concern fuel supply and financial issues. The chapter concludes with a review on co-firing experience at European level, mainly focusing on the utilization of novel solid bio-fuels, such as Solid Recovered Fuels. It is summarized that co-firing with limited biomass shares is technically feasible and has been demonstrated in several European countries. Research efforts are now focused on the deployment of co-firing applications in even more power plants and the
*
Phone: +30 210 6501 593, Fax: +30 210 6501 598; E-mail: grammelis@certh.gr
Panagiotis Grammelis and Emmanuel Kakaras

increase of solid bio-fuels shares in the mixture. For this reason, solid bio-fuel qualities have to be standardized for their use in the power sector and new fuel markets should be established.
Keywords: secondary fuels, coal, retrofitting, SRF
1. INTRODUCTION
Coal continues to be used extensively throughout the world to produce electricity and heat. However, this extensive use poses significant environmental problems. Indeed, the coalfired power plants emit substantial quantities of the known greenhouse and other acid gases, which should be reduced to the levels imposed by a few environmental legislative initiatives. One way to achieve the previous target is through the energy co-utilization of biomass in existing coal-fired power plants. Biomass is generally characterized as neutral fuel that contributes zero CO2 emissions to the environment throughout its life cycle. Its co-firing with coal represents a near-term, low-risk, low-cost, sustainable, renewable energy option that promises reduction in CO2, SOx and often NOx emissions as well as several societal benefits. Consequently, co-firing biomass with coal in traditional coal-fired boilers represents one combination of renewable and fossil energy utilization that derives the greatest benefit from both fuel types [1]. Namely, the environmental and economical benefits offered are [2], [3], [4], [5]: Conservation of fossil fuel resources and / or reduction of the dependence on fuel imports Smaller capital investment compared to new dedicated-biomass plants Utilization of agricultural and forest residues and / or recultivation of non-utilized farming areas Favorable impacts on emissions from the coal-fired plants and assistance to the European Community in meeting its Kyoto targets and reducing the CO2 emissions The establishment of a market for power and heat produced from co-firing biomass and waste
In this work, an overview of all the practical and technical considerations encountered when attempting to retrofit an existing coal-fired power plant to co-utilize biomass is presented. The biomass materials concerned are agricultural residues (straws, cereals), energy crops (miscanthus, short rotation coppice), forestry residues (wood chips), agro-industrial residues (olive and palm kernels) and waste derived fuels (sewage sludge, waste wood, Solid Recovered Fuel). The main issues under investigation are: Overview of the current techniques applied when co-firing secondary fuels with coal. Main practical issues to be taken into account when co-firing biomass with coal in existing power plants, such as the fuel availability, the power plant modifications, the legislative framework on environmental issues and the economics of co-firing.
Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal
Barriers, mostly of technical nature, that can issue a variety of constraints on cofiring implementation in large-scale power plants. A review of European operating power plants which experienced biomass co-firing at demonstration or commercial scale. Research trends for future developments in co-firing solid bio-fuels.
2. CURRENT CO-FIRING TECHNIQUES

In general terms, solid bio-fuels can be used as supplementary fuels in combination with solid fossil fuels in utility boilers or as a single fuel in combined heat and power plants of limited capacity [2]. Overall, biomass co-firing with standard fuels can be accomplished in two concepts, the direct and the indirect co-firing. The basic options [6], [7] available for the direct co-firing of biomass materials at coal-fired power stations are: The pre-blending of biomass with coal and the feeding of the mixture into the bunkers, followed by the processing of the blend through the existing coal milling and firing equipment. This approach is called the co-milling option. It is possible that the milling of biomass and / or the mixing with coal is carried out off-site, followed by the on-site handling and firing of the blended fuel. The modification of one or more of the existing coal mills in each unit to mill the biomass material on its own, and the firing of the milled material through the existing pulverized coal pipework and burners. This option is only available for a limited number of biomass materials. The installation of new, dedicated biomass mills and the introduction of the milled fuel directly into the existing coal firing system after its modification. New dedicated biomass burners may be placed in the furnace rear or side walls.
The first option [8] is the most cost-effective and favorable approach. An additional advantage is that it can be implemented easily and relatively quickly. Furthermore, it is more attractive to power plants operators, due to the uncertainties about long-term biomass suppliers. The second option is very simple and there are a number of current projects of this kind in the UK and Northern Europe. The third option demands significant construction modifications of the existing coal burners. However, there are currently projects based on this approach in continental Europe. One example of coal burner modification is at Studstrup power plant in Denmark, where chopped straw is co-fired through the core air tubes of four low-NOx burners. The pulverized coal is fired through the primary air annulus as normal. This system has been in successful operation so far. The placement of new dedicated biomass burners in the furnace may increase combustion performance of the biomass, but it is typically an expensive, complex and risky operation compared to modifying the existing coal burners. Moreover, negative impacts on operational flexibility during low load operation may arise [9]. On the other hand, the indirect option for co-firing involves the installation of a separate gasifier or boiler. In this case, the gas product from the biomass gasification or the combustion gases from separate fuel combustion can be used in a gas turbine or a common
Panagiotis Grammelis and Emmanuel Kakaras
steam cycle for heat and power generation. Up to now, indirect co-firing of biomass and coal has been demonstrated successfully at a limited number of power plants, due to the higher capital expenditure necessitated and the long-time taken to implement compared to direct cofiring. This technique is particularly suitable for biomass containing troublesome components or when the ash quality is of importance for subsequent sale or disposal [10].
3. PRACTICAL CONSIDERATIONS FOR RETROFITTING A COAL-FIRED UNIT TO CO-FIRING BIOMASS

This section deals with the main practical considerations to be taken into account when co-firing biomass with standard fossil fuels: fuel availability, power plant modifications, legislative, environmental and societal issues. A generic approach for the direct biomass cofiring implementation in a commercial coal-fired power plant is applied and described below. The selected alternative of co-firing is the co-milling option considering the plant modifications.
3.1. Fuel Availability

A wide range of different types of biomass fuels can be used for co-firing, although the qualities of the fuel can have significant impacts on the equipment required for handling, milling and combusting the material. Biomass fuel resources can be categorized as follows: Agricultural residues produced by agricultural farming practices for food production, such as straw and cereal residues etc. Forestry and wood processing residues produced from existing lumbering and coppicing operations in established forestry, such as wood chips, sawdust, bark etc. Energy crops defined as plant crop planted and grown primarily for the purpose of being used as fuel for energy production, such as willow, eucalyptus, miscanthus etc. Waste derived fuels coming from different waste streams which contain a significant portion of biogenic fraction and have an energy potential
The power plant operator will need to assess the type of biomass available that the region can provide. For example [11], in the southern part of UK, there is an abundance of cereal straw, whilst in Greece and other Mediterranean countries, olive waste is being exploited. Indicatively, a summary of solid bio-fuels available in Europe and information about their chemical and physical analysis is presented in Table 1.
Table 1. Main characteristics of solid bio-fuels [33]
Crop Dry matter per Wet matter (%) 80-90 80-90 70-80 35-45 45-55 55 - 65 55 - 65 55 - 65 55 - 65 55 - 65 45 - 55 (w) 60 65 (b) 45 - 50 45 - 50 120 - 130 30 - 40 17.0 17.3 18.4 18.9 60 - 70 16.8 17.4 3-5 2 2.5 7 - 10 7 -8 18.0 18.5 1-2 0.5 1.5 2-4 1 1.5 40 - 50 45 55 23 - 26 47 - 48 2-5 1.5 2 5-7 8-9 60-65 60-65 60-65 60-65 70-80 85 - 90 47 - 55 C/N Ratio
Main Product
Residue
No nitrogen Extracts (%) 40-50 40-50 40-45 45-50 70-80 10 - 20 35 - 40 38 - 54 39 51 53 - 63 34 - 36
Barley Oak Rice Maize
Grains Grains Grains Grains
Straw Straw Straw Straw Cob stalks Pruning Pruning Pruning
Chemical-physical analysis of the residue High Dry matter composition Calorific Proteins Lipids Fibre (%) Ashes Value (%) (%) (%) (MJ/kg d.m) 16.8-17.2 3-5 1-2 40-50 5-10 16.8-17.2 3-5 1-2 40-50 5-10 15.6-15.9 4-5 1-2 35-40 10-15 16.8-18.0 4-5 0.5-1 35-40 5-7 16.8-18.0 3-4 0.6-1 40-50 2-3 17.9 18.4 2 -3 1 2 50 60 20 - 25 18.0 18.5 18.0 - 18.5 18.0 18.5 4-5 2 -3 40 - 45 10 - 12
Sunflower Apple tree Pear tree Peach tree, nectarine tree Citrus orchard Vineyard
Pruning
Olive tree
seeds apples pear drupes (peaches) fruits (esperidi) grape (berries) Olives (drupes)
sarmenta (pruning) (w) wood, (b) branches pruned
Soybean
grains
Stalks
Panagiotis Grammelis
Once the type of biomass has been selected, the next consideration to be addressed is the availability of the selected biomass materials throughout the year. The availability is influenced by the fuels themselves and more specifically from seasonal differences and risks resulting from climatic influences. Hence, the operator has to select an alternative biomass fuel during those periods that the primary selection is not available. An alternative solution is the storage of the chosen biomass fuel, but then safety, storage conditions and economics parameters become of importance and need to be assessed [12]. The seasonal availability of biomass materials for UK and Greece is presented in Table 2. The other important consideration is biomass quantities. Due to its low energy density and its high moisture content, large biomass quantities are needed in order to maintain the thermal output level. The power operator has to ensure that the biomass suppliers can provide the necessary amount of supplementary fuel. Moreover, for financial reasons, biomass resources need to be available close to an existing power plant. Transporting biomass over a vast distance can defeat to some extend the purpose of co-firing, due to additional transportation and labor costs and pollutants generated by the vehicles. In general, the establishment of Energy Companies associated exclusively with the collection, pre-treatment and provision of standardized-quality solid bio-fuels may contribute to improve the biomass fuels logistics. Table 2. Availability of biomass materials for UK and Greece
It should be emphasized that the fuel supply of power plants is the most important issue to be addressed when intending to apply co-firing on a large-scale power plant. In other words, the feeding rates of the fuels must be guaranteed so as to maintain continuous operation. In the case of coal, securing the fuel supply is easy. As regards biomass materials, the discussion above illustrates that the power plant operator should examine a number of key factors in order to minimize any uncertainties that could arise and affect the supply chain. The establishment of institutions associated with the standardization of renewable fuels is expected to provide valuable assistance in reliably supplying power plants with the proper quantities of qualified solid bio-fuels.
3.2. Plant Modifications

Before implementing any specific set of retrofitting actions, the power plant operator has to decide on the best course of action in order to reach a specified goal. For this reason, a set of criteria [13] are established to facilitate decision making related to retrofitting coal fired power plants. This approach is used for plant modifications aiming to improve performances both in terms of plant efficiency and/or emissions reduction. The criteria, as applied in the case of co-firing, are presented below: Criterion 1: Conformity with the company objectives As mentioned previously, co-firing is a cost-effective way to reduce greenhouse gases, especially carbon dioxide. However, a series of retrofitting actions are always required for a power plant. Hence the company will have to clarify and quantify the objectives of the modernization based on both the experienced operating status and on the anticipated role of the power plant into the new energy market. Depending on the biomass species and the boiler technology, the needs to replace or even upgrade the existing equipment have to be identified. Criterion 2: Beneficial impacts of the proposed action As far as co-firing is concerned, the benefits of any related retrofitting action have already been mentioned and include carbon dioxide neutrality, conservation of fossil fuel resources and potential financial benefits. Those benefits should be qualified and quantified for all co-firing scenarios under investigation. Criterion 3: Applicability of the proposed action Any proposed retrofit measure should be examined for its applicability to the power plant under consideration. In the biomass co-firing case, the power plant site should be assessed to ensure that there is sufficient space for the construction of a storage facility for the biomass fuels. In addition, if new pipe work is needed to feed the biomass into the furnace, a feasibility study should be carried out. Criterion 4: Technical risk For any retrofitting activities, the technical risk must be kept at minimum. In the case of biomass co-firing, an initial investigation of the operational difficulties provoked from biomass introduction in the boiler must be carried out and the associated problems must be identified and minimized prior to the implementation phase. As an example, slagging, fouling and corrosion related problems that affect the power plant performance must be examined and confronted. Criterion 5: Investment cost From the plant operator point of view, the investment cost is perhaps the most important factor when deciding for retrofitting activities. In the case of co-firing with biomass, the total investment costs - such as building of new storage facilities to house the biomass, the new burners, new feeding systems etc - should be less than revenue generated with the remaining lifespan of the power plant. The high cost of dedicated biomass equipment and the accompanying plant modifications is one of the main reasons for the prevalence of direct co-firing techniques over indirect ones. Criterion 6: Requested idle period for the unit Since the unit will not be operational during the retrofitting period, a loss of revenues is expected. For biomass co-firing, as well as any other retrofitting
measures, the plant operator should ensure that the outage time is kept at a minimum. Therefore, the coordinators will have to assess the most suitable periods for retrofitting activities, ensure that the project proceeds as planned and try to organize the subtasks in parallel rather than in series. Once the above mentioned criteria are met and the decision for retrofitting is taken, practical issues of implementing the renovation activities are considered. In the case of cofiring biomass, the retrofitting activity is restricted due to cost and time factors. The basic idea is to take advantage of the existing equipment at the power plants without major modifications. As mentioned above, the most preferable approach for co-firing biomass is the co-milling option. This involves the pre-mixing of biomass with coal at low co-firing ratios, generally less than 10% on a heat input basis, in the coal yard or on the main coal conveyors and then processing the blended fuel through the existing coal milling and firing system [3], [6]. In comilling, coal and biomass are pulverized together at the existing coal mills to achieve size reduction and drying of both feedstocks. This presupposes that the solid bio-fuels can be easily handled using equipments similar to those employed for coal handling and feeding. Therefore, there are limitations on the range of biomass materials that can be co-fired in this way. For example, baled biomass materials are not suitable. Otherwise, pelletised and dried forms can be utilized [8]. Despite the low levels of capital expenditure, the major disadvantage of co-milling is that the presence of the biomass degrades the grinding performance of the coal mills [14]. In addition, there is the direct co-firing option which involves the pneumatic injection of the milled biomass into the pulverized coal pipework after by-passing the coal mills or into modified coal burners. This approach allows co-firing of different biomass fuels with coal and permits operation at higher co-firing ratios as well. In both cases, both additional air and additional fuel are being introduced to the milling system, thus the mill primary air and coal flow rates have to be reduced accordingly. This should be done in such a way so that the mills and the burners are maintained within their normal operating envelopes, when the biomass co-firing system is either in or out of service. However, there can be significant interferences with the mill control system and risks of interference with mill operation. In general, there are two different locations of the biomass injection into the pulverized coal pipework, both of which are described below [8]: Into the mill outlet pipes local to the mill outlet, which tends to be simplest to engineer since the routing of the biomass pipework is usually easier and the mill outlet pipework is cold. Into the pulverized coal pipework just upstream of the burner, which tends to be relatively more congested and the injection point is into hot pipework. In particular, at this point the pulverized coal pipework moves with the expansion of the furnace, and a suitable degree of flexibility has to build in to the biomass pipework.
Alternatively, the injection point of solid bio-fuels can be located directly into the coal burners. The third option of direct co-firing involves the injection of biomass directly into the same boiler as the coal, but via a separate feeding system. This method is more costly than
10
co-milling, since it requires greater modifications to existing coal plant. A direct injection mechanism can take six months to become fully operational and repay the initial investment in six months. Direct injection also has the advantage that the biomass does not affect the flow of the pulverized coal or the performance of the coal milling equipment. In addition, direct injection can employ separate burners for each fuel type. This allows the biomass and coal to be fired in the same boiler but ensures separate combustion of each fuel. The separate injection mechanism also provides additional control since biomass with poor milling qualities can be rejected without affecting the coal feedstock. Conclusively, the most cost-effective approach in co-firing is the co-milling option. Once the solid bio-fuels have been selected the next practical consideration would be to check the suitability of the existing equipments and the possible modifications to be done at the power plant. The final objective is the safe, reliable and effective operation of the modified power plant. The lower energy density of biomass compared to coal, necessitates a relatively high volume flow into the boiler [15]. The different properties of biomass will impact on the performance of the boiler and the associated equipment. The areas of potential impact [2] of the fuel blend and its product are analyzed below.
3.2.1. Biomass Handling, Storage and Fuel Preparation The fuel handling is important to the operation of any power plant and therefore it needs to be trouble-free. Where there was inappropriate handling of the supplementary fuel, the mill power requirements were raised by 10 - 15% and this led to increased operating costs which were reflected in the cost of electricity [15]. The characteristics of biomass materials are different from those of coal. A major concern for all biomass fuels is that the energy density is only about 10% of that for coal. Moreover, biomass is moist and has the strong tendency to create hydrogen bonds with water. Consequently, all the above restrict the maximum proportion of biomass in the mixture and demand on-site fuel preparation technologies disproportionately high compared to its heat contribution. In any case, specific pre-treatment is required when co-firing biomass with coal so as to minimize the negative effects. The degree of severity of the problems encountered during their handling is strongly dependent on the type of biomass, as well as on its moisture content. For example, the handling of straw requires different techniques compared to granular materials. The former are generally free flow and must be stored in a dry condition, while the storage time is kept at a minimum. The latter can be handled easier and their flow behavior can be characterized by the existing test methods. Woody biomass in the form of chips, chunks and sawdust is more difficult to characterize because of the wide range of particle sizes, moisture contents and the degree of elasticity they exhibit. Herbaceous biomass is usually handled, transported and stored in baled form, which is difficult and more expensive to cope with. The pelletised form of these materials can be handled easily, but when the pellets break up this necessitates problems, due to the liberised fine particles [16]. Concerns are also raised, when biomass pellets getting wet due to the moisture in the atmosphere. The generation and accumulation of dust is another important issue, especially when handling pellets in bulk. The proper equipments for dust suppression include explosion vents and fire suppression systems because the wet removal configurations tend to increase the humidity in the store, increasing thereby their cohesion. The storage area must be large enough to accommodate sufficient quantities of biomass and to ensure continuity of feeding rates. Current methods of storage are open and covered
11
piles, closed bunkers and silos. It is known that when storing large amounts of biomass for long periods of time, there is a serious threat of spontaneous heating of the stored material, due to the biological activity. In addition, wet biomass has a higher tendency for biological self-heating than dry biomass. Thus, storage times should be minimized, especially for the types of biomass which are most vulnerable to humidity. A careful arrangement of piles and sufficient natural drying is required, in order to achieve a good control of stored biomass and avoid flame incidents. The provision of enclosed and ventilated storage can prevent the solid bio-fuels from moulding [16]. In order to avoid the self-heating of the stored material and possible fire ignitions, biomass should be stored as dry as possible or with moisture content below 20%. A storage pile should not be compacted and not contain heterogeneous materials. A combination of temperature and gas measurements can offer the proper control of the biomass storage [12]. Another method of handling biomass is road transport. Lorries with large containers at the back which can be raised at one end to tip out the contents are the type of trucks used mostly for road transport. These tipper trucks require a high free space at the point of tipping. Thus, the reception facilities should be designed in such a way so as to ease the unloading of the fuel. An approach to this is by housing them inside a suitably tall portal frame building. Alternatively, a reception hopper can be installed inside the edge of a smaller building allowing the trucks to back up to an open in the building when tipping [14]. The preparation of solid bio-fuels is mainly associated with the removal of oversize materials. The separation can be completed in two stages. The reception hopper grid can help with the removal of large materials, while a moving screen or a cyclone can remove the smaller ones [14].
3.2.2. Preparation of the Blended Fuel and Modifications at the Coal Mills As regards the pretreatment processes, biomass produces a non-friable and fibrous material during grinding. According to many researchers [1] it is unfeasible and unnecessary to reduce biomass to the same size or shape as coal. In general, biomass particles with average aspect ratios (the ratio of the particles longer dimension to its shorter one), ranging from three to seven create challenges when they are pneumatically or otherwise transported. In cases where the co-grinding of fuels is needed, pulverized coal mill capacities may be reduced, especially when co-milling wood with coal. This has the effect of limiting the degree of co-firing that is possible in retrofits and removes the normal margin of mill redundancy. Besides, large and non-spherical particles induce fuel conversion problems. Biomass of proper size and under the heating rates typical of pulverized coal co-firing yields a much higher fraction of its mass through devolatilization than coal. The preparation of the fuel blend is primarily required when the co-milling option is adopted and it can be conducted either on-site or off-site. If the preparation is going to be carried out on-site, separate reception and handling facilities are required up to the point where biomass and coal are blended into one stream. Off-site blending results in a single fuel stream to the power plant, which can be handled in a similar way to coal. According to commercial-scale co-firing applications, problems like extreme bunker flow profiles with deposits building up at the bunker walls were encountered due to the blending of different origin fuels. Blending small proportions of biomass with coal can affect the properties of the mixture. Coal and biomass materials have also different heating values. Therefore, mixing the fuels has the effect of reducing the energy density and hence the volumetric feed rate of the
12
mixed fuel. In turn, this reduced energy density will have an impact on the feeder calibration and possible on the maximum mill throughput. The outermost objective of the fuel blending is the achievement of a homogeneous fuel blend. In order to provide this, biomass must be reasonably distributed within the volume of coal feedstock. Attention has to be paid to avoid the operational transients due to fuel quality changes and on the effort to keep its properties within safe limits. Also, the avoidance of problems associated with possible loss of coal mill flow is strongly demanded. A typical approach in order to achieve a typical continuous fuel delivery is by layering the biomass with coal on a stockpile. However, biomass degrades when exposed to air or moisture and thus it should be combusted in a short time. Another approach includes mixing after storage at a point slightly downstream of the coal feeder, but this requires duplication of the coal handling and storage systems employed for the primary fuel. In the biomass co-firing retrofits in UK [14], the continuous mixing of the fuels en route to the boiler, either on a conveyor or in a transfer chute, is preferred instead. On the other hand, this approach introduces inconveniences when mixing small quantities of biomass. The screw conveyors used for the control of the biomass flow create voids due to the low flow rate. Derived from the experience gained from co-firing activities in UK, the use of existing coal chutes and conveyors belt is problematic due to spillage and dust, especially where the biomass is first mixed with the coal. However, the previous is based on the fuel type and plant design. In any case, the construction of windshields, enclosures and dust suppression systems can help solve these problems. The effective and safe operation of coal mills when grounding solid bio-fuels is necessary so as to achieve a maximum level of co-firing. Special equipment, such as spike rolls disc screens, hammer mills or cutting equipments, might be needed to reduce the particle sizes to proper dimension. Problems are encountered due to the fibrous texture and the low particle density of biomass materials resulting in poor grinding, large-sized particles of the mill product and even an increase in carbon in ash. Also, by increasing the co-firing ratio, a corresponding increase in power consumption and levels of differential pressure for the coal milling equipments may be observed. The primary hot air employed to dry the coal in the mill increases the temperature and pressure. This, in turn, can lead to a release of the biomass volatile content, which starts at lower temperature than those occurring in the mill (compared to coal). Hence, necessary steps must be taken to ensure minimum risk of overheating of the fuel mixture. The combustion behavior of the fuel blend during co-firing processes is acceptable when the mill product, especially the biomass particles, has a top size in the range of 2 to 5 mm. Then, the unburned carbon levels in both the bottom and fly ashes are similar if not marginal lesser than solely burning coal [8]. Despite the difficulties and limitations, however, the co-milling option has been carried out successfully on a fully commercial basis in a large number of coal-fired power plants all over the world.
3.2.3. Impact of Co-firing on the Boiler Performance Biomass co-firing with coal can have a strong influence on the operation and reliability of the power plant, hence, considerations such as steam generator performance, slagging / fouling and corrosion of the boiler must also be taken into account for each individual power plant. The solid bio-fuels with a proper particle size can not influence negatively the combustion behavior of the fuel blend. Concurrently, the residence time of the particles in the
13
furnace is another important factor for a complete burnout. If the time is insufficient the burnout deteriorates with additional biomass share. Moreover, the moisture content of the biomass can have an impact on maximum achievable boiler load and efficiency [17]. The impacts on the boiler load have been minimized with low co-firing ratios and utilization of dry biomass materials. As far as efficiency is concerned, it is calculated by summing the losses due to unburned material, radiation and heat losses from the chimney. The heat losses from the chimney are owed to both the water emitted from it and the water generated in the combustion process. The former is dependent on the moisture content of the fuels (as fired) and the latter on their chemical composition. According to a scenario taken from literature [8], assuming a biomass fuel like wood with moisture content less than 10% co-firing with high volatile bituminous coal with 12% moisture and 15.7% ash contents, as received at a ratio of 5% by mass, the calculations result in an efficiency loss of around 0.1%. Biomass materials with high amount of alkali metals in ash have the tendency to lower the melting point of the ash depositions. In the case of wood, the high amount of calcium and potassium in ash indicates that the melting temperature of ash of this kind will be high. Therefore, the deposits will be loose even on the radiation boiler surfaces. Co-firing wood with coal produces minerals with high fouling tendency, resulting in the deterioration of boiler efficiency [18]. However, the boiler efficiency loss is much lower than the decrease in coal consumption and thus the biomass co-firing is profitable. In the case of solid bio-fuels such as straw, the high chlorine compounds in ash compared to coal tend to create strong fouling phenomena. Particularly in the superheater area, the deposits of this kind are strongly corrosive due to the high steam temperatures, creating a layer which deteriorates the heat transfer coefficient. Thus, in supercritical boilers which operate with increased steam parameters this phenomena are determinant in order to implement co-firing at them. Nevertheless, in co-firing with sulfur containing fuels like coal, this combination may help mitigate the chlorine-based corrosion of the boiler walls and steam superheaters [19]. This happens due to the reaction of SO2 generated from coal with alkali chlorides generated from biomass, to form alkali sulfates, which agglomerate on the convention surfaces and protect the tube surfaces. The promoting agent for this phenomenon is the oxidizing conditions. Thus, the conversion rates of chlorine deposits to sulfates decreases under reducing conditions or lower temperatures. At the convective heating surfaces fouling and erosion can occur. Fuels with low ash melting points enhance the fouling potential. A crucial factor that can bring about corrosion is the high flue gas rates. The high moisture content of supplementary fuel compared to primary fuels encourages variations in the wet flue gas flow rates. Then, the latter will surpass the boiler specifications resulting in changes of the convective heat transfer [2].
3.2.4. Impacts of Co-firing on Emissions In general terms, the impact of biomass co-firing with coal on environmental issues is similar to or slightly positive to the sole burning of coal. The main issues that must be examined are the air emission constraints, and constraints concerning the quality of the solid by-products produced. The environmental regulatory framework set by both the commission and national level must in any case be observed. As far as pollutant emissions are concerned, biomass co-firing has the benefit of being regarded as CO2 neutral; moreover it and has already demonstrated evidence of reduced NOx emissions and the potential for SO2
14
reductions. A description of effects of co-firing on the pollutant formations is presented in the following paragraphs. The measured NOx emission levels when co-firing is similar to those measured when firing coal. However, in some instances, there has been an improvement of up to 10% in NOx emission reduction when co-firing is implemented. Depending on the type of biomass used and the developed temperature during combustion, the corresponding NOx emissions vary. For example, some of the most commercially mature biomass materials, notably wood, contain relatively little fuel nitrogen; the nitrogen concentration of straw on the other hand is of the same magnitude with the nitrogen in coal. Also, the higher moisture content of biomass than coal decreases peak temperatures and leads to proportional decreases in NOx [1]. The generation of SO2 emissions is dependent strongly on the sulphur in the fuel if sulphur capture is not included in the plant design. Almost all coal types show higher sulphur levels than does biomass materials. In a pure coal flame a percentage of 80 to 100% of the fuel sulphur is converted into SO2 [18].When co-firing biomass with coal, the SO2 emission levels are similar to or lower than that of firing with coal only [20]. This happens because the SO2 generated from coal reacts with the alkali metals present in biomass ash and that results in the reduction of this pollutant. Particulate ash collection is an important part of plant operation to minimize dust emissions. According to several plants reports, there is a slight increase in particulate emissions, mainly, when co-firing with wood [15]. Concerns about fugitive dust emissions and odors during the biomass delivery and handling process are raised, as well. In some cases, the effectiveness of the flue gas treatment systems such as the Electrostatic Precipitators (ESPs) might also be affected, due to the increase in flue gas volume during co-firing. On the other hand, the low ash levels of solid bio-fuels relieve the ESPs performance. Besides, the different fly ash composition can alter its electrical conductivity and, hence, the behavior of the mixed ash in the precipitators. The impacts on the DeNOx catalyst system depend on its placement within the flue gas treatment system. If it is installed upstream of the electrostatic precipitator, compounds of the alkali metals can lead to its deactivation. Otherwise, if it is installed between the FDG system and the chimney the flue gas contains negligible amounts of components, which could damage the catalyst. For this reason, the low dust configuration (second one as mentioned) is particularly more suitable for co-firing. The amounts of trace elements in solid bio-fuels are lower than in coal. Thus, when cofiring with fossil fuels this lead to reduced input of these into the boiler. As a consequence, measurements of trace elements in different places, air heater inlet and emitted flue gases, showed decreased levels in the first case and little impact in the latter. None adverse impact was found on dioxins and furans emissions, as well [14]. According to the indications of some tests performed, the dioxins that would be expected to be produced during co-firing appear to be destroyed in the high temperatures within the furnace and the heavy metals appear to be trapped within the ash [15]. The ash disposal or utilization of solid by-products of biomass co-firing with coal is another area of concern due to the fact that the incorporation of solid bio-fuels may change the properties of coal-derived ash. Indeed, the environmental issues that may arise associated with the further disposal or utilisation of solid by-products could pose serious obstacles in the implementation of co-firing techniques. The changes to the properties of the solid by-products should be assessed in the light of the European standard EN 450 and the ASTM specifications
15
for the use of fly ash in construction products. Some preliminary results [1] regarding the impact of mixed fly ash on concrete are presented below. However, these results constitute nothing but indications and should not be accepted as general rule without further research. Only herbaceous biomass materials were used in the tests, since many woody fuels contain so little ash that co-firing is likely to have negligible effects on fly ash properties. According to the tests, the amount of aerating agents (surfactants) which are basic for the establishment of concrete air entrainment was increased with increasing the biomass content. This impact is of little economic concern but is of major process concern, due to the likely fail under freeze thaw cycles of the resulting concrete if the mixed fly ash that displace cement was treated the same way as coal-derived fly ash. As a general result of these data, the mixed fly ash behaves qualitatively similar to coal-derived one in terms of structural and performance properties. Nevertheless, further research work on this subject has to be done in order to ensure the mixed fly ash utilization in cement manufacture. According to the analysis presented, the overall conclusion which emerged naturally is that the impacts on performance, integrity and environmental performance of the power plants, where co-firing of a variety of biomass materials at low co-firing ratios implementing, have been small.
3.3. Legislative Framework on Environmental Issues

The aim of the environmental legislation that has been developed over the years is to ensure that any impact on the environment of industrial facilities stays within acceptable limits. The legislative framework for the utilization and promotion of solid bio-fuels into energy sector through co-firing activities with fossil fuels is the subject examined in this section. It is known that the majority of legislative issues is in close relation with economic considerations and always have financial extensions. This means that there is a penalty for those who activities are perceived as environmental threats for exceeding emissions limits or do not conform to the policy considerations in general. There are a few EU Directives of direct relevance to the power industry. For example, the Integrated Pollution Prevention and Control (IPPC) Directive specifies that Best Available Techniques (BAT) for minimizing pollution should be determined for various industry categories, including Large Combustion Plant (LCP). The European Pollutant Emission Register (EPER), established by a Commission Decision of 07/17/00, which is based on Article 15(3) of 96/61/EC Directive, requires Member States to produce a triennial report on the emissions of listed pollutants of industrial facilities into air and water [21]. The selection of biomass as a fuel for the purpose of recovering its energy content is due to these naturally-occurring materials are viewed as CO2-neutral fuels. The introduction of biomass into energy activities was strengthened by 2001/77/EC Directive whose purpose is to promote an increase in the contribution of renewable energy sources to electricity production in the internal electricity market. Moreover, co-firing applications offer an environmentally sound and a more economic approach to existing coal-fired power plants compared to dedicated biomass boilers, since no other practical technology exists except for fuel switching. Before burning biomass materials, all combustion plants operating under IPPC [19] need approval from the Environment Agency (EA) in order to proceed in co-firing
16
applications. In order to avoid possible inconsistency in approach to approvals given the EA produce a protocol to standardize the approvals for operators. The revised 2001/80/EC Directive on the limitation of emissions of certain pollutants into the air from large combustion plants establishes emissions limit values (ELVs) for new and existing plants. Under this legislation, existing combustion plants must either observe lower emission limits or achieve equivalent emission reductions via a national emission reduction plan by 2008 unless it is intended to close the plant after a further 20,000 operating hours between 2008 and the end of 2015. Namely, the ELVs for existing and new plant with installed capacity greater than 300 MWth are given in the Table 3. Concerning the Greenhouse Gas mitigation in the power sector, the EC confirmed its strong commitment to ensuring that the EU and Member States achieve their greenhouse gas emission targets under the Kyoto Protocol. The latter was approved on behalf of the Community by Council Decision 2002/358 of 28 April 2002. In accordance with the 2003/87 Directive, the EC activate a system called the European Emissions Trading Scheme (EUETS), in which an overall cap is placed on CO2 emissions. Thus, industrial facilities are allocated a number of CO2 emissions permits according to each Member States National Allocation Plan. Facilities that exceed their permitted emissions must purchase surplus permits from other operators. In this way, every ton of CO2 emitted by a power station has a market price reflecting either the cost incurred from the purchase of emissions permits or the income the plant could otherwise have earned from the sale of the permit. As biomass is considered carbon neutral, it can be burned without using up emissions permits. Therefore, a generator who co-fires biomass with coal effectively receives an additional source of income, beyond the price of electricity itself, the carbon price of the coal emissions foregone. [21] The following example is characteristic of how the legislative frameworks are connected with economic considerations. The data source is mainly the Department of Trade and Industry in UK. Therefore, on 27/10/2006, the carbon price was 11.85 per ton of CO2. In order to offer incentives for co-firing, a carbon price of at least 20 is needed, as Table 4 shows. The carbon price exceeded 20 during 2005, reaching a peak of almost 30 in April 2006, before falling to a price of 10-15 (DTI, 2006). On this basis, it appears that co-firing in the UK is close to becoming an economically viable activity. However, a permanent increase in the carbon price is required. This means that any future tightening in the emissions permits issued under EU-ETS will have a significant effect on the viability of co-firing in the UK or elsewhere. Table 3. ELVs for large combustion plant (>300 MWth)
Fuel Type SO2 NOx Dust Ref. O2 dry Solid 400 500 50 6% Existing1 Liquid 400 400 50 3% New Liquid 200 200 30 3%
Gas 35 200 5 3%
Solid 200 200 30 6%
Gas 35 100 5 3%
Valid until 31 December 2015
Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal Table 4. Calculated example of carbon price needed for biomass co-firing to be competitive
Coal Cost of fuel () Carbon Price (/tCO2) Carbon Emission (tCO2/MWh) Carbon price needed for biomass co-firing to be competitive () 20.11 11.85 0.85 Biomass 37.36 11.85 0 20.3 Data Source DTI. 2006d. Scottish Executive. 2004 http://www.pointcarbon.com Commission of the European Communities. 2006
17
Table 5. The cumulative ROCs issued for biomass co-firing for all of the large coal-fired power plants in Britain [5]
Station Aberthaw Alcan Lynmouth Cockenzie Cottam Didcot Drax Eggborough Ferrybridge Fiddlers Ferry Ironbridge Kingsnorth Longannet Ratcliffe Rugeley Tilbury West Burton Capacity (MWe) 1,455 420 1,200 2,000 2,100 4,000 1,960 2,035 1,995 970 2,034 2,400 2,010 1,000 1,085 1,980 Generator RWE npower Alcan Scottish Power Electricity de France RWE npower Drax Power British Energy SSE SSE E.on UK E.on UK Scottish Power E.on UK International Power RWE npower Electricity de France Total ROCs/MWh Cumulative ROCs 91,290 45,127 25,958 124,431 50,303 748,527 140,778 1,246,550 606,864 135,740 237,113 269,915 37,501 252,085 25,946 64,677 4,102,805
Another implication of the legislative framework which is enforced in UK and it is worthwhile to be mentioned is the Renewables Obligation (RO) [22]. The RO was established in 2002 and constituted the base for the development of co-firing on a commercial scale in the country. This is a system of tradeable permits in which electricity suppliers are obliged to produce Renewables Obligation Certificates (ROCs2) to guarantee that an increasing proportion of the total electricity supply was generated from renewable sources. The level of
2
A ROC is equivalent to 1MWh of electricity generated of renewable sources
18
the RO began at 3% in 2002/3 rising to 10% by 2010 and has been extended to 15% by 2015. One of the results of the introduction of the RO has been the increase in biomass co-firing involving all of the large coal-fired power plants. A list of the cumulative ROCs issued, up to February 2006, is presented in the following Table 5. Apparently, the electricity supply industry has responded rapidly to the financial incentives, such as the extra source of income from the ROC value.
3.4. Financial Evaluation

In general, co-firing biomass in a coal-fired power plant is a cost-effective implementation as it requires a relatively small capital, while offering simultaneously immediate benefits to the power plant operators and the inhabitants of the region. However, the profits available to the company are defined to a large extent by some key parameters [6], such as: The delivered price and availability of the solid bio-fuels The market value of possible subsidies named as green energy generation The costs of the pre-blending route option
From an economic point of view, in order for biomass to be more attractive as a substitute fuel it has to be first and foremost cheaper than the fossil fuel it displaces. It is clear that the market value of biomass materials is strongly influenced by the type, source, availability and storage/transport requirements. Biomass in general has too low an intrinsic value to justify transportation over long distances. This acts as an economic barrier to cofiring, if biomass materials are not in close proximity to the coal-fired power plants. In Figure 1, the current final costs for the biomass fuels in UK are presented. The transport costs, which were estimated assuming a collection radius of 40 km around the power plant, play a large part in the final market value of the fuels as a result of the smaller volumes compared to mass transport [8]. Table 6. Energy cost of a number of solid bio-fuels [3], [23]
Fuel Coal Forestry residues Cereal (straw) Olive kernels Sawdust pellets Energy cost-UK (/GJ) 1.8 2.7 3 3-3.7 5.2-6.7 Energy cost -GR (/GJ) 2.23 3 n.a 3.75 4.48 0.97 4
3 4
Brown coal Considering a heating value of 17.28 GJ/tn and fuel cost of 16.7 /tn
19
Figure 1. Final costs of biomass fuels in UK, Source DTI-2006.
Furthermore, seasonal differences and risks resulting from climatic influences affect the availability and price of solid bio-fuels. Also, the involved pre-processing conditions of the delivered biomass to power plants have their own share in the price. For example, in UK these prices are in the order of 3 - 7.5 /GJ, with agricultural residues lying at the bottom of this range and highly processed (dried and pelletised) biomass materials at the top, respectively. In Table 6, the delivered costs of a number of relevant biomass materials both in UK and a Mediterranean country, Greece, are given. Moreover, the dispersed resource of indigenous bio-fuels can have adverse effects in the price. In the case of coal, the supply is easy and can be guaranteed to maintain continuous operation. In addition to the above, the coal-fired power plants have historically been sited at convenient locations for the receipt of coal. On the other hand, biomass has low bulk energy density which exaggerates the transport costs and the price itself. Thus, the intention of applying biomass co-firing in existing energy plants should be sized according to locally available biomass resources or within reasonably transport distances. In cases of increased market value of biomass due to increased utilization or seasonal shortage, the coal-fired power plants have the possibility to temporarily suspend the co-firing activities and minimize the fuel costs. At this point, environmental regulation authorities have to play a significant role, regulating a ceiling on the biomass price and keeping the same distance between the energy sector and the solid bio-fuels suppliers. In the section related to legislative framework, the connection between the legislative framework and the economics was emphasized. The relation was illustrated through a calculated example of the carbon price required to consider biomass co-firing a competitive alternative. Such renewable certifications, like ROCs in UK, or other subsidies are in a strong place to affect profitably the biomass co-firing implementations.
20
Generally, costs for co-firing projects applied in existing power plants are lower compared to those of new dedicated biomass plants [23], [24]. The pre-blending route option comprises the pre-blending of biomass with coal and the processing of the blend through the existing milling equipment followed by the feeding of the mixture into the bunkers. The capital cost is relatively modest due to the already existing infrastructure. However, there are other capital expenditure requirements which depend on the nature of the solid bio-fuel, the arrangements needed for the reception and the on-site storage of it, the blending facilities and other possible additional modifications. Specifically, the associated systems include the biomass reception area, the roads and the turning areas for the delivery trucks, the storage hall and hoppers, the feeders and conveyors. Also, there are the development costs which are associated with more paper-works and their respective costs. The selection and securing of biomass supply, the performance of the necessary co-firing test run and the preparation of the application for the authorization to permit commercial operation, are some of the above mentioned. Finally, a part of the operating costs of co-firing activities are mainly associated with the additional direct labor and contract services costs. These are highly variable and dependent on site-specific factors. In the end, the energy utilization of biomass must secure a premium price for the electricity produced in terms of power generation. The cost-savings or revenue-increase benefits must be sufficient to cover the cost of capital for additional infrastructures and risk involved in the technical retrofitting of the plant to accommodate solid bio-fuels. The absence of funding often acts as a financial barrier to co-firing applications, which always require capital investments. Namely, the specific barriers are due to the risk-perception by the suppliers of funding. In general terms, there are two types of financiers, the external commercial financiers and the internal plant operators. Both of them require minimisation of the risk, the first on the repayment of their loans and the latter for their revenue generating operations. As discussed, fuel supply, technology and economics are the major sources of risk. It should be stressed out that a good relation of these factors is essential in order to find funding for a co-utilisation action.
3.5. Societal Issues

Although co-firing is adopted in order to alleviate the environmental issues that fossil fuel power plants create, its implementation may face reactions from the public. The nature and the intensity of these objections usually depend on the location of the plant and the type of supplementary fuel used. For biomass, which is considered a green fuel, the public objections are likely to be less intense and rather enthusiastic. The opposite reaction can arise if the intention is co-utilisation with waste-derived fuels, even for the biodegradable fraction of them. Thus, the use of public informing campaigns for the environmental positive impacts of co-firing technology is beneficial [23].
21
4. REVIEW ON CO-FIRING EXPERIENCE AT EUROPEAN LEVEL

4.1. Current Status
There has been remarkably rapid progress over the past 5-10 years in the development of the co-utilization of biomass materials in coal-fired boiler plants. Several plants have been retrofitted for demonstration purposes, while new plants are already being designed for involving biomass co-utilization with fossil fuels. A study on the application of co-firing worldwide [25], [26] shows that over 80 coal-fired plants have experienced co-firing biomass or waste. Typical power stations where co-firing is applied are in the range from approximately 50 MWe to 700 MWe. The majority are equipped with pulverized coal boilers (tangentially fired, front-wall fired, back-wall fired, dual-wall fired). Tests have been performed with every commercially significant coal type (lignite, sub-bituminous coal, bituminous coal) and every major category of biomass (herbaceous and woody fuel types generated as residues) [1]. However, the final selection of the fuel type used for energy purposes is strongly depended on its availability in the region. Namely, in Northern European countries, such as Finland, Sweden and Denmark, a large amount of small scale biomass-based combustion steam boiler plants for process and district heating purposes are in operation. This is due to the fact that these countries have a large amount of low cost biomass which can be used for primary fuel for heat or combined heat and power production. This is accomplished with relatively simple technical ways. On the other hand, in Central European countries such as Germany, The Netherlands and Belgium biomass-based energy conversion systems are much less applied due to the high availability of indigenous fossil fuels resources. Furthermore, the corresponding biomass quantities in this region are limited, so that whenever these countries want to use the latter as substitute or main fuel, they depend on the use of a variety of relatively more expensive domestic bio-fuels or imported energy crops. Thus, the implementation of biomass-based energy conversion systems is more difficult both from a technical and economic point of view. However, as a consequence of CO2 emission reduction targets, renewable energy policies and tax incentives imposed by the national governments, the energy utilization of biomass is a major topic for the power production sector. Because of the existing energy infrastructure in these countries the attention is mainly focused on direct and indirect co-firing in already existing coal-fired power plants. Several significant case studies of co-firing coal with solid bio-fuels accomplished in European countries are reviewed briefly below.
4.2. Demonstration of Co-firing at Commercial Level in Finland

The Case Study of Naantali-3 Power Plant Co-firing tests with sawdust and coal have been conducted at FORTUMs Naantali-3 CHP power plant. The Naantali-3 is a tangentially-fired pulverized-coal unit with a Sulzer once-through boiler that produces 79 MW electricity, 124 MW district heat and 70 MW steam. The basic fuel, Russian coal, is pulverized in three mills (Loesche roller mills). Each of them feeds one burner level which consists of four burners. The boiler is also equipped
22
with low-NOx burners (IVO RI-JET) and over-fire air (OFA) system. The system for the fuel gas treatment includes electrostatic precipitators and a flue-gas desulphurization facility. The co-firing tests were carried out for three months during April 1999 to April 2000 period and without any investments. The solid bio-fuel used in the tests was pine sawdust, which was received from the saw mill, located about 50 Km from the power plant. The proportion of the supplementary fuel in the mixture started from 2.5% and came up to 8% of fuel input. The coal to Naantali was delivered by sea. The blending of coal and sawdust was accomplished in the coal yard. Then, the mixture was loaded to the hopper by a bulldozer and farther to the existing coal silos by the belt conveyors. During this process, the behavior of the blend was normal without self-ignitions problems. Afterwards, it was fed into the boiler via the existing coal mills. The behavior of the blend in the coal mills caused some problems. The simultaneous grinding of the fuels in the roller mills affected negatively the coal fineness result. Specifically, the amount of bigger particles was increased while the corresponding of the smaller ones reduced. Both had adverse side effects on the burn-out efficiency of the coal and on the boilers efficiency or on the quality of fly ash, successively. When coal was cofired with sawdust and if the latters proportion was high, the grinding of the coal was worsened, increasing the amount of unburnt carbon in the fly ash. Furthermore, increased smoke formation in the coal mills was noticed during the co-firing tests. This was probably due to the roller mills are not designed to handle fibrous materials like the wood sticks and chips, which came with the sawdust. Thus, the bigger particles were not ground in the mills but dried and flattened. Then they met with the primary hot air and caused smoke formations. Therefore, the simultaneous feed is not a suitable solution for long-term use. Concerning the combustion behavior, the ignition and the flame stability were normal, even with a high substitute percentage, due to the special burners used in the plant. All the values which affect the heat transfer in the furnace plus the slagging and fouling behavior were at normal levels. The SO2 emissions were lower in co-firing than in the sole coal combustion, due to the higher sulphur content of the coal compared to biomass. No important changes occurred in the ESP performance, while the dust emissions were lower during the cofiring process. The latter is probably connected with the low concentration of ash in the sawdust. In general, the co-firing tests at Naantali were successful. However, Fortum concentrated on developing a new concept for co-firing, in order to solve the problems encountered with the coal mills. Generally, co-firing a solid bio-fuel in existing pulverized-coal fired boilers sets high demands for its grinding technology. In order to achieve the desirable size of the biomass, a complex and expensive micro-milling system is needed. However, according to the gained experienced from the full-scale co-firing tests, some new ideas arose. It was decided that the existing simple grinding system was sufficient provided that modern burners are to be used. The new co-firing concept is shown in the Figure 2. According to this, the biomass is screened and ground in the crusher process and the grinded biomass particles are fed to the boiler through special bio-coal burners (RI-BIO). There two possible alternatives for the mixing of biomass and coal. In the first case the ground biomass flow can be introduced to the coal flow near the burner area, so that the coal and biomass are mixed before they enter the furnace. In the latter case the biomass can be fed to the furnace via a special centre pipe so that this and coal particles are mixed in the flame. Applying this concept, it is possible to substitute 5 30% (by heat) coal by solid bio-fuels. [27]
23
Figure 2. Fortum's new concept for the direct co-firing of bio-fuel and coal.
Figure 3. Direct co-firing of straw with coal in Studstrup Power Plant.
24
Panagiotis Grammelis Table 7. Danish power plants experienced co-firing
Plant name Studstrupvaerket 1 Studstrupvaerket 4 Amager 3 Esbjergvrket
Location Aarhus Aarhus Amager Esjberg
Output (MWe) 150 350 250 150
% substitution 20% heat 20% heat Up to 20% heat Up to 25%
Status Commercial (1996-1998) Commercial (since 2001) n.a n.a
Primary fuel Pulverized coal Pulverized coal Pulverized coal Pulverized coal
Suppl. Fuels straw straw straw / miscanthus straw /wood
4.3. Demonstration of Co-firing at Commercial Level in Denmark

Major projects aiming at large scale commercial or semi-commercial operation are under way in Denmark. These projects are important parts of the energy and environmental policies of the country. Some of these demonstrations of co-firing tests are reported in Table 7.
The Case Study of Units 1 and 4 at Studstrup Power Plant Two of the above mentioned co-firing activities have been conducted at Units 1 and 4 at Studstrup Power plant. The latter is located near Aarhus in Jutland and operated by Elsam. Based on the experience gained over a number of previous projects, Danish engineers converted the 150 MWe installed capacity pulverized coal-fired Unit 1 to straw co-firing. A diagram of the system, including the handling and the pre-processing equipment, is presented in the Figure 3. The system was installed and commissioned during 1994-5. A demonstration program started in 1996, focusing on the performance of the straw-handling and firing system, the boiler performance, the slagging / fouling and residue characteristics, and the pilot testing of the system for the flue gas treatment (DeSOx and DeNOx equipment). This involved long-term operation at 10 and 20% straw on a heat input basis. The option followed in this case was the direct injection of biomass into the boiler through the existing coal burners. The principle of that process is thoroughly described below. Initially, the straw bales were delivered to the Station by lorries. An automatic overhead crane unloaded and placed the bales in the store or feed the two-tier conveyor system. The crane was also equipped with load cells and microwave pads for the automatic recording of the weight and the moisture content of each bale. Afterwards, the two-tier conveyor and transport system automatically feed four straw pre-processing lines. Upon entry to these, the bale cord was cut and removed. The loose bale then fell into a size-reducing unit. This had a cutting table with two shafts fitted with a number of curved blades that broke up the straw bale and reduced originally the straw to a size of 100 to 150 mm. The shattered straw was then dropped onto an inclined belt conveyor. This helped to balance the flow of straw and the speed of the belt to control the straw feed rate to the burner. From the top of the belt, the straw was dropped to an air classifier, which removed any undesirable material that may have been included in the bales. Then, a short pneumatic system conveyed the straw into the hammer mill. The hammer mill farther reduced the straw size depending on the open in the
25
outlet screen. The screens used at Studstrup have diameters size of 50 and 75 mm. From the hammer mill outlet, the ground straw dropped into a collection bin, which was fitted with an aspiration chamber and a rotary airlock. This dropped the straw into the pneumatic conveying system to the burners. Twelve pulverized fuel (PF) burners existed into the boiler, which were arranged in three rows of four on the furnace back wall. The boiler produced 139 kg/s of steam at 540C and 143 bar. The burners that had been converted to allow straw-coal cofiring were those of the centre row. The latter was achieved with the ensuing arrangement. The ground biomass was arrived at the burner at a primary air velocity of 25 m/s and reduced to 15 m/s at the burner. In sequence, the straw was fired down the burner core air tube, with the coal being fired through the primary air annulus. With this settlement, it was possible to co-fire straw at up to 20% of the total furnace heat input at full boiler design load. Despite the successful operation of Unit 1, some problems related to the handling and the conveyance of the straw was encountered. The wet nature of the solid bio-fuel created inconveniencies at the mentioned system. Namely, the system was sensitive to relatively small quantities of straw with moisture content in excess of 25%. In order to set out this unfavorable situation, an improvement of the quality and consistency of the delivered fuel is needed. This can be accomplished with a better collaboration with the straw providers so as to secure a long term quality controlled biomass supply scheme. Concerning the combustion system, its overall performance was good, as long as the straw injection velocity through the burner was below 15 m/s. In addition, a good burnout was achieved. No particular problems associated with ash deposition, high temperature corrosion, or the environmental performance of the plant were identified at straw co-firing rates up to 20% on a heat input basis [28]. Particularly, the positive effect of biomass addition on NOx emissions were confirmed by reduced levels compared to operation on coal alone. An increase in HCl emissions was noted though, due to the higher input of chlorine contained in the fuel mixture. After the deployment of co-firing technology in Unit 1 from 1996 to 1998, there was the intention to convert to straw co-firing an additional Unit of Studstrup Power Station. This was the Unit 4, a 350 MWe boiler. When the problems concerning the utilization of the mixed ash as a component of cement and concrete product were solved in 2001, Unit 4 was converted to straw in the same way. As of January 2002, Studstrup Power plant was capable of co-firing 250,000 tn of straw plus 100,000 tn of other biomass fuels (maize, sunflower seed husks) with coal, annually.
The Case Study of Unit 1 at Esjberg Power Plant The same concept of the direct injection of the supplementary bio-fuel into the boiler via dedicated biomass burners was followed at another Danish hard coal-fired Power Plant as well. The co-firing tests of straw and wood with coal were performed in the 150 MWe pulverized coal Unit 1 of Elsam at Esjberg. The approximately 15,000 tn of baled straw were fed in two trains to a shredder, followed by a cutter. The density of biomass was 500 kg/m3, the heating value amounted to 19 MJ/kg and the ash content to 0.45%. The corresponding quantities of coal were 63,000 tn. Its heating value ranged from 24.8 to 26.9 MJ/kg, while the ash content was set in high levels of 18%. The feeding system presented difficulties and operated well only when the moisture content of the solid bio-fuel was held below 15%.Then, the straw was transported with air to separate burners into the boiler. The structural modification which discriminated these burners from those of coal was the fixed cones instead of impellers to spread the injected straw. The ensuing configuration included two
26
straw burners in conjunction with ten coal burners, which were installed above and to either side of the straw burners. The aim of the above arrangement was to moderate the air flow and avoid the collision of straw particles with the boiler wall. Another inhibitory factor was the foreign material in the straw stream which can cause breakdowns. Concerning boiler operation, there were no deposits on boiler tubes, superheaters and preheaters. Although slag accumulated on straw burners, the deposits were easily removed. Moreover, the slagging / fouling and corrosion effects were of particular interest to the utility due to the higher levels of both chlorine and potassium presented in the straw, compared to coal. Also, the boiler load was typically 80% of rating. Finally, indications of unburned straw and carbon in the slag ash for co-firing proportions in excess of 10% by heat obstructed the mixed ash to be utilized as raw material in the cement sector [15].
The Case Study of Unit 3 at Amager Power Plant An additional co-firing test in the country was carried out at the Amager Power Plant Unit 3. The wall-fired pulverized coal Unit 3 was chosen to substitute its main fuel, hard coal, used for the production of 250 MWe with 20% straw by thermal energy. The biomass was delivered in forms of cylindrical pellets of approximately 9% moisture content so as to avoid the need of shredding. The concept followed was that of the mixing of biomass with coal prior to entering the mills for co-grinding. In the boiler itself, the combustion process was performed well and sustained in stable levels, while no depositions were observed in the heat transfer surfaces. The Flue Gas Desulphurisation system operated with no problems except for a decrease in the limestone reactivity [15].
4.4. Demonstration of Co-firing at Commercial Level in Germany

In Germany, the demand for alternative waste treatment concepts, the CO2 reduction potential coupled with savings in fossil fuels resources are currently the issues of the greatest interest. One problem arising is the suitability of existing power plants for the co-firing of biomass and / or waste. The locations and additional information about the respective power plants are listed in Table 8. The main achievements gained so far from the accumulated experience in German coal /lignite fired power plants in using supplementary fuels are summarized as follows [2]: The fuel supply has to be scrutinized as regards the transport costs and the availability of the supplementary fuels. The former is referred to the long distances between the locations of the power plant and the source of the additional fuels. The latter is referred to the high local concentration of potential users. The fact that existing facilities primarily designed for coal are used for the preparation and the handling of the supplementary fuels limits their share on the thermal load. However, by appropriate process settings any differences of the fuel characteristics are acceptable. Quality standards for the fuel suppliers are necessary in order to confine the concerns that induce the nature and the characteristics of the fuels themselves.
Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal Table 8. German power plants experienced co-firing
Plant name Location Afferde El.Werke Wesertal Bayernwerke AG Bavaria Berrenrath Rheinbraun Heilbronn EnBW Janschwalde VEAG Lubbenau VEAG Magdeburg VEAG Moabit BEWAG Saarberg Saaberwerke AG Schwandorf Bayerwerke B Schwandorf Bayerwerke D Weisweiler RWE Firing System Fluidized bed PF (direct) Fluidized bed Output Output (MWe) (MWth) n.a 124 % heat % mass Status Operation since 1997 n.a 1999 tests Feasibility study 1999 2002 Tests 1993 tests Primary fuel Hard coal Suppl. Fuels Wood, straw straw wood Wood straw Wood straw Wood straw wood Wood straw
27
Up to n.a 10% n.a n.a
108 n.a
n.a 235 1933 n.a n.a
Tests with approx.8t/h 8% n.a n.a n.a n.a 7%
Pulverized coal Brown coal Hard coal Brown coal Brown coal coal Hard / Brown coal Pulverized coal Brown coal Brown coal Brown coal
Dry bottom n.a Dry bottom n.a PF (direct) 100 Wall fired Dry bottom PF (direct) 350 Fluidized bed PF (direct) n.a
n.a 240
n.a 1 to 13% n.a
n.a n.a
n.a 1995 tests
75
n.a
n.a
n.a
Sewage sludge Straw pellets Wood chips SRF /paper pulp
Grate n.a (direct) Dry bottom Grate 316 (direct) Dry bottom PF (direct) 2 x Dry bottom 600
280
Up to n.a 20% 12 8
1996 tests
n.a
Operation since 6/1999 Cont.operati on starting 2000
n.a
2 to 8%
n.a
As regards the corrosion and slagging of the boilers, for co-firing at ratios less than 10% in a thermal heat basis no problems appear. Exceptions arise in the case of fuels with high chlorine content and for boilers with supercritical characteristics. Changes in the flue gas composition show a direct correlation to the fuel analysis. Higher concentrations of SO2 or HCl are separated in the flue gas purification system. The NOx emissions which are strongly dependent on the combustion conditions and the fuel compositions decrease during co-firing of straw and wood due to the reducing effect of the high volatile release of the additional fuels. The amounts of trace elements such as mercury are lower in supplementary fuels than in brown coal. Hence, no increased emissions for them were observed. This can
28
Panagiotis Grammelis be also attributed to the small proportions of the additional fuels or to the detection limits of available measuring techniques. With respect to dioxins and furans, no negative effects of the co-firing were observed. This can be attributed to the destruction of all the organic intermediates, due to both the high temperatures in the flame zone and the sufficient oxygen availability. Overall, the experience gained from the co-firing tests has shown that the quality of the co-firing residues are not altered , except for fly ashes of straw co-firing at high thermal shares
Two cases studies for co-firing biomass and waste, respectively, are described below:
The Case Study of VEAGs Lubbenau Power Plant The co-firing tests were conducted at VEAGs Lubbenau Units with installed capacity of 100 MWe. The test program involved co-firing 7% (by mass) of pine wood chips with lignite using two boilers pulverized fuel Units. According to the co-firing results, a single recycle of wet ash was needed so as to reduce the unburned carbon in the coarse ash. On the other hand, the fly ash showed no additional carbon even without this recycling. The route of recycling comprised the addition to the fuel before it reached the feed mills. When this recycling was applied the CO emissions remained within the limits set by German environmental legislation. Considering the combustion process, it carried out successfully even when using 10,000 tn of wood chips [15]. The Case Study of Weisweiler Power Plant A summarized description of the co-firing application in a German lignite-fired power plant is given farther down. The full-scale co-combustion trials took place in the RWE Powers site at Weisweiler, which is located 50 km westwards from Cologne in the Rheinish brown coal area. The Weisweiler site consists of six units with total installed capacity of 2,100 MWe. The trials took place in Units G and H, with a nominal capacity of 600 MWe each, which already utilize paper sludge for co-combustion, Figure 4. The substitute fuel is produced by REMONDIS; more specifically it is the solid recovered fuel (SBS Substitutbrennstoff in German nomenclature) coming from the high calorific fractions of municipal solid wastes via positive sorting methods. The co-firing tests in Weisweiler were carried out in two stages. During the pre-trials in July 2004, the focus was mainly on fuel handling, conveying and feeding issues. The share of SBS to the thermal input started from 2% and was increased up to 8.5%. The results of them showed no severe operational problems. Namely, the unloading of the fluffy alternative fuel was successful due to the good weather conditions. In the case of stronger winds, the construction of a windshield could prevent the low density fuel to be spread around. The feeding process to the boilers was accomplished through the conveyor belts together with lignite, that followed the coal bunkers where the SBS from the unloading area was first transported. Some blockage of SBS in parts of the transportation parts and some layers from metal aluminium and aluminium oxides on the second row of pre-beater arms in the coal mills were observed. The low density of SBS was to be blamed for the first problem, while the high content of metal aluminium in it created the appearance of the aluminium deposits.
29
On the contrary, a small increase in power production was observed, owing to the higher heating value of the fuel mixture. Moreover, no significant changes in the flue gas emissions were noticed. Finally, the amount of the unburnt SBS in the wet bottom ash was negligible [29]. The share of SBS was kept to the level of 2% of the thermal capacity and doubled to 4% for specific time periods during the main combustion trials. The total (Unit G and H) feeding rates of SBS fluctuated from 21 to 24 tn per hour at the first thermal share and 40 to 46 at the latter. The feeding rate of paper sludge and lignite came up to 54 tn per hour and 750 tn per hour, respectively. The total feedstock consumption was 4,200 tn of SBS, 13,300 tn of paper sludge and 345,000 tn of lignite. A special unloading station has been built at the Weisweiler site for the co-firing of paper sludge. It consists of a double 180m3 bunker and a single 90m3 one together with two intermediate storage areas summing up to 4,000 tn. The feeding capacity amounts to 125 m3/h of each of the three screw feeders resulting to a feeding capacity of 62.5 tn per hour for paper sludge or 25 tn/h for SBS. The existing infrastructure could be successfully applied during the SBS co-firing tests. The measurements were performed at two different levels at the furnace exit and in front of the air preheater- and mode -lignite / paper sludge operation and co-firing with SBS mode-. The mixing of the fuels took place into the coal conveyor belt before the pre-crushers. In case to secure a permanent alternative operation, a separate pneumatic system for the SBS transportation with an unloading station is needed. The accrued conclusions of the tests are summarized below. Concerning the combustion process and the overall power plant behavior, the operation of the boiler, namely parameters such as steam production and temperatures, was not influenced. As far as emissions are concerned, CO levels were slightly increased while SO2 emissions could not be detected, nor dioxins/ furans nor dust and NOx were slightly lower. Also, no unburnt SBS in the ash was found. Though, HCl concentrations were observed before the air preheater, they were not monitored in the stack measurements after the FGD system. As regards corrosion issues, the experimental data indicated no clear influence of SBS co-firing. Relating to the quality of the residues, the results of the leaching values of the mixed fly and bottom ash did not exceed the permission limits and it is suitable for disposal. [30]
4.5. Demonstration of Co-firing at Commercial Level in the Netherlands

The main goal of Dutch Renewable Energy Policy is that in 2020 10% of the total energy consumption has to be provided by renewable energy sources. Further to that, in order to meet the CO2-reduction goal (Kyoto Protocol), the fossil fuel related CO2 emissions of the Dutch Power Plants have to be decreased significantly. Based on the agreement drawn up between the power production sector and the Dutch government, a large part (6 out of 25 Mtn CO2 equivalents) of this target has to be realized at coal-fired power plants [31]. Co-firing of biomass in existing coal-fired power plants along with other activities (efficiency improvements, fuel switching to natural gas) is expected to contribute substantially to this target. Moreover, it is already applied in many Dutch coal-fired plants, due to the high availability of solid bio-fuels. The co-firing percentage is still less than 5% by thermal heat input. A list of these applications that had been successfully demonstrated at commercial or pilot scale in units of up to 645 MWe is given in Table 9. The direct co-firing is the preferable
30
concept, due to the high electric conversion efficiency and low additional investment costs. Furthermore, it is not expected to cause any problems concerning the air emission regulations of the EU and the quality of produced solid byproducts. The potential high-temperature corrosion of heat transfer surfaces and slagging or fouling problems are among the most important technical constraints. Although emissions of dust particles and heavy metals may be problematic in certain cases, these problems have to be assessed experimentally for a variety of biomass materials.
Figure 4. SRF/ paper sludge supply system to the boiler, Source: RWE Power AG.
Table 9. Dutch power plants experienced co-firing

Plant name/ Location Gelderland / Nijmegen Borssele 12 Owner EPON EPZ (=Essent) Output (MWe)net 602 403 % substitution 4.5% heat n.a Status commercial commercial Primary fuel Coal Coal Suppl. Fuels wood Kernels, paper sludge, shells, fibers Paper sludge Biomass pellets
Amer 8/ Geertruidenberg Massvlakte1/2,Rotterdam
EPZ (=Essent) EZH (Germany)
645 2 x 518
0.8% heat 5% heat 4% heat
commercial commercial tested
Coal Coal
31
Figure 5. The wood handling system at the Gelderland Power Plant.
The Case Study of Gelderland-13 Power Plant Co-firing was demonstrated at Gelderland Power Plant in Nijmegen during 1995 and is now in full commercial operation. The selected Unit was a 602 MWe (net electrical capacity) pulverized coal-fired boiler, which was converted to burn almost 60,000 tn/yr of processed waste wood. Moreover, the Unit is equipped with a wet limestone FGD system and a SCR system for SOx and NOx emission control respectively. The goal of this conversion was the utilization of waste wood as boiler fuel so as to avoid the greenhouse gas released into the environment during their decomposition at the landfill sites. The co-firing concept is based on the separate size reduction and drying of the supplementary fuel. The waste wood material is collected and processed first into raw wood chips, which are then delivered to the power plant. The wood handling system at the Gelderland Power Plant is shown in Figure 5. According to this, the wood chips are unloaded in the reception area and conveyed to the grinding area. Magnetic separation and air classification equipment are employed to provide additional cleaning of the wood chips. The chips are then fed to a hammer mill, which reduces them to a top size of 4 mm. After screening out the fines, the oversize material is sent to the milling plant. The mill product and the fines from the hammer mill are combined and powdered in a wood-handling system. There are two mills, each with a capability of producing around 1.8 tn/h of final mill product, whose moisture content is below 8% in dry basis. The powdered wood is conveyed pneumatically to a storage silo, of 1,000 m3 capacity, located adjacent to the boiler. A powdered wood metering system delivers the fuel, via air blowers, to four burner injection systems, each capable of delivering 1.1-3.5 tn/h of fuel. The four burners, each rated at 20 MWth, are located in the side-walls of the furnace. The boiler furnace is opposed-wall-fired with three rows of six burners in both the front and back walls. The wood burners and the wood combustion control system were independent of the coal-
32
firing system, and there was no interference with the boilers capability to burn coal at full load. The wood-firing capacity mounted to 10 tn per hour, which equals to 3-4% of the heat input [28]. At this co-firing ratio the impacts on the operation, the environmental performance, and the availability of the boiler have been small. The coal substitution for the used quantities of biomass mentioned above come up to 45,000 tn annually. Concurrently, the corresponding annual CO2-emission reduction is 110,000 tn. On the other hand, the investment costs, mainly for the biomass handling and feeding system, drew near to 13.6 M [31].
The Case Study of Borssele-12 (BS-12) Power Plant The co-firing test was carried out at one of Essents Power Plants, known as Borssele-12. The BS-12 is a tangentially-fired pulverized-coal unit, a view of which is depicted in Figure 6. The boiler produced 365 kg/s of steam at 543C and 183 bar. The supplementary fuel tested was processed waste wood, including olive residues, palm kernels, citrus pellets and friable wood. One of the most significant parameters that affect the co-firing activity is the rate of biomass grindability. According to experiments, the particle size distribution of three different fuels after milling showed size variations even in the own fuel. Specifically, a proportion of 70% of ground coal particles sized from 0 to 0.2mm. On the other hand, only the 35% of milled palm kernels had a top size of 0.5mm, while wood presented a 40% percentage of its particles having diameter of 0.2 up to 0.9mm. Other prerequisites of fuels for co-firing are the volatile content and self-ignition temperature in order to ward off fires in potential hot primary air environment. A high silicon content of the fuels can create congregations on the rotating pulverizer table, as well. Moreover, the ratio of weight to surface area of the pulverized bio-fuel powder determines classifying at the pulverize outlet. Essent Energy proposed four options so as to apply co-firing at Borssele-12. One of them was accomplished via the coal yard, where coal and wood were disposed. Mixing of fuels was carried out on the conveyor belt and the mixture headed toward the boilers. Afterwards, new covered storage was constructed for the biomass with sufficient room in four separate bunkers. The other two options included the presence of separate silos and the conveyance of biomass via the central feeding pipe of the coal pulverizers to the boiler or via separate hammermills and crumblers to dedicated biomass burners in the boiler. The fourth option which was founded at BS-12, the so-called Multi-fuel co-firing installation, is presented in Figure 7. The storage of biomass took place at four silos that lifted close to the boiler. Each section of the silos were transported by trucks from port and assembled at the power plant. After the installation was turned over, it was waiting the delivery of the solid bio-fuels. The biomass was transported to the boiler via enclosed chain conveyors [32]. An additional initiative to be executed in the short term constitutes the direct co-firing of 6,000tn dry base sewage sludge. The increasing of the total c-firing potential to 120,000 tn per year partly by sewage sludge and 7,000 tn/yr CO2-reduction are some of the most important results. The corresponding investment cost comes up to 1.2 M [31].
33
Figure 6. The tangential fired boiler as is in Borssele-12.
Figure 7. Multi-fuel co-firing installation at Borssele-12.
34
Figure 8. Flow chart of the direct co-firing concept at Amer-8.
The Case Study of Amer-8 Power Plant A direct co-firing activity was performed at Unit Amer-8 as well. The number 8 unit of the Amer power plant of Essent in Geertruidenberg has a total capacity of 645 MWe. The unit has been co-firing 75,000 tn/year of paper sludge commercially for a number of years. Thereby, it substituted almost 4,500 tn/year of its main fuel, coal. This represents 0.8% of the heat input to the plant. In addition the fossil fuel substitution leads to the mitigation of 11,000 tn per year CO2 emissions [30]. The co-firing concept is presented in Figure 8. No adverse operational problems were noticed. Since 2003, one of the coal mills has been converted to grind wood pellets. The Case Study of Maasvlakte-1/2 Power Plant The pulverized coal power plant of EZH near Rotterdam has been applying direct cofiring activities in the Units 1 and 2 for several years. Each of the Units has a net electrical capacity of 518 MWe. Maasvlakte 1 is utilized 150,000 tn per year of biomass pellets. The pellets are produced at the power plant site from a mixture of paper sludge, wood residues, and woody particles remaining from a compost sieve overflow for a number of years. Namely, the solid bio-fuels are composed of 60% paper/cardboard, 24% waste wood and 16% compost. This equals about 5% of energy input. Biomass co-firing with the main fuel results in the substitution of 30,000 tn coal per year. This corresponds to the reduction of 78,000 tn CO2 per year. On the other hand, the investment cost comes up to 7 M. Aiming to increase the co-firing capacity, Maasvlakte Power Plant increased the biomass quantity to 288,000 tn per year. This implies coal substitution 10% by weight. Further, improving the quality of bio-pellets by pre-drying makes an advance on the co-firing capacity. Also, direct co-firing of 40,000 tn per year of poultry litter (4% of heat input) is
35
tested at Maasvlakte 2. This equals about 4% of energy input, with a coal substitution of 23,000 tn per year. The corresponding CO2-emission reduction amounts to 60,000 tn per year. [31]
4.6. Demonstration of Co-firing Tests at Commercial Level in UK

At the present time, the great majority of large coal-fired power plants are pursuing cofiring projects. The current situation is summarized in Table 10. The number of coal plants co-firing has increased following the introduction of the Renewables Obligation, from two in 2002/03 to sixteen in 2005/06. [8] Most of the UK coal-fired Power Plants are co-firing biomass materials on a commercial basis by pre-mixing the biomass with the coal, either on or off-site. In general terms, the biomass co-firing projects at the power plants have been reasonably successful, both technically and environmentally. On the other hand, the majority of the problems were associated with the fuel supply arrangements, and with the on-site storage and handling of the biomass materials. Some operational [9] problems occurred when co-milling biomass with high volatile content. This caused fire hazards when hot primary air was used. Also, in some cases, the high moisture content of biomass disturbed the mill heat balance and the larger biomass particles left the mill or increased mill power consumption and pressure drop. Some demonstrations of co-firing tests are described [6] below. The specific co-firing option adopted at each power plant depends on the geographical location, on the security of supply of suitable quantities of bio-fuels, either imported or from indigenous sources. Also, the majority of the power plant operators are focused almost exclusively on the ROCs income. After the success of the co-firing tests, the electricity supply industry proceeded to commercial co-firing of biomass materials, afterwards.
The Case Study of Tilbury Power Plant A co-firing test was conducted at Tilbury coal-fired Power Plant, with installed capacity of approximately 1.085 MWe, located in the Thames Estuary. The supplementary fuel was sawdust, a high grade biomass material in a pelletised and dried form that was delivered to the energy plant by sea at a price of 6 /GJ. During testing, the thermal share of solid bio-fuel in the overall thermal input was adjusted to 3%. The installation of a separate unloading, storage and conveying system was required for the biomass. Moreover, the co-firing option applied involved the installation of a new, dedicated biomass mill under the existing (redundant) bunker and the supply of the milled biomass through a pipework system to one of the existing coal burner rows. For this reason the capital expenditure was significant and more specifically in the order of several million pounds, while the time for the project implementation surpassed twelve months. Consequently, the Tilbury case study turns out not to be a very robust business case for the preferred biomass co-firing option. The relatively high delivered fuel price, the high capital expenditure requirement for the fuel reception, storage, new milling, handling and firing facilities were among the reasons for that.
36
Panagiotis Grammelis Table 10. British power plants experienced co-firing
Plant name / Firing Company technology Aberthaw/RWE Direct injection Cockenzie/SP Co-milling Cottam/EDF Co-milling
Output Status [4] % (MWe) substitution 1,553 Commercial n.a 1,200 2,000 Commercial n.a Commercial 5%
Primary fuel coal coal coal
Suppl. Fuels Wood, sawdust Wood pellets Forestry/ sawmill wastes Energy crops Milled palm nuts Palm kernels Olive residue, palm nuts, citrus, wood Olive residue, palm nuts, citrus, wood Wood, palm kernels Cereal residues Sewage sludge, wood pellets sawdust
Didcot A/ RWE Co-milling Drax / Drax Direct injection Eggborough/BE Co-milling Ferrybridge C / Direct SSE injection Fiddlers Ferry / SSE Ironbridge/ E.on UK Kingsnorth / E.on UK Longannet/SP Direct injection Co-milling Co-milling Co-milling
2,100 4,000 1,960 2,035
Commercial 2% Commercial 3% Commercial 3.5% Commercial 10%
coal coal coal coal
1,995
Commercial 10%
coal
970 2,034 2,400
Commercial 10% Commercial 10% Commercial n.a
coal coal coal
Tlibury / RWE
Co-milling
1,085
Commercial 3%
coal
The Case Study of Kingsnorth Power Plant Another co-firing test was performed at the Kingsnorth coal-fired Power Plant of 2.034 MWe installed capacity. The biomass material, cereal straw residues, contributes to the substitution of the main fuel by 10% on a heat input basis. Due to the high delivered fuel price (6 /GJ), a low capital cost option was favored, which was subsequently implemented relatively quickly. For the selected option, the existing coal milling, handling and firing system for the pre-blending and co-firing of the pelletised bio-fuel with coal were utilized. All the relevant potential impacts of straw co-firing were assessed by employing suitable modeling techniques. Thus, the impacts of biomass co-firing on the boiler operation and on the environmental performance of the power plant indicated that co-firing, at ratios less than 10% on a heat input basis, was technically viable with modest impacts on plant performance.
The Case Study of the Co-firing of Wood at a Scottish Power Plant The third case study is concerned with the co-firing of wood in a large coal-fired power plant in Scotland. The wood fuel derived from forestry residues and other wood materials. In
37
order to reduce the capital expenditure on storage and handling facilities, the energy company delegates all these responsibilities to the fuel suppliers, thereby including the relevant costs into the delivered fuel price. Thus, the solid bio-fuel was pre-blended with coal and the mixture was delivered to the power plant by train. As a consequence, this approach offered a higher degree of control over the blend ratio and the quality of the blend compared to what can be achieved with most of the on-site blending options. In sequence, the pre-blended fuel was handled and fired through the existing coal infrastructures. There was no requirement for any significant modification of the installed equipment on-site. This approach to co-firing has been adopted on a short-term trial basis but has not been demonstrated in long-term operation. Moreover, the co-firing ratio that can be achieved is very unlikely to be greater than 10% on a mass basis. At these low co-firing ratios, and with a clean, low ash content wood fuel, the impacts on the technical and environmental performance of the plant will be modest.
4.7. Co-firing Experience in Mediterranean Countries [33]

In general, no large-scale co-firing plants are in commercial operation in the Mediterranean region. Consequently, the acquired experience is derived from pilot-scale activities solely. Both the available type of each bio-fuel and the low-rank coal used for combustion should be taken into account when the optimum percentage of biomass in the mixture is defined. The typical ratio biomass/coal ranges from 5 to 10% in conventional systems of pulverized fuel. Furthermore, despite the high availability of biomass in all Mediterranean countries, only 40% of this is potentially available for co-firing, due to logistic constraints.
4.7.1. Co-firing Experience in Greece In general, Greek biomass characteristics comprise the average biomass characteristics, such as high volatile and oxygen content and low ash content. Also, the sulphur content is low, but significant Cl percentage is present instead. A significant factor affecting the biomass quality is the weather conditions during harvesting. The substitution of conventional fuels from biomass in existing power plants is a promising alternative option for the country, as well. Nonetheless, these biomass co-firing applications are limited to few numbers and only on pilot scale implementations. Up to date, the concern has focused on utilizing biomass residues (olive kernels) and wood, as additional fuels for the production of electricity or steam. Specifically, short-term co-firing activities have been successfully attained in two pulverized-fuel Units at the Megalopolis Thermal Power Plant and long-term ones have been realized in an industrial boiler with a moving grate firing system [33]. Each case study is described afterwards. The environmental and socio-economic impacts of biomass co-firing with brown coal in an existing thermoelectric unit in Northern Greece were also evaluated through a comparative assessment of its pure combustion in a new plant. The study was conducted at a 125 MWe lignite-fired power plant in Ptolemais power plant. Concerning the environmental impacts, the co-firing activity offers cleaner electrical power with lower emissions of pollutants. Moreover, it acts beneficially on the saving of natural resources. In terms of economics, biomass co-firing with lignite is counted as a low-risk investment due to the exploitation of
38
the existing infrastructure with fewer modifications that requires small capital. In addition to this, the previous will bring about fewer jobs than the construction of a biomass-dedicated plant, but it will secure the existing jobs for several years more. Beyond the positive social impacts, the application of co-firing activities utilizing environmentally neutral fuels such as biomass is necessary for the government in order to meet the standards of the Kyoto Protocol or comply with the EU regulations. Aiming to achieve the above, there is a strong prerequisite for the development of fuel supply networks, the classification and standardization of solid bio-fuels produced in the country [35].
The Case Study of Megalopolis Lignite-fired Power Plant [34] As it is reported, direct co-firing trials took place in Units I and II of Megalopolis Thermal Power Plant, each one of 125 MWe power output. The aim of the co-firing tests was to examine the efficiency and availability establishment in the existing boiler infrastructure. Attention was addressed to the investigation of lignite substitution percentage by biomass in the fuel blend. Although initially the biomass percentage in the mixture was 2% by mass, it was gradually increased at 7%. For this percentage, the available biomass quantities in Greece are enough to cover the needs for co-firing application in large power plants. Concerning the boiler operation, an obvious increase in the steam generation and the load was noted. The high volatile content of olive kernels contributed to a stable flame and its complete combustion resulting in unaffected unburnt particles in the ash. Based on financial and environmental issues, there are apparent advantages owning to SO2 and CO2 reduction. In order to assure proper blend feeding in the boiler, some constructive modifications are necessary in the existing boiler installations.
The Case Study of Pindos Industry [36] In the Greek industry Pindos it was possible to utilize the existing combustion installation for conventional domestic fuels (lignite) in combination with thermal utilization of waste wood. The industrial boiler was of lower capacity, equal to 13.8 MWth and equipped with a moving grate firing system. Referring to the emissions, the values of CO, NOx and SO2 were very satisfactory, dioxin and furan emissions were recorded lower than the legislative limit while the concentrations of heavy metal were affected in a similar way. Furthermore, it was observed that even the extensive use of waste wood does not bring any additional burden to the maintenance cost of the combustion equipment. 4.7.2. Co-firing Experience in Italy In Italy, co-combustion of coal and biomass is, at present, applied only in the Fusina power plant (units 3 and 4). Since 2004, SRF co-firing has been continuously performed at the two 320 MWe coal-fired boilers. The average share of SRF is about 2.5% of thermal input with a typical heating value ranging between 16.5-18.5 MJ/kg and the organic fraction being higher than 60%. Information described below was obtained regarding two experimental Enel (National Electric Energy Board) installations, namely in Pietrafitta (Umbria) working from 1958 and in S. Gilla (Sardinia) working from 2001. In both cases, the basic fuel was lignite
39
initially and coal afterwards, excavated from the mines close to the plants locations. The biomass utilized as secondary fuel was, in both plants, olive husk.
Co-firing in Pietrafitta Enel Power Plant The total installed capacity of this power plant is equal to 244 MWe. From 1973 to 1988 the main fuel was integrated with a small quantity of olive husk. The aim of the co-firing activities was to increase the plant residual life and to offset the reduced availability of the lignite deposits. The problems in realizing the olive husk co-firing with coal were related to difficulties concerning the proper heat contribution of each fuel as well as some inconveniences that occurred during the tests. More specifically, the additional wear of the existing mills mechanical components associated with the grinding of biomass posed problems in obtaining there were difficulties to obtain a good and homogeneous size rate. Due to the lack of homogeneity, the heaviest particles were thrown on the boiler bottom, having only realized partial combustion. The lighter particles on the other hand were burned in the mills ducts, before entering the combustion chamber. In addition, problems associated with the automatic removal of the ashes because of their weight and size occurred in the combustion chamber, on the bottom grid and on the transporting strap. Moreover, whenever the proportion of the mixture was not rightly set, particularly when an excess quantity of olive husk was present, the combustion chamber temperature was raised with a consequent increase of heavy ashes production. Co-firing in S. Gilla ENEL Installation The S. Gilla Plant (Cagliari) consists of a 5 MWth boiler and was installed for experimental tests on co-combustion of olive husk and coal, in order to verify its feasibility on thermoelectric plants of a large size. The first test concerned the optimization of the grinding system and the feeding system of the secondary fuel while the second one conducted a series of combustion tests with mixtures of coal and olive husk. Based on the results of the first test, the only problem to be solved regarded the reduction of the mills yield, compared to the coal grinding. According to the chemical analyses in both fuels, preceding the co-firing tests, the olive husk ash was rich in silicon (47%), potassium (17%), calcium (9%), phosphorus (3.5%) and sodium (4.2%). All these elements are unfavorable due to problems related to fouling and slagging of the boilers. Moreover, the presence of Si with a high percentage of K could cause the formation of low melting point substances in the ashes. Also, a change in the ratio to a percentage of 60/40 for olive husk could increase the potential of fouling in the convection section of the boiler because of the presence of high alkaline metals content. In relation to traces, almost all of these elements in olive husk, with the exception of Cr, Ni, Zn, were in a lower concentration than in the coal. The steps of the co-firing application in Cagliari followed at each phase are mentioned below. The granular nature of olive husk makes it easy to be transported. Except for the possible drying in the presence of humidity higher than 10 -15%, no other particular pretreatment was used. As regards the grinding and the feeding procedure, the olive husk was mixed with the coal in a silo and grinded within the main mill of the plant. Then the mixture was sent to the burners. Combustion problems were not observed due to the high percentage of volatiles in biomass, which helped produce good flame stability. Fouling and slagging problems were not presented. Some inconveniences caused by deposits of dusty nature were easily removed by the soot-blowers systems. The most significant advantage of combustion
40
with olive husk was observed in the emissions, with a generalized reduction of the total amount. Concerning the ash quality, the process of olive husk working can affect and render the mixed ash improper for future exploitation without additional pre-treatment.
4.7.3. Co firing Experience in Portugal In Portugal, the energetic utilization of biomass residues represents a great contribution for the improvement of renewable energies. The promotion of biomass in the country is especially related to forest residues improvements. This situation is mainly due to the fact that forest is covering a significant area of the country and on the other hand, forest is threatened every year by forest fire. However, the use of these residuals is limited in co-firing with solid fossil fuels plants as the country has only two coal-fired power plants. One of them, Sines power plant with an installed capacity of 1.200MWth, is located by the sea, where there are not enough residues in a radius of 50km to provide 10% of fuel replacement. Therefore, the larger amounts of biomass residuals are not located close to these plants. The implementation of biomass-fired plants is suitable in dense forest areas as in the centre of the country. Thus, different alternatives have to be followed. One alternative for the use of biomass residues can be their exploitation in decentralized power plants, like the first biomass fired power plant located in Mortgua. Another is their utilization in biomass boilers existing in pulp industries. Due to their specific location, there is enough capacity to use all residues produced in a radius of 50 km to replace 10% of the biomass feed in the boiler.
5. RESEARCH TRENDS FOR BIOMASS CO-FIRING IN EUROPE

Nowadays, research development activities in Europe are focused on extending applications to a wider range of biomass materials [39]. Demonstration projects should aim at medium to large scale bio-electricity systems, covering the whole process chain from sustainable feedstock supply over energy conversion to the recovery of by-products. The overall scope is to decrease the electricity generation costs of biomass from the range of 0.050.08 /kWh to below 0.04 /kWh in 2020. Although several R&D activities have been accomplished in the past for the operational problems met during co-firing biomass or waste, there are still some technical difficulties to be confronted. The development of on line monitoring tools to control slagging and corrosion phenomena during biomass co-firing could contribute towards that direction. On the other hand, more demonstration activities have to be undertaken for co-firing biomass in existing large-scale fossil fuel fired power plants with high net electric efficiency and enhanced biomass shares. The production of new, tradable solid biofuels for electricity generation purposes will certainly assist the deployment of cofiring technologies. Such biofuels may be fully or partially based on unconventional and difficult resources such as e.g. straw and/or organic wastes. Especially concerning the latter, there is a developing market for Solid Recovered Fuel (SRF) in the EU, enforced by the implementation of the waste landfill directive in 2005. The markets enlargement in the future is strongly dependent on the standardisation needs of SRF, as well as the barriers occurred in its co-utilisation in coal fired power plants, such as the logistics, legislation and social impact.
41
6. CONCLUSIONS
The following conclusions can be drawn from this study on retrofitting coal fired power plants for co-firing biomass and the accumulated experience from co-firing projects in Europe. 1. Co-firing biomass with coal can offer the best short- and long-term solution to greenhouse gas emissions from power generation. The effectiveness of the co-firing application, combined with its low cost and low technical risk, classify it high on a priority list of the large scale power plant operators and institutions considering mitigation of CO2 emissions and saving of diminishing fossil fuels. 2. The economic evaluation of biomass co-firing with coal is complex. The evaluation must include several components. The price of solid bio-fuels is frequently a very important determinant of a plants economic viability. Straw delivered to the coal/straw co-fired Grenaa plant in Denmark costs three to four times more than coal [36]. Moreover, biomass co-firing is commonly slightly more expensive than dedicated coal systems concerning the cost for the additional infrastructures, not to mention the higher operating cost. Consequently, subsidies providing to the generators of electricity from renewable sources, such as the Renewable Energy Certificates (ROCs) in Britain, or other financial incentives like the Emission Trading, are necessary in order to initiate co-firing activities. 3. Concerning the technical situation of co-firing, a multi-criteria analysis needs to be addressed. The involved criteria take into account the following subjects: Fuel supply and handling The case of fuel supply is essential and has to be investigated thoroughly. Particularly, even if the renewable fuel is cost free at the point of its generation, its transportation and preparation increase its effective cost per energy unit, which sometimes exceeds the corresponding of the reference fuel. Additionally, biomass availability is another key parameter which depends on seasonal variation and additional market of potential users. Thus, the preparation and handling of supplementary fuels in existing installation, originally designed for coal, is limited to small proportions. Low energy density, high shares of fines, high moisture content and unpleasant odors may cause inconveniences. In such cases, a separate biomass handling and feeding system is preferred, which raises the cost significantly. The proper selection of fuel supply system and its coupling with the firing system is probably the most serious retrofitting action to be realized in a co-firing project. Technical modifications and effects on boiler operating conditions Generally, the addition of biomass to a coal-fired boiler does not impact or at worst slightly decrease its overall efficiency [37]. Co-firing with shares of less than 10% of the thermal output appears to be no problematic regarding corrosion and slagging of the boilers. By appropriate process settings any differences of the fuel characteristics can be overcome. With respect to dioxins and furans, no negative effect were observed. The high temperatures in the flame zone together with sufficient oxygen availability or the addition of some nitrogen compounds
42
Panagiotis Grammelis in the post combustion zone can effectively reduce the concentration of PCDD/F in the flue gases [2], [36]. Environmental and financial issues In addition to greenhouse emissions, pollutants like SO2 are decreased when proper choices of biomass and operating conditions are made. NOx emissions, which are strongly reliant on combustion conditions, decrease due to the reducing effect of the higher volatile release of the solid bio-fuels. Furthermore, available experience has shown that changes in the flue gas composition are associated with the fuel analysis. In relation with the economics, the quality of fuel for the energy supply determines the type and size of the equipment and consequent capital costs. Operating and maintenance costs are dependent on the technology used to store, process, and burn the fuels and the potential impact of fuel characteristics on plant performance, including efficiency. 4. Beyond the technical concerns to co-firing, legislation issues should be also taken into account. The previous are referred to the political goals that have launched by the European Commission. These issues include actions on supporting and promoting the energy utilization of alternative solid bio-fuels. Apart from that, the use of biomass in co-firing activities with fossil fuels can make a real contribution to Governments renewable targets and obligations. Thus, administrative expenditures should be estimated especially when introducing supplementary fuels which are subject to other intended markets.
Conclusively, co-firing with limited biomass shares is technically feasible, even though operational problems and issues on the variability of residues composition need to be further clarified. Generally, there is a wide range of secondary fuels co-utilised in fossil fuel fired power and industrial plants in Europe throughout the years. In several countries, bio-fuels such as wood, straw or even wastes and other biomass resources are used in modern and older power plants for energy generation, in particular when an economical advantage exists. However, the implementation of large scale co-firing activities within the European power industry has to be more expanded. For this reason, standardized fuel qualities for the use in power plants have to be defined and fuel markets to be established. Especially concerning the waste derived fuels, methods should be developed aiming to determine their biogenic share in terms of the CO2 certificate trading system. Such tools will encourage and facilitate the implementation of secondary fuel production/utilization as an integral part of an integrated waste management concept.
ACKNOWLEDGMENTS
I would like to thank Mrs. A. Malliopoulou, Mr. M. Agraniotis, Mr. M. Karampinis and Dr. D. Chua for their valuable contribution in collecting the data, without which the accomplishment of this study would not have been possible.
43
REFERENCES
[1] [2] [3] L. Baxter, J. Koppejan, Opportunity for affordable renewable energy, IEA Bioenergy Task 32: Biomass-coal Co-combustion, Switzerland 2006. K.R.G. Hein, W. Scheurer, Co-combustion of biomass, wastes and residues with coal, Proceedings of the EU seminar, Cottbus, October 2000. P. Grammelis, G. Skodras and E. Kakaras: Thermal xploitation of Waste Woods with Lignite for Energy Production, Journal of the Air & Waste Management Association, Volume 53, November 2003, pp. 1301-1311. Hein KRG, Bemtgen JM. EU clean coal technology co-combustion of coal and biomass. Fuel Process. Technol 1998; 54:15969. Baxter L. Biomass-coal co-combustion: opportunity for affordable renewable energy. Fuel 2005;84:1295302. W. R. Livingston, Development of low cost systems for co-utilisation of biomass in large power plants, Mid-term review report, 2003. Tillman D. A., Biomass cofiring: the technology, the experience, the combustion consequences, Biomass and Bioenergy 19, pp. 365-384, 2000. Mitsui Babcock, IPA Energy consulting, The economics of co-firing, DTI, Final report URN 06/1959, July 2006. W.R. Livingston Mitsui Babcock Technology, Thermal Net Workshop on Biomass co-processing and co-firing, France, April 2006. F. Rohan, Experience of indirect co-firing of biomass and coal, CCC64 pp.39, IEA Coal Research, October 2002. AEA Technology, VTT Energy, Sydkrat Konsult AB, European co-combustion of coal, biomass and wastes, Final report Project DIS-0506-95-UK, January 2000. R. Meijer, C.H. Gast, Spontaneous combustion of biomass: Experimental study into guidelines to avoid and control this phenomena, Kema Power generation and Sustainables. E. Kakaras, P. Grammelis and J. Jacobs: Existing electricity generation capacity using solid fuels within the enlarged EU: market size and renovation options, VGB PowerTech (2002), Nr.4, P.81-87. DTI, Co-firing of biomass at UK Power Plant, Carbon Abatement Technologies, Programme BPB009, August 2005. European Commission Directorate-General for Energy and Transport, Opportunities and markets for co-utilisation of biomass and waste with fossil fuels for power generation, Final report version 4. W.R. Livingston, A review of the recent experience in British with the co-firing of biomass with coal in large pulverized coal-fired boilers, The Future of energy, Coutilisation of coal and biomass for bioenergy and biofuel production, Fossil Fuel Foundation of Africa, Johannesburg, June 2005. Tillman D. A., E. Hughes & B. A. Gold, Co-firing of biofuels in coal fired boilers: Results of case study analysis, 1st Biomass Conference of the Americas, Burlingtion, VT, 1994. M. Pronobis, The influence of biomass co-combustion on boiler fouling and efficiency, Fuel, Vol.85, Iss.4, pp.474-480, 2006.
[4] [5] [6] [7] [8] [9] [10] [11] [12]
[13]
[14] [15]
[16]
[17]
[18]
44 [19] [20] [21] [22]
Panagiotis Grammelis Robinson AL, Junker H, Baxter LL. Pilot-scale investigation of the influence of coalbiomass cofiring on ash deposition. Energy Fuels 2002;16:34355. J. Wood, Feel the burn The truth behind co-firing biomass, IEE Power Engineer, Oct/Nov, pp.18-21, 2004. Best practice brochure, Co-firing of biomass at UK Power Plant, prepared by the British Department of Trade and Industry (DTI), August 2005. M. Perry, F. Rosillo - Calle, Co-firing Report UK, IEA Bioenergy Task 40: Sustainable International BioEnergy Trade: Securing supply and demand, Switzerland, November 2006. Hughes E. Biomass cofiring: economics, policy and opportunities. Biomass Bioenergy 2000;19:45765. Rudiger H, Kicherer A, Graul U, Splietho. H, Klein K. Investigations in combined combustion of biomass and coal in power plant technology. Energy Fuels 1994;10:78996. J. Koppejan, Overview of experiences with co-firing biomass in coal power plants, IEA Bioenergy Task 32: Biomass-coal Co-combustion, 2004. Ekmann J. M., Winslow J. C., Smouse S. M., Ramezan M., International Survey of cofiring coal with biomass and other wastes, Fuel Processing Technology 54, pp. 171188, 1998. K. Savolainen, Co-firing of biomass in coal-fired utility boilers, Applied energy 74 p.369-381, 2003 Handbook of Biomass Combustion, IEA Bioenergy Task 32: Biomass Combustion and Cofiring E. Kakaras, P. Grammelis, M. Agraniotis, W. Derichs, H.P. Schiffer, J. Maier, Th. Hilber, Solid recovered fuel as coal substitute in the electricity generation sector, Thermal Science Journal, Vol.9, Iss.2, 2005. Th. Hilber, M. Agraniotis, J. Maier, G. Scheffknecht, P. Grammelis, E. Kakaras, Th. Glorius, U. Becker, W. Derichs, H.P. Schiffer, M. de Jong, L. Torri, Advantages and possibilities of Solid Recovered Fuel (SRF) co-combustion in the European Energy Sector, Journal of Air and Waste Management Association (to be published). R.van Ree, R. Korbee, T. De Lange, Biomass co-firing potential and experiences in the Netherlands, ECN-RX0-035, November 2000. A. Weststeijn, Co-firing in the Netherlands, Proceedings of 3rd Project meeting on RECOFUEL, June 2005. ALTENER 4/1030/C/00-022/2000, Biomass residues utilisation as fossil fuel substitute for power generation and district heating in the Mediterranean region, Final Report, September 2002. P. Grammelis, P. Vourliotis, E. Kakaras, Co-combustion of different waste wood species with lignite in an industrial steam boiler with a moving stoker firing system, in Progress in Thermochemical Biomass Conversion (2-Volume Set): Bioenergy, ed. Bridgwater, A. V. 2001/09 Blackwell Science Inc, ISBN:0632055332. P. Grammelis, G. Skodras, E. Kakaras, An economic and environmental assessment of biomass utilisation in lignite- fired power plants in Greece, Int. J. Energy Res, Vol.30, Iss.10, pp.763-775, January 2006.
[23] [24]
[25] [26]
[27] [28] [29]
[30]
[31] [32] [33]
[34]
[35]
Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal [36]
45
[37]
[38] [39]
E. Kakaras, P. Vourliotis, P. Grammelis, Utilization of Wood matter from Pressed Oil-Stone (WPOS) as Lignite Substitute for Power Generation, 1st World Conference and Exhibition on Biomass for Energy and Industry, Spain, June 2000. Skodras G., Grammelis P., Samaras P., Vourliotis P., Kakaras E. and Sakellaropoulos G.P., Emissions Monitoring During Coal Wastewood Co-Combustion in an Industrial Steam Boiler, Fuel, Vol.81, pp.547-554,2002. R. Stenteville, The State of Garbage, Biocycle, April 1994. European Commission, Work Programme 2007-2008 Cooperation Theme 5: Energy, 2007.
Chapter 2
MICRO-CHP POWER GENERATION FOR RESIDENTIAL AND SMALL COMMERCIAL BUILDINGS

Louay M. Chamra* and Pedro J. Mago
Department of Mechanical Engineering, Mississippi State University, 210 Carpenter Engineering Building, P.O. Box ME, Mississippi State, MS 39762-5925, USA
ABSTRACT
The traditional structure of the electrical utility market, green engineering issues, and environmental acceptability have resulted in a relatively small number of electric utilities. However, today's technology permits development of smaller, less expensive power systems, bringing in new, independent producers. Competitions from these independent producers along with the re-thinking of existing regulations have affected the conventional structure of electric utilities. The restructuring of the electric utility industry and the development of new "onsite and near-site" power generation technologies have opened up new possibilities for buildings, building complexes, and communities to generate and sell power. Competitive forces have created new challenges as well as opportunities for companies that can anticipate technological needs and emerging market trends. Micro-cooling, heating, and power (micro-CHP) is decentralized electricity generation coupled with thermally activated components for residential and small commercial applications. A micro-CHP system consists of a prime mover, such as a reciprocating engine, which drives a generator, which produces electrical power. The waste heat from the prime mover is recovered and used to drive thermally activated components and to produce hot water or warm air through the use of heat exchangers. Micro-CHP holds some of the answers to the efficiency, pollution, and deregulation issues that the utility industry currently faces. A review of micro-CHP systems, specific
Phone: (662) 325-0618; Fax:(662) 325-7223; E-mail: chamra@me.msstate.edu E-mail: mago@me.msstate.edu
48
Louay M. Chamra and Pedro J. Mago

types of distributed power generation, and thermally-activated technologies are introduced and discussed in this chapter.
INTRODUCTION
Micro-cooling, heating, and power (micro-CHP) is decentralized electricity generation coupled with thermally activated components for residential and small commercial applications. micro-CHP systems can simultaneously produce heat, cooling effects, and electrical power. The micro regime is typically designated as less than fifty kilowatts electric (< 50 kWe). The concept of micro-CHP is illustrated in Figure 1. A prime mover, such as a reciprocating engine, drives a generator which produces electrical power. The waste heat from the prime mover is recovered and used to drive thermally activated components, such as an absorption chiller or desiccant dehumidifier, and to produce hot water or warm air through the use of heat exchangers. Cooling, heating, and power (CHP) has proven beneficial in many industrial situations by increasing the overall thermal efficiency, reducing the total power requirement, and providing higher quality, more reliable power. Applying CHP technology to smaller scale residential and small commercial buildings is an attractive option because of the large potential market. The residential and small commercial sectors account for 40% of the electrical usage in the U.S. As can be seen from Figure 2, the residential and small commercial sectors make up the largest portion of the utility electricity market. The residential energy consumption is not only the largest portion of the pie, but it is also the fastest growing segment. Between 1978 and 1997, the number of U.S households has increased by over 30%. At the same time, space heating expenditures have increased by 75%, air conditioning by 140%, and water heating by 184%. The largest increase in household energy expenditures was for home appliances, which increased by 210%.
Waste Heat
Reject Heat
Prime Mover Absorption Chiller Heat Exchanger Desiccant Device Heat Exchanger
Generator/ Power Conditioner
Electrical Power
Building Cooling
Building Heating
Building Humidity Control
Hot water
Figure 1. Schematic of a micro-CHP System.
Micro-CHP Power Generation
49
Figure 2. 2001 U.S. Electrical Consumption by Building Sector. (Available at http://www.eia.doe.gov)
As a result of such large increases, residential energy consumption is projected to increase by 25% from 2001-2025. The question is: Why should micro-CHP be considered a viable option to meet the needs of the U.S. residential and small commercial market? The basis of this answer can be found by applying the wells to wheels analysis concept to the energy production for a single residence. The idea of wells to wheels is that the whole system must be considered from fuel harvesting to the energy (in some final form) that is used. In addition, each time that fuel is converted, packaged, or transported, there is an associated loss of energy. The more conversion and transportation steps in a process, the greater the associated energy losses. In the U.S. as of 2004, electricity is generated by coal (50%), nuclear (20%), natural gas (18%), hydro (7%), petroleum (3%), and various renewable energy methods (2%). The traditional method of electrical power generation and distribution is based on large, centrallylocated power plants. Central means that the power plant is located on a hub surrounded by major electric load centers. Once the electricity is produced, the power must be delivered to the end user. Delivery is achieved by a utility transmitting the electricity to a substation through a high-voltage electrical grid. At the substation, the high-voltage electricity is transformed, or stepped down, to a lower voltage to be distributed to individual customers. The electricity is then stepped down a final time by an on site transformer before being used by the customer. The number of times that the electricity must be transformed depends largely upon the distance the power is transmitted and the number of substations used in distributing the electricity. Inefficiencies are associated with the traditional methods of electrical power generation and delivery. To begin, the majority of the energy content of the fuel is lost at the power plant through the discharge of waste heat. Traditional power plants convert about 30% of a fuels available energy into electric power. Highly efficient combined-cycle power plants convert about 50% of the available energy into electric power. Further energy losses occur in the transmission and distribution of electric power to the individual user. Inefficiencies and pollution issues associated with conventional power plants have provided the motivation for new developments in on-site power generation. The overall efficiencies of central power generation and distributed combined-cycle power generation are shown in Figures 3 and 4.
50
100%
91%
32%
31%
~30%
Natural Resource
Power Plant
Transmission
Distribution
End Uses
Figure 3. Efficiency of Central Power Generation.
Once the electric power reaches the end user, the electricity is used to run central heat and air conditioners, appliances, lighting, and in some cases, water heating. These are the same end uses that could be provided by a micro-CHP system at a greater overall thermal efficiency. Micro-combined heat and power units utilize waste heat while simultaneously producing electric power for a residence or building. The waste heat is used to meet space and water heating requirements and to provide cooling if an absorption chiller is incorporated into the system. Heating and cooling are major end uses of residential energy. Because the heating and cooling loads of the space are being met without total dependence on electrically-driven thermal components, the overall electric load of the residence will be reduced.
100%
91%
50%
~48%
Natural Resource
Power Plant
Distribution
End Uses
Figure 4. Efficiency of Combined-Cycle Power Generation.

~75%
51
100%
91%
Natural Resource
CHP Power Plant
Thermal Energy
End Uses
Figure 5. Efficiency of micro-CHP System.
Another advantage of micro-CHP is that there are no losses associated with power distribution and transmission as opposed to the traditional power generation method. MicroCHP systems can utilize about 75% of the fuels available energy to provide electric and thermal energy. A micro-CHP system can produce an overall efficiency of about 75% while a modern combined-cycle power plant will have an overall efficiency of around 50%. The overall efficiencies of a micro-CHP system is shown in Figure 5. Larger homes, higher energy costs, volatile fuel markets, electricity blackouts, power security, power quality, and increasing concern for environmental issues have all helped open the door for micro-CHP. RKS, a leading market research firm, found that more than 38% of high-income households, (i.e., incomes greater than $50,000) are interested in generating their own electricity. (Micro-CHP Technologies Roadmap, U.S. DOE, 2003).
History
Combined heat and power generation, or cogeneration, is a well established concept dating back to the 1880s when steam was a primary source of energy in industry and electricity was beginning to be used for both power and lighting. As electrical power and electrical motors became more widely used, steam driven mechanisms were replaced creating a transition from mechanically powered systems to electrically powered systems. In the early 1900s, an estimated 58% of the total power generated in the United States by on-site industrial power plants was cogenerated power. The development of central power plants and reliable utility grids drove electricity costs down, and industrial plants began buying electricity from utility companies and ceased generating their own power. On-site industrial cogeneration declined in the United States and accounted for only 15% of total electricity generation by 1950 and dropped to about 5% by 1974 (Knight and Ugersal, 2005). Increasing regulatory policies, low fuel costs, and advances in technology also contributed to the decline of cogeneration. In the last forty years systems that are efficient and have the ability to utilize alternative fuels have begun to appear because of energy price increases and the uncertainty of fuel supplies. In addition, CHP has gained attention because of decreased fuel consumption and
52
lower emissions. Today, many industrialized countries are taking leading roles in establishing and promoting the use of cogeneration in the industrial, residential, and other market sectors.
THE MICRO-CHP SYSTEM

Micro-cooling, heating, and power combines distributed power generation with thermally activated components to meet the cooling, heating, and power needs of residential and small commercial buildings. The success of CHP systems for large-scale application coupled with the development of power generation equipment and thermally activated components on a smaller scale have contributed to the development of micro-CHP applications. Distributed power generation technologies and thermally activated components will be introduced and briefly discussed.
Distributed Power Generation

A number of technologies are commercially available or under development for generating electric power (or mechanical shaft power) onsite or near site where the power is used. Distributed power generation is a required component of micro-CHP systems. Fuel cells, reciprocating engines, Stirling engines, Rankine cycle engines, and microturbines are prime movers that have the most potential for distributed power generation for micro-CHP systems. When discussing various prime movers for micro-CHP systems, a primary method of comparison is to examine the efficiency of each prime mover. The efficiency of a micro-CHP system is measured as the fraction of input fuel that can be recovered as power and heat. The remaining energy is rejected as low-temperature heat. There are three primary efficiencies that are associated with micro-CHP systems: electrical efficiency, thermal efficiency, and overall efficiency. These efficiencies are defined as
Electical efficiency =
electrical output fuel input
(1)
Thermal efficiency =
Thermal output fuel input
(2)
Overall efficiency =
useful thermal + electrical output fuel input
(3)
53
PRIME MOVERS
Reciprocating Engines
Reciprocating engines can be used to produce shaft power. The shaft power can then be used to drive a generator to produce electrical power. The shaft power can also be used to operate equipment such as compressors and pumps. The application of reciprocating engines is widespread and highly developed. Reciprocating engines use natural gas, propane, gasoline, diesel and biofuels to produce 0.5 kW to 10 MW of power. A diesel fuel engine generator set is shown in Figure 6. Reciprocating engines exhibit characteristics that are advantageous for micro-CHP applications. Reciprocating engines used for power generation have proven reliability, good load-following characteristics, low capital cost, fast startup, and significant heat recovery potential. Recent advances in combustion design and exhaust catalyst have also helped reduce overall emissions of reciprocating engines. Currently, reciprocating engines are the most widely used distributed energy technology. Typical electrical conversion efficiencies are in the range of 25% to 40%. The overall thermal efficiencies of these systems increase with the incorporation of thermally activated components. The thermal energy in the engine cooling system and exhaust gases from reciprocating engines can often be recaptured and used for space heating, for hot water heating and for driving thermally activated components. Shaft power from the engine can also be used to power thermal components, such as gas vapor compression chillers. Such chillers are very similar to electric-driven chillers with the exception that the compressor is driven by the reciprocating engine rather than an electric motor.
Figure 6. Model D13-2, 12-kW Diesel Engine Generator Set from Caterpillar (Available at http://www.cat.com).
54
Louay M. Chamra and Pedro J. Mago Table 1. Overview of Reciprocating Engine Technology (Available at http://www.energy.ca.gov/distgen/)
Reciprocating Engines Overview Yes 0.5 kW 7 MW Natural gas, diesel, landfill gas, digester gas 25 45% Emission controls required for NOx and CO Cogeneration (some models) Products are widely available
Commercially Available Size Range Fuels Efficiency Environmental Other Features Commercial Status
Emissions of reciprocating engines tend to be higher than that of other distributed generation equipment. Due to the emissions and noise emitted by these engines, care must be exercised in the location of the engine with respect to the occupants of the building. In some areas, local air quality standards may limit the use of reciprocating engines.
Application Reciprocating engine generator sets are the most common and most technically mature of all distributed energy resources (DER) technologies. Reciprocating engines can be used for a variety of applications due to their small size, low unit costs, and useful thermal output. Applications for reciprocating engines in power generation include continuous or primepower generation, peak shaving, back-up power, premium power, remote power, standby power, and mechanical drive use. An overview of reciprocating engine characteristics is presented in Table 1. Reciprocating engines are an ideal candidate for applications in which there is a substantial need for hot water or low pressure steam. The thermal output can be used in an absorption chiller to provide cooling. Comparatively low installation costs, suitability for intermittent operation, and high temperature exhaust make combustion engines an attractive option for micro-CHP. Internal combustion engines utilize proven technologies with a well established infrastructure for mass production and marketing. The development of combustion engines has also formed a maintenance infrastructure with certified technicians and relatively inexpensive and available parts are available. Due to the long history and widespread application, internal combustion engines are a more developed technology than most prime movers considered for micro-CHP. Heat Recovery Traditional large-scale electric power generation is typically about 30% efficient, while combined cycle plants are typically 48% efficient. In either case, the reject heat is lost to the atmosphere with the exhaust gases. In an internal combustion engine, heat is released from the engine through coolant, surface radiation, and exhaust. Engine-driven micro-CHP systems recover heat from the jacket water, engine oil, and engine exhaust. Low pressure steam or hot water can be produced from the recovered heat, and can be used for space heating, domestic hot water, and absorption cooling.
55
Figure 7. Heat Balance for a Representative Reciprocating Engine (Knight and Ugersal, 2005).
Heat from the engine jacket coolant is capable of producing 200 F (93 C)hot water and accounts for approximately 30 % of the energy input from the fuel. Engines operating at high pressure or equipped with ebullient cooling systems can operate at jacket temperatures of up to 265 F (129 C). Engine exhaust heat can account for 10 30 % of the fuel input energy and exhaust temperatures of 850 F 1200 F (455 C 649 C) are typical. Because exhaust gas temperatures must be kept above condensation thresholds, only a portion of the exhaust heat can be recovered. Heat recovery units are typically designed for a 300 F 350 F exhaust outlet temperature to avoid corrosive effects of condensation in the exhaust piping. Lowpressure steam (~15 psig) and 230 F (110 C) hot water are typically generated using exhaust heat from the engine. The combined heat recovery of the coolant and exhaust in conjunction with the work produced by combustion can utilize approximately 70 80% of the fuel energy. Figure 7 shows a heat balance for a representative reciprocating engine. Cost Reciprocating internal combustion (IC) engines are the traditional technology for emergency power all over the world. They have the lowest first costs among DER technologies. The capital cost of a basic gas-fueled generator set (genset) package ranges from $300-$900/kW, depending on size, fuel type, and engine type. Generally speaking, the overall engine cost increases as power output increases. The total installed cost can be 50100% more than the engine itself. Additional costs include balance of plant (BOP) equipment, installation fees, engineering fees, and other owner costs. Installed costs of micro-CHP projects using IC engines typically range between $800/kW - $1500/kW. The maintenance costs over the life of IC engines can be significant. The core of the engine maintenance is in the periodic replacement of engine oil, coolant, and spark plugs (if spark ignition). Routine inspections and/or adjustments are also necessary. Maintenance costs
56
of gas and diesel IC engines range between $0.007-$0.015/kWh and $0.005-$0.010/kWh respectively. (Available at http://www.energy.ca.gov/distgen/equipment/html)
Advantages and Disadvantages Reciprocating engines are generally less expensive than competing technologies. They also have start-up times as low as ten seconds, compared to other technologies that may take several hours to reach steady-state operation. Through years of technology advancements, reciprocating engines have climbed in efficiency from under 20% to over 30%. Today's most advanced natural gas-fueled IC engines have electrical efficiencies (based on lower heating value, LHV) close to 45% and are among the most efficient of the commercially available prime mover technology. Lower heating values neglect the energy in the water vapor formed by the combustion of hydrogen in the fuel. This water vapor typically represents about 10% of the energy content. Therefore the lower heating values for natural gas are typically 900 950 Btu per cubic foot. Advantages of reciprocating engines include:
Available in a wide range of sizes to match the electrical demand. Fast start-up and adjustable power output on the fly. Minimal auxiliary power requirements, generally only batteries are required. Demonstrated availability in excess of 95%. In load following applications, high part-load efficiency of IC engines maintains economical operation. Relatively long life and reliable service with proper maintenance. Very fuel flexible. Natural gas can be supplied at low pressure.
Disadvantages of IC engines are: Noisy operation. Require maintenance at frequent intervals. Relatively high emissions to the atmosphere.
Microturbines
Microturbines were derived from turbocharger technologies found in large trucks or the turbines in aircraft auxiliary power units (APUs) and have a capacity range of 25 kW to 500 kW. Microturbines utilize a variety of fuels including natural gas, propane, and biofuels. Electrical energy efficiencies of 25% to 30% are capable with the use of regenerators. Microturbines have fewer moving parts than other generation equipment of similar capacity, creating the potential for reduced maintenance intervals and cost. Though the generating capacity of microturbines is above the range defined in the micro-CHP regime, microturbines have considerable potential in on-site power generation applications such as apartment complexes and clusters of small commercial buildings.
57
Figure 8. Capstone C30 Microturbine (Available at www.capstone.com).
The waste heat from a microturbine is primarily in the form of hot exhaust gases. This heat is suitable for powering a steam generator, indirect heating of a building, allocation to thermal storage devices, or use in heat-driven cooling systems. Most designs incorporate recuperators that limit the amount of heat available for micro-CHP applications. Microturbines have relatively low emissions and noise and also have low maintenance costs. Another advantage is that microturbines are relatively small in size or footprint. The fuel flexibility and quantity of hot exhaust gases make microturbines an advantageous technology for micro-CHP and cogeneration applications. The Capstone C30, a 30-kW microturbine, is pictured in Figure 8.
Application Markets for microturbines include commercial and light industrial facilities. Microturbines can be used for stand-by power, power quality and reliability, peak shaving, and cogeneration applications. In addition, because microturbines are being developed to utilize a variety of fuels, microturbines are used for resource recovery and landfill gas applications. Microturbines are well-suited for small commercial establishments such as restaurants, hotels, motels, small offices, retail stores. The development of the microturbine technology for transportation applications is also in progress. Automotive companies are interested in microturbines to provide a lightweight and efficient fossil-fuel-based energy source for hybrid electric vehicles, especially buses. Microturbines are also being developed to utilize a variety of fuels and are being used for resource recovery and landfill gas applications.
58
Heat Recovery The waste heat from a microturbine is primarily in the form of hot exhaust gases. This heat is suitable for powering a steam generator, heating of a building, allocation to thermal storage devices, or use in absorption cooling system. However, most designs incorporate a recuperator that limits the amount of heat available for micro-CHP applications. The manner in which the waste heat can be used depends upon the configuration of the turbine system. In a non-recuperated turbine, the exhaust gas typically exits at a temperature between 1000 1100 F (538 594 C). A recuperated turbine can provide waste heat for heating and operating an absorption cooling system at exhaust temperatures around 520 F (271 C). The recovered heat can also be used to drive a desiccant dehumidification device. The use of the recovered heat influences the selection of the microturbine with or without a recuperator. Cost The capital costs of microturbines range from $700 - $1,100/kW when mass produced. These costs include all hardware, associated manuals, software, and initial training. Adding heat recovery components increases the cost by $75- $350/kW. Installation and site preparation can increase the capital costs by 30-50%. Manufacturers are striving for future capital costs of microturbines to be below $650/kW. This goal appears feasible if the market expands and sales volumes increase. With fewer moving parts, vendors hope their microturbines can provide higher reliability and require less maintenance than conventional reciprocating engine generators. The singleshaft design with air bearings will not require lubricating oil or water. Microturbines that use lubricating oil should not require frequent oil changes as the oil is isolated from the combustion products. Manufacturers expect microturbines to require maintenance once-a-year when the technology matures and are targeting maintenance intervals of 5,000 8,000 hours. Actual maintenance costs and intervals for mature microturbines are less well known since there is a limited base of empirical data from which to draw conclusions. Forecasted maintenance costs for microturbines range from $0.005 $0.016 per kWh, slightly lower than the costs for small reciprocating engine systems. Advantages and Disadvantages The operation of a microturbine offers several advantages. Microturbines have fewer moving parts than IC engines. The limited number of moving parts and the low lubrication requirements allow microturbines long maintenance intervals. Accordingly, microturbines have lower operating costs in terms of cost per kilowatt of power produced. Another advantage of microturbines is their relatively small size for the amount of power that is produced. Microturbines are also light weight and have relatively low emissions. Potentially, one of the greatest advantages of microturbines is their ability to utilize a number of fuels, including waste fuels or biofuels. Microturbines have great potential in cogeneration applications because microturbines produce a large quantity of clean, hot exhaust gases compared to other distributed generators. The primary disadvantages of microturbines are that they have a low fuel to electrical efficiency. Also, with higher elevation and increased ambient temperatures, microturbines
59
experience a loss of power output and efficiency. The ambient temperature directly affects the temperature of the air at the intake. A gas turbine will operate more effectively when colder air is available at the intake. A gas turbine cycle must compress the inlet air and the greater the compression, the greater efficiency. Another potential disadvantage is that microturbines experience more efficient operation and require less maintenance when operated continuously.
Stirling Engines
The Stirling engine is a type of external combustion piston engine which uses a temperature difference to produce motion. The cycle is based on the behavior of a fixed volume of gas. The heat source used to provide the temperature difference can be supplied by a wide variety of fuels or solar energy. The Stirling engine has only seen use in specific and somewhat limited applications. However, recently many companies have begun research and development related to Stirling engines due to their potential for micro-CHP applications and solar power stations. Stirling engines typically have an electrical efficiency in the range of 12% to 25%. This efficiency can be increased with the use of recuperators. The operation of a Stirling engine requires that one side of the engine remain hot while the other side remains cool. This requirement makes heat recovery an integral part of the operation of a Stirling engine. Heat can be recovered from dissipation of the heat source and through the use of heat exchangers on the cool side of the engine. Stirling engines have low emissions and create low noise levels. These engines are also mechanically simple, and because there is no internal combustion, the maintenance requirements of Stirling engines are relatively low. However, due to design, Stirling engines are heavy and large for the amount of power generated. Stirling engines also have one of the higher capital costs of distributed power generation technologies. The SOLO 9-kW Stirling engine based micro-CHP unit is shown in Figure 9. Stirling engines use a displacer piston to move the enclosed gas back and forth between the hot and cold reservoirs. The gas expands at the hot reservoir and displaces a power piston, producing work while at the same time forcing the gas to move to the cold reservoir. At the cold reservoir, the gas contracts, retrieving the power piston and closing the cycle. The operation of a Stirling engine can be best understood by examining the operation of a twocylinder (or alpha) Stirling engine. A step-by-step diagram of a two-cylinder Stirling engine and further explanation on the operation of the Stirling engine are shown in Figure 10. In the two-cylinder Stirling engine, one cylinder is kept hot while the other is kept cool. In Figure 10, the lower-left cylinder is heated by burning fuel. The other cylinder is kept cool by an air cooled sink. In a two-cylinder Stirling engine each piston acts as both a power piston and displacer piston.
60
Figure 9. SOLO 9-kW Stirling Engine (Available at http://www.stirling-engine.de/engl/index.html
Application Stirling engines are an old technology; however their use became limited with the improvement of steam engines and the invention of the Otto cycle engine. Recent interest in distributed energy has revived interest in Stirling engines. Stirling engines can be used in a variety of applications due to their thermal output, simple operation, low production costs, and relatively small size. Applications for Stirling engines include power generation units for space crafts and vehicles, small aircraft, refrigeration, micro-CHP, solar dish application, and small scale residential or portable power generation. An overview of Stirling engine characteristics is presented in Table 2. Commercially available Stirling engines can produce between 1 kW 25 kW. Stirling engine technology has been widely used in England and Europe with great success, particularly in the micro-CHP arena. However, a very small percentage of the worlds electrical capacity is currently provided through the use of Stirling engines.
Table 2. Overview for Stirling Engine Technology (Available at http://www.energy.ca.gov/distgen)
Stirling Engine Overview Commercially Available Size Range Fuel Efficiency Environmental Other Features Yes, mostly in Europe <1 kW - 25 kW Natural gas primarily but broad fuel flexibility is possible 12 20% (Target: >30%) Potential for very low emissions Cogen (some models)
61
Cool
Cool
Hot Step 1: Compression. At this point, the majority of the gas has been shifted to the cool cylinder. As the gas cools and contracts , both pistons are drawn outward.
Hot Step 2: Heat Transfer. The now contracted gas is still located in the cool cylinder. Flywheel momentum carries the crank another 90 degrees, transferring the gas to back to the hot cylinder.
Cool
Cool
Hot Step 3: Expansion. Now, most of the gas in the system has just been driven into the hot cylinder. The gas heats and expands driving both pistons inward.
Hot Step 4: Heat Transfer. At this point, the gas has expanded (about 3 times in this example). Most of the gas (about 2/3rds) is still located in the hot cylinder. Flywheel momentum carries the crankshaft the next 90 degrees, transferring the bulk of the gas to the cool cylinder to complete the cycle.
Figure 10. Two-cylinder Stirling Engine Diagram (Available at www.keveny.com/vstirling.html).
Heat Recovery Traditional large-scale electric power generation is typically about 30% efficient, while combined cycle power plants are approximately 48% efficient. In either case, the reject heat is lost to the atmosphere with the exhaust gases. Stirling engines are typically in the range of 15 30% efficient, with many reporting efficiencies of 25 30%. The overall efficiencies of these systems can be greatly increased by recovering the waste heat.
62
A high percentage of the Stirling engine heat losses will go to the cooling fluid instead of into the exhausts, which makes the Stirling engine suitable for combined heat and power generation. Typical operating temperatures range from 1200 1470 F (650 800 C), resulting in electrical engine conversion efficiencies of around 30% to 40% when a recuperator is included in the engine system. These high operating temperatures can convert into high quality waste heat. The reject heat can be recaptured through piping the cooling fluid through a heat exchanger and by ducting the exhaust gases through a heat exchanger to produce hot water.
Cost The capital costs of Stirling engines are comparably high and vary greatly depending upon manufacturer ($2,000 - $50,000). And, overall cost increases with size. However, several companies have targeted Stirling technology for micro-CHP units and have achieved relative success. These companies include PowerGen, WhisperTech, Sunpower, and ENATEC. Advantages and Disadvantages Some advantages of Stirling engines are:
The burning of the fuel air mixture can be more accurately controlled due to the external heat source. Emission of unburned fuel can be eliminated as a continuous combustion process can be used to supply heat. Less lubrication is required leading to greater periods between overhauls because the bearings and seals are placed on the cool side. Simplicity of design; no valves are needed, fuel and intake systems are very simple. Low noise and vibration free operation. Low maintenance and high reliability. Multi-fuel capability. Long service life.
Some disadvantages of Stirling engines are: High costs. Low efficiencies. Require both input and output heat exchangers which must withstand the working fluid pressure and resist corrosion effects. Relatively large for the amount of power they produce due to the heat exchangers. Cannot experience instantaneous start-up. Power output is relatively constant and rapid change to another level is difficult to achieve.
Rankine Cycle Engines Rankine cycle engines are based upon the well known thermodynamic cycle that is used in most commercial electric power plants. The shaft power from a Rankine cycle engine is
63
used to drive an electric generator in the same manner as reciprocating or Stirling engines. Rankine cycle engines have relatively low electrical conversion efficiency. However, as micro-CHP technologies are designed to follow the thermal load, this low electrical efficiency becomes less of a drawback because significant thermal energy that can be recovered from a Rankine cycle engine. The durability and performance characteristics of Rankine cycle engines are also well known, and low production costs are a potential benefit. The construction of a Rankine cycle engine allows heat to be recaptured easily through the use of a condenser, which is already a component in the engine cycle. Currently, Rankine cycle engines for micro-CHP applications are in the development stage. As a result, cost and specific performance characteristics are not yet defined. A Cogen Microsystems 2.5-kW micro-CHP unit based on a Rankine cycle engine is pictured in Figure 11.
Application The advantage of the Rankine cycle for power plants is that the working fluid is a liquid. Many times this liquid is water, which is a cheap and readily available resource. Currently, companies such as the Baxi Group, Enginion, and Cogen Microsystems are exploring the possibility of using Rankine Cycle engines for micro-CHP. Heat Recovery As the Rankine cycle is a closed-loop, which incorporates a condenser, heat recovery can be achieved easily at the condenser. However, as most of the Rankine cycle engine technologies for micro-CHP are still under development, the quality and quantity of heat that can be recaptured is currently not well defined.
Figure 11. Cogen Microsystems 2.5-kW Rankine Cycle micro-CHP Unit (Available at www.cogenmicro.com).
64
Cost Unfortunately, little is known about potential costs of small-scale Rankine cycle units for micro-CHP. As more of these technologies complete the field trial stage of development, more information concerning capital and maintenance costs will be available.
Fuel Cells
Fuel cells are electrochemical energy conversion devices that produce electrical power rather than shaft power. Unlike the technologies discussed previously, fuel cells have no moving parts and, thus, no mechanical inefficiencies. Four major types of fuel cells will be discussed: proton exchange membrane (PEMFC), solid oxide (SOFC), phosphoric acid (PAFC), and molten carbonate (MCFC) fuel cells. Each of these fuel cell types operate differently and exhibit different performance characteristics. In general terms, fuels cells combine a hydrogen based fuel input and gaseous stream containing oxygen in the presence of a catalyst to initiate a chemical reaction. The products of this reaction vary for each type of fuel cell but typically are electrical power, heat, and water. In some instances, other product gases such as carbon dioxide are formed. As a pure hydrogen-rich fuel is required by most fuel cells, hydrogen reformers are often included in a fuel cell system. Like batteries, fuel cells produce direct current (DC) electrical power. This requires that an inverter and power conditioner be used to transform the DC current into alternating current (AC) at the appropriate frequency for use in the majority of applications.
Figure 12. Plug Power Fuel Cell Unit (Available at www.plugpower.com).
65
Fuel cells can achieve high electrical efficiencies as compared to other distributed power generation equipment. Fuel cells exhibit quiet operation and low emissions. Also, the absence of mechanical components decreases maintenance. Unfortunately, the costs of fuel cells are relatively high as compared to other technologies. The fuel flexibility of fuel cells is also low as very pure streams of hydrogen are the only suitable fuel for certain types of fuel cells. In some instances, the energy required to reform the input fuel greatly decreases the overall efficiency of a fuel cell system. Still, fuel cells are a promising technology that hold potential for micro-CHP applications. A Plug Power fuel cell is shown in Figure 12. Typically, a fuel cell electrochemically reacts hydrogen and oxygen to form water and in the process produces electricity. The process consists of a hydrogen-based input fuel passing over an anode where a catalytic reaction occurs, splitting the fuel into ions and electrons. Consider the reactions at the anode and cathode of an acid electrolyte fuel cell. At the anode, the hydrogen gas ionizes, releasing electrons and creating H+ ions in an exothermic reaction. The reaction is
2 H 2 4 H + + 4e
Ions pass from the anode, through the electrolyte, to an oxygen-rich cathode. At the cathode, oxygen reacts with the electrons taken from the electrode and the H+ ions that have traveled through the electrolyte to form water according to the following reaction:
O2 + 4 H + + 4e 2 H 2 O
Figure 13. PAFC Electrochemistry (Available at http://www.fctec.com/fctec).
66
For the reactions to occur simultaneously, electrons produced at the anode must pass through an electric circuit to the cathode while the H+ ions pass through the electrolyte. A schematic diagram of the reaction occurring in a phosphoric acid fuel cell (PAFC) is illustrated in Figure 13. The performance of a fuel cell can be analyzed by using the Gibbs function or Gibbs free energy of reaction. The Gibbs free energy of reaction is an indication of the maximum work that can be achieved from combining two substances in a chemical reaction. The maximum theoretical work (WMAX) is the difference in the Gibbs free energy (G) of the reactants and the products of the chemical reaction.
WMAX = Greact G prod
(4)
The Gibbs free energy is a function of the enthalpy (h), entropy (s), and temperature (T) associated with the chemical compound of interest at a specific state of matter and is defined as
G = h T s
(5)
The temperature-entropy product represents the loss due to changes in entropy. The maximum work that can be achieved by a fuel cell is
WMAX = ( hreact hprod ) T ( sreact s prod )

And the thermal efficiency (th) is defined as
(6)
th =
Greact G prod hreact hprod
(7)
The thermal efficiency of a fuel cell typically ranges from 82 94% in ideal cases. The thermal efficiency of a fuel cell as defined in Equation 4-45 is not the actual electrical conversion efficiency achieved by the fuel cell. Due to ohmic losses resulting from concentration polarization and activation polarization the electrical conversion efficiency of a fuel cell (fc) generally falls between 40 60%. The ohmic losses are accounted for using a voltage efficiency (v) and a current efficiency (i). The electrical conversion efficiency of a fuel cell is defined as the product of the thermal, voltage, and current efficiencies.
fc = th v i
(8)
The components needed for the operation of a fuel cell system vary depending on the type of fuel cell used and the fuel. Common major components for a fuel cell system include a fuel reformer (processor), the fuel cell stack, and a power conditioner. The fuel reformer, also known as a fuel processor, generates hydrogen-rich gas from the supply fuel and removes poisons from the fuel stream. Poisoning and fuel flexibility are two major
67
considerations when selecting a fuel cell. Nitrogen, phosphorous, arsenic, antimony, oxygen (in specific instances), sulfur, selenium and tellurium (Group VA and VI A on the Periodic Table) are typical poisons for fuel cells. Table 3 displays the fuel requirements of each fuel cell type. Fuel reformers are needed because the fuel is often in the form of hydrocarbons, such as methane, and the hydrogen concentration is not at a level suitable for operation. The fuel cell stack consists of layers of the cathode, anode, and electrolyte mentioned earlier. A power conditioner converts the direct current (DC) electricity generated by the fuel cell into alternating current (AC) electricity at the appropriate voltage and frequency. Fuel cells are categorized by the electrolyte used to transport ions between the cathode and the anode (or vise versa). Fuel cell technologies include molten carbonate (MCFC), alkaline (AFC), proton exchange membrane (PEMFC), solid oxide (SOFC), direct methanol (DMFC), and phosphoric acid (PAFC). While PEMFC are arguably the most advanced of the fuel cell technologies, the low operating temperatures are not ideal for CHP applications, particularly if the waste heat is required to drive thermally activated components. Because high temperature or high quality waste is heat needed for micro-CHP applications, the solid oxide (SOFC) and the molten carbonate (MCFC) configurations are two likely candidates. Direct methanol fuel cells also hold some potential because a hydrogen reformer is not required, thereby reducing the system costs. Currently, the PEMFC and the SOFC are regarded as having the most potential for micro-CHP. Table 4 shows the characteristics of the five major types of fuel cell technologies.
PEMFC Proton exchange membrane fuel cells (PEMFC) use an ion-conducting polymer as the electrolyte. The electrolyte works well at low temperatures, typically around 175 212 F (80 100 C), which allows for fast start-up. The polymer membrane construction varies depending on manufacturer; however a standard practice uses a modified polymer known as polytetrafluoroehtylene (PTFE), or as it is commonly know, Teflon. Because the electrolyte is a solid polymer, electrolyte loss is not an issue with regard to stack life. Most PEMFCs use platinum as the catalyst for both the anode and the cathode. Hydrogen is used as the fuel at the anode, and air or oxygen is supplied to the cathode. The reactions that occur within a generic PEMFC are as follows:
Table 3. Fuel Requirements for Fuel Cells (Laramie et al., 2003)
Gas Species H2 CO CH4 CO2 and H2O S (as H2S and COS) PEMFC Fuel Poison (>10 ppm) Diluent Diluent Few studies, to date PAFC Fuel Poison (>0.5%) Diluent Diluent Poison (>50 ppm) MCFC Fuel Fuel Diluent Diluent Poison (>0.5 ppm) SOFC Fuel Fuel Diluent Diluent Poison (>1.0 ppm)
68
Louay M. Chamra and Pedro J. Mago Table 4. Characteristics of Fuel Cells (Available at http://www.oit.doe.gov)
Anode Reaction: H 2 2 H + + 2e CathodeReaction: O2 + 4 H + + 4e 2 H 2O Overall Reaction: 2 H 2 + O2 2 H 2O

The electrode reactions are shown in Figure 14. Hydrogen ions and electrons are produced from the fuel gas at the anode. The hydrogen ions travel through the electrolyte to the cathode. Electrons pass through an outside circuit to join the hydrogen ions and oxygen atoms at the cathode to produce water and product gases. The solid electrolyte does not absorb the water. The operation of a PEMFC requires a certain level of water be maintained
69
in the stack; however, too much water, or a flood of water, will shut the fuel cell down. The optimum water level creates water management issues in that the amount of water remaining in the fuel cell stacks and the amount of water rejected must be carefully controlled. The primary advantage of the PEMFC is that extensive development has resulted in increased electrical efficiency and decreased size and material requirements. However, there are several disadvantages of using a PEMFC for micro-CHP applications. The operating temperature is too low to drive the thermally activated components of the micro-CHP system. Water management and humidity control are critical issues in the operation of a PEMFC. There must be sufficient water content in the polymer electrolyte as the proton conductivity is directly proportional to the water content. However, there must not be so much water that the electrodes are bonded to the electrolyte, blocking the pores in the electrodes or the gas diffusion layer. Also, due to poisoning issues, relatively pure hydrogen is needed at the anode and a hydrogen reformer is necessary, which increases the overall costs of the system. The American Council for an Energy Efficient Economy (ACEEE) Emerging Technologies & Practices (2004) performed a theoretical study on a residential CHP system using a 2-kW PEMFC as the power generation device. The economics indicated that with the estimated costs of the PEMFC unit and maintenance cost, the cost of the electricity generated would be $0.18/kWh. Therefore the installation of the 2-kW PEMFC unit would only be advantageous in an area that has an electrical cost higher than $0.18/kWh, or in an area where grid electricity is unavailable. Thermally activated components were included only implicitly in the study, and the systems were not sized large enough to have excess power either to store or sell to the electricity grid. The study listed several major market barriers: dwindling natural gas supplies, introducing and integrating a new technology and overcoming the inertia of the established market, and uncertain system reliability. Also to be considered is how and who will provide system maintenance. The Entergy Centre of the Netherlands (ECN) experimented with a 2 kW PEMFC microCHP system. The system uses natural gas, and the fuel cell operates at approximately 65 C. For system operation, natural gas is desulphurised and converted to hydrogen rich gas in the reformer. The study showed that a start-up time of 2.5 hours was needed when starting from cold conditions to steady operating conditions where the rejected heat could be used to aid the reforming process. This start-up time was reduced to 45 minutes when transitioning from hot stand-by conditions. Characteristics demonstrated by the system include: Gross (electrical + thermal) efficiency of fuel processor varies from 70% at 1 kW to 78% at full load (10 kW) Stack electrical conversion efficiency of 40% at 2 kW and 42% at 1 kW Recovered 53% of the waste heat (LHV basis)
The results of this study predicted a payback period of five years for a 1 kW micro-CHP system costing 1000 1500 EUR ($1300 - $2000), taking into account the market value of Dutch natural gas, electric rates in the Netherlands, and local energy tariffs. Note that system costs of $1300-$2000 is a theoretical value the manufacturers hope to achieve.
70
SOFC Solid oxide fuel cells use an oxide ion-conducting ceramic material as the electrolyte. The anode of a SOFC is usually a cermet composed of nickel and yttria-stabilised zirconia. A cermet is a mixture of ceramic and metal. The cathode is a porous structure typically made of lanthanum manganite. All of the materials used to construct a SOFC are solid state. SOFCs operate in the temperature range of 800 1100 C. Either hydrogen or methane can be supplied at the anode, and a SOFC can accommodate both oxygen and air at the cathode. The reactions that occur within a generic SOFC are illustrated in Figure 14. The reactions that occur within a generic SOFC are as follows:
Anode Reaction: H 2 + O 2 H 2O + 2e CO + O 2- CO2 + 2e CH 4 + 4O 2- 2 H 2O + CO2 + 8e Cathode Reaction: O2 + 4e 2O 2 Overall Reaction: 2 H 2 + O2 2 H 2O

In the SOFC reactions, hydrogen or carbon monoxide in the fuel stream reacts with oxide ions traveling through the electrolyte. These reactions produce water and carbon dioxide and disperse electrons to the anode. The electrons pass through the exterior load and return to the cathode. At the cathode, the electrons are used to ionize the oxygen molecules from the air. The oxide ions then enter the electrolyte and the process begins again.
Figure 14. SOFC Reactions (Available at http://www.fctec.com/fctec).
71
Solid oxide fuel cells have several advantages from an micro-CHP point of view. First, the high operating temperatures are attractive for driving thermally activated components and for heating domestic hot water. The construction negates any electrolyte management issues or water management difficulties. The presence of nickel at the cathode can be used as an internal reforming catalyst, eliminating the need for a reformer and reducing system cost. The SOFC has challenges and disadvantages. The high temperatures result in construction and material difficulties. The elevated temperatures also decrease the opencircuit voltage achieved by a SOFC because the Gibbs free energy of formation of the products (water) tends to be less negative. The reduced open-circuit voltage leads to a decrease in electrical efficiency. However, the advantage of the quality of the waste heat overcomes the small decrease in the system efficiency. Solid oxide fuel cells hold perhaps the most promise for micro-CHP applications. The European Commission has recognized the natural gas powered, high-temperature fuel cell as being among the most efficient and environmentally friendly means of achieving distributed cogeneration. Several micro-CHP systems using SOFC technology have been developed and tested. Ceramic Fuel Cells Limited (CFCL) has been developing flat plate SOFC technology for thirteen years. In 2004 CFCL announced an all-ceramic stack design capable of volume fabrication for a 1-kW micro-CHP fuel cell system targeting the residential market. The SOFC stack developed by CFCL has been designed with commercial requirements in mind. The stack has the ability to reform fuel at the anode without the use of an external reformer and operates at temperatures of 1470 1600 F (800 870 C). The SOFC stack utilizes a modular design that allows stacks to be arranged in parallel to generate power in a range of 1 10 kW. An example of an SOFC installed for residential micro-CHP use is at the Canadian Centre for Housing Technology in Ottawa. The unit is a 5-kW system that operates on natural gas, using a tubular arrangement developed by Siemens Westinghouse Power Corporation. The unit has the ability to use low fuel pressures which decreases operational costs. The system also includes an inverter that meets residential standards and output control that allows the system to output/receive electricity to the grid. The SOFC unit has been placed in a research house that incorporates simulated occupancy and an intensively monitored real world environment. Researchers intend to monitor heating, ventilation, and air-conditioning conditions to develop methods for better controls systems and grid connections. The Energy Research Centre of the Netherlands (ENC) has also experimented with the use of a 1-kW SOFC micro-CHP unit. The system operates with an HXS 1000 Premiere fuel cell system installed in September 2002. The unit has a thermal output of 2.5 kW and is equipped with an auxiliary heater. The unit is fueled by natural gas, and the operating temperature is approximately 900 C. The SOFC cells are manufactured by InDEC and are arranged in a stack assembly. The system is pictured in Figure 15. The operation of the system begins with the desulphurization of the natural gas which is mixed with de-ionized water vapor. The mixture is internally reformed at the anode, and carbon monoxide and hydrogen are consumed at the anode. Waste heat is recovered by the combustion of the mixture of anode off gas (water vapor and carbon dioxide) and cathode exhaust. The combustion provides sensible heat to the storage vessel. (Oosterkamp et al., 1993)
72
Figure 15. SOFC Stack Assembly (one cell shown) (Oosterkamp et al., 1993).
The SOFC micro-CHP system operates in a heat following mode and has demonstrated the following performance: Electrical efficiency of 25-32% at full load Thermal efficiency of 53-60% on LHV basis Total efficiency of approximately 85%
However, the study also found that the electrical efficiency does decrease over time. The efficiency decrease is rather fast for strongly dynamic system operation, but is also noticeable for a more stable and steady operation.
PAFC Phosphoric acid fuel cells are generally considered the first generation technology. PAFC fuel cells typically operate at a temperature of 390 F (200 C) and can achieve 40 % to 50 % fuel to electrical efficiencies on a lower heating value basis (LHV). The PAFC operates similar to the PEMFC. The PAFC uses a proton-conducting electrolyte (phosphoric acid), and the reactions occur on highly-dispersed electrocatalyst particles supported on carbon black. Phosphoric acid (H3PO4) is the only common inorganic acid that has enough thermal, chemical, and electrochemical stability and low volatility to be considered as an electrolyte for fuel cells. Typically, platinum is used as the catalyst at both the anode and cathode. The reactions that occur within a generic PAFC are as follows:
Anode Reaction: H 2 2 H + + 2e Cathode Reaction: O2 ( g ) + 4 H + + 4e 2 H 2O Overall Reaction: 2 H 2 + O2 2 H 2O
73
Figure 16. PAFC Reactions (Available at http://www.fctec.com/fctec).
In the PAFC reactions, hydrogen is ionized at the anode to produce two hydrogen ions and two electrons. The hydrogen ions pass through the electrolyte to the cathode while the electrons travel through a load in an external circuit. At the cathode, the hydrogen ions, electrons, and oxygen react to form water. The reactions that occur in the PAFC are shown in Figure 16. The PAFC has experienced success in the large-scale CHP arena and holds potential for the micro-CHP market. However, currently few manufacturers are pursuing development of PAFC micro-CHP based system but have expressed more interest in the SOFC and PEMFC fuel cells. One reason for manufacturers lack of interest in PAFC fuel cells is that the PAFC fuel cell uses only pure hydrogen as fuel and other fuel cell types have the capability to utilize a fuel other than a pure hydrogen stream. This selection could also result from potential hazards the electrolyte (phosphoric acid) could pose to inhabitants in the event of a leak. The PAFC requires an external fuel reformer, increasing the production cost.
MCFC Molten Carbonate Fuel Cells (MCFC) are high-temperature fuel cells and promise the highest fuel-to-electricity efficiencies for carbon based fuels. The higher operating temperature allows the MCFCs to use natural gas directly without the need for a fuel reformer. MCFC have also been used with low-Btu fuel gas from industrial processes and other fuel sources. Developed in the mid 1960s, improvements have been made in fabrication methods, performance and endurance. The MCFC operates differently than the fuel cells previously discussed. MCFCs use an electrolyte composed of a molten mixture of alkali metal carbonate salts, which is retained in a ceramic matrix of LiAlO2. Two mixtures are currently used: lithium carbonate and potassium carbonate or lithium carbonate and sodium carbonate. To melt the carbonate salts and achieve high ion mobility through the electrolyte, MCFCs operate at temperatures of
74
600-700 C. When heated to a temperature of around 650 C, the salts melt and become conductive to carbonate ions (CO32-). These ions flow from the cathode to the anode where they combine with hydrogen to give water, carbon dioxide and electrons. Electrons are routed through an external circuit back to the cathode, generating electricity and by-product heat. Unlike other fuel cells, carbon dioxide as well as oxygen needs to be delivered to the cathode. The carbon dioxide and oxygen react to form carbonate ions, which provide the means of ion transfer between the cathode and anode. At the anode, the carbonate ions are converted back into carbon dioxide. There is a net transfer of CO2 from the cathode to the anode; one mole of CO2 is transferred for every two moles of electrons. The operation of a MCFC is illustrated in Figure 17. The reactions occurring in the MCFC are
2Anode Reaction: CO3 + H 2 H 2 O + CO2 + 2e
Cathode Reaction: CO2 ( g ) + 1 O2 + 2e CO32 2 Overall Reaction: 2H 2 + 1 O2 + CO2 (cathode) 2 H 2 O + CO2 (anode) 2
The higher operating temperature of the MCFC creates both advantages and disadvantages compared to fuel cells such as the PAFC and PEMFC with lower operating temperatures. The higher operating temperature allows fuel reforming of natural gas to occur internally, eliminating the need for a fuel reformer. The MCFC can also utilize standard materials for construction, such as stainless steel and nickel-based catalysts. The by-product heat from an MCFC can be used to generate domestic hot water, space heating and cooling, and even high-pressure steam.
Figure 17. MCFC Reactions (Available at http://www.fctec.com/fctec).
Micro-CHP Power Generation Table 5. Overview of Fuel Cell Characteristics (Available at www.energy.ca.gov/distgen/)
Fuel Cells Overview SOFC MCFC Yes Yes 1 kW - 10 MW Natural gas, hydrogen, landfill gas, fuel oil 45-60% Nearly zero emissions Cogen (hot water, LP or HP steam) Commercializati on likely in 2004 250 kW - 10 MW Natural gas, hydrogen
75
Commercially Available Size Range Fuel
PAFC Yes 100-200 kW Natural gas, landfill gas, digester gas, propane 36-42% Nearly zero emissions Cogen (hot water) Some commercially available
PEMFC Yes 3-250 kW Natural gas, hydrogen, propane, diesel 25-40% Nearly zero emissions Cogen (80C water) Some commercially available
Efficiency Environmental Other Features
Commercial Status
45-55% Nearly zero emissions Cogen (hot water, LP or HP steam) Some commercially available
The high operating temperatures and the electrolyte chemistry of the MCFC also lead to disadvantages. The high temperature requires significant time to reach operating conditions and correspondingly slow response time to changing power demands. These characteristics make the MCFC less attractive for dynamic power applications and restrict it to constantpower supply applications. The carbonate electrolyte can also cause electrode corrosion problems. Due to the use of the carbonate ions as the charge carrier, the supply of carbon dioxide to the cathode must be carefully controlled in order to achieve optimum performance. Each fuel cell type has both advantages and disadvantages over its counterparts. The selection of the best fuel cell will depend upon the details of the application. An overview of fuel cell characteristics is given in Table 5.
Application Fuel cells are being developed for stationary power in small commercial and residential markets and as peak shaving units for commercial and industrial uses. Some fuel cells, such as PEMFC, are currently undergoing development for use in automobiles and portable power applications. Phosphoric acid fuel cells have been installed at medical, industrial, and commercial facilities throughout the country, and the 200-kW size is a good match for distributed generation applications. The operating temperature is about 400 F, which is suitable for cogeneration applications. Developers are targeting commercial and light industrial applications in the 100-200 kW power range, for both electric-only and cogeneration applications. The high efficiency and high operating temperature of MCFC units makes them most attractive for base-loaded power generation, either in electric-only or cogeneration modes.
76
Potential applications for the MCFC are industrial, government facilities, universities, and hospitals. Solid oxide fuel cells are being considered for a wide variety of applications, especially in the 5 250 kW size range. These applications include small commercial buildings, industrial facilities, micro-CHP, and base load utility applications. Proton exchange membrane fuel cells are currently undergoing the most rapid development of any fuel cell type. Part of this development has been driven by the desire of automotive manufacturers to develop a fuel cell powered automobile. This surge in development has led to breakthroughs for stationary power applications as well. Research is aimed at commercial-sized power generation (e.g., Ballard's 250 kW unit) and residential power generation (e.g., Plug Power's 3-5 kW units). For the units to achieve market potential, natural gas is selected as the fuel of choice. Reject system heat in the form of hot water makes them particularly attractive for cogeneration, which is included in almost all products currently under development.
Heat Recovery The type of fuel cell determines the temperature of the heat rejected during operation and directly influences the fuel cell types suitability for micro-CHP applications. Low temperature fuel cells create waste heat suitable for producing hot water and in some cases, low pressure steam. Lower temperature fuel cells such as the PAFC and PEMFC produce lower quality waste heat and are suitable for small commercial and industrial cogeneration applications. The MCFC and the SOFC operate at high temperatures and are capable of producing waste heat that can be used to generate steam for use in a steam turbine, or combined cycle microturbine. If space cooling is considered and an absorption chiller is to be used, the recaptured heat should be at a temperature of at least 185 F (85 C). Cost The initial cost of fuel cells is higher than those of other electricity generation technologies. The only product available commercially today is the PureCell 200 (formerly PC-25) built by UTC Power. The cost of the unit is approximately $4,000/kW. The installed cost of the unit approaches $1.1 million. At a rated output of 200 kW, this translates to about $5,500/kW, installed. However, on January 3, 2005, Delphi Corp., in partnership with the DOEs advanced fuel cell development program, reported that researchers have exceeded the governments $400/kW power cost goal for fuel cells. At this price, fuel cells could compete with traditional gas turbine and diesel electric generators and become viable power suppliers for the transportation sector. Table 6 shows the uninstalled projected longterm costs of fuel cell technologies. The price of $400/kW is not included in Table 4-7 as the information was only recently released and currently, the development has not been proven for production costs in a fuel cell of substantial size.
Micro-CHP Power Generation Table 6. Projected Long-Term Costs of Fuel Cell Technologies (Available at www.energy.ca.gov/distgen/)
Emerging Fuel Cell Technologies Projected Cost (Long-term, Uninstalled) $1,200-1,500/kW $1,000-1,500/kW Initially $5,000/kW Long term $1,000/kW
77
Technology MCFC SOFC PEMFC
Table 7. Advantages and Disadvantages of Fuel Cell Types (Available at www.energy.ca.gov/distgen/)

PAFC Advantages Quiet Low emissions High efficiency Proven reliability Advantages Quiet Low emissions High efficiency Synergy with Automotive R&D Disadvantages High Costs Low energy density PEMFC Disadvantages High Costs Limited field test experience Low temperature waste heat may limit cogeneration potential SOFC Disadvantages High Costs Planar SOFC's are still in the R&D stage but recent developments in low temperature operation show promise MCFC Advantages Quiet Low emissions High efficiency Self reforming High Costs Need to demonstrate long term reliability Disadvantages
Advantages Quiet Low emissions High efficiency High energy density Self reforming
As no combustion is occurring, and there are no moving parts, fuel cells are expected to have minimal maintenance requirements. The primary maintenance will be focused on preventing poisoning of the catalyst and periodic inspection and maintenance to the fuel
78
supply system and fuel reformers. The cell stack itself will not require maintenance until the end of its service life. Fuel cell system maintenance requirements vary with the type of fuel cell, size, and maturity of the equipment. Major overhaul of fuel cell systems involves shift catalyzer replacement, reformer catalyzer replacement, and stack replacement. The maintenance and reliability of the system still needs to be proven in a long-term demonstration. Maintenance costs of a fuel cell are expected to be comparable to that of a microturbine, ranging from $0.005-$0.010/kWh (based on an annual inspection visit to the unit). (Available at http://www.energy.ca.gov/distgen/)
Advantages and Disadvantages Each fuel cell type will have advantages and disadvantages in certain areas, both as compared to other fuel cell technologies and other DER equipment. Table 7 displays some of the advantages of the four primary types of fuel cells. Fuel cells convert chemical energy directly into electricity without the combustion process. As a result, a fuel cell does not incur losses resulting from mechanical inefficiencies. Fuel cells can achieve high efficiencies in energy conversion terms, especially where the waste heat from the cell is utilized in cogeneration. A high power density allows fuel cells to be a relatively compact source of electric power, a benefit in applications with space constraints. In a fuel cell system, the fuel cell itself is often smaller than the other components of the system such as the fuel reformer and power inverter. Fuel cells, due to their nature of operation, are extremely quiet in operation. This allows fuel cells to be used in residential or built-up areas where the noise pollution is undesirable. Unfortunately, the primary disadvantage of the fuel cells is the cost. The two basic reasons are high component costs compared to other energy systems technology and fuel cell operation requires a continuous, highly selective, expensive fuel supply.
Heat Recovery
Electrical power generation devices do not convert 100% of an energy source potential into usable energy. Electrical efficiencies of reciprocating engines, microturbines, Stirling engines, and fuel cells are about 50%, 30%, 30%, and 60%, respectively. This means that the electrical power generation devices fail to utilize 40% - 70% of the available energy. Energy that is not converted to electrical or shaft power is rejected from the process in the form of waste heat. In order to utilize more of the energy stored in a given fuel and increase the overall thermal efficiency of a system, heat recovery must be incorporated into a system. Heat recovery converts waste heat to useful energy and is primarily accomplished through the use of heat exchangers. Distributed energy generation prime movers possess waste heat that can be recovered as useful energy. The type of prime mover determines the characteristics of the waste heat and the effectiveness with which useful energy can be recovered. Waste heat is typically released in the form of hot exhaust gases, process steam, or process liquids/solids. The usable temperature for heat recovery is listed for various prime movers in Table 8. Potential uses for the waste heat are hot water, space heating and cooling, and process steam.
Micro-CHP Power Generation Table 8. Waste Heat Characteristics of Prime Mover Technologies
Usable Temp. for micro-CHP (F) 180 - 900 300 - 500 800 - 1000 140 - 750 400 - 650
79
Diesel Engine Natural Gas Engine Stirling Engine Fuel Cell Microturbine
Recovered heat that is utilized in the power generation process is internal heat recovery, and recovered heat that is used for other processes is external heat recovery. Recuperators, turbochargers, and combustion pre-heaters are examples of internal heat recovery. Absorption chillers, desiccant dehumidification devices, and heat recovery steam generators are examples of external heat recovery components. Recovered heat is classified as low-temperature when less than 445 F (<230 C), medium-temperature 445 F 1202 F (230 C 650 C), or hightemperature when greater than 1202 F (>650 C). The majority of heat exchangers used in micro-CHP are used for external heat recovery.
Technology Overview Because combustion exhaust gases or process liquids cannot be used directly in many applications, heat exchangers are used to facilitate the transfer of waste heat thermal energy to heat recovery applications. Devices that transfer energy between fluids at different temperatures by heat transfer modes are known as heat exchangers. Heat exchangers come in a wide variety of sizes, shapes, and types and utilize a wide range of fluids. Applications for heat exchangers are vast, ranging from heating and air-conditioning systems, to chemical processing and power production in large plants. Heat exchanger classification is based upon component and fluid characteristics. Several classification schemes have been proposed for heat exchangers. Hewitt et al., (1994) suggests the following four-tiered system:
5. 6. 7. 8. Recuperator/Regenerator Direct-contact/Transmural heat transfer Single phase/Two-phase Geometry
A recuperator is based on a continuous transfer of heat between two fluid streams. A regenerator is a device which uses a heat-absorbing material, alternately cooled and heated in a batch mode, to transfer heat between two streams; these are often rotary devices. Directcontact heat exchangers, such as a feedwater heater used in a power plant, allow the two fluids to come into contact with each other. Transmural heat exchangers separate the two fluid streams using a wall or series of walls. Single phase/two-phase refers to the physical state of the fluids flowing in the heat exchanger. Single-phase flow implies that both fluids are either completely gaseous or liquid. If evaporation or condensation is taking place, the device involves two phases. Geometry refers to the basic shape of the heat-exchanger passages that contain the fluid.
80
This section will focus on transmural recuperators with fluids in single-phase; however, other arrangements will also be discussed. Transmural recuperators with single-phase flow can be divided further by flow arrangement into the following categories: Counterflow Cross-flow Parallel flow
Counterflow and parallel flow tube-within-a-tube and cross-flow configurations are shown in Figure 18. The temperature versus area diagrams for the parallel flow and counterflow arrangements are shown in Figure 19. In the parallel arrangement, the fluid temperatures approach each other so that the temperature difference T1 is much greater in magnitude than T2. If the length of the heat exchanger were extended long enough in parallel flow, the exit temperatures of each of the fluids would be approximately equal. The counterflow average T will be larger than the parallel flow average T. Cross-flow heat exchangers are arranged so that the two fluids flow perpendicular to each other as shown in Figure 20. An important concept when discussing cross-flow heat exchangers is mixing. A fluid is said to be unmixed if the passageway contains the same discrete portion of fluid throughout its traverse of the exchanger. If the fluid passageways are such that fluid from one passageway can mix with fluid from others, the fluid is mixed. In Figure 20, mixed and unmixed arrangements are illustrated. Cross-flow heat exchangers are classified as both fluids unmixed, both fluids mixed, or one fluid mixed and one fluid unmixed. When several tubes are placed inside a shell, the classical single-pass shell-and-tube heat arrangement results. A schematic of a shell-and-tube heat exchanger with one shell pass and one tube pass is illustrated in Figure 21. Baffles are often placed inside a shell-and-tube heat exchanger to promote higher heat transfer rates and increase the effectiveness of the heat exchanger.
Figure 18. Main Types of Transmural Recuperators with Fluids in Single-phase.
81
T h in
T h in
Temperature
Temperature
T h o ut
T 1
Tc
out
T 1
T 2
T c out
T
T 2
T h o ut
T c in
T c in
Area
Area
(a) Parallel Flow
(b) Counter Flow
Figure 19. Temperature-area Diagram of Parallel and Counterflow Arrangements.
Figure 20. Cross-flow Heat Exchangers.
Figure 21. Shell-and-tube Heat Exchanger.
82

Gas-to-Gas Waste Heat Recovery
Plat-fin and primesurface heat exchangers
Heat pipe heat exchangers
Rotary Regenerators (a) Gas-to-Liquid Waste Heat Recovery
Convection Recuperators
Liquid coupled heat exchangers
Waste heat boilers
Economizers (b)
Fluidized bed heat exchangers
Heat pipe heat exchangers
Liquid-to-Liquid Waste Heat Recovery
Shell-and-tube heat exchangers (c)
Plate heat exchangers
Figure 22. Classification of Waste Heat Recovery Heat Exchangers (Recreated from the CRC Handbook of Energy Efficiency).
Application Waste heat recovery heat exchangers may be classified as gas-to-gas, gas-to-liquid, and liquid-to-liquid. The various types of these classifications are displayed in Figure 22. Gas-to-Gas Heat Exchangers Gas-to-gas waste heat recovery exchangers can be used as recuperators to preheat combustion air in IC engines and microturbines. Rotary regenerators are often used in Stirling engines to recover and store heat. Gas-to-gas heat exchangers find many applications in micro-CHP systems. Metallic radiation recuperators, convection recuperators, and a runaround coil are the three primary types of gas-to-gas heat exchangers. A metallic radiation recuperator is a tubein-tube heat exchanger that consists of two concentric metal tubes. Hot exhaust (flue) gas flows through the inner tube, and the air to be preheated flows through the outer tube or annulus. This type of recuperator can act as a part of a chimney, flue, or exhaust line. The majority of the heat is transferred from the hot gas to the inner wall through radiation. The heat transfer in the outer tube then takes place through convection. Air and gas flowing in counterflow is the most desirable arrangement because this arrangement has a high performance. Metallic radiation recuperators typically achieve an effectiveness of 40% or lower. Though metallic radiation recuperators could be used in micro-CHP applications, their use will likely be limited by the relatively low effectiveness which would result in large heat exchangers.
83
Convection recuperators are cross-flow heat exchangers with flue gases flowing normal to a tube bundle with air to be preheated flowing in the tubes. Convection recuperators can be used in low-temperature applications such as space heating, direct-fired absorption chillers, or return air heating in a desiccant dehumidification system. A runaround coil consists of two connected heat exchangers that circulate a working fluid. The working fluid is heated by the waste gas with one heat exchanger and then circulated to the other heat exchanger where the fluid heats a stream of cool air. Runaround coils are used in HVAC applications and can be coupled with distributed generation components to produce warm air for district heating, to heat return air for desiccant dehumidification devices, or to fire an absorption chiller (Shah, 1997).
Gas-to-Liquid Heat Exchangers Gas-to-liquid heat exchangers include economizers, waste heat boilers, heat-pipe heat exchangers, fluidized-bed, and thermal fluid heaters. Economizers and thermal fluid heaters are used for low- to medium-temperature waste heat recovery while waste heat boilers, heatpipe heat exchangers, and fluidized bed heat exchangers are used for medium- to hightemperature heat recovery. Economizers are most often used with boilers to preheat the boiler feedwater. In other applications, economizers are referred to as a process fluid or water heaters. An economizer is an individually finned-tube bundle, with gas flowing outside normal to the finned tubes and water flowing inside the tubes. Thermal fluid heaters are double-pipe heat exchangers that use waste heat gases to heat a high-temperature organic heat transfer fluid. Thermal fluid heaters operate on the same principle as a domestic warm-water system, except that the water is replaced by a hightemperature organic heat transfer fluid. The heat transfer fluid can be circulated and used for heating and heat-driven absorption chillers. Fluidized-bed heat exchangers utilize water, steam, or another heat transfer fluid heated by waste heat gases that flow over a bed of finely divided solid particles. As the waste heat fluid reaches a critical velocity, the particles begin to float, and the resulting mixture acts as a fluid. Economizers and thermal fluid heaters are more likely to be used in micro-CHP systems than either boilers or fluidized-bed heat exchangers due to the amount of thermal energy available for heat recovery and the size requirements of fluidized-bed heat exchangers. Process steam is also less likely to be needed for micro-CHP applications, eliminating the need for waste heat boilers. Liquid-to-Liquid Heat Exchangers Liquid-to-liquid waste heat recovery heat exchangers are typically used in industrial applications. Shell-and-tube heat exchangers and plate heat exchangers are typically used for liquid-to-liquid heat recovery. Liquid coolant systems in reciprocating engines offer liquid-toliquid heat recovery opportunities from hot oil and other liquid coolants. However, liquid-toliquid heat recovery will not be as significant a contributor in micro-CHP applications as are gas-to-gas and gas-to-liquid waste heat recovery exchangers. (Shah, 1997)
84
THERMALLY ACTIVATED DEVICES

Thermally activated devices are technologies that utilize thermal energy rather than electric energy to provide heating, cooling, or humidity control. The primary thermally activated components used in micro-CHP systems are desiccant dehumidifiers and absorption chillers.
Absorption Chillers
Technology Overview Absorption chiller technologies are one of a group of technologies classified as heat pumps. Heat pumps may be either heat driven or work driven. Absorption technologies are heat driven, transferring heat from a low temperature to a high temperature using heat as the driving energy. Heat pumps operate on the principle of the absorption refrigeration cycle, which is similar to the vapor-compression cycle. Both the absorption refrigeration cycle and the vapor-compression cycle will be examined to draw analogies between the two. Vapor-compression refrigeration systems are the most common refrigeration systems in use today. The vapor-compression cycle is a work-driven cycle that is illustrated in Figure 23. In the vapor-compression cycle, work is input to compress the refrigerant to a high pressure and temperature at State 2. At State 2, the refrigerant condensation temperature is below the ambient temperature. As the high-pressure and high-temperature refrigerant vapor passes through the condenser, heat is rejected to the ambient air and the refrigerant vapor condenses to a liquid to achieve State 3. The high-pressure liquid at State 3 passes through an expansion valve. As the liquid passes through the expansion valve, the refrigerant experiences a reduction in both temperature and pressure to reach State 4. At State 4, the boiling temperature of the refrigerant is lower than that of the surroundings. The low-pressure liquid refrigerant passes through the evaporator, absorbing heat from the ambient environment when boiling occurs in the evaporator and creating a low-pressure refrigerant vapor at State 1. The low-pressure refrigerant vapor at State 1 enters the compressor completing the cycle. The absorption cycle has some features in common with the vapor compression cycle. For example, the absorption cycle has a condenser, an evaporator, and an expansion valve. However, the absorption cycle and the vapor-compression cycle differ in two very important aspects. The absorption cycle uses a different compression process and different refrigerants than the vapor-compression cycle. The absorption cycle operates on the principle that some substances (absorbents) have an affinity for other liquids or vapors and will absorb them under certain conditions. Instead of compressing a vapor between the evaporator and condenser as in Figure 23, the refrigerant of an absorption system is absorbed by an absorbent to form a liquid solution. The liquid solution is then pumped to a higher pressure. Because the average specific volume of a liquid is much smaller than that of the refrigerant vapor, significantly less work is required to raise the pressure of the refrigerant to the condenser pressure. This corresponds to less work input for an absorption system as compared to a vapor-compression system.
85
Qout
3 2
Condensor
Expansion Valve Evaporator
Compressor
Win
Qin
Figure 23. Vapor-Compression Cycle Schematic.
Qout
3
Z
2
Qin
Condensor
v Expansion Valve Evaporator
Generator
Absorber
Win
4 1 Pump
Qin
Qout
Figure 24. Basic Absorption Cycle Schematic.
86
Because the absorbent used in the absorption cycle forms a liquid solution, some means must also be introduced to retrieve the refrigerant vapor from the liquid solution before the refrigerant enters the condenser. This process involves heat transfer from a relatively hightemperature source. Because the thermal energy input into the system is much higher than the work input through the pump, absorption chillers are considered to be heat driven. The components used to achieve the pressure increase in an absorption chiller are viewed as a thermal compressor and replace the compressor in the vapor-compression cycle shown in Figure 23. The components of the absorption cycle are shown schematically in Figure 24. The components of the thermal compressor are a pump, an absorber, and a (heat) generator and are shown to the right of the dashed Z-Z line. The components to the left of the dashed Z-Z line are the same as the ones used in the vapor-compression system. The operation of the absorption cycle shown in Figure 24 is as follows: At State 1, the low-pressure refrigerant vapor exits the evaporator and enters the absorber. In the absorber, the refrigerant vapor is dissolved in an absorbent and rejects the heat of condensation and the heat of mixing to form a liquid solution. The refrigerant/absorbent solution is then pumped to the condenser pressure and passed to the generator. In the generator, heat is added to the refrigerant/absorbent solution to vaporize the refrigerant, removing the refrigerant from the solution. The liquid absorbent has a higher boiling temperature than the refrigerant and, therefore, stays in the liquid form. There are two streams exiting the generator. The refrigerant exits to the condenser at a high temperature and pressure (State 2) while the absorbent passes through an expansion valve, decreasing the pressure of the absorbent to the evaporator pressure before entering the absorber again. The remainder of the operation is much the same as the vapor compression cycle. The high-temperature, high-pressure refrigerant vapor at State 2 enters the condenser with a pressure such that the ambient temperature is higher than the condensation temperature of the refrigerant. The refrigerant vapor condenses as it passes through the condenser, rejecting heat to the ambient environment to achieve State 3. At State 3, the high-pressure, low-temperature liquid refrigerant enters the expansion valve where the refrigerant experiences a decrease in pressure to the evaporator pressure. The low-pressure, low-temperature liquid refrigerant that results at State 4 is at a pressure such that the boiling temperature of the refrigerant is lower than the ambient temperature of the environment. As the liquid refrigerant passes through the evaporator, the refrigerant boils, absorbing heat from the ambient air. The refrigerant exits the evaporator as a high-temperature, low-pressure vapor to complete the cycle.
Refrigerant-absorbent Selection
Though all absorption chillers operate on the basic cycle presented in Figure 24, each chiller design is dependent on the refrigerant-absorbent selection. Current refrigerant/ absorber media for absorption chillers are either water/lithium bromide or ammonia/water. Water/lithium bromide absorption chillers utilize water as the refrigerant and lithium bromide as the absorbent. Because water is used as the refrigerant, applications for the water/lithium bromide absorption chillers are limited to refrigeration temperatures above 0 C. This combination of refrigerant and absorbent is advantageous in areas where toxicity is a concern because lithium bromide is relatively non-volatile. Absorption machines based on water/lithium bromide are typically configured as water chillers for air-conditioning systems
87
in large buildings. Water/lithium bromide chillers are available in sizes ranging from 10 to 1500 tons. The coefficient of performance (COP) of these machines typically falls in the range of 0.7 to 1.2 (Herold et al., 1996). The ammonia/water combination utilizes ammonia as the refrigerant and water and the absorbent. The use of the ammonia as the refrigerant allows much lower refrigerant temperatures (the freezing temperature of ammonia is -77.7 C); however, the toxicity of ammonia is a factor that has limited the use of ammonia/water chillers to well-ventilated areas. In commercial and residential building applications where there is insufficient ventilation, emissions from an ammonia/water absorption chiller could be harmful to occupants. The primary selling point of ammonia/water absorption chillers is their ability to provide direct gas-fired and air-cooled air conditioning. Ammonia/water absorption chillers are commonly sold as a single component in an air-conditioning system; however, use is restricted in some densely populated areas. Ammonia/waster chillers are available in capacities ranging from 3 to 25 tons with COPs generally around 0.5 (Herold et al., 1996). A schematic of an absorption cycle of ammonia/water is shown in Figure 25. The addition of a heat exchanger is common in all absorption chillers to increase the efficiency of the thermal compressor. The hot solution leaving the generator is used to preheat the refrigerant/absorbent solution entering the generator. A rectifier is also included in the system of an ammonia/water chiller. This is because the ammonia vapor leaving the generator often includes a low concentration of water vapor. This water vapor will freeze in the expansion valve, negating the operation of the system.
Qout
2 3
Qin
Condensor
v Expansion Valve Evaporator
Generator
Heat Exchanger
4 Absorber
Qin
Win
Pump
Qout
Figure 25. Ammonia/Water Absorption Cycle.
88
Types of Absorption Chillers Absorption chillers are classified as single-effect, double-effect, or triple effect. Singleeffect absorption chillers contain one stage of generation, such as the systems shown in Figures 24 and 25. Single-effect absorption chillers use low-pressure steam or hot water as the energy source. Typical temperature requirements range from 200 to 270 F (93 to 132 C). Steam-powered systems operate at pressures between 9 and 15 psig. When the supplied temperatures are below the design specifications, the chiller capacity is reduced. Double-effect absorption systems use a second generator, condenser, and heat exchanger that operate at higher temperature. A double-effect water/lithium bromide absorption system is shown schematically in Figure 26. Refrigerant vapor is recovered from the first-stage generator in the high-temperature condenser. The refrigerant/absorbent in the second-stage generator is at a lower temperature than the solution in the first-stage generator. The refrigerant vapor from the first stage generator flows through the second-stage generator where a portion of the refrigerant condenses back into liquid while the remainder remains in the vapor phase. Additional refrigerant is vaporized in the second-stage generator by the high temperature and the heat of vaporization supplied by the refrigerant from the first-stage generator. The refrigerant vapor from both generator stages flows to the condenser while the absorbent solution flows back to the absorber.
89
The purpose of the two stages that make up the double-effect absorption cycle is to increase the COP of the cycle. This is made possible through the use of the recuperative heat exchangers used in the system. Double-effect chillers yield higher COPs than single-effect chillers. The COP for double-effect absorption chillers varies from 1.0 to 1.2 for water/lithium bromide chillers (Herold et al., 1996). Triple-effect absorption chillers have been in prototype development for several years. These chillers will be direct-fired and are expected to provide a 50% thermal efficiency improvement over double-effect absorption chillers. Triple-effect absorption chillers do not feature a distinct third generator stage; rather they use internally-recovered heat to achieve high efficiencies. Triple-effect water/lithium bromide chillers can achieve COPs of 1.6 and greater (Petchers, 2003). One of the most promising absorption technologies is the generator-absorber heat exchange (GAX) cycle. GAX chillers use an ammonia/water working fluid. GAX-cycle systems hold great potential for residential and light-commercial applications and provide capacities as low as 3 tons. GAX absorption chillers have obtained COPs of approximately 0.7.
Application Absorption chillers can be directly fired or indirectly fired. Direct-fired absorption chillers utilize a natural gas burner and can supply waste heat for a desiccant dehumidification device or hot water. Direct-fired chillers are often used in areas where electric rates are high and gas utilities offer lower rates or rebate programs to replace vapor-compression chillers. Indirect-fired fired absorption chillers are utilized where there is an existing source of heat that can be recovered. The supplied heat can be in the form of hot water, steam, or exhaust gases. All of the prime mover technologies that are applicable to micro-CHP can produce waste heat sufficient to drive an absorption chiller. This coupling ability makes absorption cycle chiller systems very desirable for micro-CHP applications. An absorption chiller in an micro-CHP system may not utilize all of the waste heat that is input into the chiller. Just as with direct-fired chillers, this remaining heat may be used in a desiccant dehumidification device or to produce hot water. The temperature of the waste heat available from a power source determines the appropriate absorption configuration. Table 9 matches the waste heat temperatures typical of various prime movers with appropriate absorption configurations. The match is based on the temperature of the waste heat that could be obtained to drive the generator in an absorption chiller cycle. Absorption chillers offer many advantages over electric chillers, especially when there is a source of waste heat available. As compared to electric chillers, absorption chillers have lower operating costs, shorter payback periods, quiet operation, low maintenance, and high reliability. Absorption systems also operate at lower pressures and offer safer operation. However, absorption chillers have higher initial costs and are not as widely available as electric chillers. Vapor-compression systems are much more widely manufactured and more available than absorption chillers. Still, the fact that absorption chillers do not have mechanical compressors and have fewer moving parts gives absorption technologies an advantage over vapor-compression systems in terms of lower maintenance, higher reliability and quieter operation.
90
Louay M. Chamra and Pedro J. Mago Table 9. Matching of Power Generation and Absorption Technology
Power Generation and Absorption Technology
Power Source Microturbine Reciprocating Engine SOFC PAFC PEMFC
Temperature (F) ~600 ~180 ~900 ~250 ~140
Matching Technology Triple-, double-, or single-effect Single-effect Triple-, double-, or single-effect Double-effect or single effect Single-effect
Cost The capital cost of installing an absorption chiller is generally more than installing an equivalent electric or engine-driven chiller. The RS Means Mechanical Cost Data (2001) for a 100 ton steam or water-fired absorption chiller is presented in Table 10 (replace with 10 ton unit cost).
Table 10. RS Means Cost Data for a 100 Ton Absorption Chiller Installation
Cost Source Material Labor Total Cost ($) 110,500 7,975 118,475
Desiccant Dehumidification Technologies

Controlling humidity in a conditioned space can be important for a number of reasons. Concerns include humidity damage to moisture sensitive items, product protection from moisture degradation (foods, grains, and seeds), mildew growth, corrosion, and health issues. In the last thirty years, applications have expanded for desiccant technologies to include supermarkets, hospitals, refrigerated warehouses, ice rinks, hotels, and retail establishments. Recently, development has turned to applying desiccant dehumidification to commercial and residential buildings. Desiccants are materials that attract and hold moisture. A desiccant dehumidifier is a device that employs a desiccant material to produce a dehumidification effect. The process involves exposing the desiccant material to a high relative humidity air stream, allowing the desiccant to extract and retain a portion of the water vapor, and then exposing the same desiccant to a lower relative humidity air stream where the retained moisture is drawn from the desiccant.
91
Figure 27. Sub-cooling Dehumidification Process (Chamra et al., 2000).
This section presents an overview of desiccant dehumidification. The principles of the sub-cooling system are introduced to contrast the differences between desiccant and conventional air dehumidification and to highlight the advantages of desiccant dehumidification. Also discussed are the principles of desiccant systems, types of desiccant systems, and cost considerations for choosing desiccant systems. Desiccant dehumidification technologies are an attractive component of an micro-CHP system because desiccant regeneration provides an excellent use for waste heat.
Sub-cooling Systems vs. Desiccant Systems In traditional air-conditioning systems, such as the units installed in most U.S. homes, dehumidification is achieved by cooling a moist air stream to a temperature below the dewpoint so that water (liquid) condenses out of the air. This process is familiar to anyone who has seen moisture condense on the exterior of a glass of ice water on a humid day. An example of the sub-cooling process is illustrated on a psychometric chart in Figure 27. The processes shown are for air being cooled and dehumidified from conditions of 95 F dry bulb (db) and 75 F wet bulb (wb) to 77 F db and 58 grains/lbmda. The resulting air lies approximately in the center of the ASHRAE Summer Comfort Zone shown in Figure 28. Initially, the dry bulb temperature of the air decreases, while the moisture content remains constant. The dry bulb temperature continues to decrease as moisture begins to condense out of the air onto the cooling coil, resulting in a decrease in the moisture content. In order to deliver air at the desired drier condition of approximately 45% relative humidity, some form of reheating must be used. The reheat process path is also illustrated in Figure 27. In this example, the total net cooling load is 10.8 Btu/lbmda, and of this, 6.4 Btu/lbmda, or about 59%, is latent load.
92
Figure 28. ASHRAE Comfort Zones (ASHRAE Fundamentals, 2001).
Summary of the Principles of Sub-cooling Systems The same equipment is used for both the sensible cooling and dehumidification in the conventional system. If independent humidity and temperature control are required in a system, a provision for reheat of the cooled air must be included. In the example above, the net cooling load is 10.8 Btu/lbmda, but the load on the cooling coil is 16 Btu/lbmda with the difference (5.3 Btu/lbmda) being added back in during the reheat process. Thus, energy is used both for the extra cooling and for reheat. Another disadvantage of this approach is that the air leaving the evaporator coil is nearly saturated, with relative humidity typically above 90%. This moist air travels through duct work until the air is either mixed with dryer air or reaches the reheat unit. The damp ducts, along with the wet evaporator coils and standing water in a condensate pan (Figure 29), can generate problems with microbial growth and the associated health and odor problems.
93
Figure 29. Damp Duct Symptoms (Chamra et al., 2000).
Summary of Principles of Desiccant Systems Desiccant dehumidification systems remove moisture from the air by forcing the water vapor directly into a desiccant material. The moisture from the air is attracted to desiccants since an area of low vapor pressure is created at the surface of the desiccant. The pressure exerted by the water in the air is higher, so the water molecules move from the air to the desiccant and the air is dehumidified. The functioning of desiccant material might be compared to the action of a sponge in collecting a liquid. When the sponge is dry, it soaks up the liquid effectively. Once it becomes saturated, the sponge is taken to a different spot, the liquid is expelled by squeezing the sponge, and the sponge is ready to absorb more liquid. In a desiccant system, if the desiccant material is cool and dry, its surface vapor pressure is low, and moisture is attracted and absorbed from the air, which has a higher vapor pressure. After the desiccant material becomes wet and hot, it is moved to another air stream and the water vapor is expelled by raising the temperature (this step is called "regeneration"). After regeneration, the desiccant material is ready to be brought back to absorb more water vapor. Unlike the conventional cooling coil, the water vapor does not condense, but rather remains a vapor throughout all processes. Desiccants can be either solids or liquids. The difference between solid and liquid desiccants is their reaction to moisture. Some simply collect moisture like a sponge collects water. These desiccants are called adsorbents and are mostly solid materials. Silica gel is an example of a solid adsorbent. Other desiccants undergo a chemical or physiological change as they collect moisture. These are called absorbents and are usually liquids or solids, which become liquid as they absorb moisture. Lithium chloride collects water vapor by absorption. Sodium chloride, common table salt, is another example of an absorbent.
94
Types of Desiccant Systems

General Classifications Most commercial desiccant dehumidification systems use as their working material either a solid adsorbent or a liquid absorbent. Briefly, absorption is a process in which the nature of the absorbent is changed, either physically, chemically, or both. The change may include formation of a hydrate or phase change. An adsorbent, on the other hand, does not change either physically or chemically during the sorption process. A variety of factors dictate whether an adsorbent will be commercially useful. These include cost, long-term stability, moisture removal characteristics (rate, capacity, saturation conditions, suitable temperatures), regeneration requirements (rate of moisture surrender as a function of temperature and humidity), availability, and manufacturing considerations. Solid Adsorbents Silica gels and zeolites are used in commercial desiccant equipment. Other solid desiccant materials include activated aluminas and activated bauxites. The desiccant material choice for a particular application depends on factors such as the regeneration temperature, the level of dehumidification, and the operating temperature. Solid desiccant materials are arranged in a variety of ways in desiccant dehumidification systems. A large desiccant surface area in contact with the air stream is desirable. A way to bring regeneration air to the desiccant material is necessary. The most common configuration for commercial space conditioning is the desiccant wheel shown in Figure 30a. The desiccant wheel rotates continuously between the process and regeneration air streams. The wheel is constructed by placing a thin layer of desiccant material on a plastic or metal support structure. The support structure, or core, is formed so that the wheel consists of many small parallel channels coated with desiccant. Both "corrugated" and hexagonal (Figure 30b) channel shapes are currently in use. The channels are small enough to ensure laminar flow through the wheel. Some kind of sliding seal must be used on the face of the wheel to separate the two streams. Typical rotation speeds are between 6 and 20 revolutions per hour. Wheel diameters vary from one foot to over twelve feet. Air filters are an important component of solid desiccant systems. Dust or other contaminants can interfere with the adsorption of water vapor and quickly degrade the system performance. All commercial systems include filters and maintenance directions for keeping the filters functioning properly. Liquid Absorbents Some materials that function as liquid absorbents are ethylene glycol, sulfuric acid, and solutions of the halogen group such as lithium chloride, calcium chloride, and lithium bromide (ASHRAE Fundamentals Handbook, 2001). A generic configuration for a liquid desiccant system is illustrated in Figure 31. The process air is exposed to a concentrated desiccant solution in an absorber, usually by spraying the solution through the air stream. As the solution absorbs water vapor from the air stream, the solution concentration drops, and the weak solution is taken to a regenerator where heat is used to drive off the water (which is carried away by a regeneration air stream) and the concentrated solution is returned to the absorber.
95
Figure 30. (a) Desiccant Wheel (Meckler et al., 1995) (b) Corrugated and Hexagonal Channel Shapes (Chamra et al., 2000).
Figure 31. Liquid Desiccant System (Chamra et al., 2000).
Regeneration For solid or liquid systems, regeneration energy can be drawn from a variety of sources. In an micro-CHP system, regeneration energy is drawn from the waste heat of a powergeneration component. Due to the relatively low temperature requirements of regeneration (< 250 F or < 120 C), waste heat provided by combustion turbines, IC engines, and any of the fuel cell technologies is capable of supplying heat at regeneration temperatures. The thermal energy produced in many micro-CHP systems is sufficient to meet the input requirements for absorption refrigeration as well as desiccant regeneration.
96
Figure 32. Solid Desiccant Dehumidification System (Chamra et al., 2000).
Solid Desiccant Systems Figure 32 illustrates the components of a generic solid desiccant dehumidification system. At a minimum, the system will include separated process and regeneration airstreams for the desiccant device and some kind of heater to raise the temperature of the regeneration air. The approximate path of the process air through a desiccant device is shown in Figure 33 for the same inlet and outlet conditions as were shown for the sub-cooling system (Figure 27). Note that, as indicated by the path from point 1 to point 2 in Figure 33, the desiccant process increases the dry bulb temperature of the process air. For solid desiccant materials, this increase is a result of the "heat of adsorption" which consists of the latent heat of vaporization of the adsorbed liquid plus an additional "heat of wetting." Heat of wetting is the energy released during dehumidification, in excess of the latent heat of vaporization. The path from point 1 to point 2 is close to a line of constant enthalpy. After the dehumidification process, the process air must undergo a sensible cooling process to reach the end point. Cost Considerations In many cases, the additional benefits provided by a desiccant system will lead to greater overall capital equipment cost. However, since the latent part of the cooling load is shifted from electrical energy to thermal energy, desiccant dehumidification systems can potentially have lower operating costs, particularly if waste heat can be utilized. In micro-CHP systems, waste heat can be used to regenerate the desiccant. This makes desiccant dehumidification appealing to micro-CHP applications, especially when an absorption chiller is incorporated into the system to provide the sensible cooling load. Because the humidity and temperature can be controlled independently with a desiccant dehumidification and cooling system, the system performance is often more effective than that obtainable with conventional systems. Analysis of the sensible heat ratio (SHR) suggests the energy cost savings potential that a desiccant system may have. The SHR is the ratio of sensible cooling load to the total cooling load (sensible load plus latent load). A sensible heat ratio close to unity implies that very little moisture is removed from the air, while a sensible heat ratio close to zero indicates that most of the load is latent cooling. Air-conditioned environments often have SHR values well below unity, which results in greater energy consumption for a sub-cooling system than that of a desiccant dehumidification and cooling system that meets the same temperature/humidity requirements.
97
Figure 33. Dry Desiccant Dehumidification Process (Chamra et al., 2000).
CONCLUSIONS
micro-CHP is a developing paradigm in energy systems. The driving potential behind micro-CHP systems is the thermal efficiency that these systems can achieve and the significant portion of the energy market that lies within the micro regime. Projected system efficiencies of 80% are well above the overall thermal efficiencies produced by standard energy systems. micro-CHP systems have the potential to increase the overall thermal efficiency, reduce the total power requirement, and provide higher quality, more reliable power than conventional systems. Larger homes, higher energy costs, volatile fuel markets, electricity blackouts, power security, power quality, and lower emissions are additional characteristics that make micro-CHP attractive. A variety of distributed power generation (DPG) technologies may be selected for an micro-CHP system. Table 11 illustrates how these technologies compare in efficiency, cost, technology status, emissions, noise, and load matching flexibility. The technologies are ranked from those having the most positive characteristics to those having the most negative characteristics. For example, fuel cells have the quietest operation and, therefore, the best characteristics in the noise category, and reciprocating engines generate significant noise and, therefore, the most negative characteristics in this category. The rankings in Table 11 are based on the technologies as a whole and may vary in some instances.
98
Louay M. Chamra and Pedro J. Mago Table 11. Rankings for Distributed Power Generating Technologies
Rankings for DPG Technologies Most Positive Least Positive Fuel Cells Reciprocating Stirling Microturbine Rankine Reciprocating Microturbine Fuel Cells Stirling Rankine Reciprocating Microturbine Fuel Cells Stirling Rankine Fuel Cells Stirling Microturbine Rankine Reciprocating Fuel Cells Stirling Microturbine Rankine Reciprocating Reciprocating Fuel Cells Microturbine StirlingRankine
Category Efficiency Technology Status Cost Emissions Noise Load Matching Flexibility
The thermally activated components discussed in this chapter are non-competing technologies. The selected use of either an absorption chiller or desiccant dehumidifier is based upon the needs of the application rather than performance. A system configuration requiring only latent cooling would incorporate an absorption chiller. If both cooling and dehumidification are required, both a desiccant dehumidifier and absorption chiller can be included in an micro-CHP system. Despite the potential benefits of micro-CHP systems, numerous significant obstacles and challenges must be addressed if micro-CHP is to be successful in the United States. Research must be conducted to validate micro-CHP systems and components. Likewise, efforts must be made to reduce installed initial capital costs of micro-CHP systems. These issues can be resolved with time and the application of good engineering practices. The more difficult issues to overcome are characteristics of the market to which micro-CHP systems apply. In order for micro-CHP systems to succeed, maintenance and service support issues must be addressed. Large scale CHP systems are typically installed in large factories, hospitals, or other facilities which employ a full-time maintenance staff. Retaining a full-time maintenance staff for a single micro-CHP system is neither practical nor economically feasible. Expecting the owner to maintain the system is also impractical because of technical complexity and safety issues. Still, providing professional, timely service and maintenance will be a crucial part of successful micro-CHP system operation. More than one potential technology has become a distant memory because of poor maintenance and technical support. This situation leads to a very important question: Who will provide maintenance and support? Should this responsibility fall on utility providers, individual micro-CHP system manufacturers, or independent maintenance contractors? Another important issue to address is the inclusion and cooperation of local electrical utilities. Ideally, an micro-CHP system would provide the exact amount of electrical power needed instantaneously, continually throughout the day. While designing such a system is possible, unfortunately, the increase in system complexity and increased capital costs would be a detriment to early system installations. Initially, most micro-CHP systems will likely be designed as constant-power output or base-load systems. This implies that at some point the power requirement will not be met, or that the requirement will be exceeded. Realistically, both cases will occur within a 24-hour period. For example, in the United States, the base electrical load for the average home is approximately 2 kW while the peak electrical demand is slightly over 4 kW. (EIA) If a 3 kWe micro-CHP system were installed in this situation,
99
part of the time more energy will be provided than could be used and for a portion of the time more energy will be required than could be provided. One option is to size the system for base electrical load. This option requires that the system include an energy storage device or be connected to the electrical grid to provide the peak electrical load. Another option is to size the system for peak electrical load. This option requires that the excess energy be stored, transferred to the electrical grid, or simply rejected as a loss. In all instances, either energy storage or grid connectivity is required. Energy storage is expensive and increases the system size and initial costs. In many instances, a more economically attractive option is for the micro-CHP system to be grid connected. Grid connectivity will require the inclusion and cooperation of local electrical utilities to establish an infrastructure and pricing system to accommodate micro-CHP installations. Grid connectivity also requires the use of two-way metering. The Institute of Electrical and Electronics Engineers-United States of America (IEEE-USA) has developed a series of standards for interconnection of distributed energy resources with the electricity grid. The IEEE 1547 2003 Standard for Interconnecting Distributed Resources With Electrical Systems is the first in the 1547 series of planned interconnection standards. The standard establishes technical requirements for electrical power systems (EPS or electrical grids) interconnecting with distributed generators such as fuel cells, photovoltaics, microturbines, reciprocating engines, wind generators, large turbines, and other local generators. Additional IEEE interconnection standards activities are now designated under the 1547 series of standards. Before a full-market drive for installation of micro-CHP systems is attempted, an indepth case study of an micro-CHP system should be performed. Also, a full economic assessment including all capital costs and the potential effects of the volatile fuel market should be conducted. The selection and price of fuel to be used in an micro-CHP system is very important. A small percent change in the market demand creates a large change in the cost of many fuels.
REFERENCES
ASHRAE, 2001, 2001 ASHRAE Handbook Fundamentals, the American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., Atlanta, Georgia. California Distributed Energy Resources Guide. Available at http://www.energy.ca.gov/ distgen/ Caterpillar. Available at http://www.cat.com Capstone. Available at www.capstone.com Chamra, Louay, Parsons, Jim A., James, Carl, Hodge, B.K., and Steele, W. Glenn, Desiccant Dehumidification Curriculum Module for Engineering/Technology HVAC Courses, Mississippi State University, 2000. DOE Energy Information Administration. Available at http://www.eia.doe.gov. DOE Industrial Technologies Program. Available at http://www.oit.doe.gov. Georgia State University Physics Website. Available at http://hyperphysics.phyastr.gsu.edu/hbase/thermo/diesel.html. Harold, Keith E., Klein, Sanford A., and Radermacher, Reinhard. Absorption Chillers and Heat Pumps, New York: CRC Press, Inc., 1996.
100
Hewitt, G. F., Shires, G. L. and Bott, T. R. Process Heat Transfer, CRC Press., 1994. Keveny, Matt, 2000, Two Cylinder Stirling Engine. Available at http:// www.keveny.com/Vstirling.html Knight, I., and Ugursal, I., Residential cogeneration systems: a review of the current technologies. A report of Subtask A of FC+COGEN-SIM The Simulation of BuildingIntegrated Fuel Cell and Other Cogeneration Systems Annex 42 of the International Energy Agency Energy Conservation in Buildings and Community Systems Programme. 93 pages. First published: April 2005 Laraminie, James, and Dicks, Andrew, Fuel Cell Systems Explained, 2nd edition, West Sussex: Wiley, 2003. Meckler, M., R. Heimann, J. Fischer, and K. McGahey. Desiccant Technology Transfer Workshop Manual. American Gas Cooling Center, 1995, Arlington, Virginia. Micro Cogeneration system for residential and small commercial applications. Available at www.cogenmicro.com Plug Power. Available at www.plugpower.com RS Means Mechanical Cost Data 24th Annual Edition, Kingston, Maryland: RS Means Company Inc. 2001 Shah, Ramesh K., Recuperators, Regenerators and Compact Heat Exchangers, CRC Handbook of Energy Efficiency, New York: CRC Press, Inc., 1997. The micro-CHP Technologies Roadmap: Meeting 21st Century Residential Energy Needs December 2003 United States Department of Energy Office of Energy Efficiency and Renewable Energy Solo Stirling Engine. Available at www.stirling-engine.de/engl/ The U.S. Department of Defense (DoD) Fuel Cell Test and Evaluation Center. Available at http://www.fctec.com/fctec Van den Oosterkamp, P.F., Goorsen, A.J., and Blomen., L.J., Journal of Power Sources, 1993 41, p.239 - 252.
Additional References
Alanne, Kari, and Saari, Arto. Sustainable small-scale CHP technologies for buildings: the basis for multi-perspective decision making. Renewable & Sustainable Energy Reviews, Helsinki University of Technology, Hut, Finland. Basso, T. S. and DeBlasio, R., IEEE 1547 series of standards: interconnection issues. IEEE Transactions on Power Electronics, 2004 Vol 19, 5. Boyce, Meherwan P., Handbook for Cogeneration and Combined Cycle Power Plants, New York: ASME Press, 2002. Caton, Jerald A., and Turner, W. Dan, Cogeneration, CRC Handbook of Energy Efficiency, New York: CRC Press, Inc., 1997. Cengel, Yunus A., Heat Transfer: A Practical Apprach, 2nd edition, New York: McGraw Hill, 2003. Commercial micro-CHP Using Fuel Cells and Microturbines. Emerging Technologies and Practices: 2004 American Council for an Energy- Efficient Economy. Felder, Richard M., and Rausseau, Ronald W., ElementaryPrinciples of Chemical Processes, 2nd edition, New York: John Wiley and Sons, 1986.
101
Flin, D. Domestic CHP in Europe. Cogeneration and On-Site Power Production, Jan. Feb. 2005, pp. 43 49. Goswami, D.Y., Kreith, F., and Kreider, J., Principles of Solar Engineering, 2nd edition, Taylor and Francis Pub., 2000. Hardy, J.D, Cooling, Heating, and Power for Building Instructional Module, Mississippi State University, 2004. Harrison, J. Micro CHP in rural areas. EarthScan James & James. 1 January, 2003a. Harrison, J. Micro CHP in Europe. EA Technology. Presentation at the 2003 National Micro CHP Technology Pathways Workshop. 2003b. Harrison, J. Towards a strategy for micro CHP in the USA domestic markets. EA Technology, June 2003c. Moran, Michael J., and Shapiro, Howard N., Fundamentals of Engineering Thermodynamics, 4th edition, New York: John Wiley and Sons, 2000. Micro Gas Turbines and Heat-Driven Cooling. Australian National Training Authority, 2003. Peltchers, Neil, Combined Heat and Power Handbook: Technologies and Applications, The Fairmont Press, Inc., Georgia, 2003. Pehnt, P., Praetorius, B., Foscher, C., Schnieder, L., Cames, M., and VoB, J.P. Micro CHP a sustainable innovation. Transformation and Innovation in Power Systems. Berlin/Heidelberg, 2004. Stull, D.R., et al., JANAF Thermochemical Tables, Michigan: Dow Chemical Company, 1965. Senft, James R., Ringbom Stirling Engines, New York: Oxford University Press, 1993. Urieli, Isreal, and Berchowitz, David M., Stirling Cycle Engine Analysis, Great Britain: Adam Hilger Ltd.,1984.
Chapter 3
WIND ENERGY IN ELECTRICITY MARKETS WITH HIGH WIND PENETRATION

Julio Usaola* and Edgardo D. Castronuovo
Departamento de Ingeniera Elctrica. Universidad Carlos III de Madrid Av. Universidad 30. 28911 Legans, Madrid, Spain
ABSTRACT
The great amount of wind power recently connected to electric power systems makes necessary new grid analysis tools and control strategies. Some of these tools are short term wind power prediction programs, which have been recently developed and are already in use in many System Operators and wind power producers. Prediction programs also allow the participation of wind energy in the electricity market by keeping the economic losses due to power imbalances for wind generators within acceptable values, specially if the uncertainty of the predictions is evaluated and used in the bidding process. Such participation intends to make the power system operation easier. Distributed control of power system, and specially optimized control of clusters of wind generators is another measure that is being taken in order to maintain the levels of power system security and to minimize the consequences of possible and temporary wind power curtailments, due to grid security reasons, on the affected wind farms. The contents of this chapter include a detailed description of short term wind power prediction programs, an evaluation of the economic losses due to imbalances incurred by wind generators participating in an electricity market, and an optimization strategy aimed at minimizing the consequences of power curtailments.
1. INTRODUCTION
In the last years, a great and sudden increase of wind generation has taken place in many countries. Most of its power injects their production into the power grids, which have to face
*
e-mail: jusaola@ing.uc3m.es, ecastron@ing.uc3m.es
104
Julio Usaola and Edgardo D. Castronuovo
the problems of a different form of generation, with peculiar features. Among them, perhaps the most annoying is the variability of the primary resource, the wind, and its difficulty in dispatching this power. It is well known that electric energy cannot be easily stored, and this means that wind power fluctuations must be balanced by other types of generation. This uncertainty stresses the system management and operation, and may add extra costs to the electricity. These disadvantages are clearly compensated by the renewable nature of this source of energy, and for the null cost of the resource, that reduces the overall cost of electricity. However, the need for new analysis and control tools for the power grids is evident, if we expect to keep the reliability of these grids at a maximum level and to limit all avoidable cost to a minimum. Short term wind power prediction programs are one of the tools that have been recently developed to help power system operation. They use the output of numerical weather forecast, and produce a reasonably accurate prediction for the wind power production of the next hours or few days. With this information, system operation is substantially helped. Prediction tools also allow the participation of wind energy in electricity markets. Wind farm owners may commit themselves to produce the predicted wind power for the next hours. Because of the inaccuracy of these predictions, actual generation differs from the predicted values committed, and this imbalance generate an extra cost that must be paid by those who produce it, namely the wind generators. Nevertheless, this extra cost is reasonably small if a state-of-the-art prediction program is used, and may even be reduced with an adequate subsidy policy. The generators will be interested in reducing this imbalance cost as much as possible, and to fit the generation to the predicted value. In this incentive for adjusting the production lays the main benefit of the participation of wind generation in the markets it is a way of reducing the need of balancing energy in a power system. Another problem of wind generation is that older wind farms cannot withstand non standard grid conditions such as under/over voltage or under/over frequency. This means that much of wind generation disconnects if voltage goes below 90% of the rated value. Although this is not important when there are just a few wind farms connected to a big grid, when there is a substantial wind penetration, these disconnections may mean the switching off of hundreds on MW and may have a strong effect on power system stability. Most of the recent generation has been designed to remain connected under severe undervoltages, but a great part of wind farms cannot still deal with undervoltages produced by a short circuit, an incident quite common in the power systems. Modern system operation must tackle this problem by the use of new control strategies. In the following sections, these problems have been addressed, describing new strategies and control tools that have, or are being implemented in power systems with high wind power penetration. First of all, short term wind power prediction tools are described, their typical performance presented and examples of actual prediction programs given in order to show their principles, possibilities and limitations. Then, the economic consequences of participation of wind energy in the electricity markets are evaluated. Numerical tests show that economic losses of wind producers, due to imbalance costs, are assumable if they use a short term wind power prediction program. Knowledge and adequate use of the uncertainty of this power prediction may even reduce this imbalance cost.
Wind Energy in Electricity Markets with High Wind Penetration
105
Finally an optimized control strategy for distributed management of wind power is presented. This strategy minimize the consequences of possible power curtailments imposed by the System Operator and uses market mechanisms to allocate the power decrease to those wind farms more willing or capable to control their production. This strategy is also aimed at improving the manageability of power systems with a heavy wind power penetration.
2. SHORT TERM WIND POWER PREDICTION

2.1. General Background
Short term wind power prediction programs are tools that provide an estimation of the future power production of a wind farm, or a group of wind farms, in the next hours. For this purpose, they use meteorological forecasts coming from a Numerical Weather Prediction (NWP) tool, and sometimes real time SCADA data from the wind farms. Typically, their predictions are issued at least once a day, although many prediction programs refresh these predictions once per hour, or even with greater frequency. The output is the average hourly (typically) production forecasted for the prediction time horizon. These programs have the following main applications: They aid the power system operation by estimating the wind power production for the future hours. This helps to foresee the congestions in the system, to program the power reserve, or to prevent extreme events under adverse meteorological conditions. This is why System Operators of most countries with high wind penetration use them. They are also used by the balancing companies, i.e., those who have to buy energy for the customers connected to their grids, where there are also wind farms connected. An inaccurate prediction makes these companies to pay for their imbalances. They make possible the participation of wind power in electricity markets. Power predictions can be used to make bids in the market. Although there are deviations between the predicted power and the actual generation, prediction programs keep these imbalances small enough (around 10% of the income) to make this participation profitable. This participation makes easier the integration of wind energy in the electric systems. They are also useful for wind farm maintenance scheduling, since they estimate future production. However, accuracy of short term wind power prediction tools decrease drastically after a few days, and for this reason, their application to this purpose is limited.
A short term wind power prediction tool usually has the outline shown in Figure 2. As mentioned, the basic inputs to the program are meteorological forecasts. Additionally, there may also be inputs from a SCADA system, with data of the actual production of one or several wind farms. This allows the regular updating of prediction, based on more recent data. With these fresh data some prediction tools are able to issue new predictions each hour, or
106
even more frequently. SCADA data include power output from the wind farms most of the times, but it may include also meteorological data, such as measured wind speed and direction, temperature or atmospheric pressure. In any case, the real production data from the wind farms must be known, in order to check the accuracy of the predictions and to tune the models. In some prediction programs, this information is received regularly, but some time (one day, for instance) after real time. It is also necessary to have data of the wind farms, such as rated power, type and availability of wind turbines, etc. Also information about the wind farm site is needed, as the UTM coordinates, and some prediction tools need information about terrain: roughness topography, turbine site etc. The output of these programs is the average wind farm production for the next hours. The time scope of this prediction depends on the horizon of the numerical weather prediction available. Typically, predictions are issued for the next 48 hours, but longer time horizons are possible, sometimes at the price of a poorer resolution. An important part of the wind power prediction program is the management of past predictions that are used to revise the modelling and to assess the accuracy the performance of the prediction tool.
Real time P, v measurements
Wind farms database
Downscaling
Numerical Weather Forecasts
Prediction algorithms
Upscaling
Output processing Report 1 Report 2 ... Report N
Data storage
Statistical correction
Data presentation
Figure 2. Elements of a short term wind power prediction program.
2.2. Numerical Weather Predictions (NWP)

Numerical Weather Predictions (NWP) are the basic input to a prediction program. They are usually sent by a provider of this service, who typically are National Weather Services, or companies specialized in this field. New predictions are received only some times each day, due to the high computation requirements of the programs employed.
107
Different numerical models provide weather predictions around the world. For instance, the European Centre for Medium-Range Weather Forecasts (ECMWF) provides (among many other products) wind speed and direction predictions, for many places throughout the Earth, each six hours for up to 72 hours in advance at 3 hours intervals, and up to 240 hours in advance at 6 hours intervals, with a resolution of 0.25 x 0.25 lat/long. Predictions are produced for the whole world. This organization is funded by 28 countries, mostly European ones. There are other projects running that provide also predictions on a regional or global scale, as the project HIRLAM, Aladin, LokalModell, or the MRF models of the USA. Numerous references can be found in literature and elsewhere. These forecasts have usually low resolution, since their results are extended to a wide area in a continental scale. This is why a downscaling procedure is often used. Downscaling uses meso or microscale models to fit the wind speed predictions to the wind farms site. This is usually a difficult task, since it may require detailed modelling of the terrain, which for some farms is rather rugged, since they are placed to profit of local effects.
2.3. Prediction Methods

The core of a prediction program is the algorithm used to produce the predictions. All current prediction programs use at least one of two kinds of methods of prediction wind power physical and statistical methods. Physical methods try to model the wind farm equations according to the aerodynamical behaviour of wind turbines in the actual site, and the local effects of wind speed and directions. Statistical methods try to reproduce the behaviour of the wind farm from past data under different conditions. There are different techniques to do this: time series analysis, nonparametric methods and Artificial Intelligence based methods have been successfully used. The simplest prediction method is called persistence. This method consists of assuming that the future prediction, for all the time horizon considered, is the current production of the wind farm. Although this method has little practical value, it is considered as a lower threshold for the performance of a prediction method. Predictions may be also easily made by taking the NWP (downscaled or not) and converting this value to power by means of a P-v curve. This curve could be made as easily as to take the sum of the standard power curves of the wind farms as shown in Figure 10 a), or it may be obtained from past power and wind measurements in the wind farm, as shown in Figure 10 b). It could also be an empirical curve that expresses the relation between past wind forecasts, and power measurements. The power curve of a wind farm is different according to the wind direction, due to wake effects or to a terrain with different roughness in different directions, so a more sophisticated model could be a power surface that takes into account this different behaviour. However, actual methods are more complex than these, because these simple methods do not take into account the different conditions of the wind farm along time, nor the characteristics of the site. The physical models of most prediction programs, such as [19] or [21] receive the wind speed and direction data from a NWP, and adapt them to local characteristics. Then, this wind is transformed into output power using the wind farm power curve, which normally is a complex model. This output is modified according to the wind farm efficiency that accounts for the wake effects. A Model Output Statistics (MOS) module is normally used to improve
108
the results. This module can be applied either to the incoming wind, or the outcoming power. MOS require data of the farm for several months, and detailed information of the turbine site and terrain. Physical models usually require long computation times, and for this reason their predictions are updated just few times a day. Statistical models also use the outcome of NWP programs, as well as real time power production (and sometimes other SCADA data from the wind farm), and the power predictions are obtained from a statistical module that models the wind farm from the inputs and outputs through a long period. Statistical prediction models usually do not have high computational requirements, and the predictions can be updated typically every hour or even more frequently. Different methods can be included in this family. For instance [19] uses time series models with exogenous inputs (the NWP), with parameters that vary adaptatively with time. It also models the power curve of the wind farm according to past data and recent conditions. Other models use fuzzy artificial neural networks for this purpose [27]. Statistical models need also to work for some time since they reach their optimal performance. This time is no less than four months, although significant results can be obtained before. Actually, the best performance of a prediction program combines physical and statistical methods. As already said, physical models use statistics (MOS modules) to suppress systematic errors and to improve the model, and statistical models improve their results if a better wind input is used. In difficult terrain, to adapt the NWP to local conditions can lead to improvements of 10% in accuracy [22].
nP Power forecasted
Number of turbines
P Power forecasted
Windfarm power
Standard power curve
Standard power curve
Wind forecast
a)
Figure 3. Simple prediction methods.
Wind forecast
b)
When it is necessary to forecast the production of the wind farms in a region, but there are no data available from all the wind farms present in it, it becomes necessary to upscale the results, i.e., to forecast the prediction for all wind farms, from the prediction for few of them. This process requires data from all the wind farms, in order to find the spatial correlation among all the wind farms present in the area. This is also a complex task, and a great amount of data is needed [11].
109
2.4. Evaluation of Results

The error descriptors used for assessing the accuracy of the prediction are usual in Statistics. If we define the error in the power prediction for time t+k, produced at time t as e(t+k/t), then this error has the value
e(t + k / t ) = p(t + k ) p(t + k / t )

Where p(t+k) is the production of the wind farm at time (t+k) and p (t + k / t ) is the
power predicted at time t for time (t+k). Then, the error measures usually considered are: Normalized Root Mean Square Error (NRMSE(k))
NRMSE (k ) =
1 Pr
N
e2 (t + k / t )
t =1
Normalized Mean Average Error (NMAE(k))
1 NMAE (k ) = Pr
Skill score (R(k))
e(t + k / t )
t =1
2 P e2(t + k / t ) R(k ) = 2 P
Where Pr is the rated power of the wind farm, N is the number of predictions examined
2 along the considered time, produced k hours before the measure, P is the estimated 2 variance of the power production along the considered time, and e (t + k / t ) is the estimated
variance of the error of the predictions produced k hours before the measure. The skill coefficient tries to explain part of the variance of the data. It has a value between 0 and 1. A value of 1 means perfect prediction. A value of 0 means useless prediction the model does not give any improvement compared with the constant mean model. It may be seen that all these error measures depend on the time elapsed between the moment when the prediction is generated and when the power is measured. This error increases when this time is bigger. Values of NMAE, NMRSE and R for a wind farm of 21 MW are given in Figure 4, Figure 5 and Figure 6, for a prediction tool and persistence. It can be observed that the method always performs better than persistance, although for the first hours this difference is very small.
110
NMAE
0,3 0,25 0,2 p.u. 0,15 0,1 0,05 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 look-ahead tim e (hours) Pred. Tool Pers
Figure 4. NMAE of a typical wind farm, for a prediction tool and persistence.
NRMSE
0,4 0,35 0,3 0,25 p.u. 0,2 0,15 0,1 0,05 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 look-ahead time (hours) Pred. Tool Pers
Figure 5. NMRSE of a typical wind farm, for a prediction tool and persistence.
111
1 0,8 0,6 0,4 0,2 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 look-ahead tim e (hours)
Pred. Tool
Pers
Figure 6. Values of R of a typical wind farm, for a prediction tool and persistence.
Some other features of wind power predictions, is that they are often biased, i.e., that the average value of the errors is not zero. Figure 7 shows this effect for eighteen months data of a wind farm of 14 MW of rated power. The error is defined as above. The mean value is always negative, so the bias is towards overprediction. This bias increases quickly the first hours, and then keeps stable. Standard deviation also increases with look-ahead time, because these predictions are more inaccurate. The trend of standard deviation for the same wind farm is shown in Figure 8.
p.u.
error bias
1 0,00 -0,02 -0,04 -0,06 -0,08 -0,10 -0,12 -0,14 -0,16 look-ahead time (hours) 4 7 10 13 16 19 22 25 28 31 34 37 40 43
Figure 7. Error bias of short term wind power prediction.
MW
112
error standard deviation

4,0 3,5 3,0 MW 2,5 2,0 1,5 1,0 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 look-ahead time (hours)
Figure 8. Standard deviation of short term wind power prediction.
The distribution of error has also skewness, as shown in Figure 9. This means that there are larger errors that are greater than the average. All these results may be considered as typical, although the accuracy of wind power prediction may vary widely between different wind farms. This accuracy depends heavily on the roughness on the terrain, which makes wind predictions more difficult and inaccurate. It is also a common feature that the error decreases for a group of wind farms. Errors tend to compensate and give a better joint result. This is usually called portfolio effect that may increase the accuracy up to a 100%.
skewness coefficient
1,0 0,8 0,6 0,4 0,2 0,0 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 look-ahead time (hours)
Figure 9. Skewness coefficient of short term wind power prediction.
113
2.5. SIPRELICO as an Example of a Short Term Wind Power Prediction Program

In this section, a more detailed description of a short term wind power prediction program is given. This intends to give a deeper insight into the properties, the accuracy and the application of these tools as a help to power system operation.
2.5.1. Introduction Wind power has grown very quickly in the last years in Spain. The installed power in the Spanish peninsular system reached the value of 11854 MW in May 2007, and the growth of the power installed in the whole of Spain has been very quick, as shown in Figure 10. Due to this amount of installed power, electric system operation must take into account the wind power production for the day ahead scheduling (grid constraints solution) and for real-time operation. This is why it has been revealed necessary a short term wind power prediction for the wind production in the whole of Spain. The Spanish System Operator, Red Elctrica de Espaa uses since the year 2002 the short term wind power program SIPRELICO, developed by Universidad Carlos III de Madrid [29], [30], [36], whose main features are described below. SIPRELICO issues predictions each 15 minutes for the whole Spanish peninsular system. This means that predictions are generated for 88 farms or groups of farms, and then grouped by zones. Finally, the prediction for the whole peninsular system (more than 11.5 GW) is found from the individual predictions. SIPRELICO receives numerical weather predictions four times a day from the National Meteorological Service of Spain (Instituto Nacional de Meteorologa), who produces them with the program HIRLAM (see below). Some features of these predictions are given in Table 1.
12000 10000 8263 8000 6371 MW 6000 4000 2000 0 4500 3200 2228 1453 834 11239 10000
39 45 69 109205455
Figure 10. Installed wind power in Spain.
9 19 2 93 19 9 19 4 9 19 5 9 19 6 9 19 7 9 19 8 9 20 9 0 20 0 0 20 1 0 20 2 0 20 3 0 20 4 0 20 5 06
19
Year
114
Julio Usaola and Edgardo D. Castronuovo Table 1. Numerical weather predictions supplied by HIRLAM
Time of day (hour) Time horizon (hours) Spatial resolution ()
00 48 0.16
06 24 0.05
12 48 0.16
18 24 0.05
As SIPRELICO has to manage a great number of wind farms, no detailed description for each of them, and consequently, no downscaling is possible by the moment, due to the huge computation times that would be needed. However, the portfolio effect keeps the error for the whole peninsular Spain within adequate levels. Data of wind farms include information about the wind farm situation and rated power. There are not details about each wind farm. Wind farms are not obliged by the moment to provide the real time production to the System Operator. Hence, real time SCADA data used as input to SIPRELICO come from the distribution lines that evacuate only wind power. This is more than 80% of the total installed wind power. Due to this fact, it is not necessary any complex upscaling models. This upscaling just follows a proportional rule.
2.5.2. Prediction Algorithm The core of SIPRELICO is the time series module. The design of it is critical, since it has to deal with changing relations between wind and power, which are nonlinear, and with numerical weather predictions that may have gross errors due mainly to the low spatial resolution of the NWP. This is why SIPRELICO uses different statistical models working in parallel, and only a combination of the most accurate among them are selected for each farm and each hour. The models used are the following:
Univariant models. Autorregresive models AR(3), that take into account the wind inertia, and the daily cycle which is usually present in wind speed. These models can be used up to 10 hours before real time. Parametrical models with dependence of wind and speed direction. These models include the effects mentioned above in the univariant models, as well as the numerical predictions. Non parametrical models dependent on wind speed and direction. Nonparametric methods try to find the relations between the input to the model (the numerical weather predictions) and the output (the power produced by the wind farm). These relations depend on the time lag between the prediction and production, the wind speed and direction and the time in the day (diurnal cycle). These relations are found from past results and data, and are locally linearized. Since past predictions and outputs do not define clearly a curve, but values around each point, a probability distribution function is obtained, from which the prediction is finally issued. An illustration of this model is given in Figure 11.

P Predicted power Power curve D=0
115
Power curve D=180

0
v V
v D
v Wind prediction (V, D)
Figure 11. Nonparametric methods.
Zone 1
Zone 2
Zone M
SIPRELICO Prediction Total peninsular

Other sources of prediction Prediction
Model A
Predictin Model B
Adaptative combination
Figure 12. Scheme of SIPRELICO operation.
Final prediction Improves all of them
However, for an accurate prediction, two more aspects must be considered: adaptative estimation of the time series parameters, and the optimal combination and selection of the best prediction provided by the different tools. The parameters of the models change with time, due to seasonal variations of weather, wind farm repowering, or other reasons. Hence, these parameters must be adapted with the new data that are being obtained. This parameter updating is made using two alternative procedures, Recursive Least Squares and Kalman Filter. The final prediction is made by a final combination of the parallel predictors. This combination is based on the actual performance of each predictor. For this reason, it is necessary to implement a real time evaluation of each of them. Then, the degree of performance of each predictor in each horizon determines the optimal weighting of each forecast in the final predicted output, together with several measures and diagnostics. The evaluation of the predictions of each predictor is made with an exponentially weighted mean squared prediction error (EW-MSPE). The final prediction for a given horizon
116
is obtained using the combination of the 3 predictors with lower EW-MSPE. A more detailed description may be found in [30]. Finally, this prediction is optimally combined to prediction coming from other sources, and the accuracy of the joint prediction is higher than any of the incoming sources. This process is illustrated in Figure 12.
2.5.3. Application in Red Elctrica de Espaa The wind generation forecasts produced by SIPRELICO are already used in the following aspects of the system operation:
Improvement of the day-ahead demand prediction carried out by the Control Centre of Red Elctrica de Espaa. Determination of the reserve needs for the secondary and tertiary regulation markets. Continuous resolution of the technical constraints of the system. Transmission grid works and line disconnections scheduling
When a line disconnection takes place in a zone with many wind farms, problems may arise for the evacuation of the whole power generated in the area (this could be the case when both the generation groups and the wind farms are working at full power). As an example, it may be said that a situation such as this took place in the Northwest of Spain during the disconnection of a 220 kV line. Since wind generation forecasts were available 48 hours in advance, the system operator imposed a limitation to the technical minimum to several conventional groups in the area, when necessary. A study conducted latter showed that the decisions adopted/undertaken by the system operator thanks to SIPRELICO forecasts were corrected in a 96% of the cases. More detailed information may be found in [14].
2.5.4. Accuracy of Results In this section, the accuracy of SIPRELICO is shown. The evaluation of this accuracy for the last month before June 5, 2007 are shown in the following figures, that corresponds to the whole of the wind power connected to the Spanish peninsular grid. These results have been kindly provided by Red Elctrica de Espaa. The high accuracy of these results is partly due to the portfolio effect applied to all the generation connected at that moment.
2.6. Other Prediction Programs

Many other short term wind power prediction programs are currently being used, and are available. In this section a few of them are shortly described. WPPT [23] (Wind Power Prediction Tool) is a prediction program developed in Institut of Mathematical modelling of the Technical University of Denmark and it is running since 1996 for different users. It is a statistical model, and uses non-linear time series models, including a semi-parametric power curve model for wind farms taking into account both wind speed and direction. The models are self calibrating and adaptive. The inputs are predictions of wind speed and direction as well as real time power production measurements and other
117
meteorological variables. It produces predictions for up to 120 hours ahead, depending on the input predictions.
Coefficient R-month
1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 look-ahead time (hours
Figure 13. Value of coefficient R for the last month in the Spanish peninsular grid.
NMAE-month
0,06 0,05 0,04 0,03 0,02 0,01 0 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 look-ahead time (hours) Serie1
Figure 14. Value of NMAE for the last month in the Spanish peninsular grid.
118
RMSE-month
0,08 0,07 0,06 0,05 0,04 0,03 0,02 0,01 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 look-ahead time (hours)
Figure 15. Value of RMSE for the last month in the Spanish peninsular grid.
Prediktor [19] is a prediction tool that uses physical models and MOS. Its basic input are the NWP, and information about the wind farm. The NWP are adapted to local conditions, taking into account the turbines siting, the orography, the roughness of the terrain, etc. MOS (model output statistics) are used to correct the model. The observed wind farm productions are used to adjust the parameters of statistical filters. The final output of the model is the expected production of the wind farm every 3 hours over the next 48 hours. It has been developed by Risoe, and has been used in Denmark, the USA and elsewhere. LocalPred [21] is another tool that uses physical model that includes weather prediction using mesoscale models. Its inputs are the NWP and data for the wind farm, such as orography, roughness, turbine siting, and past data. MOS statistics and real production are also used for tuning the model. It has been developed by CENER and CIEMAT and is currently being used in Spain. Advanced Wind Power Prediction Tool (AWPT) [10] has been developed by ISET, Germany. It uses statistical models using Artificial Neural Networks. They use NWP and a few measurements from selected wind farms for producing predictions of a large area in Germany with more than 13 GW of wind connected. Off-line measurements of the production in the region are used to tune the upscaling method. eWind [3] is an USA product, developed by AWS Truewind that produces predictions using physical modelling. From NWP produced by a Regional Weather Model, mesoscale simulation tools adapt the prediction to local conditions. MOS are also used to tune the models and correct systematic errors. This model is being widely used by wind farm owners in the USA.
119
Previento [11] is another physical prediction model developed in the University of Oldenburg that takes the prediction inputs from different services. The prediction of regional production is upscaled from selected wind farms, and MOS are used to correct the model results. It is being used in Germany. ANEMOS is the result of an European project that joins different prediction tools in Europe. The project aims to develop accurate models or improve existing ones, aiming to outperform actual state-of-the-art, for onshore and offshore wind resource forecasting (statistical and physical). Emphasis is given on integrating high-resolution meteorological forecasts. For the offshore case, marine meteorology is considered, as well as information by satellite-radar images. ANEMOS itself is an integrated software that hosts the various models developed by the projects partners. The system is being used by several utilities and wind farm owners for on-line operation at onshore and offshore wind farms for local/regional/national wind prediction. The applications are suited to different terrains and climates, on-/near-/off-shore farms, interconnected or island grids. A detailed description of this project may be found in [16]. More information about the state of the art in short term wind power prediction may be found in [12]. An interesting study (in Spanish) comparing results for different wind farms and prediction tools in Spain is [2]. The current trend of research in the prediction field is focused on a better use of mesoscale models, the combination of different predictions coming from independent sources, and the use of ensembles [13]. These lines are very promising, and future increases in accuracy can be reasonably expected. Besides, the advances in computation will also lead to NWP more frequent and with greater resolution and accuracy.
3. BIDDING STRATEGIES FOR WIND FARM OWNERS

One of the applications of short term wind power prediction programs is to help the wind generators to participate in the electricity markets. As shown in the next sections, to use such programs implies important benefits to the users. Participation of wind energy in electricity market is now becoming usual, and this is why this section aims to quantify those benefits and describes the possibilities of using these tools, pointing at the consequences of this use.
3.1. Wind Power and Electricity Markets

Electricity markets are nowadays the most common paradigm for electricity trading. In many of the electricity markets, a important amount of the energy is traded in a daily market, that may account for the 20 to the 80% of the total energy, and the rest is traded bilaterally between generators and consumers. To participate in these daily and intraday markets, it is necessary to present bids for selling or buying power at a certain price, more than 12 hours before the operation time. The energy is committed to each market participant by means of an auction. To deviate from this commitment may suppose an overcost for the system, and must be paid accordingly. However, it is possible in some countries to trade the foreseen power excess or deficit in intraday markets run along the day, if this imbalance is known several
120
hours beforehand. Other countries, such as the UK, have a different market model, with a small balancing market that is closed shortly before the operation time [25]. Table 2 summarizes some of these features for different European countries. Wind energy must integrate into electricity markets, in spite of its non-dispatchable nature, and the variability of the resource. Also, wind energy, as coming from a renewable source of energy, usually benefits from subsidies. These subsidies may take the form of feedin tariffs, green certificates of other mechanisms, such as tenders, that vary widely among the different countries. It is not common that wind energy participates in electricity markets. Wind generators are usually allowed to inject all their power to the grid under preferential rules, and they are just paid the feed-in tariff for their production. In order to assess the wind power production, TSOs and balancing companies must run prediction programs for a region in order to allocate reserves, to account for the possible congestions in the system, to buy the energy need by their customers, etc. However, there is a trend to encourage the participation of wind power in the electricity markets, following the market rules as any other participant. This means that wind generators must submit bids to the Power Exchange office, the day before operation, saying how much power they want to sell, and its price. Since the power generated cannot be easily stored, the price is usually zero, and all the bids are accepted. However, the amount of power committed in the energy auction must be generated, or the system overcost due to deviations from this commitment should be paid. For this reason, wind power generators participating in an electricity market need to make power predictions between (typically) 12 and 36 hours in advance of the operation time for bidding in the daily markets. Whenever possible, it is advisable to negotiate in intraday markets using updated predictions, in order to reduce the imbalance cost. Participation of wind generators in electricity markets has several advantages. Predictions made by wind generators themselves may be more accurate than those produced at a regional scale, since they have more information about their wind farms. For instance, they know accurately the orography of the terrain, and a downscaling may improve the NWP input, and they know better the availability of the wind turbines for the next day. This information arrives with more difficulties to a regional organization. In general, a better prediction helps the system operation. Table 2. Common points in regulatory frameworks
Pool Spain Denmark France Ireland Eng.&W Greece X X X X X Intraday/balance Bilateral Gate closure markets contracts 14 h Yes (2h15m) X 12 h Yes (30 m) X 13 h Yes (2 h) X 14 h Yes X Yes (1 h) X 12 h No
121
3.2. Benefits of Using Short Term Wind Power Prediction Tools

The wind generators that participate in an electricity market obtain interesting benefits from using a short term wind power prediction tool, because the amount of power imbalance reduces significantly. In this section, this reduction is quantified under rather general assumptions. In general terms, the revenue R for a given wind farm in a pool market may be generalized as:
R = Pd , t MPt + MPi , t (Pi , t Pdi , t ) + imbalance

t t
(1)
Where
up + MPt Pgen,t Plast ,t imbalance = down Plast ,t Pgen,t MPt
Pgen,t > Plast ,t Pgen,t < Plast ,t
(2)
The meaning of the different terms of the equations is: Pgen,t Pd,t Power actually generated by the wind farm in the hour t Power committed to the wind farm in the daily market for the hour t. It coincides with the prediction available at the gate closure of the daily market. Power committed to the wind farm in the intraday market for the hour t It coincides with the prediction available at the gate closure of the intraday market. Power committed to the wind farm in the last update for the hour t Marginal price of energy in the daily market for the hour t Marginal price of energy in the intraday market for the hour t Marginal price of energy in the spot market for selling energy in the hour t Marginal price of energy in the spot market for buying energy in the hour t
Pi,t
Plast,t MPt MPi,t MPtup MPtdown
In markets without intraday markets, the second term of equation (1) does not apply. In these cases Plast,t = Pdt. Where there is a possibility for updating bids in an intraday market, Plast,t = Pit. An example of time schedules for daily and intraday markets is shown in Figure 17. The imbalance term (2) is simpler for some markets, because the price in both senses of the imbalance is the same. In other markets, this price depends on the system deviation and may be zero in some cases. Most times the imbalance term represents a decrease in revenue, and an accurate prediction program should reduce as much as possible this term. In order to simplify the equations, for a better understanding of the results, some simplifications are going to be made to them. Equations (1) and (2) are simplified to:
122
R = Pgen,t MPt Pdev,t MPt

t t
where Pdev,t = |Pgen,t- Pd,t|
penalty factor.
Hence, the following simplifications have been assumed: The intraday market has not been considered. Since the overpredictions and the underpredictions could be assumed to be more or less unbiased, and the intraday MP might be equally higher or lower than the daily MP, the weight of this term is not high. The assumption of unbiased predictions seems not to be completely correct as shown later. The imbalance term has been changed into a penalty proportional to the absolute value of the deviations incurred, and the amount of the penalty is a fraction of the marginal price. This simplification has the following effects in different markets:
For those markets where the penalties are paid only if the imbalances have the same sense than the overall system deviations, this assumption is a pessimistic one. Better results must be expected. For those markets with different penalties (or market prices) for up deviations and down deviations, the effect is small, due only to the bias of the prediction. The results presented in this section compare the performed prediction and persistence with perfect prediction. The comparisons are made between the revenues of a producer that would perform a perfect prediction and those of a producer using SIPRELICO as a prediction tool. This comparison has been made also for predictions performed using persistence. The value of the relative revenue (RR) that compares these values is given by the following equation.
Pgen,t MPt Pdev,t MPt t RR = t Pgen,t MPt

t
(3)
The lower part of the equation represents the revenues obtained with perfect prediction. The upper part is the loss in revenues due to inaccuracies of the prediction tool. Therefore, the term RR will be always less or equal than 1. A further simplification is to try to make equation (3) independent of the price. If we consider that it leads to the same results for a long period of time to be paid every hour the MP or to be paid the average MP along that period, then the term RR of equation (3) changes into RRapp of equation (4).
123
Gate closure Operation time RW=4h, 8h,24h Time ahead (h)

Figure 16.
Pgen,t Pdev,t t RRapp= t Pgen,t

t
(4)
In this equation it may be seen that the reduction income depends proportionally of the penalty . Results have been obtained for different values of this parameter. The study has been performed for different rolling windows (RW), and for different times between gate closures and operation time (Ta). An illustration of these times is given in Figure 16. Some additional assumptions concerning market rules are made: A pool system is considered. Results under bilateral contracts would depend on agreed conditions, difficult to know. However, bilateral prices and spot prices could be assumed as correlated in the long term, so that the conclusions could be extrapolated for markets based on bilateral contracts. The use of past market prices implies to assume that wind power has no influence on market prices. This is a too simplifying assumption, especially for countries with a high wind power penetration. However, it must be considered that most farm owners are price takers in a competitive market, and for them the prices are independent variables. The conclusions are intended to be independent of the price level. There is no difference between wind generators and conventional generators in the market. That is to say that no subsidies are considered for the wind generation. The losses due to imbalance costs then are smaller in relative terms than those shown here. Wind generators present their bids for the amount given by the prediction program, at zero price, so they are always accepted. The prediction tool makes new prediction from available data (wind forecasting and real-time production) every hour. SIPRELICO has been the prediction tool used for performing the prediction. This prediction tool may be considered as representative of the state of the art.
These equations have been applied to offline data of two different real wind farms, with the following features:
124
Julio Usaola and Edgardo D. Castronuovo WF 1: 19 months of power measurements, farm situated on flat terrain WF 2: 3 months power measurements, farm placed on mildly rough terrain
The Spanish market prices have been used for the studies. However, the results are fairly independent of the price level. According to equation (4), the revenues loss is an amount that depends on the accuracy of the prediction tool, and independent of the actual market prices. The average value for all considered values of time ahead and rolling window could be considered as a characteristic parameter of that farm that gives a measure of the advantages of a prediction tool at a glance. These results are given in Table 3. Table 3. Losses as a percentage of the penalty factor
WF1 39.37 79.21 WF2 23.52 45.19
SIPRELICO Persistence
In this table, the number 39.37 means that for a penalty of 0.4 times the marginal price, the loss for WF1 would be 0.4x39.37=15.748% of the revenue, when compared to the revenues with prediction. As this number is the average for very different values of look ahead times and rolling windows, the losses may be smaller. More details of this study may be found in [34]. The effect of trading in the intraday market the difference between old predictions for the daily market and the new ones is also to reduce the cost of the imbalance. If, for instance, we consider that prediction for the daily market are produced at 1100 of the previous day, the look-ahead time for the predictions is between 13 and 16 hours. If there is an intraday market each 4 hours, with a market gate closure 3 hours before operation time, the updated predictions can be used with a look-ahead time between 3 and 6 hours, with a much better accuracy. This time schedule is shown in Figure 17. Under these assumptions, the reduction of the imbalance costs when trading also in the intraday markets is shown in Figure 18.
DAY D-1 Hours 11 12 13 14 15 16 17 18 19 20 21 22 24 1 Daily ID1 ID2 ID3 ID4 ID5 ID6 P P 2 3 4 5 6 7 8 DAY D 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 D D D D D D D D D D D D D D D D D D D D D D D D ID1 ID1 ID1 ID1 P ID2 ID2 ID2 ID2 P ID3 ID3 ID3 ID3 P ID4 ID4 ID4 ID4 P ID5 ID5 ID5 ID5 P ID6 ID6 ID6 ID6
Figure 17. Time schedule of the different markets. D = daily market; ID = intraday market.
125
16,00 Cost (% max revenue) 14,00 12,00 10,00 8,00 6,00 4,00 2,00 0,00 20 25 30 35 40 45 50 Penalty (%MP) MD MID MID+TID MD-MID
Figure 18. Effect of intraday markets on revenue. Results for WF2.
In this figure, the reduction in revenue due to imbalances is shown against the penalty, expressed as a percentage of the marginal price (MP). Different cases are considered only bids in the daily market (MD), bids in the daily and updates in the intraday markets (MID), bids in the daily and updates in the intraday markets, plus the price of purchasing energy in the intraday market (MID+TID). The difference of the income reduction between MD and MID is also shown. The positive difference between (MID+TID) and (MID) comes from the bias that tends to overpredict rather than to underpredict, and then most of times energy must be bought in the intraday market. It is apparent that the decrease in lost revenue may be quite high and obviously increases with time and frequency of intraday markets. For very low penalties, and if there is a bias for overpredictions in the prediction program, to participate in the intraday markets may imply only a small advantage, since the expenses in buying more energy are more or less the same amount than the reduction in penalty for imbalance.
3.3. Uncertainty of Wind Power Predictions

The results shown in the previous sections have been obtained by considering only one value produced by the prediction program used. However, the predictions provided by a short term wind power prediction program are uncertain, and it is interesting to estimate this uncertainty to have more information about the future production of a wind farm. As we will see later, this knowledge may have economic consequences for the wind farm owner. Let p be the random variable associated with the power output of a wind farm. Then, the probability of producing p MW, having predicted p* MW k hours before, is given by the probability density function fp*,k(p). The uncertainty, and hence the probability density function, changes with the range of the wind farm power output, since this value is bounded between zero and the rated power. Besides, the power curve of a wind turbine or wind farm is nonlinear. If we assume that the
126
wind power predictions have gaussian uncertainty , then the probability density functions of the power predictions will not be gaussian [18]. The shape of these probability density functions is also affected by the time lag elapsed between the prediction and the operation times. As shown before, predictions with a shorter time lag are more accurate, and their variance is smaller than those predictions produced longer before. To obtain analytically, or in real time, the uncertainty of this prediction is difficult, but approximate estimations can be made from past data, and some research has already been made in this field [18]. Given the past predictions and wind production for these predictions, the accuracy of these predictions can be tabulated, and then their frequency can be used as an approximation of these probability density functions. If the power range of a wind farm is comprised between 0 and Pmax, and this range is divided in Q intervals, the power p is included in the interval q, if
q 1 q Pmax < p Pmax Q Q

The probability density function fp*,k(p) may change into fq*,k(p), where q* is the interval in which the predicted power p* is included. As an example, the following figures give the frequency distributions of the produced powers for different values and time lags of the prediction. Figure 19 reflects the frequency distribution when a low power had been predicted 7 hours before real time, while Figure 20 shows the frequency distribution when the power level is near the average.
Figure 19. fq*,k(p) for q* = 2 and k = 7. Q = 14.
127
128
Figure 21, on the other hand, shows the frequency distribution for a prediction in the high range of power 7 hours before the operation, while Figure 22 gives the frequency distribution for the same power range, but 36 hours before operation. It can be seen that this last distribution has a larger variance. All these values have been obtained from real production of three months of a wind farm whose rated power has been normalized to 1.
3.4. Bidding Wind Power under Uncertainty

The evaluation of the uncertainty of the predictions can be used advantageously in the bids submission to the electricity markets. Due to the fact that the cost of imbalances are usually higher when the generator is short (it produced less than predicted, and energy must be bought) than when it is long (the production is greater than expected, and extra power is sold), to bid a different power than the average of the probability distribution (the usual output of a prediction program) may yield greater revenues. The following sections quantify this effect, that should be taken into account by System Operators and Regulators.
3.4.1. Problem Formulation Since the predictions are more accurate when the prediction and the operation times are closer, a possible use of this uncertainty evaluation could be to bid the output of the prediction program (average of the distribution) in the daily market (13 hours before
129
operation), and to correct these bids in the intraday, according to the expected price of the energy in this markets, the expected penalty, and the uncertainty of the more recent predictions. This is developed in this section. As explained in a previous section, the revenues of the wind farm may be calculated by the formulas (1) and (2). When bidding in the intraday markets, the bid to the daily market (the committed power) is already known, as well as the price for the energy in this market. The price of the intraday markets is unknown, although they can be predicted, as well as the price of the imbalance. At the time of the intraday market, a prediction is generated for the next hours, and this prediction has an uncertainty estimated from past data as explained in last section. Then, the revenue for a given time could be expressed as a function of the power bid (or power traded) in the intraday markets, and the power actually generated, as R(Plast,Pgen), where Plast, and Pgen are vectors of bid and generated powers for a whole period (one day, or a few hours). The aim of the problem is to obtain the value of Plast that maximizes the revenue for a given set of intraday energy prices, imbalance costs and the uncertainty of the prediction. This problem may be formulated for each hour as in eq. (5). The formulation for more hours is straightforward. Since time couplings between the different hours are not considered, this formulation is general.
max R Plast , Pgen , j f q * , k Pgen , j

Plast j =1
j =Q
{(
)}
(5)
In equation (3), Q is the number of intervals of the power range of the wind farm that have been considered in the uncertainty probability density function, and fq*,k(Pgen,j) is the probability of the generated power to take the value Pgen,j when the prediction k hours before is in the interval q*. The assumed hypotheses for the study have been the following: A pool system has been considered. Wind producers make their bids for a given amount of power at price zero. This means that bids are always accepted. The prediction of the prices of the intraday market is perfect. The subsidies for wind energy are not considered. The prediction tool makes new prediction from available data (wind forecasts and real-time production) every hour. SIPRELICO has been the prediction tool used for performing the prediction.
3.4.2. Study Case In order to show the consequences of taking into account the uncertainty of the prediction into the power bidding, an example has been run with real data from wind farms and electricity prices. The data of wind farm come from the actual production of a wind farm of 14 MW of rated power during three months. The probability functions of the wind farm have been obtained from these same production data and predictions performed for this wind farm for these three months.
130
Figure 23. Prices for the period considered.
Although the study conditions do not follow the Spanish market rules, the prices of the Spanish market between January and March 2003 have been used for this study. The level of the prices is given in Figure 23. The average price for this period was 23.678 c/MWh. The intraday market prices for this period where also used. The average intraday market price used was 22.4791 c/MWh. The study performed includes a comparison between three different assumptions: 9. OPTIMAL: The described method of maximizing the revenues taking into account the uncertainties of the wind power prediction tool. 10. BEST PREDICTION: When the best prediction available is used to modify the bid in the intraday markets. 11. NO INTRADAY: When no updating is produced in the intraday market. Therefore, only one bid per day is produced. The study has been performed for eighty days. Longer studies, however, do not lead to very different results.
3.4.3. Results For the three assumptions, and for values of MPsell = 0,5MP and MPbuy = 1,5MP, the revenues, the average errors and the average of the absolute value of the errors along the whole period are presented in Table 4. It can be seen there that the most profitable option
131
presents larger errors than the most accurate. Errors and absolute errors for instant t are defined as:
errort = Plast , t Pgen, t abs _ errt = Plast , t Pgen, t

Table 4. Errors for the different assumptions
Optimal 290 112 -0.3661 1.8722 Best prediction 280 500 0.1412 1.4132 No intraday 269 046 0.1619 2.0198
(6)
Revenues () Error (MW) Absolute error(MW)
Although the values of MPsell and MPbuy are realistic in the sense that, usually, to buy energy at the last moment is more expensive than to sell it, real systems have more complex methods to calculate the cost of the imbalances. Figure 24 shows the differences between the three assumptions. The asterisks (*) show the difference between the OPTIMAL and the BEST PREDICTION assumptions. The diamonds show the difference between the OPTIMAL and the NO INTRADAY assumptions. The dotted line shows the difference between the BEST PREDICTION and the NO INTRADAY assumptions. It may be observed that the differences, especially between OPTIMAL and BEST PREDICTION are almost always greater than zero. Only in some cases where the predictions had been bad, the results for the BEST PREDICTION were better than the OPTIMAL assumptions. However, if we check the error between the actual generated power and the different powers, as shown in Table 4, we can see that, even if the errors of the OPTIMAL assumptions are greater, the revenues obtained are higher, and then, the bids provided by the wind farm owners aiming to maximize their revenues are not the most accurate. In Figure 25 a sample of the power error in a series of 30 hours is shown. It may be seen there, that the error of the best prediction is lower than the optimal, and also lower than the case where only bids to the daily markets are presented. Although there are cases where this does not happen, the general behaviour follows this pattern. Table 5. Revenues for different sell and buy prices ()
Buy/Sell prices Optimal Best prediction No intraday 1.5/0.5 290 112 280 500 269 046 1/1 348 834 311 751 313 039 0.5/1.5 467 178 343 002 357 032
132
Figure 24. Results for one month of the cases studied.
Figure 25. Power errors for a series of 30 hours.
133
From Table 4 we can deduce that the prediction tool has a trend for overprediction as expected, because the errors, defined as in (4) are positive. This is more apparent when the buy and sell prices change from those assumed in the previous results. Although the following discussion is mainly of theoretical interest, it gives a good insight into the properties of the process. The Table 5 gives different results for different values of sell and buy prices. The second column has the same values as the first row of Table 4, but it is included here for an easier comparison. We can see in this table that: 1. The revenue when bidding the optimal power is always the highest of the three possible bids. 2. The revenue is higher when the sell price becomes greater. This means that the system tends to overpredict, and then, power must be bought at the last moment most of the times. This need of purchasing is an additional cost that reduces the overall revenues. 3. When the sell price is equal or lower than the buy price, the revenues are larger when no updates are made in the intraday markets. This is also a consequence of the tendency to overpredict, since the error is higher in this last case, as shown in Table 4. In order to show the bias of the errors and absolute errors, these errors are given in Table 6 and Table 7. In them, the values for the BEST PREDICTION and NO INTRADAY assumptions given in Table 4 are also included for an easier comparison. From these tables we may conclude also that: The lowest errors are always given by the BEST PREDICTION case. This better performance, however, does not lead to the highest revenues, even compared with the NO INTRADAY option, if the sell price were higher than the buy price. The errors from the OPTIMAL case show that with this strategy, the trend to overpredict of the prediction tool is compensated. When the sell price is higher than the buy price, the bid tends to be higher than the actual generation. The fact that the average error is negative when both prices are equal is another consequence of this trend to overprediction. Table 6. Average errors for different sell and buy prices (MW)
Buy/Sell prices Optimal Best prediction No intraday 1.5/0.5 -0.3661 0.1412 0.1619 1/1 -0.2423 0.5/1.5 0.6255
Table 7. Average absolute errors for different sell and buy prices (MW)
Buy/Sell prices Optimal Best prediction No intraday 1.5/0.5 1.8722 1.4132 2.0198 1/1 4.2822 0.5/1.5 5.5945
134
4. OPTIMAL DISTRIBUTED MANAGEMENT OF WIND POWER

Geographic distribution of wind farms depend mainly on best wind resource locations and environmental regulations. Therefore, wind parks are generally concentrated in zones of the country, where these installations are more profitable. In addition, wind farms should be connected to the transmission system in buses which can accept non-dispatchable generation in a secure way. For this, it is frequent to find two or more wind farms injecting power to the same transmission system node. The wind farms are connected to the transmission bus through a subtransmission grid, generally including only wind farms. This is the case in Spain, where the larger installed wind power (4922 MW, 42.4% of the full wind power installed capacity) is allocated in the North-West side of the country (2007 data, [38]). In this region, up to 6 wind farms of an average power of 30 MW may inject active power to the same transmission node. The geographic closeness among the wind producers and their particular production characteristics have led to create regional Renewable Control Centres or Delegated Dispatches (DDs), to control the wind farm production. At the moment, DDs only can include wind power generators linked to the same transmission bus and to proceed in contingence situations. Spanish System Operator (REE, Red Elctrica de Espaa) had created a centre called CECRE (Centro de Control para Rgimen Especial), to manage the renewable production in contingency situations. CECRE coordinates the action of DDs and independent wind farms (not included in any DD) in contingency situations, receiving information about the production of the wind farms and setting the maximum admissible active power injections of wind power in the nodes of the transmission system. In the next sections, an optimal approach for the DD control actions in contingency situations is developed, considering the restrictions of the operation and the predictions of active power production of the wind farms.
4.1. Power Curtailments of Wind Power Production

In the Spanish electrical system, some generators can inject all the available production to the system in normal situations, independently of the hourly market price. This is the NonDispatched Generation (NDG) of the system. Wind power, solar, hydraulic without regulation dam and other producers that use fluent or variable renewable energy are included in the NDG. However, the SO may require a reduction in the production of the NDG, if it is necessary to prevent risks in the system operation. This decrease can be demanded to the wind parks if there are not other means to avoid the risks. The following problems could be cause of NDG reductions [32]: Congestions in the generation provision: the SO calculates the maximum input of NDG that the system can absorb, in situations when the power predicted by the NDG can not be injected into the system, either because of excessive wind power production or due to restrictions in the transmission grid. Stability reasons: wind parks are sensitive to temporary voltage reductions (voltage dips). Reductions in the voltage value (even for periods lower than 100 ms.) could produce disconnections of older technology wind parks, with the subsequent stability
135
risks. The SO must calculate the maximum of NDG that could be instantaneously disconnected by a voltage dip in each bus of the system, without causing stability problems. In this calculation, the SO must consider the sensitivity of each wind park to the voltage dips. Viability of the power system balances: the SO must guarantee both active and reactive power balances in the system. To perform this task, technical restrictions can be imposed by the SO to dispatched and not-dispatched generation units. The SO must consider possible generation surpluses, produced by either generations greater than forecast or demands lower than the expected loads.
Based on security analyses, the SO calculates at every moment the maximum amount of NDG that can be safely injected to the system, for each transmission bus of the grid. When a reduction of the NDG is required in some part of the system, the SO must select which producers must decrease their production. This assignment involves numerous tasks: the SO requires accurate predictions of the NDG for the following periods, the producers send the information to the SO, which collect the data, the SO calculates the required reductions in each transmission node and finally the SO informs the producers of the generation maximums. These tasks must be performed generally in critical situations, increasing the effort of the SO in these circumstances. Therefore, some SO (as REE) had proposed the creation of regional DDs, to act in such critical situations. When the SO detects the need of a NDG reduction, communicates to the related DD the maximum of acceptable NDG for the transmission buses involved. The DD calculates the optimal reduction of the generation in the transmission bus, informing the producers of the new operational points. DDs operate as intermediate entities between the SO and the non-dispatched producers, acting (as specified in the current legislation) only in critical situations. In Spain, the largest amount (54,6%, [28]) of the NDG corresponds to wind generation. Therefore, in a first step, DDs only could manage wind power producers, being expected the incorporation of other non-dispatchable producers in the next years. The first Spanish DDs are (in 2007) being tested and it is expected that they will be in operation along 2008. Under the present regulation [32], NDG producers may choose the way in which the reductions are calculated: either they follow the settings set by the SO, or they associate in a DD and follow their settings. In the last case, the producer must declare its wish to participate to the DD, signing an agreement. On the other hand, the SO calculates the required reduction by following a proportional rule (explained in the following sections of the chapter) while DDs can use a proportional rule or any other to find the optimal reduction for each of its members. In any case, the DD calculation methodology must be previously approved by the SO, and the operational constraints (including the maximum admissible NDG in the transmission node) must be satisfied. In the next sections, an alternative to the proportional rule that can be used by the DD is described, considering the different controllability of the producers.
4.2. Rate of Decrease for Non Dispatchable Generation

Once the maximum of admissible NDG in the transmission buses has been determined, the SO matches these values with the predicted power injection in the nodes. Both, admissible
136
NDG and forecasted production should correspond to the same period of the next future. Due to technological restrictions, there is a delay between the request of reduction and the effective application of the reduction. In the present analysis, the delay is 15 min. Therefore, the first period of application of the restriction will be (at least) 15 min. after the calculation. In the transmission buses in which NDG prediction exceeds the admissible injection, a reduction of the NDG is necessary. The SO calculates the reduction of the NGD production bus by bus, using a proportional rule. Suppose that in a bus the prediction for the total
Av Av production of the wind farms ( PGj , with PGj the estimation for the available active power
ng
j =1
production of the wind farm j and ng the number of wind farms connected to the transmission Max bus) in a period of the future exceeds the admissible NDG ( AdPOut ) in the bus for the same period, as shown in (7).
P
j =1
ng
Av Gj
Max > AdPOut
(7)
The rate of decrease (DR) can be calculated as (8).

DR =
Max AdPOut Av PGj j =1 ng
if 0.0 DR < 1.0 NDG reductions = normal operation if DR 1.0
(8)
In (8), when DR varies between 0 and 1 the SO may require reductions in the active power production of the NDG. The active power production of producer j can be calculated, using DR, as in (9).
Av PGj = DR PGj = Max AdPOut Av PGj
P
j =1
ng
(9)
Av Gj
4.3. Calculation of Power Reduction by the Delegated Dispatches.

When the SO communicates the possible existence of critical situations to the involved DD, implying probable NDG diminutions, the DD must calculate the reductions needed for not exceeding AdPOutMax. For this task, the DD may also use a proportional rule, as previously presented, or any other. However, the calculations performed by the DD and SO may differ in three points. a) As presented in previous sections, a reasonably accurate prediction can be obtained for the average production of the wind farms for the next 15 min. In this period, the wind power prediction can be considered a known quantity. However, if the security of the system is evaluated for periods beyond 15 min, uncertainties associated with wind power previsions could be considered. In this aspect, DD calculations may have
137
a relative advantage in relation to the prevision performed by the SO. Better communications with the producers, a better knowledge of the wind characteristics of the region and the possibility of relating the prediction of a wind farm to the others suggest that the DD forecasts could be more accurate than the SO predictions. b) The proportional rate calculated in (8) does not consider the power losses in the internal grid to the DD. The wind farms are connected to the transmission node through a distribution network, supervised by the DD. Therefore, the total production injected into the transmission node is the added production of the generators, decreased by the losses in the DD internal grid. If the amount of active power losses in this network is small, (8) is a good approximation. However, if active power losses represent an appreciable value, the wind productions would be decreased more than required, causing economic losses to the producers because the injected wind powers would be below the admissible NDG. An improved proportional rate calculation, that allows increasing the NDG production to compensate the DD internal losses, is analyzed in [4]. c) In the DD calculation, the different technological control possibilities of the wind farms may be considered. When DR is calculated by the DD, the sharing of the power decrease can be performed only by the controllable wind farms. As presented in the next section, not all the wind parks can control the active power generation continuously. The ability of the wind farms to modify the production as wished depends on the technology implemented in the wind park. Suppose that only m1 wind farms, of the m generators belonging to the DD (m1 m ng), can modify voluntarily the active power generation in the DD. Also, suppose that a reduction of
Av Max the NDG is required in the bus ( PGj > AdPOut ). Therefore, the Decrease Rate j =1 ng
considering only m1 controllable generators (DRm1) is presented in (10), where AdPDDMax is the maximum admissible input of DD in the transmission bus.
AdP
m1 j =1 Max DD Av Gj Av DR PGj j =1 Av PGj j =1 m1 m
DRm1 =
(10)
The reduction effort is only assumed by the m1 controllable generators, as shown in (11).
Max AdPDD Av Av PGj j = 1, ... , m1 DRm1 PGj = m1 Av PGj = PGj j =1 Av PGj j = 1, ..., ( m - m1) m
(11)
Av Max In (11), when the required reduction of the production ( PGj AdPDD ) exceeds j =1
m1 m Av Max Av the control capacity of the DD PGj AdPDD > PGj , the non-controllable j =1 j =1
138
Julio Usaola and Edgardo D. Castronuovo generators of the DD may be called on to switch off the complete production, to satisfy the SO constraints. The (m-m1) non-controllable units will compensate the controllable wind farms for the reduction service, after the operation. Many payment schemes can be used for this compensation, for instance, to share it proportionally among the non-controllable generators, or according to the rated power, to the production, or considering controllability prices, etc. The consideration of controllability (and interruptive) prices in the payment scheme allows to implement a method based on market rules within the DD, as shown in next sections. Firstly, a classification of the wind power generation units, based on their controllability and interruptive abilities, is presented.
4.4. Classification of Wind Power Producers

Three main technologies dominate now the wind turbine market: squirrel cage induction generators directly coupled to the grid, direct drive synchronous generators and doubly fed induction generators [31]. The controllability of the wind farm turbines depends on the technology used. Doubly fed induction generators may vary continuously their production within their operational limits. This kind of turbines can follow different control strategies, as shown in [1], [8] and [9]. However, wind producers able to control their production could decide not to participate in the DD reduction service. Their willingness for active participation in the DD may depend on the benefits of participating. Therefore, the controllable wind power producer can offer its production control to the other producers, putting a price to this service, the Controllability Price. All the wind farms have the capacity to cut off the complete production, disconnecting the wind farm from the network. In fact, this is the basic measure that the SO applies to the wind parks, switching off the parks in the connection bus under contingences. The wind farm owners can choose between allowing the DD (or the SO) to disconnect them only in very critical situation, or to put also a price to this service, the Interruptive Price. Partial curtailments (disconnecting only some of the wind park turbines) could be also possible, but they have not been considered in the model. However, the consideration of this ability can be easily included in the formulation. According to the controllability and the interruptive aptitudes of the wind farms, they can be classified into the following groups: Type 1 wind parks offer to the DD both interruptive and controllability prices, aiming to participate in the control of the DD active power production under contingencies. Their services will be used if the price they give for them is low enough. When the wind farm can not be controlled but the producer wants to participate in the reduction service, by offering its interruptive ability, he may submit a very high controllability price. In [5], it is shown that the control of the power factor is necessary to perform an adequate production control. Therefore, Type 1 wind farms can be required to set their production and their power factor, between the limits imposed by the reactive power production of the generator and the available wind power. The reactive power control is not paid in the present approach.
139
However, a payment for reactive power controllability might be included in the model without problems. Type 2 wind parks contribute to the operation of the DD by controlling the power factor of their production. As previously expressed, in the present paper the contribution of these producers to the system operation is not paid. The active power output of Type 2 wind farms is the available wind power at the moment. Type 3 wind parks are producers that cannot, or do not wish to control either their active power generation or their power factor. The active power delivered by these wind farms is the available wind power at each instant. The power factor of these wind farms is one fixed value, according to the existing incentives for power factor specifications [33]. Type 4 wind parks are those WPG that do not belong to the DD, but that are connected to the same output transmission bus. Under current regulation, this situation is possible. Type 4 Wind Parks receive the reduction setting directly from the SO, who follows a proportional rule (8). Like Type 3 wind farms, the Type 4 wind power producers usually specify a fixed power factor for the production.
Wind parks of Types 2, 3 and 4 may be interrupted, in very critical situations and as last choice, to solve the SO constraints.
4.5. Optimal Allocation of Power Curtailments under Market Rules

In the power reduction imposed by the SO, all the producers are considered controllable and with equal controllability cost. Therefore, the reduction effort is equally distributed among the NDG, using a proportional factor. However, some of the producers can not be controlled (facing interruptions in critical situations) or do not want to control the production (because of the management and maintenance costs associated). Furthermore, the technological alternatives of production introduce different charges for the control actions. Therefore, the costs for both control and interruptive actions will vary among the producers. A methodology based in bids is analysed in the present section, considering that the Type 1 wind parks offer within the DD their controllability and interruptive abilities. The prices submitted by these producers for their controllability and interruptive actions must consider the following aspects: Cost of reducing the production: the producers that reduce the generation to reach SO constraints will decrease the revenue obtained for injecting active power generation into the system. The price of the NDG depends on the regulation adopted to reward this kind of generation. In general, to calculate the price of NGD market prices and possible governmental incentives must be considered. Management costs: to control the active power generation in a wind farm requires to monitor the production, to decrease the generation in some of the turbines and probably to switch off others, when necessary. Besides this, the set-points of the generators (or electronic devices) must be handled to reach the reactive power
140
Julio Usaola and Edgardo D. Castronuovo generation required for a secure operation. Specific software, control equipments and specialized personnel must be employed. Amortization of the controllability equipments and increased maintenance costs: possible increments in the maintenance expenses for additional operations and the amortization of the control devices must be considered in the prices offered for reduction operations.
To simplify the formulation, maximum ramps for increasing/decreasing generation between consecutive periods and start-up costs were not considered. These aspects are generally considered not significant in wind turbines. The next section develops the mathematical formulation proposed.
4.6. Mathematical Formulation

The mathematical formulation used to distribute optimally the reduction effort among the DD producers for the next hours, considering prices for controllability and interruptive services, is presented in equations (12) to (26).
min
( cp
m1 j =1 m 3+ m 4
Av CR j + ip j IR j PGj )+ 2 m4
m 2 + m3
j =1
Av fnp ( SGj cos j PGj ) + ... 2 Av 2 Gj
j =1
fnc ( j f
) + fnp ( S
j =1
Gj
cos j DR P
(12)
s.t.
Max Pout AdPOut
(13) (14) (15) (16) (17)
SGi cos i PDi Pi (V , ) = 0

S Gi sin i QDi Qi (V , ) = 0
sk = 0
S G j cos j + C R j = PGav (1 IR j ) j
CR j
av PGj
1.0 IR j
(18)
cos 1 ( fpf ) for capacitive fpf f = cos 1 ( fpf ) for inductive fpf
(19)

Max AdPOut
141
DR =
P
j =1
ng
Av Gj
(20)
SGi cos i 0
cos j cos min j
(21) (22)
Vi min Vi Vi max
Tikmax Tik Tikmax i k
(23) (24)
CR j 0
(25)
IR j = {0;1}
i , k = 1...n j = 1...m1
(26)
In equations (12)-(26), cpj and CRj are the controllability price and controllability reduction factor of wind producer j, respectively; ipj is the price of wind producer j for disconnection; IRj is a binary variable, representing the connection of wind producer j; fnp and fnc are the weight coefficients for maintaining the active power production (of Types 2, 3 and 4 wind farms) and the power factor (for Types 3 and 4 wind farms) in the specified values, respectively; DR is the proportional factor, used by Type 4 wind farms to calculate their required active power production; Pout is the total active power output of the DD area; PDi and QDi are active and reactive power demands at bus i, respectively; Vi and i are the module and angle of the bus voltage, respectively, at bus i, sk is the voltage angle at the slack bus; cos jmin is the minimum power factor of the j producer; Vimin and Vimax are minimum and maximum limits of the voltage module in bus i; Tik is the apparent power transmission max between buses i and k; Tik is the maximum apparent power limit of the transmission line between buses i and k; fpf is the specified power factor in Types 3 and 4 wind farms; m (m=m1+m2+m3) is the number of wind farms included in the DD action; m1, m2, m3 and m4 (m4=ng-m) are the numbers of Types 1, 2, 3 and 4 producers; and n is the number of buses at the internal of the DD system, including the transmission node. Objective function (12) minimizes the cost for reduction actions, both for interruptive and controllability procedures. This is the primordial objective of the DD action, to distribute the reduction effort among the more adequate units. Also, objective function (12) includes some equality restriction, by using large-value multipliers fnp and fnc. In Types 2 and 3 wind farms,
142
the active power generation is not controlled; therefore the production must be equal to the available wind power in each instant. Type 2 and 3 wind farms generate at a fixed production angle, f. Finally, the generation in Type 4 producers is specified by the SO, following a proportional rule. These equality equations are included in the objective function, instead as equality constraints, to improve the convergence of the algorithm in very critical situations. In addition, violations of these equations will show possible corrective alternatives in non factible solutions. Restriction (13) includes output restriction specified by the SO for NDG in the transmission bus. The power flow nonlinear equations within the DD internal grid are shown in (14) and (15), considering a null value of the voltage angle at the slack bus (16). The discrete variable IRj specifies the unit commitment for the j producer. If IRj = 1, the j wind farm is disconnected; if IRj = 0, the j wind park is connected. In connected wind farms, the available active power can be decreased in a continuous mode, modifying CRj, as shown in (17). The continuous control CRj can only be used in connected producers with IRj = 0, as expressed in (18). The fixed production angle for Types 2 and 3 wind farms is calculated in (19), as function of the specified power factor fpf. The proportional factor DR, used by the SO in Type 4 wind farms, is obtained in (20), as specified in (8). Minimum values for the active power production and the power factors of the producers are shown in (21) and (22). Operational constraints for voltages and apparent power flowing in the transmission lines within the DD area must be also considered in the model, (23) and (24). The optimization problem (12)-(26) must be executed for each programming interval of the future. The model results in a mixed-integer nonlinear optimization problem. It can be solved by using Branch and Bound, Branch and Cut [6], [7] and [3915], Lagrangian relaxation [393], [24], [37], [39] and [40] or other methods [26]. In the present case, an exhaustive enumeration algorithm was implemented [39], [26], due to the low number of discrete variables. It is reasonable to assume that up to 6 big wind farms may be injecting power to the same transmission node.
4.7. Test Case

The method was evaluated in a real network, extracted from the Spanish National grid. In this test case, 6 wind farms inject power into the system through a transmission node. The main characteristics of the internal grid and the wind farms are shown in Table 8 and Table 9. Table 8. Internal DD System
From 1 2 2 2 3 3 4 4 To 2 3 4 9 5 6 7 8
R [pu] 0.00132 0.00360 0.00420 0.00500 0.00120 0.00180 0.00160 0.00680 XL [pu] 0.0720 0.1845 0.2153 0.2563 0.0615 0.0692 0.0615 0.0307 BC [pu] 0.0040 0.0008 0.0001 0.0007 0.0010 0.0016 0.0014 0.0020 Tikmax [pu] 14.2 4.8 4.4 2.7 2.2 2.4 2.0 2.2
Wind Energy in Electricity Markets with High Wind Penetration Table 9. Buses and Wind Farms
Bus 1 2 3 4 5 6 7 8 9 Producer Name WG2 WG5 WG6 WG7 WG8 WG9 Type 1 1 2 4 1 1
PGjMax [pu] 5.000 2.000 2.200 1.800 2.100 2.500 ipj [/MWh] cpj [/MWh]
143
B [pu]
75 72 82 78
150 150 87 84
0.899 2.100 0.000 0.000 0.000 0.000 0.000 0.049 0.000
In Table 8, the pu values of the resistance (R), series reactance (XL), parallel susceptance (BC) and maximum apparent power flow (Tijmax) in the transmission lines of the grid are shown. Apparent power and voltage bases are 10 MVA and 66 kV, respectively. In the simulations, the output and slack bus is 1 and the voltages vary between 0.95 and 1.05. As shown in Table 9, there are 6 wind generators in the test system, with rated capacities PGjMax. One of them (at bus 7) does not want to participate in the DD control, although it is injecting power to the same transmission node (bus 1). Another one (connected to bus 6) only controls reactive power, injecting all the available active wind power to the grid. The other four aim to participate in the reduction control, submitting both controllability and interruptive prices to the DD. The wind power producers at buses 2 and 5 cannot control the active power production, offering a large value for this service. It must be stressed that the prices presented in Table 9 are only orientative values, used to show the operation within the DD. In this table, the compensation susceptance (B) in the buses is also included. In Figure 26, the test system is represented graphically. Bus 1 is the transmission node, in which the wind farms inject the production into the system.
1 2 3 4 5 6 9 7 8
Figure 26. Test System.
144
9 8 Active Production [pu] 7 6 5 4 3 2 1 0 0 5
10
15 Period
20
25
30
Available Wind Power in the Type 1 producers. Available Wind Power in the Type 4 producers. Available Wind Power in the Type 2 producers. Admissible Wind Power.
Figure 27. Available and Admissible Wind Powers.
In the present formulation, the solution of optimization problem (12)-(26) provides the optimal generation profile for one period of 15 minutes. In the next simulations an interval of 8 hours is simulated, solving 32 consecutive optimization problems (one for each 15 minutes period). In Figure 27, the available and admissible wind powers for each period are shown. In Figure 27, the available wind power (classified by controllability type) and the output restriction imposed by the SO are depicted, in all the 32 programming periods. In periods 1-2, 14-16 and 31-32 there are not restrictions for injecting wind power into the transmission node. In all the others periods, the SO constraints the admissible wind power at different levels. The available wind power in Type 1 producers constitutes the large amount of renewable production, resulting of the addition for the previsions of WG2, WG5, WG8 and WG9. Also, in Figure 27 the available wind powers of Types 2 and 4 wind power producers are shown. There are not Type 3 wind power producers in the considered system (Table 9). In Figure 28, the results obtained by solving the optimization problem in each period are shown. As shown in Figure 28, SO constraints can be solved only by using the combined abilities of Types 1 and 4. If the admissible wind power is below the available production of the controllable producers (Types 1 and 4), disconnections of non-controllable Types 2 and 3 may be required. In the presented simulations, the optimization problem (12)-(26) gives the amount of generation in Type 4 producers following the SO proportional rule (8), it maintains constant (if possible) Type 2 active power production and determines the optimal production of the
145
Type 1 wind power generators. In those periods when a decrease of the total wind power production is required, it can be observed (Figure 28) that the added production is slightly higher than the admissible wind power. The optimization problem explicitly considers the active power losses in the operation, increasing the Type 1 production to closely reach the admissible wind power limit at the transmission node. The best combination of generations in the Type 1 productions depends of both controllability and interruptive abilities of each producer, and the corresponding prices for these services. In Figure 29 to Figure 33, the production of the controllable producers is shown.
9 8 A ctive Production [pu] 7 6 5 4 3 2 1 0 0 5 10 15 Period 20 25 30
Figure 28. Production and Admissible Wind Powers.

3 A ctive P roduction in W 2 [pu] G 2.5 2 1.5 1 0.5 0 Av. Wind Power Wind Power Prod.
10
15
Period
20
25
30
Figure 29. Production and Available Wind Power in WG2.
146
1 A ctive Production in W 5 [pu] G
Av. Wind Power Wind Power Prod.
0.8
0.6
0.4
0.2
10
15 Period
20
25
30
Two control strategies converge in transmission node 1. WG7 follows the proportional rule, as calculated by the SO. Therefore, this producer must modify continuously its active power generation (Figure 31) in all the periods in which restrictions on wind power are applied. The SO proportional rule (8) depends on the admissible wind power and on the wind power predictions for the wind farms connected to the same transmission node. Thus, the decrease rate DR varies in each interval and WG7 must adapt its own generation to this variation. The other controllable producers (WG2, WG5, WG8 and WG9) are managed by the DD. Wind power producers with only interruptible abilities (WG2 and WG5) are called on to switch off the generations when necessary. WG5 (Figure 30), with the lowest interruptible price, is firstly requested to suspend its generation. Larger wind power reductions demand the disconnection of WG2. Depending on the intensity of the reduction and the power generation, other controllable producers (WG8 in periods 5, 6, 27 and 28 and WG9 in periods 25 and 26) may be called to interrupt their production. The lowest controllability price was offered by WG9. Therefore, this producer is required to permanently vary the active power generation to follow, as close as possible, the maximum of available wind power accepted by the system at bus 1. Only when WG9 is called to interrupt the production (intervals 25 and 26), WG 8 (the other controllable producer) is requested to control the active power production. As expected, in the solution of the optimization problem the DD producers without continuous control abilities (WG2 and WG5) are not called on to control the production at any interval.

1 Av. Wind Power Wind Power Prod. A ctive Production in W 7 [pu] G 0.8
147
0.6
0.4
0.2
10
15 Period
20
25
30
A ctive Production in W 8 [pu] G
1 0.8 0.6
Av. Wind Power Wind Power Prod.
0.4 0.2 0
10
15 Period
20
25
30
4.8. Ex-post Analysis. Economic Balance

After the operation, the economic balance for all the participants of the DD must be settled. The producers that reduced their own generation to match SO requirements must be rewarded by the others. This operation is performed for each period. As an example, the 5th period is considered. In Table 10 and Table 11, the productions and flows in this period are synthesized.
148
1.5 Av. Wind Power Wind Power Prod. A ctive P roduction in W 9 [pu] G
0.5
10
15 Period
20
25
30
In the 3th column of Table 10, the available wind power of the wind parks is shown. DD manages 5.05 pu of available wind power and 0.63 pu (of WG7) is directly controlled by the SO, totalizing 5.68 pu. This total amount of wind power can not be injected into the transmission grid, due to the imposed restrictions, and should be reduced to 2.2 pu. Following a proportional repartition, the SO calculates a decrease rate (8), as shown in (27).
Max AdPOut
DRSO =
P
j =1
Av Gj
2.2 = 0.3873 5.68
(27)
Table 10. Active Power Balance at Interval 5. ADPOUTMAX = 2.2 p.u.

PGjAv [pu] 0.6300 1.8500 0.6800 0.8400 0.5300 1.1500 5.0500 5.6800 SO Proportional [pu] 0.2440 0.7165 0.2634 0.3254 0.2053 0.4454 1.9560 2.2000 Results of Optimization [pu] 0.2440 0.0000 0.0000 0.8400 0.0000 1.1363 1.9763 2.2203 DD Proportional [pu] -0.7240 0.2661 0.3287 0.2074 0.4500 1.9763 --
Management SO
Producer WG7 WG2 WG5 WG6
Pj [pu]
0.0000 0.7240 0.2661 -0.5113 0.2074 -0.6863 0.0000 0.0000
DD
WG8 WG9 Total
Total
Wind Energy in Electricity Markets with High Wind Penetration Table 11. Economic Balance at Interval 5. = 71 /MWh, DD = 84 /MWh
SO Proportional [] 43.3100 127.1788 46.7535 57.7585 36.4585 79.0585 347.1900 390.5000 Results of Optimization [] 43.3100 0.0000 0.0000 149.1000 0.0000 201.6933 350.7933 394.1033 Reduction Control [] -152.0400 55.8820 -107.3630 43.5540 -144.1130 0.0000 --
149
Management SO
Producer WG7 WG2 WG5 WG6
Total Remuneration [] 43.3100 152.0400 55.8820 41.7370 43.5540 57.5803 350.7933 394.1033
DD
WG8 WG9 Total
Total
Applying DR to the available wind powers, results the SO proportional reduction shown in 4th column of Table 10. WG7, controlled by the SO, produces (0.38730.63=0.244 pu) in this period. To match the SO output restriction, the total DD injection should be (0.38735.05=1.956 pu). The solution of optimization problem (12)-(26) (5th column of Table 10) explicitly considers the losses, therefore the amount generated within the DD is slightly greater, 1.9763 pu (1% of increased production). After the operation, the effective contribution of the wind farms to the reduction effort is calculated. Considering only DD producers and the real generation, the proportional decrease rate within the DD is obtained in (28).
Max AdPDD Av PGj j =1 5
DRDD =
1.9763 = 0.3913 5.05
(28)
In the 6th column of Table 10, the requested reductions (following the DD proportional rule, (28)) for the DD producers is shown. As shown in the 5th column of this table, the economic rule used in the optimization problem implies that some producers have increased reductions and other ones inject all their available wind power. The difference between the productions calculated using either DD proportional or economical rules is shown in 7th column of Table 10. In Table 11, the economic results are analysed. In the 3th column of the table, the remuneration obtained by the producers if all of them follow the SO proportional rule is shown. As an example, considering a market price for the wind power generation injected in the system of 71 [/MWh] (including possible incentives), a decrease in generation of 0.244 pu in the wind park, with base power of 10 MVA (1 pu = 10 MVA) and a 15 minutes interval (15 min. = 0.25 hour), WG7 receives (710.244100.25 = 43.31 ). After the optimization process, the payment obtained by the DD producers for injecting power to the system, at the
150
market price for wind power generation, is shown in the 4th column of the table. As an example, WG6 receives (710.84100.25 = 149.1 ). However, these amounts do not consider the reduction effort performed within the DD. In fact, WG6 should not reduce its generation, when following the economic rule implemented in the optimization problem. As shown in 7th column of Table 10, WG6 is not performing its reduction of 0.5113 pu. The internal price for the DD reduction effort is determined by the highest price unit required to reduce the production. In the present case, the internal DD reduction price is specified by the controllability of WG9, DD = 84 [/MWh]. Therefore, WG6 should pay to the DD producers that reduced the production in an amount equal to (840.5113100.25 = 107.36 ). In the 5th column of the Table 11, the payments and rewards for the reduction effort within the DD producers is represented. Adding 3th and 5th columns of the table, the total remuneration of the producers is shown in the 6th column. In synthesis, the formula for calculating the remuneration of a wind park acting in the DD is presented in (29).
Av Profit j = SGj cos j + DD ( DRDD PGj SGj cos j )
(29)
5. CONCLUSION
The high penetration of this resource in the grids has raised concerns about the best way of integrating it into the power networks. The nature of wind power implies that the power production is highly changing and that it is difficult to manage with the traditional power system analysis and control tools. This leads to the need of developing new ones and to implement operation strategies that facilitate the integration of this clean and renewable resource without decreasing the efficiency and reliability of power grids. In the previous sections, some of these new tools have been shown. The short term wind power prediction tools have been demonstrated to be necessary for the power system operation in grids with high wind power penetration. They have a reasonable accuracy, and are easy to implement and include in a power system control centre. When these predictions are produced for several wind farms, the accuracy of the joint prediction is even greater, and if predictions for only a wind farm (or a small number of them) are needed, then a higher detail in the wind farm modelling may lead to an assumable accuracy. This is still an open field for research, since more intensive computation tools (for numerical weather predictions) will lead to higher accuracies, and also new statistical and mesoscale models will give better results, according to the promising results of the research in this subject. The integration of wind energy is also helped if wind generation participates in electricity markets as any other kind of generation. This leads to a greater implication of wind power producers into the system management and operation, and drive them to minimize the imbalance between the committed and the generated powers. Of course, prediction programs play a basic role in this participation and this minimization of deviations. It has been shown that these imbalances, when a prediction program is used, have an assumable cost and that trading in intraday markets and taking into account the uncertainty of the prediction still reduces them to a lower value.
151
The need of keeping power system in a secure operation may lead to wind power curtailments at some moments. To decide which wind farms must reduce their production should be decided taking into account the technical possibilities of the wind farms, and the price that the farm owners give to this control. This is why an optimization process that uses market mechanisms is an adequate and transparent way of allocating power curtailments among a group of users. The consequences of following this strategy have been shown in the previous sections and they are reasonable and easy to accept and understand. New forms of generation demand a new management of the grid, if we want to keep the reliability of power networks and to include the renewable sources of energy that are required by the society.
6. REFERENCES
Almeida R.G., Castronuovo E. D., Peas Lopes J.A., Optimum generation control in Wind Parks when carrying out system operator requests, IEEE Trans. on Power Systems, pp. 718-725, vol. 21, n. 2, 2006. [2] Asociacin Empresarial Elica (2006) Ejercicio de Prediccin. Informe final. Available at http://www.aeeolica.es. [3] Bailey B., Brower M., Zack J., Wind Forecast: Development and Application of a Mesoscale Model. Wind Forecasting Techniques: 33 Meeting of Experts. Technical Report from the Internatinal Energy Agency. R&D Wind. Ed. S.-E. Thor, FFA, Sweden,pp. 93-116. 2000. [4] Castronuovo E.D., Usaola J., Alternatives of revenue for corrective actions of wind generators in a Delegated Dispatch, Proc. of the IEEE International Conference on Clean Electrical Power, Capri, Italy, pp. 567-573, May 2006. [5] Castronuovo E.D., Usaola J., Optimal controllability of wind generators in a Delegated Dispatch, Electr. Power Syst. Res, vol 77, n. 10, pp. 1442 1448, Aug. 2007. [6] Chen X., Bushnell M. L., Efficient branch and bound search with application to computer-aided design, Boston, USA, Kluwer Academic, 1996 . [7] Dillon T.S., Edwin K.W., Kochs H.D., Taud R.J., Integer programming approach to the problem of optimal unit commitment with probabilistic reserve determination, IEEE Trans. Power App. Syst., vol. PAS-97, pp. 2154-2166, Nov./Dec. 1978. [8] Ekanayake J., Holdsworth L., Jenkins, N. Control of DFIG wind turbines, Power Engineering Journal, vol. 17, n.1, pp. 28-32, 2003. [9] Ekanayake J., Jenkins N., Comparison of the response of doubly fed and fixed-speed induction generator wind turbines to changes in network frequency, IEEE Trans. on Energy Conversion, vol. 19, n. 4, pp. 800-802, 2004. [10] Ensslin C., Ernst B., Rohrig K., Schlgl F. Online Monitoring and Prediction of Wind Power in German Transmission System Operation Centers. Proceedings of the EWEC03. Madrid (Spain), June 2003. [11] Focken U., Lange M., Waldl H.-P. Previento- A wind Power Prediction System with an Innovative Upscaling Algorithm. Proceedings of the EWEC01. Copenhagen (Denmark), July 2001. [1]
152
[12] Giebel G. The State-of-the-Art in Short-Term Prediction of Wind Power. A Literature Overview. Available at http://anemos.cma.fr [13] Giebel G. (ed.) Wind Power Prediction using Ensembles. Report Ris-R-1527(EN). Risoe, Roskilde, 2005. [14] Gonzlez G., Daz-Guerra B., Soto F., Lpez S., Snchez I., Usaola J., Alonso M., Lobo M.G. SIPRELICO- Wind power prediction tool for the Spanish peninsular power system. Proceedings of the 2004 CIGR 40th General Session & Exhibition. Pars (France), August 2004. [15] Huang K.Y., Yang H.T., Huang C.L., A new thermal unit commitment approach using constraint logic programming, IEEE Trans. on Power Systems, Vol. 13, pp. 936-945, Aug. 1998. [16] Kariniotakis, G., et al., Next Generation Short-term Forecasting of Wind Power Overview of the Anemos Project, Proceedings of the EWEC06, Athens (Greece), March 2006. [17] Kariniotakis G. et al. What performance can be expected by short term wind power prediction models depending on site characteristics?. Proceedings of EWEC04. London (England), November 2004. [18] Kariniotakis G.N.; Pinson P. Uncertainty of short-term wind power forecasts a methodology for on-line assessment. Proceedings of the International Conference on Probabilistic Methods Applied to Power Systems. Ames (USA), September 2004. [19] Landberg, L.: Short-term Prediction of Local Wind Conditions. PhD-Thesis, Ris-R702(EN), Ris National Laboratory, Roskilde, Denmark 1994, ISBN 87-550-1916-1. [20] Lai S.Y., Baldick R., Unit commitment with ramp multipliers, IEEE Trans. on Power Systems, vol. 14, n. 1, pp 58-64, Feb 1999. [21] Mart Perez, I.: Wind Forecasting Activities. Proceedings of the First IEA Joint Action Symposium on Wind Forecasting Techniques, Norrkping, Sweden, December 2002, pp. 11-20. Published by FOI - Swedish Defence Research Agency. [22] Mart I., Usaola J., Snchez I., Navarro J., Roldn A., Gonzlez G., Daz-Guerra B. Wind power prediction in complex terrain. LocalPred and SIPRELICO. Proceedings of the 2003 European Wind Energy Conference & Exhibition. Madrid (Spain). June 2003. [23] Nielsen T.S., Madsen H., Tfting J. WPPT: A Tool for On-line Wind Power Prediction. Wind Forecasting Techniques: 33 Meeting of Experts. Technical Report from the Internatinal Energy Agency. R&D Wind. Ed. S.-E. Thor, FFA, Sweden,pp. 93116. 2000. [24] Nowak I., Relaxation and Decomposition Methods for Mixed Integer Nonlinear Programming, Birkhuser Verlag, Switzerland, 2005 [25] OFGEM, 2000. An Overview of the New Electricity Trading Arrangements. Available at http://www.ofgem.gov.uk [26] Padhy N.P., Unit commitment A bibliographical survey, IEEE Trans. on Power Systems, vol. 19, n. 2, pp 1196-1205, May 2004. [27] Pinson, P., Siebert, N., Kariniotakis, G., "Forecasting of Regional Wind Generation by a Dynamic Fuzzy-Neural Networks Based Upscaling Approach.", CD-Rom Proceedings of the European Wind Energy Conference & Exhibition EWEC 2003, Madrid, Spain, June 16-19, 2003.
153
[28] Red Elctrica de Espaa, El Sistema Elctrico Espaol, avance del Informe 2006, available on-line in www.ree.es . [29] Snchez I., Usaola J., Ravelo O., Velasco C., Domnguez J., Lobo M. G. SIPRELICO- a wind power prediction system based on fexible combination of dynamic models. Application to the Spanish power system. Proceedings of the 2002 World Wind Energy Conference & Exhibition. Berlin (Germany), July 2002. [30] Snchez I. "Short-term prediction of wind energy production", International Journal of Forecasting (2006), 22, 43-56. [31] Slootweg J.G., Polinder H., Kling W.L., Representing wind turbine electrical generating system in fundamental frequency simulations, IEEE Trans. on Energy Conversion, vol. 18, n. 4, pp. 516-524, Dec. 2003. [32] Spanish Secretary of Energy, Operational Procedure 3.7, Programming of the NonDispatched Renewable Generation (in Spanish), BOE n. 254, pp. 37020-37022. [33] Spanish Ministry of Energy, RD 661/2007 (in Spanish), BOE n. 126 (2007), pp. 2284622886, available on-line in www.boe.es . [34] Usaola J., Angarita J. Benefits of short term wind power prediction programs for the integration of wind energy in electricity markets. Proceedings of the EWEC06. Athens (Greece), March 2006. [35] Usaola J., Angarita J. Bidding wind energy under uncertainty. Proceedings of the International Conference on Clean Electric Power. Renewable Energy Resources Impact. Capri (Italy). May 2007. [36] Usaola J., Ravelo O., Snchez I., Velasco C., Domnguez J., Lobo M. G., Gonzlez G. SIPRELICO, a wind power prediction toll for the Spanish peninsular power system operation. Proceedings of the 2002 Global Wind Power Conference. Paris (France), April 2002. [37] Wang C., Shahidehpour S.M., Ramp-rate limits in unit commitment and economic dispatch incorporating rotor fatigue effect, IEEE Trans. on Power Systems, vol. 9, n. 3, pp. 1539-1545, Aug. 1994. [38] Wind Power Observation, Spanish Wind Energy Association, available on-line at http://www.aeeolica.org/english/ [39] Wolsey L.A., Integer Programming, Wiley-Interscience Series, John Wiley and Sons, New York, USA, 1998. [40] Zhuang F. and Galiana F.D., Towards a more rigorous and practical unit commitment by Lagrange Relaxation, IEEE Trans. on Power Systems, vol.3, n. 2, pp 763-773, May 1988.
Chapter 4
LEAKAGE CURRENT ON HIGH VOLTAGE CONTAMINATED INSULATORS

P. T. Tsarabaris and C. G. Karagiannopoulos*
National Technical University of Athens, School of Electrical and Computer Engineering, 9 Iroon Polytechniou Str., GR 157-80 Athens, Greece
ABSTRACT
The present chapter will present a short review of research work regarding leakage current. In addition, measurement system of the leakage current is introduced using a high sampling frequency analogue/digital converter. The application of this system in 20 kV insulators contaminated from a compound of salt and kaolin are described. Using this system measurement in a time frame of one period (50 Hz) were done and an investigation of the observed phenomena is attempted with the assistance of i-u characteristic curves plotted for one cycle of voltage application. The fact that existence of partial discharges on the surface of the contaminated insulators, beyond threshold field intensity leads to radiation emission, is also be examined. An estimation of the free electrons energy has been done and the corresponding emitted radiation which seems to include acoustic waves, radio waves, microwaves and infrared waves is also examined. The classification of the leakage current values of a typical porcelain insulator of 20 kV, contaminated by salt and kaolin, is presented. The classification is based on the collaboration between the above high precision data acquisition system having high sampling rate and an unsupervised self-organized neural network. In addition a simulation model for contaminated insulators is presented. The proposed model will be provided, together with a mathematical function that simulates the behavior of the dry band resistance as a function of time, even in cases where arcs or partial discharges occur. The models parameters of a typical porcelain insulator of 20 kV, contaminated by salt and kaolin, are presented.
Email: ckarag@central.ntua.gr
156
P. T. Tsarabaris and C. G. Karagiannopoulos
1. INTRODUCTION
The flashover is probably the major visible undesirable effect on an insulator, which results in power loss. This effect usually appears on insulators that operate under adverse conditions, like industrial areas, near the sea, in a desert or in snow [1-10]. The probability of a flashover to occur on a contaminated insulator depends on the level and nature of the insulator surfaces contamination, and also on the wetting conditions [11]. The flashover is the result of a series of other effects, which occur in the following order: insulator contamination, wetting of contaminants, generation of a thick electrolytic solution, increase of the leakage current, increase of the temperature, formation of a dry band, occurrence of partial discharges, occurrence of arcs, and finally the flashover [1]. Research on this phenomenon has been carried out for many years by scientists all over the world. Test methods have been developed and many papers published. Researchers have been studying the flashover on contaminated insulators, especially from salt contaminants, and most of them have worked on the correlation of the voltage level at which the flashover occurs either with the number of units on the insulator string or with the equivalent pollution density [2, 6, 7, 11-15]. Other researchers have been working on the correlation of the precipitation rate with the equivalent salt deposit density [14]. A large part of the bibliography [3, 8, 9, 14, 16-20] concerns the leakage current. There, has been studied the leakage current of the contaminated insulator in relation to the salinity [8], the leakage current for salt pollution in relation with time[17], as well as the exposure time in a foggy place [14]. There has also been recorded the leakage current in insulators polluted by a combination of salt and desert dust [9], and also in insulators exposed in a coastal area having snow deposited on their surface [16]. Furthermore, a measurement method of the leakage current has been proposed, for the calculation of the rms value of the leakage current [3]. The form of interference in the radio and TV sets in relation with the form of the leakage current under DC [20] and AC [17, 18] voltage has also been examined.
2. LEAKAGE CURRENT MEASUREMENT AND RECORDING SYSTEMS

In order to record and study the behaviour of parameters what are related with the contaminated insulator behaviour, many researchers have used various recording and monitoring systems. The leakage current, speed and direction of wind, relative humidity, environment and insulators surface temperature, the rainfall and finally the conductivity of polluted are the parameters commonly recorded because it is accepted that these parameters has influence on the leakage current behaviour. The presentation of sensors used for the measurement of speed and direction of wind, relative humidity, environment and insulators surface temperature, rainfall and conductivity of polluted surface is believed that doesnt fit the scope of this chapter. For this reason only the common used leakage current sensors would be presented. A leakage current recording system composes by the sensor and the recording unit. The systems used, among other, for the leakage current measurements are given in the following paragraphs.
Leakage Current on High Voltage Contaminated Insulators
157
Habib et al [21] and Khalifa et al [22] have proposes a monitoring system which gives an alarm signal when the number of leakage current pulses exceeding a threshold value, exceeds a predetermined limit. The system uses a toroidal coil, a current pulses comparator, a counter, a pulses number comparator and finally a circuit for the production of the alarm signal. The toroidal coil, which operation is based on Hall effect, is used for the current pulses measurement. The pulse comparator is used for the pulse width comparison concerning a predetermined threshold (which varies from 45 mA until 1 A or even more). The counter measures the number of current pulses (for a time interval 4-30 min) which have exceeded the threshold. Then the comparator compares if the number of current pulses has exceeded the in advance determined number of pulses and if yes it gives command for alarm. Vlastos et al [23] have proposed a system for the current pulses recording. The system uses leakage current measurement sensors, a voltmeter for the measurement of applied voltage, as well as measurement instruments for speed and direction of wind, atmospheric pressure as well as rainfall. The outputs of above instruments are connected in a sampling system of 100 channels. The sampling speed is 30 Hz and each input is sampled once each 10 seconds roughly. The leakage current measurement sensors give as output the biggest price of current for all the time interval of sampling. Marrone et al [24] have proposed a recording system for the quantity of insulator surface pollution. The system is constituted by a surface conductivity recording system. Richards et al [25] have proposed a system of pollution recording in the insulator surface. The system is constituted by a leakage current recording system, as well as sensors for the measurement of relative humidity, the environment temperature, the temperature of insulator surface (radiation sensor), the speed and the direction of wind, the level of condensation and the quantity of rain. The current recording measures the biggest value of leakage current for each half- period of applied alternative voltage. Gilbert et al [26] have proposed a system of current pulses recording. The system uses a very small resistance (5 ) for the leakage current measurement, and an integrating circuit which integrates the current values over a fixed period (10 seconds). The integrated current values (each 10 seconds) are stored in memory. The system checks also the current values in the case, where their value exceeds, for three sequentially time intervals ( of 10 seconds), a predetermined limit, then the system produces a alarm signal for the insulators washing procedure start up. Thalassinakis et al [27] have proposed a leakage current monitoring system. The system is constituted by a leakage current sensor connected with a recorder. The sensor is a high accuracy Hall effect current transducer. The recorder incorporates an analogue-to-digital (A/D) converter and a microprocessor data acquisition system, specially designed to record the leakage current activity on nine different insulators. All captured data are stored in the system and can be retrieved via an RS232 serial port or via an external dial-up modem. The sampling rate of the A/D converter of the instrument is 20 kHz and the whole system accuracy is 0.5% of full scale. The unit is also equipped with 3 voltage inputs in order to record the three phase to ground voltages through appropriate voltage transformers, 160/3 kV to 120/3 V. The recording system is additionally equipped with weather sensors for recording humidity, temperature, wind-speed and rainfall. Williams et al [28] and Sebo et al[29] have proposed a leakage current measurement systems constituted by a measuring resistance and a digital storage oscilloscope. The leakage current was observed on the oscilloscope as voltage drop across the measuring resistance.
158
Sorquist et al [30] have used a digital multimeter connected with a PC for the leakage current measurement. The multimeter is directly connected to the nonenergized size of the insulators. Fernardo et al [31] have proposed a leakage current measurement systems constituted by a shunt (varied from 2 to 4.7 k), a digital storage oscilloscope and a PC. The leakage current were observed and saved on the oscilloscope as voltage drop across the shunt. Tsarabaris et al [3236] have proposed a leakage current recording system. The measurement circuit is provided schematically in figure 1. The voltage elevation transformer is supplied by a Variac, thus resulting in a controlled gradual voltage increase. The transformers secondary coil is connected in parallel with a capacitive voltage divider composed by capacitors C1 and C2 respectively. In parallel with the voltage divider the series connection of the insulator and the measurement resistance Rm is coupled. The voltage across C2 and Rm is measured with the use of an A/D converter. The data transfer from the A/D converter is achieved via an RS 232 connection to a computer where it is recorded and visually displayed with the use of appropriate software. The technical characteristics of the capacitors C1 and C2 were 224 pF, 100kV / 50 Hz and 2 F, 1000V / 50 Hz respectively. The values of the above-mentioned capacitors determine the appropriate voltage division ratio, as the maximum voltage across C2 must not exceed the permitted input voltage of the A/D converter. In this case, this maximum value was 5 V. The resistance Rm, has zero selfinduction and its maximum permitted power is considerably higher than the one imposed by the measured leakage current. The thermal factor a of the resistance was practically zero. Applying high voltage techniques the electromagnetic interference from the high voltage components was eliminated and with a careful screening, low induced voltages caused by stray magnetic fields, were minimised. The whole sampling system was specifically designed and released for measuring fast rising voltage profiles and was capable of detecting voltages as low as 1mV with a typical error 0,1% when the input measured voltage was 1V. The employed 12 bit A/D converters were of successive approximation type with fast conversion rate of 0,020sec (2x20 MSPS). The above system has successfully used to measure leakage current pulses.
Figure 1. Measurement circuit 1: variac, 2: high voltage transformer, C1C2: voltage divider, Rm: measuring resistance.
159
It results that the recording systems proposed among others concern mainly the leakage current, speed and direction of wind, relative humidity, environment and insulators surface temperature, the rainfall and finally the conductivity of polluted. From these recordings useful conclusions can be exported regarding the form and the value of leakage current and the effect of factors such as the speed and the direction of wind, the relative humidity, the environment temperature, the rainfall and the conductivity of surface in the leakage current. Also conclusions can be exported regarding the effect of speed and the direction of wind, relative humidity, temperature of environment, and rainfall in the conductivity of surface. These conclusions can help on one side in the planning and the manufacture of the high voltage insulators and on the other side in the better understanding of mechanisms of pollution deposition, the leakage current behaviour and flashover.
3. LEAKAGE CURRENT BEHAVIOUR

In the extensive bibliography for the insulator behaviour, a number of researchers [4-6] for the case of porcelain or glass insulators have studied in detail the form of leakage current. Woodson and McElroy [4] have used photographic camera for the study of arcs on surface of polluted porcelain insulator. Their results showed that in the surface arcs are presented having radial direction. The duration of these arcs depends considerably from the amount of humidity on the pollution layer. For humidity density in the surface equal to 3 mg / cm 2 the time duration of arcs is varied between 0.5 and 1 seconds while for density equal to 16 mg / cm 2 time duration of arcs is varied between 2 and 6 seconds. The intensity of above arcs is of the order of tens or hundreds mA. Lambeth et al[6], studying the current behaviour during salt fog tests, found that the current waveform is constituted by regions where the current has very low value (almost zero) and from parts with sinusoidal form, which are reported as current pulses, the duration and the width of which depend considerably from the density of salinity contained in the pollution layer. The time duration of sinusoidal form parts is varied from 5 ms to 7 ms for salinity density variation from 5 kg / m 3 to 40 kg / m 3 . The corresponding width variation is roughly between 112.5 mA and 300 mA . The current study at the phase of ice fusion on ice covered insulators, showed that the current presents pulses of width of the order of 10 mA and duration roughly 50 seconds as well as pulses having width of the order of 450 mA and duration of roughly one min. The vibrations that had width of the order of 10 mA were attributed corona streamers while the pulses with considerably bigger width in the white arc occurrence [5] between a number of disks. According to bibliography [8] the leakage current during one period of time (50 Hz) is initially practically zero and retains a prominent value only close to the maximum of the applied voltage. It is also documented that its waveform always consists of pulses appearing randomly, without discrimination among the pulses [19]. It has been reported [31] that the leakage current may have sinusoidal form as well as small high-f spikes and long duration discharges. At the application of the system in the leakage current measurement on porcelain insulators of 20 kV, Tsarabaris et al [32] have observed that the leakage current waveform is
160
constituted of linear regions as well as of small and high amplitude pulses. By observing the waveforms of the leakage current in relation with the applied voltage (figures 2a and 3a) as well as the corresponding i-u characteristic curves (figures 2c and 3b), we can distinguish that the leakage current is zero for values of the applied voltage near zero (figure 2b point A). This is expected, as the potential difference at the edges of the insulator is too low and thus the corresponding field is also low.
Figure 2. a,b) Waveforms of leakage current and applied voltage in relation to time. c) i-u characteristic curves which have been obtained after the elimination of time from the previous measurements.
161
Figure 3. a) Waveforms of leakage current and applied voltage in relation to time. b) i-u characteristic curves which have been obtained after the elimination of time from the previous measurements.
The current rises linearly with the increase of the applied voltage up to a point B. This linearity is obvious in the linear part of the i-u characteristic curves and is represented by symbol I in figure 2b. This implies that in the A-B area conductivity probably appears at the surface of the insulator [21]. The conductivity can be probably attributed to the low conductivity of the solid insulating material of the insulator (porcelain), and also to the conductivity of the deposits on the insulators surface. The deposits conductivity can be attributed to the conductivity of the dense electrolytic solution (such as NaCl), which is formed among the substances of the insulators surface pollution and the water concentrated on its surface. The ions of the diluted substance obtain higher mobility under the influence of the potent electric field, causing the appearance of the conductivity[21]. There is a point in the leakage current characteristic, where the current increases instantaneously, obtaining the shape of high amplitude - negligible width, pulses. The pulses vary in number and amplitude among different periods (50 Hz). They dont appear periodically. The leakage current pulses which appear randomly, are assumed to be caused by partial discharges on the insulators surface, as well as by arcs on the insulators surface among different disks or across dry bands. This is rational because it is documented [22-25] that the occurrence of partial discharges is followed by quick current fluctuations of the order of a few tens of secs. Therefore, a dense spectrum of current pulses may imply that on the insulator surface there was a high activity of partial discharges and arcs. Looking at the waveforms of the leakage current on fig. 2a and 3a, we can see that the current pulses can be split into three categories. In the first category belong pulses with amplitude up to 10 mA and
162
with negligible width of the order of few sec. In the second category belong pulses with amplitude over 10 mA and with negligible width as before. In the third category we can find pulses with amplitude over 10 mA and with significant width usually of the order of a few thousands of sec. First category pulses are probably caused by the partial discharges, because their amplitude is small, in contradiction with the pulses of the other two categories, whose amplitude is quite higher and which are probably caused by arcs. Referring to the second and third category pulses, those which have the less width are caused by momentary arcs, whereas those which have significant width are caused by long duration arcs. Regarding the pulses caused by continuous arcs, they start close to the maximum value of the applied voltage. Thereafter their amplitude rises linearly up to a maximum value of approximately 40 mA. After that their amplitude decreases linearly with the applied voltage and becomes zero when the applied voltage is almost zero. This implies that on the insulators surface long duration arcs existed which die out when the applied voltage reaches a rather low value. The third part of the leakage current half period waveform (area III of figure 2b) is similar to the first, i.e. leakage current responds almost linearly to the variation of voltage, due to the existence of conductivity on the insulators surface.
4. RADIATION EMISSION
It has been shown that ageing and penetration of solid insulators during their electrical strain under high voltage are explained by methods that can be classified into two theories [37, 38]: macroscopic theory and quantum mechanical theory
According to the macroscopic theory, the efforts of research have been aimed at an equivalent electrical circuit based on the combination of electrical measurements with optical observations or changes in the surface of the material caused by ageing. From measurements taken, it was determined that there are 4 basic factors which contribute to the reduction in the molecular stability of the insulating material and to the change of its molecular structure. These factors are [37]: Joule losses Partial discharges Coulomb forces Ambient temperature
The most important factor of ageing is considered to be partial discharges. By the term partial discharge we mean each local discharge of restricted length in the solid insulator. The lack of homogeneity in the volume of the solid insulators constitutes the centre of partial discharges. Furthermore, lack of homogeneity and contaminants on the insulator surface also constitute centers for the initiation of partial discharges.
163
(a)
(b)
Figure 4. Typical waveform of discharge current and the corresponding voltage applied in relation to time and the corresponding i-u characteristic curves obtained after the elimination of time from the previous measurements. a. (temperature = 20 oC, pollution = 0.1 mg/cm2, and relative humidity = 80%). b. (temperature = 20 oC, completely clean, and relative humidity = 80%).
The quantum mechanical theory principally concerns the energy of free electrical carriers (electrons) which are produced during the ageing of the material. The lack of homogeneity within the solid insulator induces the appearance of disturbance bands between conductivity and strength bands, which, like intermediate energy levels, facilitate the transmission of electrons to the conductivity band [37, 38]. A determining factor in the creation of free electrical carriers is the value of the imposed field [37]. Above a threshold value of the electric field the electrons acquire increased kinetic energy, which they emit as electromagnetic radiation when they are absorbed at the anode [38, 39]. The spectrum of the radiation during insulator ageing ranges from acoustic waves to ultra-violet radiation. [3739]. During measurements acquired for an applied voltage of 20 / 3 kV in porcelain post insulator samples, having a leakage distance of 540 mm, and contamination (with a compound of salt and kaolin) approximately 0.1 mg/cm2, Tsarabaris et al [32] have found that one region of the leakage current waveform (around the point where the applied voltage obtains its maximum value), where rapid current fluctuations appear. They presumed that these fluctuations are generated by non-self-supported discharges [40], where the required electrical field for the production of free electrical carriers is provided by the field that has evolved across a dry band. In figure 4a someone can also notice that the current jump (starting of instabilities) takes place at a distinct threshold value of the applied voltage. This phenomenon has been observed in solid dielectrics and has extensively analyzed in the bibliography [40]. Specifically, it is widely accepted that among the quantities of primary interest, concerning electronic transport in solids, are the carrier mobilities, the majority carrier sign,
164
the carrier deep trapping lifetime, the efficiency of photogeneration (the number of free carriers produced per absorbed photon), the surface and bulk recombination lifetime and the bulk conductivity [41]. In disordered insulating materials, these properties are generally strong functions of variables such as the applied electric field, the temperature and in some case the overall geometry. The current-voltage characteristics of dielectrics stressed by high electric fields is often characterised by current instabilities which may appear beyond a threshold field intensity [42]. A widely used physical model describing charge transport in dielectrics is based upon the presence of deep electronic states in their energy gap. It is believed that electron traps are induced in the band-gap of these materials as a result of various kinds of defects, e.g. chain folding and branching, carbonyl groups, cavities and entanglements, unsaturated bonds and molecular oxygen [43]. The current flow (in the low field region) has mainly been associated with electron hopping conduction and quantum mechanical tunneling between adjacent electron sites. For higher applied fields, electrons can gain enough energy to be excited towards wide energy bands allowing kinetic energies of the order of l0 eV, in which case impact ionization phenomena may become of primary importance [37, 44]. Theoretical studies, based on quantum mechanics, prove that for an electron hopping conduction process in a narrow impurity band [45], or for a low-density electron gas in random potential [46], the electric field will induce composite traps leading to exponential conductivity decay and N-type negative differential conductance (NDC). For even higher field values, the free electron concentration becomes larger than the trap density and electron-electron interaction increases their escape probability, changing N-type NDC to superlinear or even S-type NDC. Although the above theories were initially developed to explain ionization phenomena in high field domains of III-V semiconductors [47], they also appear to be in remarkable agreement with the data obtained experimentally so far [40] of a large area mechanically contacted MIM structure, when excited by high voltages. It seems that the dry solid contamination on the insulators surface (dry band), behaves as a solid dielectric between two electrodes (conducting wet areas on the insulators surface). This assumption is verified by the fact that non-linear i-u characteristics are not presented in figure 4b, which represent completely clean insulators. The current-voltage loop observed in Figure 4b can be attributed to the dry band capacitance. Besides, the hysteresis effect is frequently noticed on the i-u characteristics obtained to electrode gaps bridged by dielectrics or vacuum and it has been attributed to the capacitive nature of the structure or even to the production of photoelectrons by the electrodes or the insulator's surface [48]. According to the previously mentioned model, during the free electrical carriers incidence onto the positive electrode, their kinetic energy is given in the form of radiation. This energy is given by the following expression [40]:
W=
where
u 2
(eV)
(1)
u = i R
(2)
165
Figure 5. Emitted radiation energy in the case of figures 4a non-self-supporting discharges.
By substituting the term u into equation (1) we obtain the following expression:
W =R
i 2
(eV)
(3)
The emitted radiation energy, in the case of figure 4as non-self-supporting discharges, is given in the histogram in figure 5. From the calculated emitted radiation energy values, it results that the range during the non-self-supported discharges mentioned above includes acoustic waves, radio waves, microwaves and infrared waves. This range is in accordance with the range of the radiation emitted in the case of partial discharges. Consequently, the fluctuations that appear in figure 4as current waveform are probably due to partial discharges. They could probably take place on the insulators surface and/or in the bulk of the insulators material. But by observing figure 4bs current waveform, we can see that no fluctuations appear. This reasonably implies that partial discharges do not take place in the bulk of the insulators material.
5. CLASSIFICATION OF LEAKAGE CURRENT USING NEURAL NETWORKS

The application of neural networks in the field of the insulators has been reported a few years ago. Various systems have been developed for the forecasting of the flashover [49-51], for the analysis of the surface tracking on solid insulators [52], for the identification of faulty insulators using corona discharge analysis [53], for the modeling of partial discharges on epoxy resin post insulators [54], for the inspection of the insulators on a string based on the texture analysis of the images of the strings units [55],for the modeling of partial discharge inception and extinction voltages of sheet samples of solid insulating materials [56], for the
166
ESDD modeling[57] and for the forecasting of number and the place of faults due to insulators contamination in a distribution network [58]. Fernardo et al [31] have used neural networks for the evaluation of the harmonic contents (3rd and 5th harmonics) of leakage current waveforms of contaminated polymeric surfaces. The neural networks which have been used are based on the following training algorithms: back - propagation [ 31, 50-51, 54-58 ] and learning vector quantization [ 31]. From these algorithms, back - propagation requires supervision while learning vector quantization doesnt [59]. Regarding leakage current Fernardo et al [31] have used neural networks for the classification of the leakage current half-period waveforms on contaminated polymeric surfaces. Tsarabaris et al [35] have used neural networks for the classification of the leakage current pulses on contaminated insulators. The proposed neural network (NN) is based on the known from the bibliography as self-organized feature mapping which has been developed by Kohonen[59, 60]. The neural network consists of the input layer and the output layer. The input layer includes a single neuron. The output layer is one-dimensional and consists of a variable number of neurons. The single neuron in the input layer corresponds to anyone of the input patterns. Each of the output layer neurons corresponds to each class. The output layer neurons number is variable and is defined by the user each time the training procedure is executed. The Euclidean norm between the input vector and the weight vector of the corresponding neuron is used as the criterion for the selection of the winning neuron. The normalized values of momentary leakage current values have been considered to be the input patterns of the neural network. The normalization is carried out using the highest value among all samples of leakage current that are held for the training. A class is defined as a fixed region of normalized values of leakage current momentary values which has clearly defined limits and which doesnt overlap another class. Since a normalized value of a leakage current momentary value corresponds to a single momentary value of the leakage current, the input patterns and the classes in which they are classified come from and univocally correspond to momentary values of the leakage current. Initially, a large number of measurements of leakage current momentary values were acquired using the system previously described [32-34]. The contamination of the insulators surface was in the range of 0.10.14 mg/cm2 [61]. Those measurements, approximately 1.000.000 leakage current momentary values, were stored in file. These measurements correspond to approximately 100 periods of the leakage current waveform. Then the normalization of all samples of that file took place, and the normalized samples were stored in a new file. The neural network was trained using as input patterns the normalized patterns of the momentary leakage current values. A summary of all the training parameters of the network is given in Table 1. It is required for the network training that the number of output neurons is defined. This number expresses the different categories of the leakage current momentary values. By observing the waveforms of the leakage current, drawn from measurements, someone can see that no more than five categories are needed, whereas less than three would be insufficient. This admission will be checked below. For that reason the NN was trained separately for three, four and five output neurons. The rest of the parameters remained unchanged, and are given in Table 1. After the completion of each training process, the
167
convergence weights that resulted were stored in a file. The above procedure was conducted in order to choose the appropriate number of output neurons. It decided [35] to adopt the three output neurons approach. The limits that determine each class, actually determine which current values belong to each class. In this way, a categorization of the levels of the leakage current values is achieved. The case where a large number of classes derived would imply significant fluctuations in the current values. On the other hand, the existence of only one class would imply absence of such fluctuations. The leakage current waveforms seem to present regions with negligible, as well as with intense current fluctuations. The objective is to make a clear distinction between the current values that are provocative or non-provocative of fluctuations and if they are, to make a distinction among the various levels of fluctuation. So, the right determination of the limits of the classes is particularly important. In order to achieve our objective, the contribution of the neural network is considered of great importance. The proposed neural network is able to determine the precise limits of the classes and to do this dynamically. This means that if in the future we wish to enlarge the training patterns base, then the new training of the network would result in the recalculation of the classes limits. From the analysis that was achieved with the use of the NN, it seems that there are three classes in which the leakage current values can be categorized, which correspond to three regions of values. Comparing the results with optical observations and observing one typical waveform of the leakage current (like the waveform in Figure 3a), we can see the existence of regions where the current presents very small fluctuations, as well as regions with significantly large fluctuations. In order to investigate this subject, we plotted Figure 6, in which typical forms of pulses are presented, as they resulted from the measurements that were performed. In this figure we can distinguish pulses of small and large amplitudes, as well as parts of the waveform where no fluctuations are present. Based on the above, it is possible to correlate the classes (in which the leakage current is classified), with the pulses that are present in the leakage current waveform. Based on the current value when a pulse occurs, we see that it is possible to discriminate the pulses into those of small and those of large amplitude. The classification results that class 2 represents pulses of small amplitude. Similarly, class 1 represents pulses of large amplitude. Class 3, represents the linear region of leakage current waveform (Figure 3) where no fluctuations appear. Table 1. The neural networks training parameters
Number of neurons in the output layer Initial neighborhood Minimum neighborhoods Radius Initial weights Initial learning rate Amount to change learning rate Minimum learning rate wlimit Maximum number of epochs Variable All output layers neurons 0 Between 0 and 1 0.9 0.005 0.005 0.001 2000
168
Figure 6.Typical forms of pulses resulting from the measurements.
Pulses having high amplitude can be separated into two subcategories, one with short width and one with long width pulses. This neural network classifies both pulses in the same class. That was expected, as the classification was based on the amplitude of the leakage current with the same range of variation. Short and significant width pulses, are probably caused by momentary and long duration arcs, respectively, due to the large current value. Therefore, in order to discriminate clearly the pulses according to their width, other additional parameters of the leakage current waveform must be sought for, which characterize time duration. This can be a topic for further investigation. This NN in cooperation with the measurement system previously described (figure 1) can be used in the approximate calculation of the number of arcs and partial discharges that occur on the insulators surface for a specific time interval. This can be achieved by calculating the mean value of the number of measured samples that are acquired throughout the duration of an arc or partial discharge, and from that the mean time of arc and partial discharge duration, since we know the sampling rate. From this calculation one can make assumptions about the insulators deterioration during the presence of arcs and partial discharges, since in the area, in which an arc is created, erosion on the insulators surface is observed [1] (due to the heat release). Moreover the system can be used in the calculation of power loss on each insulator, since both the momentary value of the leakage current and the momentary value of the applied voltage is known. The power loss can be calculated both for cases of individual arcs and partial discharges, and also during a specific time period of leakage current. The system can also be used as alarm device for the washing of insulators, which are in service in substations. This can be done either by only checking the existence of arcs, or by counting the number of arcs (exceeding a limit) for a given time period. Finally, the system can be applied for the study of leakage current in the case of insulators manufactured from
169
other material (e.g. composite), where the phenomena probably differ, as well as for insulators which are in service in different high voltage networks (66 kV, 150 kV, 400 kV) placed in different environments, where the level and the kind of pollution varies.
6. MONTELLING
In the bibliography [62-65], regarding the study of pollution flashover phenomenon measurements have became and various models for the insulator surface behaviour have been proposed. A review of the main mathematical models for pollution flashover has been done by Rizk [62]. As results from this review the polluted surface can be simulating from an arc in line with a resistance which represents the resistance of its not short circuit part of the pollution layer. For the study of longrod insulator behaviour a model of composed layer of pollution (two layers of pollution in line where one corresponds in the pollution of the cylindrical form shanks and the other in the pollution of disks) was used. During the study of the resistance of pollution layer it was considered that in a small region round the points of inception and extinguishment of the arc the resistance of pollution layer is different from that of rest of layer. It has been proposed a new relation for the calculation of mean resistance of pollution layer per unit of creep length which takes into consideration the occurrence of multiple arcs. Improvement factors of the pollution layer conductivity have also been proposed. Farzaneh et all [63] and Farzaneh et al [64] using models, based on that of Obenaus [62], for the simulation of surfaces covered by ice and for the arc that is propagated on them, they proposed a model for the propagation of discharges on the insulator surface covered by ice under DC voltage. In their study they used triangular samples of ice. This model allows among others the resistance determination of the non short circuit part of ice. Farzaneh et al [65], used a model, similar with that of Obenaus, for the simulation under AC voltage of surfaces that is ice covered.
Figure 7. Equivalent circuit of a polluted insulator: a) Before the formation of dry bands. b) After the formation of a dry band. c) The proposed model.
170
Vosloo and Holtzhausen [66] recently proposed a model of an equivalent circuit of a contaminated insulator, which is based on daily measurements of the leakage current. Specifically, the leakage current waveform at which the daily maximum value occurs is stored. A theoretical model for flashover voltage forecasting has also been proposed [67], which takes into consideration the re-ignition characteristics of the dry band arcs. In the above models, the equivalent circuit is mainly composed of the resistance of the conductive layer on the contaminated insulators surface, which appears to be partly short-circuited along an area, implying an arc occurrence [66-68]. Tsarabaris et al [36] have proposed a simulation model for contaminated insulators. Experimentally investigating the behaviour of an insulator, they have seen that, when the insulator is clean and has no pollution on its surface, it behaves as a capacitor. As pollution is deposited on its surface, without humidity, it still behaves almost as a capacitor. When, the humidity level of the pollution on its surface rises because of, for example, hoarfrost, mist, drizzles, and frost, then an electrolytic solution is formed. This results in the development of a leakage current. Consequently, at that time the insulator behaves as a capacitor in parallel with a resistance (Figure 7a). The leakage current through the conductive layer on the insulators surface results in an increase of temperature near the path of the leakage current. This causes evaporation and therefore the formation of dry bands. In this case, the equivalent circuit shown in Figure 7b simulates the insulator. In this circuit C Db , R Db are the insulators dielectric capacity and surface resistance, respectively, in the dry band region, while C1 , C 2 and R1 , R 2 are the insulators dielectric capacities and surface resistances, respectively, in the leakage regions on either side of the dry band region. Dry bands tend to retain the leakage current and at the same time undergo stress from the high voltage drop along them. It has been suggested that the voltage, which a dry band is stressed with, is stabilized at a slightly lower value than the threshold required for discharge development in air. If this unstable equilibrium is perturbed then a discharge will appear along the length of the dry band and the current pulse intensity will be reduced by the resistance of the residue polluted surface. The extension of arcs along the length of the dry bands depends on many factors, mainly the non-uniformity of the surface pollution, the deposit's resistivity and the peak leakage current value [3,7,69-71]. The development of an arc is simulated by a small value parallel resistance R(t), valid for the period of the arcs duration. This resistance practically short-circuits the dry band resistance. Considering that not all the dry band area is short-circuited, although a considerable part is, the resistance R(t) for the period of the arcs duration must be the resistance of the non-short-circuited section of the dry band. For the time when no arcs occur, R(t) must be large, so that its effect in the equivalent circuit will be practically negligible. From the above, the equivalent circuit of the insulator is as shown in Figure 7c, where Ceq and R eq are the capacitance and resistance of the insulator not involving the dry band and C Db and R Db are the capacitance and resistance of the dry band region, respectively. On the insulators surface and specifically on the dry band not subject to arcing, partial discharges also occur. The effect of partial discharges on the equivalent circuit of the insulator is also represented by R(t), which short-circuits for this occasion a rather small part of the dry band resistance. Because not all dry band is short-circuited, but only a part, R(t) at the time when partial discharges occur represents the non-short-circuited part of the dry band.
171
Resistance R(t) that short-circuits a small or considerable part of the dry band, due to partial discharges or arcs respectively, is a non-linear time dependent resistance. For R(t) the following formula has been proposed [36]:
2 2 R ( t ) = K1 - [K1 - (rarc + r1)] e a1*(t - t1) - [K1 - (rpd + r 2)] e a 2*(t - t 2 )
(4)
where K1 is a constant, usually of the order of 1012 , r1 and r2 are the resistances of the nonshort-circuited part of the dry band in the case of an arc or a partial discharge respectively, and rpd , rarc are the resistances of the partial discharges and arcs, respectively, when one of them is in progress. The constants a1 and a2 determine the time duration of the pulses, and have values of the order of 1014. Time parameters t1 and t2 correspond to the initial time when partial discharges and arcs occur, respectively. The models evaluation, using representative samples of 20 kV porcelain insulators having contamination on their surface in the range of 0.10.14 mg/cm2 , results that resistance R eq was found to be of the order of 0.5 M. Resistance R Db found to be of the order of 5 M. Resistance r1 found to be of the order of 50 k. Resistance r2 found to be of the order of 1.6 . The resistances rpd , rarc defined to be of the order of 10 [72]. The total capacity of the insulator, Csum , derived from the bibliography [12] and from the measurements performed using a Schering Bridge, is of the order of 10 pF. The capacity C Db is of the order of 100 pF. and Ceq in the order of 11 pF. This model satisfactorily simulates 20 kV porcelain insulators subjected to partial discharges and arcs. To further improve the model, a study could be undertaken regarding the time partial discharges and arcs occur on the dry band. In addition, the behaviour of the proposed model could be studied for insulators installed in 150 kV and 450 kV networks.
REFERENCES
[1] [2] Looms J.S.T., Insulators for high voltages (IEE Power Engineering Series 7, London 1990). Kawai M. and Milone D. M.: Tests on salt-contaminated insulators in artificial and natural wet conditions, IEEE Trans. Power Apparatus and Systems, vol. PAS-88, 1969, 1394-1399. Mcelroy A. J., Lyon W. J., Phelps J. D. M., Woodson H. H.: Insulators with contaminated surfaces, part I: Field conditions and their laboratory simulation, IEEE Trans. Power Apparatus and Systems, vol. PAS-89, no 8, 1970, 1848-1858. Woodson H. H. and McElroy A. J.: "Insulators with contaminated surfaces, pt. II: modeling of discharge mechanisms," IEEE Trans. PAS-89,No.8, pp. 1858-1867.,1970 Hara M. and Luan Phan C.: Leakage current and flashover performance of iced insulators, IEEE Trans. Power Apparatus and Systems, vol. PAS-98, No.3, pp.849859, May/June 1979.
[3]
[4] [5]
172 [6] [7]
P. T. Tsarabaris and C. G. Karagiannopoulos Kawai M. and Milone D. M., Flashover test at project UHV on salt-contaminated insulators, IEEE Trans. Power Apparatus and Systems, vol. PAS-89, 1970, 755-760. Fujitaka S., Kawamura T., Tsurumi S., Kondo H., Seta T., and Yamamoto M.: Japanese method of artificial pollution test on insulators, IEEE Trans. Power Apparatus and Systems, vol. PAS-87, 1968, 729-735. Lambeth P. J., Looms J. S. T., Sforzini M., Cortina R., Porcheron Y. and Claverie P.: The salt fog test and its use in insulator selection, IEEE Trans. Power Apparatus and Systems, vol. PAS-92, No.6, pp.1876-1887, November/December 1973. Akbar M. and Zedan F.: Performance of HV transmission line insulators in desert conditions, IEEE Trans. Power Delivery, vol. 6, 1991, 429-438. Kawai M.: AC Flashover tests at project UHV on ice-coated insulators, IEEE Trans. Power Apparatus and Systems, vol. PAS-89, 1970, 1800-1804. Kawai M.: Research at project UHV on the performance of contaminated insulators, part I, IEEE T72 552-8, 1972. Kawai M.: Flashover test at project UHV on salt-contaminated insulators,part II, IEEE Trans. Power Apparatus and Systems, vol. PAS-89, 1970, 1791-1799. Kawai M.: Research at project UHV on the performance of contaminated insulators, part II, IEEE T72 553-6, 1972 Fujimura T., Naito K. and Irie T.: Performance of semiconducting glaze insulators under adverse conditions, IEEE Trans. Power Apparatus and Systems, vol. PAS-97, 1978, 763-771. Kimoto I., Fujimura T. and Naito K.: Performance of insulators for DC transmission line under polluted conditions, IEEE T72 556-9, 1972. Takasu K., Shindo T. and Arai N.: Natural contamination test of insulators with DC voltage energization at inland areas, IEEE Trans. Power Delivery, vol. 3, 1988, 18471853. Macchiaroli B. and Turner F. J.: A new contamination test method, IEEE Trans. Power Apparatus and Systems, vol. PAS-88, 1969, 1400-1411. Bernardelli P. D., Cortina R. and Sforzini M.: Laboratory investigation on the radio interference performance of insulators in different ambient conditions, IEEE T72 1936, 1972. Sawada Y., Fukushima M., Yasui M., Kimoto I. and Naito K.: A laboratory studu on RI, TVI and AN of insulator strings under contaminated condition, IEEE Trans. Power Apparatus and Systems, vol. PAS-93, No 2, 1974, pp. 712 - 719. Fukushima M., Sanaga Y., Sasano T. and Sawada Y.: AN, RI and TVI from single unit flashover of HVDC suspension insulator strings, IEEE Trans. Power Apparatus and Systems, vol. PAS-96, 1977, 1233-1241. Habib S. E. D. and Khalifa M.: "A New Monitor for Pollution on Power Line Insulators - Part 1: Design, Construction and Preliminary- Tests", IEE Proc. Gen. Trans, and Distrib., Vol. 133, Pt. C, No. 1, 1986, pp. 105-108. Khalifa M., El-Morshedy A., Gouda O. E. and Habib S. E.-D.: "A New Monitor for Pollution on Power Line Insulators Part 2; Simulated Field Tests", IEE Proc. Gen. Trans, and Distrib., Vol. 135, Pt. C, No.1,1988, pp. 24-34. Vlastos A. and Orbeck T.: Outdoor leakage current monitoring of silicone composite insulators in coastal service conditions, IEEE Transactions on Power Delivery, Vol. 11, No.2, 1996, pp.1066 -1070.
[8]
[9] [10] [11] [12] [13] [14]
[15] [16]
[17] [18]
[19]
[20]
[21]
[22]
[23]
Leakage Current on High Voltage Contaminated Insulators [24] [25]
173
[26] [27]
[28]
[29] [30] [31]
[32]
[33]
[34]
[35]
[36]
[37] [38]
Marrone G and Marinoni F.: New apparatus set at ENEL to monitor pollution deposit and pilot cleaning operations on outdoor insulators, Cigre, Paper 33-302, 1996. Richards C. N. and Renowden J. D.: Development of a remote insulator contamination monitoring system, IEEE Transactions on Power Delivery, Vol. 12, No.1, 1997, pp. 389 -397. Gilbert R., Gillespie A., Eklund A., Eriksson J., Hartings R. and Jacobson B.: "Pollution flashovers on wall bushings at a coastal site", Cigre, Paper 15-102, 1998. Thalassinakis E. and Karagiannopoulos C.G.: Measurements and interpretations concerning leakage currents on polluted high voltage insulators, Meas.Sci.Technol, Vol.14, 2003, pp.421-426. Williams D. L., Haddad A., Rowlands A. R., Young H. M. and Waters R. T.: Formation and Characterization of Dry Bands in Clean Fog on Polluted Insulators, IEEE Trans.on DEI, Vol. 6, No. 5, 1999, pp. 724-731. Sebo S. A. and Zhao T.: Utilization of Fog Chambers for Non-ceramic Outdoor Insulator Evaluation, IEEE Trans.on DEI ,Vol .6 No. 5, 1999, pp. 676-686. Srqvist T. and Gubanski S. M.: Leakage Current and Flashover of Field-aged Polymeric Insulators, IEEE Trans.on DEI,Vol. 6 No. 5, 1999, pp. 744-753. Fernando M. A. R. M. and Gubanski S. M.: Leakage Currents Patterns on Contaminated Polymeric Surfaces, IEEE Trans.on DEI ,Vol .6 No.5,1999, pp. 688694. Tsarabaris P.T., Karagiannopoulos C.G., Bourkas P.D., Theodorou N.J.: An experimental investigation of leakage current on high voltage contaminated insulators, Iranian J. Electrical and Computer Engineering, Vol. 2, No. 1, 2003, pp. 30-34. Tsarabaris P.T., Karagiannopoulos C.G., Polikrati A.D., Paisios M.P., Theodorou N.J.: An investigation regarding leakage current on hight voltage non-ceramic insulators, 3d Int. Conf. Power and Energy Systems (EuroPES 2003), Marbella, Spain, 3-5 September 2003, pp. 717-720. Tsarabaris P.T., Halaris P.G., Polykrati A.D., Karagiannopoulos C.G., Bourkas P.D.: Radiation Emission Phenomena in 20 kV High Voltage Porcelain Insulators, 4th Int. Conf. Power and Energy Systems (EuroPES 2004), Rhodes, Greece, 28-30 June 2004, p.p.512-515. Tsarabaris P.T., Karagiannopoulos C.G., Bourkas P.D., Theodorou N.J.: A classification of the leakage current pulses of high voltage contaminated porcelain insulators using a self-organised neural network, Int. J. Engineering Intelligent Systems, Vol. 14, No. 1, 2006. Tsarabaris P.T., Karagiannopoulos C.G. and Theodorou N.J.: A model for high voltage polluted insulators suffering arcs and partial discharges, Journal of Simulation Modeling Practice and Theory, Vol.13, No.2, 2005, p.p.157-167. Zeller H. R.: Breakdown and pre-breakdown phenomena in solid dielectrics, IEEE Trans. Electrical Insulation, vol. EI-22, 1987, pp.115-122. Bourkas P. D., Kayafas E. A., Dervos C. and Stathopulos I. A.:Eine mogliche Erklarung der Glimmentladungen in festen Isolirstoffen bei Stosspannungsbeanspruchung , Etz- Archiv Bd, 11, H.5, 1989, s. 163-165.
174 [39]
P. T. Tsarabaris and C. G. Karagiannopoulos Bourkas P. D., Kayafas E. A., Dervos C. and Stathopulos I. A.:Enhanced partial discharges due to temperature increase in the combined system of a solid-liquid dielectric, IEEE Trans. Electrical Insulation, vol. EI-25, 1989, pp.469-474. Bourkas P.D., High voltage applications, NTUA Publication, Athens, 1996. Mort J. and Pfister G., Editors,: Electronic Properties of Polymers, Wiley/Interscience, New York, 1982. Dervos C., Bourkas P.D. and Kayafas E.A.: High-frequency current oscillations in solid dielectrics, J. Phys. D: Appl. Phys. Vol. 22, No.2, 1989, pp.316-322. Lowell J. and Rose-Innes A. C.: Contact electrification, Advances in Phys. Vol. 29, No.6, 1980, pp.947-1023. Sparks M., Mills D.L., Warren R., Holstein T., Maradudin AA., Sham LJ., Loh E. and King D.F.: Theory of electron-avalanche breakdown in solids, Phys. Rev. B., Vol.24 , No.6, 1981, pp.3519- 3536. Nguyen Van Lien and Shklovskii B.I.: Hopping conduction in strong electric fields and direct percolation, Solid State Commun., Vol.38,1981, pp. 99-102. ELL Levin and Shklovskii B.L.: Negative differential conductivity of a low density electron gas in random potential, Solid State Commun.,Vol. 67, No. 3, 1988, pp. 233237. Gelmont B.L and Shur M.S.: S-type current-voltage characteristic in Gunn diodes, J. Phys.TfcAppL Phys.,V, 1973, pp. 842-850. Jaitly N.C. and Sudarshan T.S.: X-ray emission and pre-breakdown currents in plain and dielectric bridged vacuum gaps under DC excitation, IEEE Trans. Electr. Insul, Vol.23, No.2,1988, pp. 231-242. Cline P., Lannes W. and Richards G.: Use of pollution monitors with a neural network to predict insulator Flashover, Electric Power Systems Research Vol. 42, 1997, pp. 27-33. Ghosh P. S., Chakravorti S. and Chatterjee N.: Estimation of Time to Flashover Characteristics of Contaminated Electrolytic Surfaces using a Neural Network, IEEE Trans.on DEI, Vol.2, No.6, 1995, pp. 1064-1074. Lavalle M. M. and Ortiz G. R.: Flashover Forecasting on HV Insulators with a Back Propagation Neural Net, Can. J. Elec. And Comp. Eng., Vol.21 No.1, 1996, pp. 29-32. Ugur M., Auckland W., Varlow B. R.and Emin Z.: Neural Networks to Analyze Surface Tracking on Solid Insulators, IEEE Trans.on DEI, Vol.4, No.6, 1997, pp. 763766. Li C. R., Shi Q., Cheng Y. C., Yu C. and Yichao Y.: Identification of Faulty Insulators by Using Corona Discharge Analysis Based on Artificial Neural Network, Proc. Int. Symp. on EI, Arlington, USA, 1998, pp. 382-385. Ghosh S. and Kishore N. K.: Modeling PD Inception Voltage of Epoxy Resin Post Insulators using an Adaptive Neural Network, IEEE Trans.on DEI,Vol.6, No.1, 1999, pp. 131-134. Hamed M. and El Desouky A.: A computerized inspection for the high voltage insulating surfaces, Electric Power Systems Research, Vol.53, 2000, pp. 91-95. Ghosh S. and Kishore N. K.: Modeling partial discharge inception and extinction voltages of sheet samples of solid insulating materials using an artificial Neural Network, IEE Proc.-Sci. Meas. Technol., Vol.149, No.2, 2002, pp. 73-78.
[40] [41] [42] [43] [44]
[45] [46]
[47] [48]
[49]
[50]
[51] [52]
[53]
[54]
[55] [56]
Leakage Current on High Voltage Contaminated Insulators [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67]
175
[68]
[69] [70]
[71] [72]
Arabani M.P. and Shirani A.S.: An artificial neural network modeling of esdd, CIGRE Paper 33-101, 1994. Tsanakas A.D., Papaefthimiou G.I. and Agoris D.P.: " Pollution flashover fault analysis and forecasting using neural networks", Cigre, 15-105, 2002. Haykin S.: Neural Networks- A Comprehensive Foundation, Prentice-Hall, New Jersey, 1994. Pandya A. S. and Macy R. B.: Pattern Recognition with Neural Networks in C++, CRC Press / IEEE Press, Boca Raton, 1996. IEC 60507 : Artificial pollution tests on high-voltage insulators to be used on a.c. systems, Geneva, 1991. Rizk. F.A.M.,: Mathematical models for pollution flashover, Electra, No.78, 1981, pp.71-103. Farzaneh M., Zhang J. and Chen X.:Modeling of the AC arc discharge on ice surfaces, IEEE Transactions on Power Delivery, Vol.12, No.1 ,1997, pp.325 -338. Farzaneh. M. and Zhang. J.: Modelling of DC arc discharge on ice surfaces, IEE Proc.-Gener, Transm. Distrib., Vol. 147, No.2, 2000, pp. 81-86. Farzaneh. M., Fofana. I., Tavakoli. C. and Chen. X.,: Dynamic modeling of dc arc discharge on ice surfaces, IEEE Trans. DEI, Vol. 10, No. 3, 2003. Rizk F. A. M. and Nguyen D. H.: Ac source interaction in HV pollution tests, IEEE Trans. on Power Apparatus and Systems. Vol. PAS-103, No. 4, 1984, pp. 723-732. Holtzhausen J. P.: Application of a reignition pollution flashover to Cap and Pin insulator strings, Proc. 6th IASTED International Conference: Power and Energy Systems, Rhodes - Greece, 2001, pp. 411-415. Vosloo W. L. and Holtzhausen J. P.: A model for electrical discharge and leakage current development on high voltage insulators, Proc. 2nd IASTED International Conference: Power and Energy Systems, Crete-Greece, 2002, pp. 640-643. CIGRE WG 33-04 Taskforce 01: A review of current knowledge: Polluted Insulators, 1998. Mizuno Y., Nakamura H. and Naito K.: Dynamic simulation of risk of flashover of contaminated ceramic insulators, IEEE Trans. on Power Delivery, Vol.12, No. 3, 1997, pp. 1292-1298. Rahal A. and Huraux C.: Flashover mechanism of high voltage insulators, IEEE Trans. Power Apparatus and Systems, Vol. PAS- 98, No. 6,1979, pp. 2223-2231. Psomopoulos C. S. and Karagiannopoulos C. G.:Measurement of Fusible elements during current interruption and interpretation of related phenomena, Measurement, Vol.32, 2002, pp.15-22.
Chapter 5
USE OF ORGANIC RANKINE CYCLES TO PRODUCE ELECTRIC POWER FROM WASTE HEAT SOURCES
Pedro J. Mago* and Louay M. Chamra
Department of Mechanical Engineering, Mississippi State University 210 Carpenter Engineering Building, P.O. Box ME Mississippi State, MS 39762-5925, USA
ABSTRACT
Several industrial processes have low-temperature waste heat sources that can not be efficiently recovered. Low grade waste heat has generally been discarded by industry and has become an environmental concern because of thermal pollution. A solution for this problem is the use of Organic Rankine Cycles (ORC) which can make use of low temperature waste heat to generate electric power. This chapter presents an ORC analysis using different types of organic fluids and cycle configurations. Some of the working fluids under investigation are R134a, R113, R245ca, R245fa, R123, Isobutane, and Propane, with boiling points between -43C and 48C. These organic working fluids were selected to evaluate the effect of the fluid boiling point temperature on the performance of ORCs. The results are compared with those of water under similar conditions. In order to improve the cycle performance, modified ORC are also investigated. Regenerative ORCs are analyzed and compared with basic ORCs in order to determine the configuration that presents the best thermal efficiency with minimum irreversibility. The evaluation for both configurations is performed using a combined first and second law analysis by varying certain system operating parameters at various reference temperatures and pressures. Some of the results show that ORC using R113 as working fluid has the best thermal efficiency while those using Propane show the worse efficiency. Also, it is shown that the organic-fluid boiling point has a strong influence on the system thermal efficiency and the electric power production. In addition, results from these analyses show that regenerative ORC produces higher efficiency compared to basic ORC.
*
E-mail: mago@me.msstate.edu; Phone: (662) 325-6602 - Fax:(662) 325-7223 E-mail: chamra@me.msstate.edu
178
Pedro J. Mago and Louay M. Chamra

Furthermore, regenerative ORC requires less waste heat to produce the same electric power with a lower irreversibility. ORCs are good candidates to produce electric power from low temperature heat sources.
(a)
(b)
Figure 1. Different ORC configurations: (a) basic cycle and (b) regenerative cycle.
Use of Organic Rankine Cycles to Produce Electric Power
179
INTRODUCTION
Low temperature waste heat has been generally discharged in several industrial processes due to the lack of effective methods to recover it. In addition, the discharged waste heat contributes to thermal pollution and therefore becomes an environmental problem [Hung et al., 1997]. Similarly, the increased consumption of fossil fuels has also caused environmental problems such as ozone depletion, global warming, and air pollution [Yamamoto et al., 2001; Cong et al., 2005]. A solution for these problems is the use of alternative renewable technologies that are environment friendly and also are able to recover low temperature waste heat to generate electricity. One of these technologies can be Organic Rankine Cycles (ORC) which uses an organic fluid to recover low-temperature waste heat to generate electric power. The basic principles of ORC are similar to those of the conventional Rankine cycle. The working fluid is heated to boiling, and the expanding vapor is used to drive a turbine. This turbine is used to drive a generator to convert the work into electricity. The working-fluid vapor is condensed back into a liquid and fed back through the system. Schematics of a simple and a regenerative ORC are shown in Figure 1. One advantage of using ORC instead of steam Rankine cycles in the energy industry is that the thermal efficiency becomes economically feasible when organic fluids are used to recover low grade waste heat [Hung et al., 1997]. ORC can be used in geothermal plants with low energy content, solar applications, in the recovery and use of waste heat such as Combined Heat and Power (CHP) plants (especially those utilizing biomass as fuel), in the efficiency improvement in power stations [Verschoor and Brouwer, 1995], and general heat recovery applications from many potential sources. Therefore, the use of ORC to recover waste heat is favorable in many aspects: better and economical use of the energy as well as reduce the emission of CO2. Some example of low grade waste heat (80-200C) are: industrial waste streams, solar heat trapped in the collectors, cooling water streams of stationary diesel engines, the exhaust of diesel engines and biomass, among others. Another important aspect of the operation of ORCs is the selection of the working fluid. The organic working fluid must be carefully selected based on safety and technical feasibility. There is a wide selection of organic fluids that could be used in ORC. Performances and characteristics of different working fluids for waste heat recovery system can be found in Maizza and Maizza [1996], Vijayaraghavan and Goswami [2005], and Maizza and Maizza [2001]. Generally, a good working fluid should exhibits low toxicity, good material compatibility and fluid stability limits [Niggeman et al., 1978], and low flammability, corrosion, and fouling characteristics. Refrigerants are good candidates for ORC applications due to their low toxicity characteristics [Lee at al., 1993]. Another characteristic that must be considered during the selection of organic fluid is its saturation vapor curve. This characteristic affects the fluid applicability, cycle efficiency, and arrangement of associated equipment in a power generation system [Hung, 2001]. The slope of the saturation curve in the T-s diagram depends on the type of fluid employed. A dry fluid has a positive slope; a wet fluid has a negative slope; while an isentropic fluid has infinite large slopes. This is explained in more detail later in this chapter. Several researchers have investigated the application and performance of ORC. Some of them are: Hung [2001], Hung et al. [1997], Gurgenci [1996], Yamamoto et al. [2001], Lee et al. [1988], Larjola [1991], Larjola et al. [1987], and Mago et al. [2007], among others. Mago
180
et al. [2007] presented an analysis of the performance of ORC using R113 and R134a in which it was shown that organic fluids can be used to generate power using low-temperature waste heat. They have also shown that organic fluids must be operated at saturated conditions to reduce the total irreversibility of the system. Most of the investigations presented by the above-mentioned researchers have been focused in ORC using basic configuration. In order to improve the performance of ORCs, different configurations such regenerative ORC have to be analyzed and compared with basic ORC. Reheat ORCs were not included in this chapter since the authors determined that the performance of this type of cycles is very similar to the performance of basic ORC, therefore the analysis presented in this chapter is focused on regenerative ORCs. The evaluation of regenerative ORCs includes the effect of this configuration on the overall thermal efficiency of the cycle, the cycle total irreversibility, amount of waste heat needed to operate the cycle, and system second law efficiency. Different procedures to perform second law analysis of power cycles can be found in Bejan [1996], Moran and Shapiro [2003], among others. This chapter presents a study to evaluate the use of ORC to produce electric power from low temperature waste heat sources. This study includes the analysis of the change in thermal efficiency and irreversibility of ORC using different working fluids varying system-operating parameters at various reference temperatures.
ANALYZING ORGANIC RANKINE CYCLES

To evaluate the performance of electric power generation systems, ORC in this specific case, thermodynamic models of the different components have to be developed. Although the analysis of simplified thermodynamic models generally leads only to qualitative conclusion about the cycle performance, these models allow to evaluate how the changes in operating parameters affect the actual cycle performance. Also, different parameters can be evaluated to improve the cycle overall performance by utilizing these models. The basic principles of ORC are similar to those of the conventional Rankine cycle. The major difference is that the working fluid in the ORC is an organic fluid which has a lower boiling point and a higher vapor pressure than that of water, which improves the total performance of the cycle using low temperature heat sources. The equations used to determine the cycle efficiencies as well as the cycle irreversibility for basic and regenerative ORC configurations are presented in this section. Using the First and Second Laws of Thermodynamics, the performance of ORCs can be evaluated under various working conditions for different working fluids. The thermodynamic model presented in this chapter assumes the following: steady state conditions, no pressure drop in the evaporator, condenser, and pipes, and isentropic efficiencies for the turbine and the pump. The components of a basic and a regenerative ORCs for converting waste heat into useful electrical power are shown in Figure 1. As observed in Figure 1(a), basic ORC consists of four different processes: Process 1-2 (pumping process), Process 2-3 (constant-pressure heat addition), Process 3-4 (expansion process), and Process 4-1 (constant-pressure heat rejection). For the regenerative cycle, Figure 1(b), a feed-water heater is incorporated into the ORC. The vapor extracted from the turbine mixes with the feed-water exiting the pump. Ideally the mixture leaves the heater as a saturated liquid at the heater pressure. Figure 2 shows a
181
temperature-entropy diagram corresponding to the ORC configurations shown in Figure 1(a) and (b). For each individual component, the first and second laws of thermodynamic are applied to find the work output, the heat added or rejected, and the components and system irreversibility. The energy balance equation can be expressed as:
& E +Q = E
i i o o
& +W
(1)
& & where Ei and Eo are the energy rate in and out, Q is the heat transfer rate, and W is the power.
The irreversibility rate for uniform flow conditions can be expressed as: dS & & = To m I = To dt
exit
inlet
dssystem + dt +
T
j
qj j
(2)
where the subscript j stands for the heat transfer for different reservoirs and the term dssystem dt = 0 for steady state conditions.
Basic ORC Figure 1(a)

a. Process 1-2 (Pump) The liquid leaving the condenser at Point 1 is pumped into the evaporator. Taking a control volume around the pump, using Equation (1) and assuming a pump isentropic efficiency, the pump power can be expressed as:
& W p ,ideal m(h1 h2 s ) & & Wp = =
(3)
& & where W p ,ideal is the ideal power of the pump, m is the working fluid mass flow rate, p is
the pump isentropic efficiency, and h1 and h2s are the enthalpies of the working fluid at the inlet and outlet of the pump for the ideal case. Using Equation (2), the pump irreversibility rate can be determined as:
& & I p = To m(s2 s1 )
(4)
where s1 and s2 are the specific entropies of the working fluid at the inlet an exit of the pump for the actual conditions, respectively.
182
b. Process 2-3 (Evaporator) The evaporator heats the working fluid at the pump outlet to the turbine inlet condition. Taking a control volume enclosing the evaporator, the heat transfer rate from the energy source into the working fluid is given by:
& & Qe = m(h3 h2 )
(5)
where h3 and h2 are the enthalpies of the working fluid at the exit and inlet of the evaporator, respectively. Using Equation (2), the evaporator irreversibility rate can be determined as:
h h & & I e = To m (s3 s2 ) 3 2 TH
(6)
where s3 and s2 are the specific entropies of the working fluid at the inlet an exit of the evaporator, respectively, and TH is the temperature of the high-temperature heat source. This temperature is considered to be equal to TH = T3 + TH
c. Process 3-4 (Turbine) Vapor from the evaporator at Point 3, with high temperature and pressure, expands trough the turbine to produce mechanical work and then is discarded to the condenser at Point 4. For a control volume around the turbine and assuming a turbine isentropic efficiency, the turbine power is given by:
& & & Wt = Wt ,idealt = m(h3 h4 s )t
(7)
& where Wt ,ideal is the ideal power of the turbine, t is the turbine isentropic efficiency, and h3
and h4s are the enthalpies of the working fluid at the inlet and outlet of the turbine for the ideal case. The turbine irreversibility rate can be expressed as:
& & I t = To m(s4 s3 )
(8)
where s3 and s4 are the specific entropies of the working fluid at the inlet an exit of the turbine for the actual conditions, respectively.
d. Process 4-1 (Condenser) The condenser heat rejection rate can be expressed as:
& & Qc = m(h1 h4 )
(9)
The condenser irreversibility rate can be determined from Equation (2) as follows
183
h h & & I c = To m (s1 s4 ) 1 4 TL
(10)
where s1 and s4 are the specific entropies of the working fluid at the inlet an exit of the condenser, respectively, and TL is the temperature of the low temperature reservoir. This temperature is considered to be equal to TL = T1 TL
e. Cycle Efficiency The thermal efficiency is defined as the ratio between the net power of the cycle to the evaporator heat rate. It gives a measure about how much of the waste heat input to the working fluid passing trough the evaporator is converted to net work output. It can be expressed as:
cycle =
& & Wt + W p & Q

e
(11)
Substituting Equations (3), (5), and (7) into Equation (11) the thermal efficiency for a basic ORC can be write as:
cycle =
(h3 h4 s )t + (h1 h2 s ) 1 p (h3 h2 )
(12)
f. Total Cycle Irreversibly The total irreversibility can be obtained adding Equations (4), (6), (8), and (10) as follows:
& I cycle =
I&
j
h h h h & & & & & =I p + I e + I t + I c = To m 3 2 1 4 TH TL
(13)
g. Second Law Efficiency The second law cycle efficiency can be calculated using the following equation:
II =
& Wnet & T Qe 1 L T H
(h3 h4 s )t + (h1 h2 s ) 1 T p 1 L = TH (h3 h2 )
(14)
h. Electric Generator The mechanical power produce in the turbine is converted to electric power in the generator. The total electric power can be expressed as:
184
& & & & Welectic = Wnet generator = Wt + W p generator

where generator is the generator efficiency.
(15)
Regenerative ORC - Figure 1(b)

a. Feed-water Heater In order to determine the fraction of the flow rate that goes into the feed-water heater and the fraction that goes into the condenser a mass and energy balance is applied in the feedwater heater. The fraction of the flow rate that goes into the feed-water heater is given by:
X1 = h3 h2 h6 h2
(16)
b. Pump: (Processes 1-2 and 3-4) The pump power for Pump I (Process 1-2) can be expressed as:
& W p,ideal (1 X 1 )m(h1 h2 s ) & & W p,1 = =
(17)
and the pump power for Pump II (Process 3-4) can be expressed as:
& W p ,ideal m(h3 h4 s ) & & Wp,2 = =
(18)
& & where W p ,ideal is the ideal power of the pump, m is the working fluid mass flow rate, p is the pump isentropic efficiency, h1 and h2s are the enthalpies of the working fluid at the inlet and outlet of Pump I for the ideal case, and h3 and h4s are the enthalpies of the working fluid at the inlet and outlet of Pump II for the ideal case. Adding Equations (17) and (18) the total pump power can be determined as:
(1 X 1 )(h1 h2 s ) + (h3 h4 s ) & & Wp = m p Using Equation (2), the total pump irreversibility rate can be expressed as: & & I p = To m[(1 X 1 )(s1 s2 ) + (s3 s4 )]
(19)
(20)
where s1 and s2 are the specific entropies of the working fluid at the inlet an exit of Pump I for the actual conditions, respectively, and s3 and s4 are the specific entropies of the working fluid at the inlet an exit of Pump II for the actual conditions, respectively.
185
c. Evaporator: (Process 4-5) The evaporator heat transfer rate can be expressed as:
& & Qe = m(h5 h4 )
(21)
where h4 and h5 are the enthalpies of the working fluid at the inlet and exit of the evaporator, respectively. The evaporator irreversibility rate can be expressed as: (h h ) & & I e = To m (s5 s4 ) 5 4 TH
(22)
where s4 and s5 are the specific entropies of the working fluid at the inlet and exit of the evaporator, respectively, and TH is the temperature of the high temperature heat source. Similarly to the other configurations, this temperature is considered to be equal to TH = T3 + TH
d. Turbine: (Processes 5-6 and 5-7) The turbine power for this configuration is given by:
& & & Wt = Wt ,idealt = mt [(h5 h7 s ) + X 1 (h7 s h6 s )]
(23)
where Wt ,ideal is the ideal power of the turbine, t is the turbine isentropic efficiency, h5 is the enthalpy of the working fluid at the inlet of the turbine, and h6s and h7s are the enthalpies of the working fluid at the exit of the turbine for the ideal case. The turbine irreversibility rate can be expressed as:
& & I t = To m[(s7 s5 ) + X 1 (s6 s7 )]
(24)
where s5 is the specific entropy of the working fluid at the inlet of the turbine, and s6 and s7 are the specific entropies of the working fluid at the exit of the turbine for the actual conditions.
e. Condenser: (Process 7-1) The condenser heat rate can be expressed as:
& & Qc = m(1 X 1 )(h1 h7 )
(25)
where h1 and h7 are the enthalpies of the working fluid at the exit and inlet of the condenser, respectively. The condenser irreversibility rate can be expressed as:
186
Pedro J. Mago and Louay M. Chamra h h & & I c = To m(1 X 1 )(s1 s7 ) 1 7 TL
(26)
where s1 and s7 are the specific entropies of the working fluid at the inlet an exit of the condenser, respectively, and TL is the temperature of the low temperature reservoir. This temperature is considered to be equal to TL = T1 TL
f. Cycle Efficiency The thermal efficiency can be determined as:
cycle =
& & Wt + W p & Q

e
(27)
Substituting Equations (19), (21), and (23) into Equation (27) the thermal efficiency of regenerative ORC can be expressed as
cycle = t [(h5 h7 s ) + X 1 (h7 s h6 s )] + (1 X 1 )(h1 h2 s ) + (h3 h4 s ) 1 p
(h5 h4 )
(28)
g. Total Cycle Irreversibly The total irreversibility can be obtained adding Equations (20), (22), (24), and (26) as follows:
(h h ) h h & & I cycle = mTo 5 4 (1 X 1 ) 1 7 T TH L
(29)
h. Second Law Efficiency The second law cycle efficiency for the regenerative cycle can be expressed as:
1 t [(h5 h7 s ) + X 1 (h7 s h6 s )] + (1 X 1 )(h1 h2 s ) + (h3 h4 s ) p = (h5 h4 )1 TL T H
& Wnet II = & T Qe 1 L T H
(30)
i. Electric Generator similar to the basic ORC the total electric power generation can be determined as:
& & & & Welectic = Wnet generator = Wt + W p generator
(31)
where generator is the generator efficiency.
187
WORKING FLUIDS FOR ORC APPLICATIONS

As was mentioned before, one of the most important parameters of the operation of ORCs is the selection of the organic working fluid. The organic working fluid must be carefully selected based on safety and technical feasibility. There is a wide selection of organic fluids that could be used in ORC, however the selection of the working fluid and other operating conditions has to be taken very seriously in order to have a good system performance. The proper choice of the working fluid is as important as the waste heat source temperature for an effective ORC power plant [Liu et al., 2004]. Information regarding different working fluids that could be used in ORC applications can be found in Maizza and Maizza [1996], Vijayaraghavan and Goswami [2005], and Maizza and Maizza [2001]. A good working fluid for an ORC system should have the following desirable characteristics:
(a) (b Figure 2. Temperature-entropy diagram corresponding to the ORC configurations shown in Figure 1(a) and (b), respectively.
Small specific volume Low vapor superheat requirement Moderate vapor pressure in the heat exchange units Low viscosity Surface tension High thermal conductivity Suitable thermal stability limits Compatible with engine materials and lubricating oil Non corrosive Non inflammable Non toxic
188
Low in cost
a wet fluid has a negative slope (dT ds < 0 ) ; while an isentropic fluid has an infinite large
In fact no fluid will satisfy all of the above characteristics. Prediction of thermodynamic capability of substances could help to select potential working media from numerous compounds. To predict the thermodynamic properties of organic working fluids thermodynamic models, such as the model proposed by Lee et al. [1993] or commercially available software, such as REFPROF, could be applied. A detailed list of environmentally acceptable working fluid with their different characteristics is given by Maizza and Maizza [1996]. Refrigerants are good candidates for ORC applications due to their low toxicity characteristics [Lee et al., 1993]. However, because of zero ozone depletion potential (ODP) hydrofluorocarbons (HFCs) have been predominantly chosen as alternative refrigerants replacing chlorofluorocarbons (CFCs) and hydrochlorofluorocarbons (HCFCs). Since HFCs have a high global warming potential (GWP) there is still search for next generation refrigerants which might have a better cycle performance. It has been found that the presence of hydrogen bonds in certain molecules such as water, ammonia, and ethanol may result in wet fluid conditions due to larger vaporizing enthalpy and is regarded as inappropriate for ORC systems. Another important characteristic that must be considered during the selection of organic fluid is its saturation vapor curve. This characteristic affects the fluid applicability, cycle efficiency, and arrangement of associated equipment in a power generation system [Hung, 2001; Mago et al., 2007]. The slope of the saturation curve in the temperature-entropy (T-s) diagram depends of the type of fluid employed. A dry fluid has a positive slope (dT ds > 0 ) ;
slope (dT ds ) . Generally, dry and isentropic fluids are better working fluids for an ORC since they do not condensate after the fluid goes through the turbine [Mago et al., 2007; Lee et al., 1993]. Also, dry fluids are superheated after isentropic expansion thereby eliminating the concerns of impingement of liquid droplets on the turbine blades [Liu et al., 2004]. The comparison of the temperature-entropy diagram for dry, wet, and isentropic fluids (for a basic ORC) is presented in Figure 3. Properties and classification of some of the organic working fluids employed in this investigation are presented in Table 1.
Table 1. Properties of some organic working fluids and water
Fluid Density (kg/m3) 511.9 560.0 550.0 523.6 220.5 224.4 517.0
Type
Molecular Weight (kg/mole) -26.07 47.59 27.82 25.13 -42.09 -11.61 14.90 4.059 3.392 3.662 3.925 4.248 3.64 4.25 101.06 214.06 183.68 174.42 96.70 134.7 154.05
Boiling Point (C)
Critical properties P T (MPa) (C)
R134a 1 R113 2 R123 3 R245ca 4,5 Propane 6 Isobutane 6 R245fa 7
Wet Dry Dry Dry Wet Dry Dry
102.03 187.38 152.93 134.05 44.10 58.12 134.05
Range of application Minimum Maximum Temperature Temperature (C) (C) -103.3 180.0 -36.2 251.9 -107.2 326.9 -73.2 226.9 -187.3 326.9 -159.6 326.9 -73.2 226.9
Maximum Pressure (MPa) 70 200 40 60 733 35 60
Tillner-Roth, R. and Baehr, H.D. 1994. Marx, V., Pruss, A. and Wagner, W. 1992. 3 Younglove, B.A. and McLinden, M.O. 1994. 4 Huber, M.L. and Ely, J.F. 1994. 5 Defibuagh, D.R., Gillis, K.A., Moldover, M.R., Schmidt, J.W., and Weber, L.A. 1996. 6 Younglove, B.A. and Ely, J.F. 1987. 7 Defibaugh, D.R. and Moldover, M.R. 1997.
190
The effect of various working fluids on the thermal efficiency and on the total heat recovery efficiency has been studied by Liu et al. [2004]. The results included: 1) The presence of hydrogen bond in certain molecules such as water, ammonia and ethanol results in wet fluids due to larger vaporizing enthalpy and is regarded as inappropriate for the ORC systems. 2) Thermal efficiency for various working fluids is a week function of the critical temperature regardless of the fact that the thermal efficiency for the working fluids with the lower critical temperature is lower. 3) In general, the maximum value of total heat recovery efficiency occurs at the appropriate evaporating temperature that is between the inlet temperature of waste heat and the condensing temperature. The maximum value of total heat recovery efficiency increases with the increase of the inlet temperature of the waste heat and decreases it by using working fluids of the lower critical temperature. The properties of the working fluid affect the cycle efficiency and capital cost. Therefore, the selection of a proper working fluid for an ORC is very important since the purpose of the ORC is to recover low grade heat to produce electric power. The purpose of using ORC will not be served if proper care is not taken in the selection of the working fluid.
Figure 3. Comparison of the working fluids: (a) dry, (b) isentropic, and (c) wet.
191
RESULTS AND DISCUSSIONS

This section presents a study of a basic ORC and the comparison of the basic ORC performance with the performance of regenerative ORC.
Basic ORC
For the purpose of this study seven organic fluids with different boiling points ranging from -43C to 48C were employed. These organic fluids are: R134a, R113, R245ca, R245fa, R123, Isobutane, and Propane. These organic fluids were selected to investigate the effect of the boiling point and the effect of the fluid classification on the performance of basic ORC. The selected fluids are dry and wet. Isentropic fluids such as R12 and R22 were not included since they are being phase-out and replace with alternative refrigerants. The results for the different organic fluids were compared with those of water under similar operating conditions. Some of the properties of the fluids used in this section are presented in Table 1. To investigate the effect of the turbine inlet temperature on the cycle efficiency, system specific irreversibility, and second law efficiency the following operating conditions were used. The evaporator pressure and condenser temperature were kept constant at 1.5 MPa and 298 K, respectively. The isentropic efficiencies of the turbine and pump were 80% and 85%, respectively, while the temperature differential was kept constant at 15 K for all cases. Figure 3 shows the variation of the system thermal efficiency with the turbine inlet temperature. Basically, this figure shows the effect of superheating of the working fluid over the thermal efficiency of the cycle. The range of temperature used to analyze each fluid varies from the saturation temperature to the critical temperature. This figure illustrates that the cycle thermal efficiency for the evaluated organic fluids is a weak function of the turbine inlet temperature since it remains approximately constant or slightly decreases with the increment of the turbine inlet temperature. This reflects the fact that organic fluids do not need to be superheated to increase the cycle thermal efficiency as oppose to water that increasing the inlet turbine temperature increases the thermal efficiency. Even though only one evaporator pressure is shown in Figure 3, the authors verified that the performance for other pressures is similar to the one presented here. Figure 3 can also be used to analyze the influence of the fluid boiling point temperature on the system thermal efficiency. The boiling point temperatures are shown next to each fluid in Figure 3. It can be observed that the organic fluid that shows the best thermal efficiency is R113, which has the highest boiling point among the selected fluids (47.59C), while the fluid with the worst thermal efficiency is propane, which has the lowest boiling point temperature (-42.09C). Similar trend is observed for the remaining working fluids selected in this investigation. Therefore, it can be concluded that the higher the boiling point temperature of the fluid the better the cycle thermal efficiency. From Figure 3, it can also be observed that R113 shows the maximum efficiency among the organic fluids for temperatures above 430 K, while R123, R245ca, and R245fa show the best efficiencies for temperatures between 380 K and 430 K. For temperatures below 380 K, Isobutane shows the best efficiency, while water is the best fluid when temperatures go above 470 K for the analyzed conditions. This figure also illustrates that the selected organic dry fluids (R113,
192
R123, R245ca, R245fa, and Isobutane) show better performance than the selected organic wet fluids (R134a and propane). As was mentioned before, one of the reasons dry fluids show better thermal efficiencies is because they do not condense after the fluid goes through the turbine as opposed to wet fluids that present condensate after the turbine. However, it is important to point out here, that independently of the fluid classification, organic fluids can be used to produce power from low temperature waste heat. However, organic fluids are restricted to a small range of applicability depending on their thermodynamic conditions. Figure 4 shows the variation of the system specific irreversibility with the turbine inlet temperature. It can be observed that the total system irreversibility increases with the increment of the turbine inlet temperature for all the fluids. The results presented in this figure show the importance to perform a second law analysis. According to the results presented in Figure 3 the thermal efficiency is approximately constant with the increment of the turbine inlet temperature. However, a combine first and second law analysis shows that the best case scenario is obtained when the fluid is operated at saturated conditions before the turbine. This yields the same thermal efficiency with lower irreversibility that operating under superheated conditions. Figure 4 also illustrates how the system with higher (R113) and lower (propane) thermal efficiencies present the lower and higher irreversibility, respectively. Water shows higher specific irreversibility compared with the evaluated organic fluids. The effect of the boiling point temperature in this figure does not present a consistent trend as the one showed in Figure 3.
30 R113 R123 R245ca R245fa Isobutane R134a Propane Water
TBP = 99.97C
System Thermal Efficiency (%)
25
20
TBP = 47.59C
TBP = 27.82C
15
TBP = 25.13C TBP = 14.90C TBP = -11.61C TBP = -26.07C
10
TBP = -42.09C
0 300 350 400 450 500
Turbine Inlet Temperature (K)

Figure 3. Variation of the system thermal efficiency with the turbine inlet temperature (Pe = 1.5 MPa and Tc = 298 K).

600
193
System Specific Irreversibility (kJ/kg-K)
500
400
R113 R123 R245ca R245fa Isobutane R134a Propane Water
300
200
100
0 300 350 400 450 500

Figure 4. Variation of the system irreversibility with the turbine inlet temperature (Pe = 1.5 MPa and Tc = 298 K).
100 R113 R123 R245ca R245fa Isobutane R134a Propane Water
Second Law Efficiency (%)
80
60
40
20
0 300 350 400 450 500

Figure 5. Variation of the system second law efficiency with the turbine inlet temperature (Pe = 1.5 MPa and Tc = 298 K).
194
30
25
20
15
R113 R123 R245ca R245fa Isobutane R134a Propane Water
10
0 0 1 2 3 4 5
Turbine Inlet Pressure (MPa)

Figure 6. Variation of the system thermal efficiency with the turbine inlet pressure (Tc = 298 K).
Figure 5 illustrates the effect of the turbine inlet temperature on the system second law efficiency. It can be observed that for all the fluids the second law efficiency decreases with the turbine inlet temperature. These results agree well with the results presented in Figure 4, since an increment in the system irreversibility yield a decrease in the system second law efficiency. For temperatures between 430K and 525K, R113 shows the best second law efficiency, for a range of 400K to 430K, R123 shows the best efficiency; R245ca and R245fa present the best second law efficiency for temperatures between 380K and 400K. Isobutane shows the best efficiencies for a temperature range of 360K to 380K, R134a for a temperature range of 330K to 360K, while propane shows the low second law efficiency among all the evaluated fluids. The effect of the turbine isentropic efficiency on the system second law efficiency was also evaluated. It was found that an increment of the turbine isentropic efficiency represent and increase of the second law efficiency for all the evaluated fluids. The next parameter to be evaluated is the effect of the turbine inlet pressure on the system performance. For this case the condenser temperature was kept constant at 298 K, while the maximum pressure used for each fluid was the critical pressure. The isentropic efficiencies of the turbine and pump were 80% and 85%, respectively, while the temperature differential was kept constant at 15 K. Figure 6 illustrates the variation of the system thermal efficiency with the turbine inlet pressure while keeping the turbine inlet temperature at saturated conditions. The results are consistent for all the fluids, since the system thermal efficiency increases with the increment of the turbine inlet pressure for all of them. This can be explained since with the inlet turbine pressure increase both the net work and the evaporator heat increase. However, the percentage of increase of the net work is higher than the percentage of increase
195
of the evaporator heat. Therefore the ratio of the net work and the evaporator heat increases with the turbine inlet pressure. Similar to the results presented in Figure 3, this figure shows that R113 has the best performance among the organic fluids for pressures below 3.4 MPa, R123 and R245ca present the best efficiencies for a range of pressures between 3.4 and 3.6 MPa, while R134a presents the best efficiency for a range of pressure between 3.6 and 4.2 MPa. Water shows the best thermal efficiency under the conditions analyzed in this case. The trend observed with the boiling point described in Figure 3 is also consistent with the results presented in this figure. Figure 7 showed the total specific irreversibility versus the turbine inlet pressure for the same conditions used to generate Figure 6, for R113, R123, R245fa, and Isobutane. Some of the fluids were omitted from this figure to improve its readability. It can be observed how the irreversibility increases with the increment of the turbine pressure for all the fluids. Water shows the highest irreversibility values and they are not shown in the figure in order to be able to observe the values for the remaining fluids. Among the organic fluids, Isobutane and R123 show the highest and lowest irreversibility values, respectively. Similar to Figure 4, the effect of the boiling point temperature in this figure does not present a consistent trend as the one showed in previous figures.
100
System Specific Irreversibility (kJ/kg-K)
80
R113 R123 Isobutane Propane
60
40
20
0 0 1 2 3 4 5

Figure 7. Variation of the system irreversibility with the turbine inlet pressure (Tc = 298 K) for R113, R123, R245fa, and Isobutane.
196
10
Mass Flow Rate (kg/s)
R113 R123 Isobutane Propane
0 0 1 2 3 4 5

Figure 8. Mass flow rate needed to produce 100 kW of electric power versus turbine inlet pressure (Tc = 298 K) for R113, R123, R245fa, and Isobutane.
40 R113 R123 R245ca R245fa Isobutane R134a Propane
30
20
10
0 275 280 285 290 295 300 305
Condenser Outlet Temperature (K)

Figure 9. Variation of the system thermal efficiency with the condenser outlet temperature (Pe = 3 MPa).
197
The variation of the mass flow rate needed to generate the same electric power output with the turbine inlet pressure is evaluated in Figure 8. This figure was generated using the same conditions described in Figure 6 and for a fixed electric power output of 100 kW and using a generator efficiency of 82%. It can be seen that for all the fluids the mass flow rate needed to produce the same electric power decreases with the increment of the turbine inlet pressure. This is due to the increase in the net work of the cycle with the increment in turbine inlet pressure. From this figure it can be observed that propane requires the highest mass flow rates among the organic working fluids, while for pressures below and above 1.5 MPa, R113 and Isobutane require the lowest mass flow rates, respectively. Figure 9 shows the variation of thermal efficiency with the condenser outlet temperature. This figure was generated keeping the evaporator pressure constant at 3 MPa. The isentropic efficiencies of the turbine and pump were 80% and 85%, respectively. From this figure it can be observed that for all the working fluids, the system thermal efficiency decreases linearly with the increase in condenser outlet temperature. The trend observed in this figure is consistent with the results shown in Figures 3 and 6 where R113 and R123 show the best thermal efficiencies while Propane the worse among the evaluated organic fluids. This figure demonstrates once more the influence of the boiling point on the system thermal efficiency, since fluids with higher boiling point temperatures have the best thermal efficiency and vice versa. Figure 10 illustrates the variation of the system second law efficiency with the condenser outlet temperature for R113, Isobutane, and Propane. Some fluids were left out of this figure in order to make it clear to read. It can be observed that the second law efficiency decreases for all the fluids with the increment of the condenser outlet temperature. The results presented in Figures 9 and 10 indicate that ORC will be more beneficial in places with annual low ambient temperatures, since it will have higher first and second law efficiencies.
60
50
40
30
20 R113 Isobutane R134a Propane
10
0 275 280 285 290 295 300 305
Condenser Outlet Temperature (K)
Figure 10. Variation of the system second law efficiency with the condenser outlet temperature (Pe = 3 MPa) for R113, Isobutane, and Propane.
198
Regenerative ORC
Previous results showed that dry fluids have a better performance than wet fluids for basic ORC applications. Therefore, this section focuses on the use of regenerative ORC to produce electric power using only dry fluids. Four organic dry fluids, with boiling points ranging from -12C to 48C, were selected: R113, R245ca, R123, and Isobutane. Figure 11 shows the variation of the system thermal efficiency with the turbine inlet pressure while keeping the turbine inlet temperature at saturated conditions for the different ORC configurations (basic and regenerative). For both configurations the condenser temperature was kept constant at 298 K, while the maximum pressure used for each fluid was the critical pressure. The isentropic efficiencies of the turbine and pump were 80% and 85%, respectively, while the temperature differential was kept constant at 15 K. Figure 11 demonstrates that the system thermal efficiency increases with the increment of the turbine inlet pressure, the results are consistent for all the fluids used in both configurations. This figure illustrates that for both configurations, the fluid that shows the best thermal efficiency is R113, which has the highest boiling point among the selected fluids (47.59C), while the fluid with the worst thermal efficiency is isobutane, which has the lowest boiling point temperature (-11.61C). Similar trend is observed for the remaining working fluids selected in this investigation. Therefore, it can be concluded that the higher the boiling point temperature of the fluid the better the cycle thermal efficiency for both basic and regenerative cycle.
25
Thermal System Efficiency (%)
20
15
10
R113 R245ca R123 Isobutane
0 1 2 3 4
(a)

25
199
20
15
10
R113 R123 Isobutane R245ca
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
(b)
Figure 11. Variation of the system thermal efficiency with the turbine inlet pressure for different ORC configurations (Tc = 298 K): (a) basic and (b) regenerative.
Figure 12 shows a comparison of the system thermal efficiency for the two cases presented in Figure 11. This comparison was performed for R113 and Isobutane since they represent the best and worst case, respectively. The regenerative ORC shows better thermal efficiencies than basic ORC for both fluids. Regenerative ORC using R113 shows an increase of 13.4% to 16% for the lowest and highest turbine inlet pressure, respectively. On the other hand, isobutane shows an increase of 6% to 8.5% for the lowest and highest pressure, respectively. These results clearly demonstrate that using regenerative ORC the system thermal efficiency of the basic ORC can be increased. It is also important to point out, that the use of regenerative ORC is not justified for all fluids from the thermal efficiency point of view, since the difference between this configuration and basic ORC is sometimes negligible for some organic fluids. However, there are other parameters to be considered in this analysis, such as: the total irreversibility of the system, second law efficiency, and the amount of heat required to produce the same electric power, that can add additional benefits to the use of regenerative ORC. Figure 13 shows the variation of the system specific irreversibility using R113 with the turbine inlet pressure under the same conditions used to generate the results shown in Figure 11. R113 was the only organic fluid included to improve the readability of the figure. It can be observed that the total system irreversibility increases with increasing turbine inlet pressures for both configurations. A comparison between the two configurations shows that regenerative ORC has approximately 42% less specific irreversibility than basic ORC for the pressure range evaluated in this chapter.
200
25
Thermal System Efficiency (%)
20
15
10
R113 - Basic R113 - Regenerative Isobutane - Basic Isobutane - Regenerative
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Figure 12. Comparison of the two ORC configurations for R113 and Isobutane (Tc = 298 K).
70 Basic ORC Regenerative ORC 60
Irreversibility (kJ/kg K)
50
40
30
20
10 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 13. Variation of the specific system irreversibility with the turbine inlet pressure for R113 (Tc = 298 K).
201
100 Basic ORC Regenerative ORC
80
60
40
20
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 14. Variation of the system second law efficiency with the turbine inlet pressure for R113 (Tc = 298 K).
Figure 14 illustrates the variation of the system second law efficiency with the turbine inlet pressure for R113 for both ORC configurations. It can be observed that for both configurations the second law efficiency decrease with the increment in the turbine inlet pressure. The results presented in this figure agree well with the results presented in Figure 12, since an increment in the system irreversibility represent a decrease in the system second law efficiency. It can also be seen that the second law efficiency obtained for regenerative ORC are higher than those obtained for basic ORC. For low inlet turbine pressures the second law efficiencies for regenerative ORC using R113 are approximately 12% higher than those obtained for basic ORC, while for high inlet turbine pressures the second law efficiencies for regenerative ORC are approximately 17% higher than those obtained for basic ORC. Figures 13 and 14 exhibit the importance of performing a second law analysis. According to the results presented in Figure 12 using regenerative ORC, employing R113, the thermal efficiency can be increased by approximately 15% compared with basic ORC. However, a combined first and second law analysis shows that using regenerative ORC not only increases the system thermal efficiency but also reduces the system irreversibility by 42%, or increases the second law efficiency by an average of 14.5% compared with basic ORC. The heat transfer rate required in the evaporator to generate the same power output with the turbine inlet pressure is evaluated in Figure 15. This figure was generated using R113 for the different configurations using the same conditions described in Figure 3, and for an electric power output of 100 kW, with a typical generator efficiency of 82%. The generator
202
efficiency is defined as the ratio of the electric power and the net power produced by the ORC. It can be seen that for R113 the evaporator heat rate needed decreases with increasing turbine inlet pressures. This is due to the decrease in the mass flow rate needed and the increase in the net work of the cycle with the increment in turbine inlet pressure. Figure 15 demonstrates that regenerative ORC required lower heat rates compared to basic ORC to produce the same power. The amount of heat required for regenerative ORC is 7.5% to 9.7% less than the heat required for a basic ORC, for the lowest and highest turbine inlet pressures, respectively. From Figures 12, 13, 14, and 15 it can be concluded that regenerative ORC not only increases the first and second law efficiencies but also decreases the amount of heat required and system irreversibility to produce the same power output as compared with basic ORC. Figure 16 shows the mass flow rate needed for R113, to produce the same electric power, for the cases analyzed in Figure 15. It can be seen that the required mass flow rate decreases with the increment of the turbine inlet pressure. This is due to the increase in the net work of the cycle with the increment in turbine inlet pressure. Another interesting point is that the basic configuration requires lower mass flow rates compared to regenerative ORC. This can be explained since basic ORC generates higher specific net works than regenerative ORC. However, the ratio between the net work and the total heat input (thermal efficiency) is lower for this configuration than for the regenerative ORC.
800
Heat Rate Required (kW)
700
600
500
400
300 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 15. Variation of the heat transfer rate required in the evaporator to produce 100 kW of electric power with the turbine inlet pressure for R113 (Tc = 298 K).

5 Basic ORC Regenerative ORC 4
203
Mass Flow Rate (kg/s)
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 16. Variation of the mass flow rates required to produce 100 kW of electric power with the turbine inlet pressure for R113 (Tc = 298 K).
Figure 17 shows the variation of the system thermal efficiency with the turbine inlet temperature. To generate this figure, the evaporator pressure and condenser temperature were kept constant at 2 MPa and 298 K, respectively. Similar to the previous case, the isentropic efficiencies of the turbine and pump were maintained at 80% and 85%, respectively, and the temperature differential was kept constant at 15 K for all cases. This figure shows the effect of superheating the working fluid on the thermal efficiency of the cycle. The temperature range for each fluid used to analyze the different configurations is from the saturation temperature to the critical temperature. Figure 17 illustrates that the efficiency of the cycle slightly decreases for some fluids or remains approximately constant for others with the increment of the turbine inlet temperature for all the configurations. This reflects the fact that organic fluids do not need to be superheated to increase the cycle thermal efficiency as opposed to water where increasing the inlet turbine temperature increases the thermal efficiency. The trend observed with the boiling point described in Figure 12 is also consistent with the results presented in this figure. Figure 17 demonstrates that regenerative ORC has the best thermal efficiency for the analyzed organic fluids. Among the three organic fluids included in this figure, R113 shows the maximum efficiencies for temperatures above 450 K. R123 shows the best efficiencies for temperatures between 420 K and 450 K, while Isobutane is the best fluid for temperatures between 375 K and 410 K. It is important to point out here, how organic fluids can be used to produce power from low temperature waste heat; however, organic fluids are restricted to a small range of applicability depending on their thermodynamic conditions.
204
25
System Themal Efficiency (%)
20 R123 15 Isobutane 10
R113
R113 - Basic R113 - Regenerative R123 - Basic R123 - Regenerative Isobutane - Basic Isobutane - Regenerative
0 340 360 380 400 420 440 460 480 500

Figure 17. Variation of the system thermal efficiency with the turbine inlet temperature (Pe = 2 MPa).
80
60
40
20
0 450 455 460 465 470 475 480 485

Figure 18. Variation of the system second law efficiency with the turbine inlet temperature for R113 (Pe = 2 MPa).
205
Figure 18 shows the variation of the system second law efficiency with the turbine inlet temperature using R113 for both configurations. It can be observed that for both configurations the second law efficiency decreases with the turbine inlet temperature. Regenerative ORC shows an increase on the second law efficiency of about 14% compared with the basic ORC. According to the results presented in Figure 17, the thermal efficiency is approximately constant with the increment of the turbine inlet temperature. However, a combined first and second law analysis shows that the best case scenario is obtained when the fluid is operated at saturated conditions before the turbine since this produces the same thermal efficiency with lower irreversibility and higher second law efficiencies than operating under superheating conditions.
CONCLUSIONS
This paper presents an analysis of the performance of basic and regenerative ORC to produce electric power. This analysis was based on the first and second laws of thermodynamics, and parameters such as thermal efficiency, irreversibility, and second law efficiency were evaluated and compared with the results for water under the same conditions. It was shown that the examined organic fluids could be used to generate power using lowtemperature waste heat. Organic fluids do not need to be superheated since the cycle thermal efficiency remains approximately constant when the inlet temperature of the turbine is increased. However, using the second law analysis it can be seen that superheating organic fluids increases the irreversibility. Therefore, organic fluids must be operated at saturated conditions to reduce the total irreversibility of the system. It can also be concluded, that the thermal efficiency of ORC increases when the condenser temperature is decreased. Therefore, using ORC in locations with low ambient temperatures will be more effective. Organic dry fluids (R113, R123, R245ca, R245fa, and Isobutane) show better performance than wet fluids (R134a and Propane). This is due because they do not condense after the fluid goes through the turbine as opposed to wet fluids that present condensate after the turbine. It was shown that regenerative ORC not only has higher first and second law efficiencies than basic ORC but it has lower irreversibility, and lower heat required to produce the same power. Dry organic fluids do not need to be superheated since the cycle thermal efficiency remains approximately constant when the inlet temperature of the turbine is increased. Moreover, second law analysis showed that superheating organic fluids increases the irreversibility and decreases the second law efficiency. Therefore, organic fluids should be operated at saturated conditions to reduce the total irreversibility of the system. The influence of the boiling point temperature on the system thermal efficiency for both configurations was determined. The fluid that shows the best thermal efficiency is the one that has the highest boiling point among the selected fluids (R113, Tbp = 47.59C), while the fluid with the worst thermal efficiency has the lowest boiling point temperature (Propane, Tbp = -42.09C). Therefore it can be concluded that the higher the boiling point temperature of the organic fluid the better the thermal efficiency that will be achieve by the ORC. For the different scenarios analyzed in this investigation, ORC using R113 show the best thermal efficiency while those using Propane show the worse efficiencies. However, it is
206
important to point out that some organic fluids show better performance within a range of temperature. Therefore, designers have to closely monitor the operation conditions in order to select the right organic fluid.
NOMENCLATURE
h & I & m q & Q S s T TBP TH TL
T & W
specific enthalpy, kJ/kg Irreversibility rate, kW mass flow rate, kg/s efficiency, % specific heat, kJ/kg heat rate, kW entropy, kJ/K specific entropy, kJ/kg K temperature, K Boiling Point Temperature, C temperature of the high-temperature reservoir, K temperature of the low-temperature reservoir, K temperature differential, K power, kW condenser cycle evaporator exit conditions generator second law isentropic case inlet conditions pump ambient turbine
Subscripts: c cycle e exit generator II ideal inlet p o t
REFERENCES
Bejan, A., Tsatsaronis, G., and Moran, M. Thermal Design and Optimization, New York, USA: John Wiley & Sons; 1996. Cong, C.E., Velautham, S., Darus, A. M. Sustainable Power: Solar thermal driven organic rankine cycle, Proceedings of the International Conferences on Recent Advances in Mechanical and Materials Engineering, Paper No. 91, pp. 424-429, Kuala Lumpur, Malaysia, May 2005. Defibuagh, D.R., Gillis, K.A., Moldover, M.R., Schmidt, J.W., and Weber, L.A. Thermodynamic properties of CHF(2)-CF(2)-CH(2)F, 1,1,2,2,3-pentafluoropropane. International Journal of Refrigeration, 1996 19(4), 285-294.
207
Defibaugh, D.R. and Moldover, M.R. Compressed and saturated liquid densities for 18 halogenated organic compounds. Journal of Chemical Engineering, 1997 42(1),160-168. Gurgenci, H. Performance of power plants with organic rankine cycles under part-load and off-design conditions. Solar Energy, 1986 36(1), 45-52. Huber, M.L. and Ely, J.F. A predictive extended corresponding states model for pure and mixed refrigerants including an equation of state for R134a. International Journal of Refrigeration, 1994 17, 8-31. Hung, T.C., Shai, T.Y., and Wang, S.K. A review of organic rankine cycles (ORCs) for the recovery of low-grade waste heat. Energy, 1997 22(7), 661-667. Hung, T.C. Waste heat recovery of organic rankine cycle using dry fluids. Energy Conversion & Management, 2001 42, 539-553. Larjola, J. Electricity from waste heat using the organic rankine cycle (ORC). International Journal of Production Economics, 1991 41, 227-235. Larjola, J., Sarkomaa, P., and Suistoranta, S. New technology ORC-plant for heat recovery of diesel engines. 17th International Congress on Combustion Engines, June 8-11, 1987, Paper D-12, Warsaw (CIMAC87). Lee, M.J., Tien, D.L., and Shao, C.T. Thermophysical capability of ozozno safe working fluids for an organic rankine-cycle system. Heat Recovery System & CHP, 1993 13, 409418. Lee, K.M., Kuo, S.F., Chien, M.L., and Shih, Y.S. Parameters analysis on organic rankine cycle energy recovery system. Energy Conversion and Management, 1988 28(2), 129136. Liu, B.T, Chien, K.H, Wang, C.C. Effect of working fluids on organic rankine cycle for waste heat recovery, Energy, 2004 29, 12071217. Maizza, V., and Maizza, A. Working fluids in non-steady flows for waste energy recovery systems. Applied Thermal Engineering, 1996 16(7), 579-590. Maizza, V., and Maizza, A. Unconventional working fluids in organic rankine-cycles for waste energy recovery systems. Applied Thermal Engineering, 2001 21(3), 381-390 Mago, P.J., Chamra, L.M., and Somayaji, C. Analysis and optimization of organic rankine cycles. IMechE Journal of Power and Energy, 2007 221 (3), pp. 255-263. Marx, V., Pruss, A. and Wagner, W. Neue zustandsgleichungen fuer R 12, R 22, R 11 und R 113. beschreibung des thermodynamishchen zustandsverhaltens bei temperaturen bis 525 K und druecken bis 200 MPa. Duesseldorf: VDI Verlag, 1992, Series 19 (Waermetechnik/Kaeltetechnik), No. 57. Moran, M., and Shapiro, H. Fundamentals of Engineering Thermodynamics. 5th Edition. New York, USA: John Wiley & Sons; 2003. Niggeman, R.E., Greenlee, W.J., and Lacey, P. Fluid selection and optimization of an organic rankine cycle waste heat power conversion system. ASME 78-WA, January 6, 1978. REFerence fluid PROPerties (REFPROP), National Institute of Standards and Technology (NIST), Version 7.0. Tillner-Roth, R. and Baehr, H.D. An international standard formulation of the thermodynamic properties of 1,1,1,2-tetrafluoroethane (HFC-134a) covering temperatures from 170 K to 455 K at pressures up to 70 MPa. Journal of Physical Chemistry,1994, Ref. Data 23, 657729. Verschoor, M.J.E. and Brouwer, E.P. Description of the SMR cycle, which combines fluid elements of steam and organic rankine cycle. Energy, 1995 4(20), 295-304.
208
Vijayaraghavan, S., and Goswami, D.Y. Organic working fluids for a combined power and cooling cycle. ASME Journal of Energy Resources Technology, 2005 127, 125-130. Yamamoto, T., Furuhata, T., Arai, N., and Mori, K. Design and testing of the organic rankine cycle. Energy, 2001 26(3), 239-251. Younglove, B.A. and McLinden, M.O. An international standard equation-of-state formulation of the thermodynamic properties of refrigerant 123 (2,2-dichloro-1,1,1trifluoroethane). Journal of Physical Chemistry, 1994, Ref. Data 23, 731-779. Younglove, B.A. and Ely, J.F. Thermophysical properties of fluids. II. Methane, ethane, propane, isobutane and normal butane. Journal of Physical Chemistry, 1987, Ref. Data 16, 577-798.
Chapter 6
REAL TIME CONGESTION MONITORING AND MANAGEMENT OF POWER SYSTEMS

Jizhong Zhu*
AREVA T&D Corporation, 10865 Willows Road, NE Redmond, WA 98052, USA
ABSTRACT
Power systems operation data could include a large list of variables. The power system uncertainties as well as restructuring are introducing new class of dynamic phenomena into the power grid and substantially increasing the data requirements for real time congestion monitoring and management (RTCMM). Generating units and customer load respond to market signals that are affected by grid conditions, such as congestion. One of the most visible impacts is the increased volatility in the system, which results in the increased volumes of data to manage. The increasingly severe data overwhelm problem results in many challenges in terms of development of effective analytical and software tools. The information availability is also an issue. Secure operations under open access require the widespread availability of real time information about physical, as distinct from financial, variables. Unfortunately, the competitive environment is not conductive to information sharing. For example, data for state estimation is becoming increasingly difficult to obtain because of some uncertain factors or the ability of that data to reveal sensitive financial information. Consequently, state estimation may be failed or couldnt be solved, which increases the difficult to analyze the behavior of real time system operations. Another challenge in real time congestion monitoring and management is to design real time controls to maintain the security and reliability of the grid under the market condition and the contingency condition as well as uncertain environment. The implementations of real time congestion monitoring and management are discussed in the chapter.
jizhong.zhu@areva-td.com
210
Jizhong Zhu
I. INTRODUCTION
The electric power industry is widely changing under the pressure of various driving forces and we can foresee that those changes will become even wider in the near future. This is mainly because of the impact of many uncertainty factors and external factors related to the environment of this industry. Therefore, modern power systems are facing many new challenges, owing to environment and market pressures, as well as more uncertainties or/and inaccuracies [1-4]. Environment pressure implies more loaded networks, market pressure increases competition, while uncertainty and inaccuracy increase the complexity of operation and planning. Consequently, these new challenges have huge and direct impacts on the security and reliability of power systems. Especially, August 14s collapse of the electric grid in 2003 caused the largest blackout in US history, knocking more than 100 power plants offline and leaving 50 million customers in the US and Canada without power. An extensive investigation on what happened (and why) began immediately after the blackout. The joint U.S. Canada task force led the effort, including support from the electric utility industry and several federal agencies such as the U.S. Department of Energy. The North American Electric Reliability Council (NERC) supported the task force. The overall blackout investigation team drew expertise from a large number of organizations, assembled into teams to address specific attributes of the blackout [5]. The related investigation reports were issued [6-10]. The old transmission systems that were built in the 60s or 70s are blamed for this outage. Certainly, it is time to upgrade and modernize the transmission system. But it is more important to predict and avoid ahead of real time risks to the existing grid, or help the analysis of the real time situation in order to minimize the damage or provide the useful information on how to upgrade the grid. In addition, several blackouts also happened in the other countries at the same year [11-12]. Thus, blackouts occur in the electric power grids for a variety of reasons that include lack of investments in transmission network, control equipment failure, human errors, atmospheric phenomena, etc. A number of discussions and analysis on the causes of the blackout and on how to improve the secure operation of power system were presented in the papers and reports. [5-12]. The new challenges in modern power systems and recent blackouts raise some high requirements for modern power systems operation and control, i.e. d) A stronger expectation from customers for higher reliability and quality of supply owing to the uncertainty factors as well as the increase of the share of electrical power in their overall energy consumption. e) More electricity exchanges across large geographical areas resulting from a greater cooperation in the electricity market and greater competition in the energy market. There also exist lots of uncertainties both in the electricity market and energy markets. f) Low production fuel cost and low price of electricity are needed in order to achieve the competitive strength in the energy market. Furthermore, we can guarantee or say only one thing with absolute certainty in the modern electrical power industry: we are living and working with many unknowns [2]. Especially in modern power system operation, the several inaccuracies and uncertainties will
Real Time Congestion Monitoring and Management of Power Systems
211
lead to deviation from operation and planning. These are mainly on the one hand the inaccuracies and uncertainties in the input information, which is needed by the operation and planning and on the other hand the modeling and solution inaccuracies. Therefore, it is very important to analyze the uncertainty in modern power system operation and to use the available controls to make sure the security and reliability of power systems. The real time congestion monitoring and management (RTCMM) of power systems is for this purpose. The RTCMM system will be a collection of analysis, processes, computing equipment, measurement devices and communications that are assembled to provide the means to maintain a secure system state. The objectives of RTCMM are to:
identify key issues of concern in real time congestion monitoring and management Identify the issues and define the research scopes to meet the needs of real time secure operation and transmission loading relief. Analyze and evaluate possible strategies for effectively meeting the challenges in RTCMM. Implement visualization based on the obtained information from the RTCMM system.
II. CHALLENGES IN RTCMM

Power systems operation data could include a large list of variables. The power system uncertainties as well as restructuring are introducing new class of dynamic phenomena into the power grid and substantially increasing the data requirements for RTCMM. Generating units and customer load respond to market signals that are affected by grid conditions, such as congestion. One of the most visible impacts is the increased volatility in the system, which results in the increased volumes of data to manage. The increasingly severe data overwhelm problem results in many challenges in terms of development of effective analytical and software tools. The information availability is also an issue. Secure operations under open access require widespread availability of real time information about physical, as distinct from financial, variables. Unfortunately, the competitive environment is not conductive to information sharing. For example, data for state estimation is becoming increasingly difficult to obtain because of some uncertain factors or the ability of that data to reveal sensitive financial information. Consequently, state estimation may be failed or couldnt be solved, which increases the difficult to analyze the behavior of real time system operations. Another challenge in RTCMM is to design real time controls to maintain the security and reliability of the grid under the market condition and the contingency condition. The focuses of RTCMM are:
Analysis and handling of the uncertainties in power systems. Developing new performance indices for congestion monitoring or assessment. Developing new indices for system stability and reliability assessment. Analysis of available transmission capacity (ATC) of system. Improving the transmission reliability through real time control and thereby increasing ATC.
212
Jizhong Zhu
Assessment and selection of the available real time controls for ensuring the security and reliability of the system. Pareto analysis and cost-benefit analysis (CBA) of the selected control strategies. Determining the best control for transmission load relief and reliability enhancement based on the benefit-cost-ratio (BCR). Data visualization for demonstrating the analysis results.
III. DEFINITION OF UNCERTAINTY

Planning process of the regulated utilities does not capture the uncertainties in the operation and planning of power systems. Especially, the factors of uncertainties are increasing as the utility industry undergoes restructuring. There are two kinds of uncertainties in power systems operation and planning [4]: 4) Uncertainty in a mathematical sense, which means difference between measured, estimated values and true values, includes errors in observation or calculation; 5) Sources of uncertainty, including transmission capacity, generation availability, load requirements, unplanned outages, market rules, fuel price, energy price, market forces, weather and other interruption, etc. These uncertainties will affect power systems planning and operation on the following aspects:
Entry of new energy producing / trading participants Increases in regional and intra-regional power transactions Increasingly sensitive loads New types and numbers of generation resources
IV. APPROACHES TO HANDLE UNCERTAINTY

Since the uncertainties affect power systems planning and operation, it is therefore very important and urgent to investigate the uncertainties and to develop efficient and reliable methods to reflect them in security assessment, reactive power planning and generator scheduling problems in power systems.
4.1. Uncertainty Power Flow Analysis

In the general power flow analysis, the input variables to the power flow problem are assumed deterministically known. The practical operation conditions with uncertainty factors are not considered. Consequently, the power flow results may not reflect the real status of
213
system operation. This limitation will be overcome if a probabilistic approach or a fuzzy approach is applied. Probabilistic power flow (PPF) provides the complete spectrum of all probable values of output variables. Like bus voltages and flows, with their respective probabilities taking into account generation unit unavailability, load uncertainty, dispatching criteria effects and topological variations [13-15]. But PPF calculation is mainly based on the mean and variance of uncertain variables. These still exists some limitations in PPF. Fuzzy power flow may be a better method to deal with the uncertainty in PPF. Especially, it is needed to develop a fuzzystochastic power flow (FSPF) method. PSPE will be an appropriate technique to study uncertainty power flow because in the practical power systems, some load is fuzzy one in some case [16], and some load is stochastic [14], or some load is both fuzzy and stochastic.
4.2. Steady-State Security Regions with Uncertainty

In steady-state security analysis, a large number of repeated load flow studies must be performed. The situation will be even more severe when load changes need to be considered. It can also be anticipated that with the uncertainty of the amount and location of the loads in the system, the complexity of the security assessment problem and required computation burden will be increased greatly. To reduce the computation amount of the steady-state security assessment, the approach of steady-state security regions is a good one for this aim [17-24]. The previous security regions, which were formed by using valid system operation constraints, were found to be too implicit and difficulty to be used in power systems operation. In order to overcome the disadvantages of the earlier methods, Wu et al [22] suggested a hyper-box to approximately express the steady-state security regions. However, Wus security region was too conservative. Liu improved the hyper-box approach having developed an expanding method to obtain the hyper-box, which tends to the maximal region [23]. The expanding speed, however, was very slow due to the adoption of a fixed expanding step. Moreover, the N-1 security constraints and uncertainty factors have not been considered in Lius work. This means that the obtained security region based on Lius approach is either impractical or still conservative. A new approach to compute the maximal steady-state security regions of power systems is presented by Zhu [24]. The method is based on fuzzy theory and optimization approach. The fuzzy branch flow constraints and N-1 security are also introduced in that paper. The maximal steady-state security region hyper-box is directly computed through a linear programming (LP) model, in which the upper and lower limits of each component forming hyper-box are taken as unknown variables, and the objective is to maximize the sum of generators power adjustment ranges.
4.3. Optimal VAR Planning with Uncertain Loads

The objective of optimal reactive power (VAR) planning is to minimize power losses with maintaining a desired level of voltages and VAR flows under various operating conditions. A number of control options such as generators, tap changing transformers, shunt
214
Jizhong Zhu
capacitors / reactors, synchronous condensers and static VAR compensators may be considered. Conventional approaches for the planning problems are based on non-linear programming (NLP) and linear programming (LP) but both have problems to deal with the limits of the variables [25-34]. In addition, the conventional methods do not take into account the fact that, in practical power systems operational planning, the data needed for VAR planning such as the load demands, voltage level at network nodes, the amounts of power generation of the generators and the operation limits of equipment contain uncertainty. For getting better VAR planning results, the uncertainty must be taken into account. Some researchers have recently carried out work along this line [16, 25]. However, their methods are time consuming and much more work needs to be performed. Up to now, most of VAR planning or VAR control methods were designed for on-line application, for which convergences reliability, processing speed and memory storages are the prime concern. For convenience, these methods are usually confined to the fixed VAR compensation locations, and may not be suitable for large-scale power systems with strict constraints and serious VAR shortage. Thus, it is needed to analyze the characteristics of uncertain reactive load in the practical power systems using fuzzy set theory, and/or stochastic methods. In order to obtain optimal VAR planning under the uncertain and /or inaccuracy environment, this chapter presents an idea: conduct the ranking of VAR compensation locations based on the fuzzy set theory, voltage/VAR sensitivities, and the voltage stability margin method. At the same time, the relative importance of different load sites will be considered in the ranking process. Due to their independent nature, results from these methods are not necessarily the same. Of course, these methods either reflect weak nodes in system voltages or the relative importance of different load nodes. But it is difficult to find a unified process for ranking results from these methods, especially some factors are uncertainty and not quantitative. The analytic hierarchical process (AHP) provides a useful means to consider comprehensively these factors in the selection and ranking of optimal VAR locations. AHP is a simple and convenient method to analyze a complicated problem (or a complex system). It is especially suitable for problems which are very difficult to analyze wholly or quantitatively, and is now widely used in society and economy areas [34-37]. It was first used in power systems area for solving combined active and reactive dispatch in 1996 [35]. AHP can help to quantify the decision-makers thinking. It can be used to compute a unified ranking for the decision problem of optimal VAR locations.
4.4. Unit Commitment and Economic Dispatch with Uncertainties

In power systems operation and control, the economy of power production, security and quality of power supply are important aspects and needed to be considered. The first aspect is related to the problems of unit commitment and economic dispatch. The objective of unit commitment is to schedule generator units on-line and off-line to meet the electric load demand in a power system. The main purpose of the optimal generation dispatch has so far been confined to minimize the total generation cost of the power system under the conditions of network security. Different methods have been proposed to reach this purpose [38-41]. However, the economy of short-term operation of thermal power systems is influenced by approximations in the operation planning methods and by the inaccuracies and
215
uncertainties of input data. At present, the uncertainty problem in economic operation has not been paid enough attention. (a) Uncertain loads The forecast loads are important input information, which are characterized by uncertainty and inaccuracy because of the stochastic nature of the load. (b) Inaccuracy fuel cost function Inaccuracy in the process of measuring or forecasting of input data. Change of unit performance during the period between measuring and operation. The inaccuracies in the cost functions for steady-state operation are caused by the limited accuracy of the determination of the thermal dynamic performance, changing cooling water temperatures, changing calorific values and contamination, erosion and attrition in boiler and turbine. These deviations lead to inaccurate values for heat inputs and fuel prices. (c) Inaccurate startup cost function In addition to the above mentioned inaccuracies, there is the uncertainty about the loss of unit life caused by heating and cooling of turbine shaft for the startup cost function. All factors taken together can mean inaccuracies of several percent in the steady-state operation cost functions and more than 10 percent in the startup cost functions. Therefore, it is necessary to develop a comprehensive fuzzy-stochastic method to study the uncertainty in system economic operation. Fuzzy theory, refined artificial intelligence (AI) approach are also needed for analyzing accurately real-time power systems operation. Because some loads in practical systems are both fuzzy and stochastic, it is very important to develop a fuzzy-stochastic optimization economic dispatch (FSED) method to conduct the calculation of power systems economic operation with uncertain loads.
4.5. Hydro-Thermal System Operation with Uncertainty

There are several complex and interrelated problems associated with the optimization of hydro-thermal systems. Long term regulation problem (1 to 2 year optimization period); Intermediate term hydro-thermal control (1 month to 6 months planning period); Short term hydro-thermal dispatch (optimization period is from 1 day to 1 week) For the short term optimization problem, the applications of deterministic methods to hydro-thermal system operation have been established, in which the water inflows and loads were considered to be deterministic. For the long term regulation problem, it is necessary to use a stochastic representation for the load and river inflow [42, 43]. Since there are the uncertainty factors in the short term hydro-thermal dispatch, the existing methods do not provide the system operators with a convincing answer on how to use the water in each separate reservoir. The following uncertainties should be taken into account in a large hydrothermal system operation.
216
Jizhong Zhu
Uncertainty of the loads Uncertainty of the unit availability Uncertainty of the river inflow
The uncertainty of the river inflows, loads and unit availability can be dealt with in a stochastic representation. Therefore, it is good to develop a new fuzzy-stochastic optimization method. In addition, the uncertainty problem for the hydro-thermal system operation is very complicated, the fast optimization method or the refined AI approach is recommended.
V. IMPLEMENTATION OF RTCMM
The blackout has provided the potential problem and new information about the electric power system. Meanwhile, it also shows the complication of real-time power system operation. Electric system monitoring and management must be responsible to make any and all efforts necessary to assure the highest level of safety and reliability of real-time system operation. According to recent research achievements [4456], the proposed implementations of RTCMM are as follows.
Step 2 for SE calculation involving whole network (Areas without measurement)
tie-line Step 1 for SE including only Areas with measurements
tie-line
Figure 1. Two-step real-time state estimation (SE) calculation.
217
5.1. Model Reduction or Equivalent

Fast and valid state estimation calculation depends on the available data and good model. Unavailable data or overwhelm data problem as well as the bad power system model easily cause the failure of state estimation calculation. The frequent SE failures will let operator lost first chance to know the system operation state change, which may hurt the grid operation. Thus, it is important to construct a valid power system model. For example, some ISO system model in the United States covers 4kv level and many of these levels data are not available. The calculation of state estimation may easily be diverged due to the low ratio of measurable data. So the model reduction or equivalent that focuses on measurable areas is an option. In this case, two-step real-time state estimation calculation may be applied that is shown in Figure 1, where the equivalent is conducted based on tie-line information that is computed from step one state estimation calculation. This is particularly useful for the large power network with the limited measurements.
5.2. Data Analysis

In modern power system operation, the several inaccuracies and uncertainties will lead to deviation from the real time operation. There are two kinds of uncertainties in power system operation mentioned in section III:
Uncertainty in a mathematical sense, that is means difference between measured, estimated values and true, practical values, includes errors in observation or calculation. Sources of uncertainty, including transmission capacity, generation availability, load requirements, unplanned outages, market rules, fuel prices, energy price, market forces, weather and other interruption, etc.
Table 1. Deterministic vs probabilistic methods
Methods Comparison Contingency Selection Contingency Probabilistic Load Levels (forecast) Unit Commitment Security Regions Criteria for Decision
Deterministic Method Typically a few probable and extreme contingencies Based on judgment Typically seasonal peaks and selected off-peak loads Traditional optimization technology Deterministic security region Well established
Probabilistic Method More exhaustive list of contingencies; Ranking based on Fuzzy / AHP methods Based on inadequate or uncertain data (ANN, Fuzzy and AHP methods) Multiple levels with uncertain factors (Fuzzy, ANN) Optimization technology & AHP / ANN Variable security regions Need a suitable method / criteria to make decision (ANN, Fuzzy and AHP methods)
218
Jizhong Zhu
It is impossible to get all available data in the real time operation due to the abovementioned uncertainties of power systems and competitive environment. Nevertheless, it is important to select an appropriate technique to handle these uncertainties. The existing deterministic methods and tools are not adequate to handle them. The new probabilistic methods such as Gray Mathematics, Fuzzy theory, and analytic hierarchy process (AHP) are very useful to compute the unavailable or uncertain data; so that the state estimation can be solvable even some key data are not available. The comparison of deterministic method and probabilistic method is listed on Table 1. Through comparing the various approaches, the following methods to handle uncertainties are recommended:
Characterization and probabilistic methods Probabilistic methods / tools for evaluating the contingencies Fuzzy or AHP (analytic hierarchy process) method to handle uncertainties (e.g. contingency ranking) Risk management tools to optimize energy utilization while maintaining the required levels of reliability Cost-benefit-analysis (CBA) for quantifying the impact of uncertainty
5.3. Topology Error Detection

Generally, blackouts often occur accompanied by lost of generators or loads as well as network topology change. The topology structure changes cause the most serious outages. Thus, it is very important to detect the network topology change in the real time power system operation in order to prevent blackout [44]-[46]. The topology error detection combined the identifying the bad data and sudden loads changes should be the correct way.
5.4. Congestion Monitoring

The proposed congestion monitoring contains the function of the traditional security monitoring. The purpose of congestion monitoring is not only to look ahead for possible contingencies that may transition the power system away from a normal operation state, but also analyze the data changes, which are from the state estimator, that may cause potential outage during the normal operation state. Thus, the several new indices for congestion analysis as well as system stability and reliability assessment will be developed in RTCMM. These indices will be used for ATC evaluation and improving transmission stability and reliability. It means that the proposed ATC calculation will consider the angle stability, voltage stability and reliability constraints [47].
219
5.5. Congestion Management

Once some operational limit is reached, some real time control may have to move variables away from the values in the normal operation state or in the agreements made in the market since the secure and reliable operation is first important factor. In this case, the congestion management should determine the minimum impact procedure to allow the proposed transactions to be undertaken. According to the recent research [48] [56], the possible real time controls in congestion management of RTCMM are: 1) Generation rescheduling: Optimal distribution of power flow among the transmission network; Loss minimization and thereby increase transmission capacity. 2) VAr Support and optimization: Improve system voltage profile; Loss minimization and increase transmission capacity and thereby improve system reliability. 6) Real-time OPF with phase shifter control: Optimal distribution of power flow among whole system network; Improve system voltage security; Change the direction of flow on the congested transmission lines and thereby enhance system security and reliability. 7) Stability controls: Voltage stability analysis; Dynamic stability analysis 8) The other remedial controls: Unit commitment; Change of operation pattern/scheme. 9) Optimal load shedding: when the above-mentioned controls are unable to maintain the security of system operation during the disturbance, optimal load shedding will be last action to make the loss of blackout minimum. The idea is to classify the system loads, and different load type has the different priority to supply. Different load area also has the different weighting factor. Different load site also has different relative importance, which is non-quantitative constraint. The proposed approach is to determine the optimal load shedding scheme through computing a unified index considering the quantitative and non-quantitative constraints. The AHP and Everett optimization methods or network flow program can be used for this purpose [55], [56].
5.6. Benefits from RTCMM

If RTCMM is implemented, the big blackout will be avoided. At least, if the optimal load shedding were applied at the beginning of disturbance, August 14s blackout wouldnt have happened in the several states. The benefits from RTCMM are:
220
Jizhong Zhu
increase system security; enhancement of reliability; enhancement of system stability; improve system voltage profile; increase capacity of the system; loss minimization savings; economic dispatch saving; unit commitment saving; reduction of outage loss cost; Reduction of zone importation cost.
5.7. Cost Benefit Analysis

Security is always ranked as first in the real time power system operation. Under the system secure operation, the cost benefit analysis will be performed for the above-mentioned all kinds of controls according the benefits to cost ratio for each control, so that the best control or the combined controls can be selected and applied to real time system operation to avoid the blackout or maintain the security and reliability of power system.
5.8. Essentials and Data Visualization

(1) Organize all relevant system cost data and calculate the new congestion, reliability, and stability indices; (2) Identify all potential benefits derived from real time control; (3) Calculate the value for any number of system controls; (4) Conduct both deterministic and probabilistic sensitivity and risk analysis of the results; (5) Data visualization for the analysis results. Use of animation and contour with different colors can produce visually appealing display of bus voltages/angles, line flow, and line loading to point out congestion or overloads, so that the operator could quickly assess how the system state has been changing over a specified time period. The proposed implementation strategy of real time congestion monitoring and management (RTCMM) of power systems is shown in Figure 2.
VI. CONCLUSION
This chapter proposes an approach of power systems real time congestion monitoring and management (RTCMM) to predict and avoid ahead of real time risks to the existing grid, or help the analysis of the real time situation in order to minimize the damage or provide the useful information on how to upgrade the grid. The chapter analyzes the challenges of the RTCMM. The proposed RTCMM system will be a collection of analysis, processes,
221
computing equipment, measurement devices and communications that are assembled to provide the means to maintain a secure system state. The implementations of real time congestion monitoring and management are discussed in the chapter.
Data measurement
Handle unavailable or uncertain data
Data error check
Yes Incorrect data? No Topology error check
Topology change? No Yes Analyze Topology and identify whether it is network outage
Correct the error
If it is not an outage but caused by data error
If it is an outage or system is congested, the best control is recommended based on cost-benefit analysis
Perform state estimation (SE) calculation
Issue Alarm to remind the operator
Visualization for result analysis
Figure 2. The implementation strategy of RTCMM of power systems.
222
Jizhong Zhu
VII. REFERENCES
[1] A.S. Merlin, Latest developments and future prospects of power system operation and control, International journal of electrical power and energy system, Vol. 16, No.3, pp137-139, 1994. N. Rau, C.C. Fong, C.H. Grigg, and B. Silverstein, Living with uncertainty, IEEE Power Engineering Review, November, pp24-26, 1994. K.H. Abdul-Rahman, S.M. Shahidehpour, and N.I. Deeb, Effect of EMF on minimum cost power transmission, IEEE Trans. On Power Systems, Vol.10, No.1, pp347-353, 1995. M. Ivey, Accommodating uncertainty in planning and operation, Workshop on Electric Transmission Reliability, Washington, DC, Sept. 17, 1999. J. E. Dagle, Data management issues associated with the August 14, 2003 Blackout investigation. Invited paper for IEEE/PES Panel on Major Grid Blackouts of 2003 in North America and Europe, IEEE PES General Meeting, Denver, CO, June 6-12, 2004. W. Sweet, Special Report: The Blackout of 2003. IEEE Spectrum, August 21, 2003. U.S.-Canada Power System Outage Task Force (2003, Nov.). Interim Report: Causes of the August 14 blackout in the United States and Canada. [Online]. Available: https://reports.energy.gov/. U.S.-Canada Power System Outage Task Force (2003, Sept.). Initial Blackout Timeline: August 14, 2003 outage sequence of Events. [Online]. Available: https://reports.energy.gov/. U.S.-Canada Power System Outage Task Force (2004, Apr.). Final Report on the August 14th Blackout in the United States and Canada. [Online]. Available: https://reports.energy.gov/. P. Fairley, The Unruly Power Grid: Advanced mathematical modeling suggests that big blackouts are inevitable. IEEE Spectrum, July, 2004. N. Dizdarevic, M. Majstrovic, N. Mandic, S. Cujic Coko, "Blackout from the system operator's perspective," Proceedings of the 2004 IEEE PES Power Systems Conference and Exposition (PSCE'04), New York, USA, October 2004. N. Dizdarevic, M. Majstrovic, and J. Benovic, "Activation of special protection system after blackout," Proceedings of the 2004 Bulk Power System Control and Dynamics VI, Cortina d'Ampezzo, Italy, August 2004. M.E. El-Hawary, and G.A.N. Mbamalu, A comparison: probabilistic perturbation and deterministic based optimal power flow solutions, IEEE Trans. On Power Systems, Vol.6, No.3, pp1099-1105, 1991. P.W. Sauer, and B. Hoveida, Constrained stochastic power flow analysis, Electric Power Systems Research, Vol.5, pp87-95, 1982. T.S. Karakatsanis, and N.D. Hatziargyriou, Probabilistic constrained load flow based on sensitivity analysis, IEEE Trans. On Power Systems, Vol.9, No.4, pp1853-1860, 1994. K.H. Abdul-Rahman, S.M. Shahidehpour, and N.I. Deeb, AI approach to optimal VAR control with fuzzy reactive loads, IEEE Trans. On Power Systems, Vol.10, No.1, pp88-97, 1995.
[2] [3]
[4] [5]
[6] [7]
[8]
[9]
[10] [11]
[12]
[13]
[14] [15]
[16]
Real Time Congestion Monitoring and Management of Power Systems [17]
223
[18] [19]
[20]
[21] [22] [23] [24] [25] [26]
[27]
[28] [29] [30]
[31]
[32]
[33] [34]
J.Z. Zhu, and C.S. Chang, A new approach to steady-state security regions with N and N-1 security, in Proc. International Power Engineering Conference, IPEC97, Singapore, 1997. E. Hnhilicza, S.T. Lee, and F.C. Schweppe, Steady-state security regions: settheoretic approach, in Proc. PICA Conference, pp347-355, 1975. R. Fischl, G.C. Ejebe, and J.A. DeMaio, Identification of Power System Steady-State Security Regions under Load Uncertainty, IEEE Summer Power Meeting, Paper A76 495-2. IEEE Trans on PAS Vol.95, No.6, p1767, 1976. J.A. DeMaio, and R. Fischl, Fast Identification of the Steady-State Security Regions for Power System Security Enhancement, IEEE Winter Power Meeting, A.76076-0. IEEE Trans. on PAS, Vol.95, No.3, p758, 1976. M.H. Banakar, and F.D. Galiana, Power System Security Corridors Concept and Computation, IEEE Trans., PAS, Vol.100, pp4524-4532, 1981. F.F. Wu, and S. Kumagai, Steady-State Security regions of Power System, IEEE Trans., CAS, Vol.29, pp703-711, 1982. C.C. Liu, A new method for construction of maximal steady-state regions of power systems, IEEE Trans. PWRS, Vol. 1, No. 4, pp19-27, 1986. J.Z. Zhu, Optimal power systems steady-state security regions with fuzzy constraints. Proceedings of IEEE Winter Meeting, New York, Jan. 27-30, 2002. K. Tomsovic, A fuzzy linear programming approach to the reactive power / voltage control problem, IEEE Trans. on Power Systems, Vol. 17, No. 1, pp. 287-293, 1992. J. Z. Zhu, W. Yan, C. S. Chang, and G. Y. Xu, Reactive power optimization using an analytic hierarchical process and a nonlinear optimization neural network approach, IEE Proceedings: Generation, Transmission and Distribution. Vol. 145 No.1, pp8996, 1998. A.D. Papalexopoulos, C.F. Imparato, and F.F. Wu, Large Scale Optimal Power Flow: Effects of Initialization Decoupling and Discretization, IEEE Trans. on Power Systems, Vol. PWRS-4, No. 2, pp. 748-759, May 1989.. O. Alsac, J. Bright, M. Prais, and B. Stott, Further Developments in LP-Based Optimal Power Flow, IEEE Trans. on Power Systems, Vol. 2, pp. 697-711, 1990. M.R. Irving, and, M.J.H. Sterling, Efficient Newton-Raphson algorithm for load flow calculation in transmission and distribution networks, IEE Proc. C, Vol.134, 1987. M.O. Mansour, and T.M. Abdel-Rahman, Non-linear VAR Optimization Using Decomposition and Coordination, IEEE Trans. on Power Systems, Vol. 103, pp. 246255, 1984. N.H. Dandachi, M.J. Rawlins, O. Alsac, and B. Stott, OPF for Reactive Pricing Studies on the NGC System, IEEE Power Industry Computer Applications Conference, PICA95, Utah, May 1995, pp. 11-17. D. Chattopadhyay, K. Bhattacharya, and J. Parikh, Optimal Reactive Power Planning and Its Spot-pricing: An Integrated Approach, IEEE PES Winter Meeting, 95 WM 204-8 PWRS, 1995. M.L. Baughman, and S.N. Siddiqi, Real-time Pricing of Reactive Power: Theory and Case Study Results, IEEE Trans. on Power Systems, PWRS, Vol. 6, pp. 23-29, 1991. T.L. Satty, The Analytic Hierarchy Process, McGraw Hill, Inc., New York, 1980.
224 [35]
Jizhong Zhu J.Z. Zhu, and M.R. Irving, Combined Active and Reactive Dispatch with Multiple Objectives using an Analytic Hierarchical Process, IEE Proceedings, Part C, Vol.143, pp344-352, 1996. J.Z. Zhu, M.R. Irving, and G.Y. Xu, Automatic contingency selection and ranking using an analytic hierarchical process, Electrical Machines and Power Systems Journal, No.4, 1998. J.Z. Zhu, and J.A. Momoh, Optimal VAR pricing and VAR placement using analytic hierarchy process, Electric Power Systems Research, Vol.48, No.1, pp11-17, 1998. O. Alsac, and B. Stott, Optimal power flow with steady-state security, IEEE Trans. on Power Systems, Vol. 93, pp. 745-751, 1974. H.W. Dommel, and W.F. Tinney, Optimal power flow solution, IEEE Trans. on Power Systems, Vol. 93, pp. 745-751, 1974. J.Z. Zhu and G.Y. Xu, A new economic power dispatch method with security, Electric Power Systems Research, Vol.25, No.1, pp9-15, 1995. T.H. Lee, D.H. Thorne, and E.F. Hills, A transportation method for economic dispatching application and comparison, IEEE Trans. on Power Systems, Vol. 99, pp. 2372-2385, 1980. T.S. Dillion, and T. Tun, Integration of the sub-problems involved in the optimal economic operation of hydro-thermal system, Proc. IFAC Symp. Contr. And Manag. of Integ. Indust. France, Sept. 1977, pp171-180. T. Tun, and T.S. Dillion, Sensitivity analysis of the problem of economic dispatch of hydro-thermal system, Proc. IFAC Symp. Autom. Contr. And Prot. Of Elec. Power System, Melb. 1977. J. C. S. Souza, A. M. Leite da Silva, and A. P. Alves da Silva, Data debugging for real-time power system monitoring based on pattern analysis, IEEE Transactions on Power Systems, Vol.11, No.3, pp1592-1599, 1998. J. C. S. Souza, A. M. Leite da Silva, and A. P. Alves da Silva, On-line topology determination and bad data suppression in power system operation using artificial neural networks, IEEE Transactions on Power Systems, Vol.13, No.3, pp796-803, 1998. L. Mili, G. Steeno, F. Dobraca, D. French, A robust estimation method for topology error identification, IEEE Transactions on Power Systems, Vol.14, No.4, pp14691476, 1999. J. A. Momoh and J. Z. Zhu, A new method to congestion analysis for aerospace power system, Proceedings of IEEE 29th North American Power Symposium, Wyoming, USA, Oct. 12-14, pp479-485, 1997. J. Z. Zhu and G. Y. Xu, A unified model and automatic contingency selection algorithm for P- and Q-Subproblem, Electric Power Systems Research, Vol.32, No.2, 1995, pp101-105. L. S. Vargas, V. H. Quintana, and A. Vannelli, A Tutorial Description of an Interior Point Method and its Application to Security-Constrained Economic Dispatch, IEEE/PES 1992 Summer Meeting, 1992. J. A. Momoh and J.Z. Zhu, Optimal generation scheduling based on AHP/ANP, IEEE Trans. on Systems, Man & Cybernetics, Part B, Vol.33, No.3, pp531-535, 2003.
[36]
[37] [38] [39] [40] [41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
Real Time Congestion Monitoring and Management of Power Systems [51]
225
[52]
[53] [54]
[55] [56]
J. Z. Zhu, and J. A. Momoh, "Multi-area power systems economic dispatch using nonlinear convex network flow programming " Electric Power Systems Research, Vol.59, No.1, pp13-20, 2001 . J. A. Momoh, J. Z. Zhu, G. Boswell, and S. Hoffman, Power system security enhancement by OPF with phase shifter, IEEE Transactions on Power Systems, Vol.16, No.2, pp287-293, May 2001. K. Tomsovic, A fuzzy linear programming approach to the reactive power / voltage control problem, IEEE Trans. On Power Systems, Vol.7, No.1, pp287-293, 1992. C. W. Richter, and G. B. Sheble, A profit-based unit commitment GA for the competitive environment, IEEE Trans. On Power Systems, Vol.15, No.2, 715-721, 2000. J. Z. Zhu, Optimal load shedding using AHP and OKA. International Journal of Power and Energy Systems, Vol. 25, No.1, pp40-49, 2005. J. A. Momoh, J. Z. Zhu and S. S. Kaddah, Optimal load shedding study of naval-ship power system using the Everett optimization technique, Electric Power Systems Research, Volume 60, Issue 3, pp145-152, 28 January 2002.
Chapter 7
POWER SYSTEMS DISTURBANCE ANALYSIS

Abhisek Ukil*
ABB Corporate Research Segelhofstrasse 1K, CH-5405 Baden-Daettwil, Switzerland
ABSTRACT
The analysis of faults and disturbances has always been a fundamental foundation for a secure and reliable electrical power supply. The introduction of digital recording technology opened up a new dimension in the quantity and quality of fault and disturbance data acquisition, resulting in the availability of a huge amount of new information to power systems engineers. Information from the analysis of digital records can provide much-needed insight into the behavior of the power system as well as the performance of protection equipments. Manual analysis of these records, however, is both time-consuming and complex. Today the challenge is to automatically convert data into knowledge, which frees the human resources to implement corrective or preventive action. Therefore, much focus is laid on automatic disturbance analysis. This also goes towards the direction of black-out prevention which is a great threat from utility perspective. This chapter will provide the insight on the recent trends on power systems disturbance and fault analysis. Basics of disturbance analysis, Fourier transform and wavelet transform-based signal processing techniques in the power systems disturbance analysis are discussed in details.
1. OVERVIEW OF POWER SYSTEMS DISTURBANCE ANALYSIS

The analysis of faults and disturbances has always been a fundamental foundation for a secure and reliable electrical power supply. The introduction of digital recording technology opened up a new dimension in the quantity and quality of fault and disturbance data acquisition, resulting in the availability of a huge amount of new information to power
*
E-mail: abhisek.ukil@ch.abb.com, abhiukil@yahoo.com
228
Abhisek Ukil
systems engineers. Information from the analysis of digital records can provide much-needed insight into the behavior of the power system as well as the performance of protection equipment. Manual analysis of these records, however, is both time-consuming and complex. Today the challenge is to automatically convert data into knowledge, which frees the human resources to implement corrective or preventive action (Ukil 2005).
1.1. Necessity of Disturbance Analysis

The disturbance analysis is intended for the following personnel.
Operations The operating personnel are charged with restoring as much of the disturbed electrical system as practical within the shortest time possible. Fault location and operational correctness are first information that they consider in order to proceed. If a line operates one time and re-closes successfully, then all that is required is to forward the fault location to line personnel for later inspection. If the re-closing was not successful then the operating personnel should use the fault location as an indication of how to isolate the problem. If the fault location indicates line problems, the operator should then use that location to return to service as much of the line as possible. If the fault location indicates that some of the major equipments (e.g., circuit breakers, transformers, switches, etc.) have been involved, then the operator proceeds to isolate that equipment and return to service other equipments. Maintenance The information requirement here is entirely different. Maintenance personnel are charged with repairing, maintenance and returning to service of the outage equipments. Therefore, they require information concerning what is damaged or operating outside the normal parameters. Then maintenance personnel can proceed to correct any problems (e.g., downed conductors, failed lightning arrestors, breakers, or transformers). Normal time requirements for maintenance personnel depends on the way a utility operates, the type of equipment involved, and its criticality in serving load. This usually is typically in the region of two hours or less. Also, properly processed maintenance data can tell us when and what needs regular services, how often, which are the particular weak spots of the equipment, possible improvements and redesigns. Protection The protection engineer is responsible to investigate about the cause of the fault. Except in extremely rare cases of catastrophic failure, normally the protection engineer has enough time to collect all data necessary for a complete evaluation of the event. The main target would be to correct the system problem that caused an erroneous system response. If fault definition was quick and all parties involved get their respective information on time, the endusers will get electricity fast and interruptions will be minimal.
Power Systems Disturbance Analysis
229
2. DIGITAL RECORDING AND INSTRUMENTATION FOR DISTURBANCE ANALYSIS

The reason why weve made separate section for this recoding equipment is that this equipment is usually a part of protection system and, as such, it has quite different overview on distribution. Most of these devices existed before digital technology was developed. Digital equipments also, allow much easier integration of protection, monitoring and control systems with huge advantages and savings. The choice of recording equipment depends on many factors including the history of the substation construction and upgrades, utility operating practices, strategic importance of the substation, etc. The following types of equipments (Grisby 2007) are typically used in modern substations:
Digital protective relays (DPRs) Digital fault recorders (DFRs) Sequence of event recorders (SERs) Remote terminal units (RTUs) of a SCADA system Intelligent Electronic Devices (IEDs) Fault locators (FLs).
2.1. Digital Protective Relays (DPRs)

It is a complex recording and measurement instrument equipped with a decision making control logic and plenty of monitoring functions, a variety of settings (both user selectable and internal), number of internally computed measurement and logic signals that may be accessed by the user. The main problem in defining a generalized model of such a relay (which may be essential for the analysis) is due to the fact that the relay designs from various vendors may be quite different and levels of data access provided for the outside use may also be significantly different. Due to a relatively low sampling rate of some of the earlier DPRs their waveform recording function provides only a limited frequency representation of the waveforms.
2.2. Digital Fault Recorders (DFRs)

DFRs are highly accurate recording instruments providing sampled waveform and contact data using relatively high sampling rate, typically above 5 kHz (a sample every 0.2 milliseconds). Their use in the analysis is quite appropriate since they provide recordings of the waveforms that were also seen by the DPRs. Problem is that various DFRs provide different triggering mechanisms, and the performance and sensitivity of the triggers may affect the ability to capture relevant waveforms. Some of the newer DFR designs allow the user to program a customized triggering mechanism, which can assist the analysis. Further potential difficulties are associated with the DFR data formats that may be proprietary not allowing implementation of an open data recording system which can easily be interfaced with the analysis function.
230
Abhisek Ukil
2.3. Sequence of Event Recorders (SERs)

These are complex recording instruments implemented today most likely using programmable logic controllers (PLCs) and analog waveform data acquisition subsystems. The SERs are capable of monitoring changes in the switching equipment status with high precision due to a high data sampling rate. Combined with measurements of analog signals, the SERs can record the status change for variety of controllers including the ones that are based on analog set points. Most of the SERs can also be set to provide control function through a number of control outputs. For the analysis purposes, existence of SERs in a substation is very important. SERs also face the problem of not having an open system design, which may reduce an ability to interface the data recorded by those systems with the data from other sources used for the analysis. If available and accessible through standard communication interfaces, SERs may be very important recording infrastructure to be used for the analysis.
2.4. Remote Terminal Units (RTUs) of a SCADA System

The RTU can be a very sophisticated recording instrument that may have a recording performance of a DFR, and at the same time may be producing a variety of pre-calculated quantities. In addition, some advanced RTUs will provide an extensive SER and some limited DPR functionalities. Due to the fact that RTUs are a part of the Supervisory Control and Data Acquisition (SCADA) system, the data is readily available for the analysis at the centralized location through a SCADA database. However, the open system design remains an issue with the RTUs as well since they are primarily designed to interface to the EMS SCADA data base using mostly customized communication protocols and database formats. Another potential problem is related to a limited opportunity for the user to access the recorded data locally before it is sent to a centralized location. In the analysis hierarchy it may be desirable that some locally recorded data is available in a substation for a local analysis, and this may be difficult to implement due to a lack of local substation user interfaces in most of the RTU designs.
2.5. Intelligent Electronic Devices (IEDs)

The modern IEDs are available today for variety of applications ranging from simple stand alone controllers and dedicated data recording systems to pretty complex integrated devices for monitoring, control and protection of the entire substation. The main issue with IEDs, when used for the analysis, is the open communication architecture and data recording performance. Since the IEDs are not standardized even regarding the functions they perform, it may be very hard to find detailed enough description that will allow generic models to be developed and used for the analysis. Notwithstanding the limitations, the IEDs are indeed a good addition to the data recording infrastructure needed for a comprehensive analysis to be performed related to substation equipment operation.
231
2.6. Fault Locators (FLs)

The stand alone FLs are designed to provide very accurate fault location, but for that purpose they may have to have a fairly advanced built-in data acquisition system. It works based on the principle that cables have constant resistance per length. However, as most modern DFRs, DPRs and even RTUs already come up with fault location calculation as an additional functionality, the specific uses of FLs are getting restricted.
2.7. Current Transformers (CTs) and Voltage Transformers (VTs)

All this analog signal monitoring and protection equipment uses current transformers (CTs) and voltage transformers (VTs) to get current and voltage recordings. Therefore, it is very important to know where the CTs and VTs are placed. They proportionally transform line voltage and current into much lower value. Usually current goes to 1A, 5A and voltage to 110V, 220V. CTs are made in various ranges of accuracy and output levels for different functions and there are usually measuring and protection core. On normal feeders we have CTs on the line, their star point earthed line side. This has the effect that power flowing out of the busbar on this line during load conditions will reflect as in phase (depending on the power factor) with the voltage. Power flowing into the busbar will be reflected as 180 out of phase with the voltage. If there is a line reactor on the line, the CTs are most commonly located on the line side of the reactor. VTs are usually located on all the transmission feeders. The voltage on the line is thus reflected at both ends. If one end of the line is open, both ends will still see full voltage.
3. COMMON-MODE INTERFACE
We have seen in section 2 that one of the main challenges in modern disturbance analysis is having a common standard of data interface among the different instruments in the system. The variety of sources of transient data, such as DFRs, DPRs and transient simulation programs (e.g., EMTP (CEA Technologies 1998)) from different manufacturers using proprietary or different standard formats, made it necessary to introduce the IEEE Standard for Common Format for Transient Data Exchange (COMTRADE) for power systems as specified by the IEEE STANDARD C37.111-1991 (IEEE 1991). The standard files should be ASCII files. Each event should have three types of files associated with it. Each of the three types carries a different class of information: header (*.HDR), configuration (*.CFG) and data (*.DAT). The intent of the header (*.HDR) is to provide supplementary information in a narrative form for the user to better understand the conditions of the transient record. The header file is not intended to be manipulated by an applications program. The following elements should be included in the header file:
Description of the power system prior to disturbance Name of the station
232
Abhisek Ukil
Identity of the line, transformer, reactor, capacitor or circuit breaker that experienced the transient Length of the faulted line Positive and zero sequence resistances and reactances Capacitances Mutual coupling between parallel lines Locations and ratings of shunt reactors and series capacitors Nominal voltage ratings of transformer windings Transformer power ratings and windings connections Positive and zero sequence impedance of the source Description of how the data was obtained Description of the anti-aliasing filters used Description of analog mimic circuitry Number of discs on which the case data is stored The format in which the data is recorded The headings of the columns of the data table.
The intent of the configuration file (*.CFG) is to provide the information necessary for a computer program to read and interpret the data values in the associated data files. Since configuration file is in a predefined, fixed format, a computer program does not have to be customized for each configuration file. The content consists of the following elements:
Station name and identification Number and type of channels Channel names, units and conversion factors Line frequency Sample rate and number of samples at this rate Date and time of first data value Date and time of trigger point File type.
Figure 1. Probability of occurrence of different faults.
233
The data file (*.DAT) contains the data values in rows and columns where each row consists of a set of data values preceded by a sequence number and the time for that set of data values. No other information is contained in the data file. The first column contains the sample number of the data set in that row. The second column gives the time of the data in microseconds from the beginning of the record. The third and remaining columns contain the data values that represent voltages, currents and status information.
4. POWER SYSTEMS DISTURBANCES

4.1. Faults
The analysis of faults and disturbances (Nagrath & Kothari 1998) during abnormal conditions in power network can provide valuable information regarding the disturbances and the remedies. Fault classification is an important topic in disturbance and fault analysis to differentiate various types of faults. Typically faults are either symmetrical (three phase shortcircuit or three phase-to-ground faults) or nonsymmetrical (single phase-to-ground or double phase short-circuit or double phase-to-ground faults). Typical probability of occurrence of different types of faults is shown in Figure 1. The faults considered in Figure 1 are three-phase short-circuit (L-L-L), three phase-toground (L-L-L-G), line-to-line short-circuit (L-L), single line-to-ground (L-G) and double line-to-ground (L-L-G) faults, where the terms L and G refer to Line and Ground respectively. The probabilities associated with a fault type depend upon the operating voltage and can vary from system to system.
Figure 2. Typical power systems oscillation (voltage profile).
234
Abhisek Ukil
4.2. Power Systems Oscillation

Power systems oscillation is an important phenomenon in the power systems disturbance analysis. Power systems under steady-state conditions operate typically close to their nominal frequency. A balance between the generated and the consumed active and reactive powers exists during steady-state operating conditions and the sending and the receiving end voltages are within limit (typically 5%). Power system transient events, like, faults, line switching, generator disconnection, and large load change result in sudden changes to electrical power, whereas the mechanical power input to the generators remains relatively constant. These system disturbances cause oscillations in machine rotor angles and can result in severe power flow swings. Depending on the severity of the disturbance and the actions of power system controls, the system may remain stable and return to a new equilibrium state experiencing what is referred to as a stable power oscillation. Severe system disturbances, on the other hand, could cause large separation of generator rotor angles, large swings of power flows, large fluctuations of voltages and currents, and eventual loss of synchronism between groups of generators or between neighboring utility systems. Large power swings, stable or unstable, can cause unwanted relay operations at different network locations, which can aggravate further the power-system disturbance and possibly lead to cascading outages and power blackouts (Kundur 1994; Taylor 1993). Figure 2 shows the voltage profile in a typical power systems oscillation.
4.3. Power Quality

Power quality is the degradation of the power systems signals: voltage and current (more prominent) due to distortion. The distortions are caused by many a reason. These are mentioned briefly below.
Voltage sag. Voltage sag is one of the prime elements causing the power quality problem. A voltage sag is a short-term reduction in, or complete loss of, root mean square (RMS) voltage. It is specified in terms of duration and retained voltage, usually expressed as the percentage of nominal RMS voltage remaining at the lowest point during the dip. Voltage sags are typically the result of what is known as a fault condition, large motor starting, or due to interaction between motor operation, faults and also due to system overloading conditions. Figure 3 shows a typical voltage sag profile due to motor starting. Harmonic frequencies. These are integral multiples of the fundamental supply frequency, i.e. for a fundamental of 50 Hz, the third harmonic would be 150 Hz and the fifth harmonic would be 250 Hz. Harmonic distorted waveform is clearly not a sine wave and that means that normal measurement equipment, such as average reading RMS-calibrated multimeters, will give inaccurate readings. Transients, surges. A transient can be a unidirectional impulse of either polarity or a damped oscillatory wave with the first peak occurring in either polarity. Transients and surges are typically caused by switching or lightning strikes on the network and
235
switching of reactive loads on the consumers site or on sites on the same circuit, voltage flickers resulting from switching, short-circuits and load changing. Unbalance. A three-phase power system is called balanced or symmetrical if the three-phase voltages and currents have the same amplitude and are phase shifted by 120 with respect to each other. If either or both of these conditions are not met, the system is called unbalanced or asymmetrical. Voltage unbalance oftentimes results in voltage dips or sags which in turn causes problem in the power quality. Interruptions. Caused by the interruptions in supply voltage for short, medium or long period. Earthing and EMC. Improper earthing of equipments and EMC incompatibility as upper IEC 61000 series also causes power quality problems.
Figure 3. Voltage sag due to induction motor starting.
Figure 4. Schematic oscillating circuit diagram of the reactor ring down phenomenon.
236
Abhisek Ukil
Figure 5. Segmented voltage recording during a reactor ring down.
4.4. Reactor Ring Down

When a line reactor is connected to a line and the circuit-breakers are opened at both ends, the voltage does not disappear. Instead, an oscillating voltage waveform can be found which slowly reduces in magnitude. This phenomenon is called reactor ring down (Ukil & Zivanovic 2007). It is a result of the interaction between the reactor and the capacitance of the line. This forms an oscillatory circuit as depicted in Figure 4. Figure 5 shows the segmented voltage recording during a rector ring down incident. The segmentations are shown as vertical dashed lines. Usually, many segments for the oscillating signals are recorded during the reactor ring down phenomenon as indicated in Figure 5. Comparison of the signal parameter values of the segments B, C, D, E and F in Figure 5 indicates that the signal is oscillating and slowly decreasing in magnitude (Ukil & Zivanovic 2007).
4.5. Magnetizing Inrush Current

Energizing of a transformer often goes hand in hand with high magnetizing inrush currents. Transformer protection must be set so that the transformer does not trip for this inrush current. This high current is the result of the remnant flux in the transformer core when it was switched out, and depends where on the sine wave the transformer is switched back in. The segmented current recording for the energizing of a transformer is shown in Figure 6.
237
Figure 6. High magnetizing inrush currents associated with the energizing of a transformer.
In Figure 6, the current signal is segmented into two segments, A and B, that is, before and after the transformer is energized. A closer look at segment B reveals the fact of high magnetizing current inrush (Ukil & Zivanovic 2007).
5. AUTOMATED DISTURBANCE ANALYSIS

5.1. Introduction
The first disturbance recording systems in electrical power systems were based on electromechanical and analogue technology (Barth et al. 2003). The measurements were recorded on metallic or photosensitive papers or magnetic tapes. The accuracy, the number of signals, the recording time and the number of data records per paper roll or magnetic tape were highly restricted. Data handling and interpretation were also extremely complex processes. With the introduction of the digital recording technology, nowadays, engineers in power utilities have more data than can be processed and assimilated in the time available. Therefore, automated disturbance analysis (Kezunovic et al. 2001) is inevitable. Users of knowledge extracted from recorded data are engineers and technicians working in operations, maintenance and protection departments. It is the task of operating personnel to return to service as much of the electric system as practical in the shortest time possible. The following questions (Ukil 2005) concern them: Where and what is the problem? Did the line reclose and stay in operation? What equipment was operating? Was everything working correctly? If so, can it be returned to service? If not,
238
Abhisek Ukil
what needs to be isolated? A complete analysis should be available for use within five minutes after the conclusion of the event. Maintenance personnel are charged with repairing and returning outage equipments to service. They require information on what is damaged or operating outside the normal parameters. The time requirement to notify the maintenance personnel is in the region of two hours or less. Protection personnel are responsible for the final assessment of the correctness of any electrical system response to a given disturbance condition. Normally the protection engineer is given adequate time to collect all the data necessary for a complete evaluation of a disturbance. Key questions that need to be answered are: Did the right thing respond in the right way? Did the wrong thing respond in the wrong way? Did the right thing respond in the wrong way? The thing in these questions might be a relay, relaying system, circuit breaker, or switch.
5.2. Analysis Implementation, Requirements

The automated analysis can be broadly grouped as follows:
fault clearing analysis power swing condition undervoltage conditions fault location information underfrequency condition dynamic conditions on generators data for relay testing.
To provide data necessary for quick analysis, the following quantities (Spasojevic 2004) should be available: A) Bus phase voltages: Depending upon the station configuration, bus and/or line voltages (phase to ground) should be monitored. A minimum of three phases per voltage class should be provided. B) Bus residual voltage (derived from phase quantities): This is necessary for the analysis of potential polarized relaying schemes. C) Line phase voltages: These are necessary for the analysis of re-closing schemes and relay operations. D) Line phase currents: Currents from all three phases on each line monitored. E) Line residual current (derived from phase quantities): These quantities are absolutely necessary to analyze any type of disturbance on a power system. Most faults can be analyzed by using this quantity and the three phase-to-ground voltages. F) Pilot channel data: The data should include both transmitter and receiver status on all power line carriers, fibber optic, audio tone, pilot wire, and microwave circuits. G) Breaker, station tripping, and blocking status data: Auxiliary contacts show when status points changed state. Breaker position, lockout relays position (differential and breaker failure) and pilot channel blocking all aid in the assessment of both correct and incorrect operations.
239
H) Control contact performance: Trip and close contact information is a must. Trip and close initiation helps to better define what happened and why (i.e. which relays initiated the trip or close). I) Alarm contacts: If possible an alarm contacts such as relay failure, breaker air and/or gas pressure and transformer temperature and cooling should be included whenever possible, in order to allow a more complete analysis. J) Relay target data: Presently, outputs with this information are not available from all relaying systems. However, any system that is to perform automated power system analysis should include target data. Microprocessor relays and relaying systems provide ASCII and contact output of target data. It would be extremely beneficial if these targets can be included in the analysis. K) Time code information: Time should be accurate to a minimum of one millisecond and it should be possible to determine the time for each data sample. This accuracy is necessary when comparing fault data from different sources. If data from different sources are to be merged to make calculations such as phase angle between sources it is necessary that time information have an accuracy of 50 microseconds or better. All data from the Sequence of Events solution and the DFR can be synchronized to one millisecond. For the SCADA solution there presently exist two different types of time data. Older SCADA systems only report time to the second, with some time tagging after receipt at the SCADA master. Some newer systems time tag to the millisecond at the RTU and should be considered as a sequence of events implementation for that reason. L) Fault duration: Used in analyzing relay and breaker performance. M) Clearing time (all phases): Used in determining pole openings of breakers and fault clearings. N) Magnitude of the fault current: Magnitude use for fault locations and determining which relays should have operated. O) Type of fault (single phase, multiphase, evolving): The analysis of faults depends on knowing how faults evolved. P) Phases involved in the fault: Useful in fault locating and relay system analysis. Table 1 shows the availability of the different information from different equipments typically used for automated disturbance analysis.
Table 1. Required data availability by different devices
X +
Data available. Data not available. Partial Data may be available in some form.
240
Abhisek Ukil
6. SIGNAL PROCESSING: FOURIER TRANSFORM-BASED ANALYSIS

6.1. Discrete Fourier Transform (DFT)
Fourier transform (Rabiner & Gold 1975; Oppenheim et al. 1996, 2001) has a key role to play in signal processing as well as power systems disturbance analysis. Originally, the Fourier transform was postulated in 1822 by French mathematician and physicist Jacques Fourier to represent any continuous periodic signal as a sum of (in general infinite number) of sinusoids. The Fourier transform for a signal f (t ) is defined as
F ( ) =
f (t ) e
j t
dt ,
1 .
(1)
where, j is the complex operator defined as j =
For all kinds of discrete-time sequences, the version of the Fourier transform is called the Discrete-Time Fourier Transform (DTFT) (Rabiner & Gold 1975; Oppenheim et al. 1996, 2001) which uses the complex exponential. DTFT for a sequence x[n] is defined as
X [ j ] =
n=
x[n]e
j n
(2)
However, DTFT is a continuous function of . Therefore, it is function of continuous variables. Moreover, as evident from (2), DTFT is defined for infinite length sequences. Hence, DTFT is not numerically computable algorithm. From the Fourier analysis we know that a periodic function (or sequence) can always be represented by a linear combination of harmonically related complex exponentials. This is the Discrete Fourier Series (DFS). We can then extend the DFS to the finite-duration sequences which lead us to a new transform, the Discrete Fourier Transform (DFT) (Oppenheim et al. 1996, 2001). DFT can actually be obtained by sampling the DTFT in the frequency domain. DFT of a sequence x[n] is defined as
X [k ] = x[n]e
n =0
N 1
2 kn N
n = 0,1, 2,K .
(3)
Comparing this with the DTFT (see (2)), we see the relationship between the DFT and the DTFT as
241
DFT = DTFT = 2 k .
N
(4)
6.2. Full and Half-cycle DFT in Power System

In the case of sampled data yk = y (tk ) , obtained from the signal y(t) over a period that corresponds to the fundamental frequency of the signal (N samples taken over one period), the fundamental frequency component of DFT is equal to
Y =
where,
2 N
y
k =0
N 1
exp( j k ) = YRe jYIm ,
(5)
YRe =
2 N
y
k =0
N 1
cos( k ) ,
(6)
YIm =
2 N 1 yk sin( k ) , N k =0
(7)
2 . N
(8)
A fundamental frequency component of DFT and fundamental frequency phasor are related as follows
Y=
j Y. 2
(9)
When the above equations are combined we obtained the estimates for Y and from DFT components YRe and YIm , as follows
Y=
1 2 2 YRe + YIm . 2 YRe . YIm
(10)
= arctan
(11)
242
Abhisek Ukil Note that the sampling frequency is N0 , and thus the input signal y(t) must be band-
limited to N0 / 2 in order to avoid aliasing (Oppenheim et al. 1996, 2001) errors. It can be shown that in the presence of normally distributed noise, the fundamental frequency component of DFT is a least squares estimate of the fundamental frequency phasor. The sine and cosine factors used in the calculation of YRe and YIm can be pre-calculated and stored in memory to speed-up the calculation. When N is the number of samples in the half of the fundamental frequency period, we use the same algorithm as for the full-cycle DFT with different sampling angle
The half cycle DFT only rejects odd harmonics because of the shorter window. It has faster response compared to the full cycle DFT.
6.3. Phasor Computation

Phasor is a quantity, usually a complex number, with magnitude and phase, measured with respect to a fixed reference. Phasors have a lot of utilizations in the power systems as they simplify computations involving the sinusoids. Figure 7 shows a simplified phasor diagram. The estimation of the fundamental frequency phasor consists of calculating first the DFT component using (3) over a data window of N samples, and scaling it by (2 / N). The real and imaginary components can be calculated separately using (6) and (7) respectively.
6.4.Uses of Phasor in Disturbance Analysis

Phasor values can be used to derive useful information from which graphs may be drawn. Examples of these are impedance- and power plots.
Figure 7. Simplified phasor diagram.
243
6.4.1. Impedance Plots Impedance plots are very useful when investigating line impedance protection relay operations since it displays what the relay observed during the power system disturbance event. All line impedance protection relays have two types of measuring elements, those measuring from line to ground and those measuring from line to line. Normally there are at least six measuring elements in a line impedance protection relay. The elements respectively measure the following quantities. For line to ground measuring elements:
Z=
I LG
VLG , + KN IN
(12)
where, K N = current is
1 Zo ( 1) and Zo and Z1 is protection setting values , and the neutral 3 Z1
I N = I R + IW + I B .
For line to line measuring elements:
(13)
Z=
VLL1 VLL 2 . I LL1 I LL 2
(14)
All above phasor quantities refers to measured values.
6.4.2. Power Plots for Dynamic System Study Dynamic studies are aimed at tuning of the dynamic circuits in power systems. This can not be done without the help of power plots. For dynamic studies, root mean square (RMS) values are of more concern instead of the measured quantities. The power transmitted in a single phase is also not used frequently and the total power transmitted is rather used. However, both formulas used are as follows. For power in a single phase circuit:
S = VI * ,
P = VLG I LG cos( ) = VLG I LG 2 2 cos( ) = VLG I LG cos( ) [WRMS ] , 2 VLG I LG sin( ) [WRMS ] . 2
(15)
(16)
Q = VLG I LG sin( ) =
VLG I LG 2 2
sin( ) =
(17)
244
Abhisek Ukil
Figure 8. Basis functions for Fourier transform (sine wave) and wavelet transform (db10).
For total power:
STOTAL = S RG + SWG + S BG ,
where,
S RG = PRG + jQ RG = V RG I RG cos( ) V I sin( ) + j ( RG RG ) [MVA 2 2
(18)
RMS
] etc.
(19)
All measured values are divided by the square root of 2 to convert it to RMS values.
7. WAVELET TRANSFORM-BASED ANALYSIS

The Wavelet transform (WT) is a mathematical tool, like Fourier transform for signal analysis. A wavelet is an oscillatory waveform of effectively limited duration that has an average value of zero. Fourier analysis consists of breaking up a signal into sine waves of various frequencies. Similarly, wavelet analysis is the breaking up of a signal into shifted and scaled versions of the original (or mother) wavelet. Figure 8 shows the basis functions for Fourier transform (the sine wave) and WT (db10: Daubechies 10 (Daubechies 1992) mother wavelet). Fourier analysis does not provide good results for nonstationary signals, unlike the stationary signal, where the signal parameters change over the time, because in transforming the complete signal to the frequency domain, the time information is lost in Fourier analysis. This deficiency in Fourier analysis can be overcome to some extent by analyzing a small section of the signal at a time - a technique called windowing the signal. This leads to an analysis technique called Short-Time Fourier Transform (STFT). But the drawback in STFT is that the size of the time-window is same for all frequencies. Wavelet analysis overcomes this deficiency by allowing a windowing technique with variable-sized regions, in other words, wavelet analysis allows the use of long time intervals where we want more precise low-frequency information, and shorter regions where we want high-frequency information. Figure 9 (Ukil 2005; Ukil & Zivanovic 2006) shows the time-based (Shannon), frequencybased (Fourier), STFT (Gabor) and wavelet views of signal analysis. Detailed mathematical
245
descriptions of WT can be referred to in the seminal works of Ingrid Daubechies (Daubechies 1992) and Stephane Mallat (Mallat 1998).
7.1.Continuous Wavelet Transform (CWT)

The Continuous Wavelet Transform (CWT) is defined as the sum over all time of the signal multiplied by scaled and shifted versions of the wavelet function . Let R be the set of real numbers. The CWT of a signal x(t) is defined as
CWT (a, b) =
where
x(t )
* a ,b
(t )dt ,
(20)
a ,b (t ) = a
1 / 2
((t b) / a ) .
(21)
Figure 9. Time, frequency, STFT and wavelet views of signal analysis.
246
Abhisek Ukil
(t ) is the mother wavelet, the asterisk in (20) denotes a complex conjugate, and
0 a R, and b R respectively are the dilation (mathematicians use the term dilation
to refer to both compression and expansion) and translation parameters respectively. The term | a |
1 / 2
in (21) is the normalization value of a , b (t ) so that if (t ) has a unit length,
then its scaled version a , b (t ) also has a unit length.
7.2. Discrete Wavelet Transform (DWT)

Instead of continuous scaling (dilation) and shifting (translation), the mother wavelet may be scaled and shifted discretely by choosing a = a0 , b = na 0 b0 , t = kT in
m m
(20) and (21), where T = 1.0 and k, m, n Z, (Z is the set of positive integers). The Discrete Wavelet Transform (DWT) is thus given by
DWT (m, n) = a0 m / 2
( x[k ]
m m [(k na0 b0 ) / a0 ] .
(22)
By careful selection of a0 and b0 , the family of scaled and shifted mother wavelets constitutes an orthonormal basis. An orthonormal basis comprises a set of vectors S such that
u.v = 0 (dot product) for each distinct pair of u, v S . We can choose a0 = 2 and b0 = 1
to constitute the orthonormal basis to have the WT referred to as a dyadic-orthonormal WT. The implications of the dyadic-orthonormal WT is that because of the orthonormal properties there will be no information redundancy in the decomposed signals. Also, with this choice of a0 and b0 , there exists a novel algorithm, known as the multiresolution signal decomposition (Mallat 1989) technique, to decompose a signal into scales with different time and frequency resolution.
7.3. Multiresolution Signal Decomposition and Quadrature Mirror Filter

The Multiresolution Signal Decomposition (MSD) (Mallat 1989) technique decomposes a given signal into its detailed and smoothed versions. Let x[n] be a discrete-time signal. Then the MSD technique decomposes the signal in the form of WT coefficients at scale 1 into c1[n] and d1[n], where c1[n] is the smoothed version of the original signal, and d1[n] is the detailed version of the original signal x[n]. These are defined as
c1[n] = h[k 2n]x[k ] ,

k
(23)
d1[n] = g[k 2n]x[k ] ,

k
(24)
247
Figure 10. Multiresolution signal decomposition realized by quadrature mirror filter banks.
where h[n] and g[n] are the associated filter coefficients that decompose x[n] into c1[n] and d1[n] respectively. Downsampling is done in the process of decomposition so that the resulting c1[n] and d1[n] are each n/2 point signals. Thus, for the original n point signal x[n], after the decomposition we have n point signal together with c1[n] and d1[n], not 2n point. The next higher scale decomposition will be based on c1[n]. Thus, the decomposition process can be iterated, with successive approximations being decomposed in turn, so that the original signal is broken down into many lower resolution components. This is called the wavelet decomposition tree (Mallat 1998). The MSD technique can be realized with the cascaded Quadrature Mirror Filter (QMF) (Mallat 1998) banks. A QMF pair consists of two finite impulse response filters, one being a lowpass filter (LPF) and the other a highpass filter (HPF). The QMF pair divides the input signal into low-frequency and high-frequency components at the dividing point, halfway between zero Hz and half the data sampling frequency. The output of the lowpass filter is the smoothed version of the input signal and is used as the next QMF pairs input. The output of the highpass filter is the detailed version of the original signal. Thus cascaded QMF pairs realize the MSD technique. Detail description about QMF can be found in the book by Strang and Nguyen (Strang & Nguyen 1996). Figure 10 shows the MSD technique and QMF pair.
7.4. Uses of Wavelets in Fault Analysis

Figure 11 shows an example of use of wavelets in power systems fault analysis (Ukil & Zivanovic 2006). In Figure 11, the original DFR recording, from the power transmission network of South Africa, for the voltage during a phase-to-ground fault in the BLUE-Phase, sampled at a frequency of 2.5 kHz (Ukil & Zivanovic 2006), is shown in the plot (i); wavelet coefficients for this fault signal and the universal threshold (dashed) are shown in the plot (ii) and the change time-instants as unit impulses computed using the threshold checking (plot ii) followed by the smoothing filtering (Ukil 2005; Ukil & Zivanovic 2006) are shown in the plot (iii). The time-instants of the changes in the signal characteristics in the plot (iii) in Figure 11 indicate the different signal segments owing to the different events during the fault, e.g.,
248
Abhisek Ukil
segment A indicates the pre-fault section and the fault-inception, segment B the fault, segment C opening of the circuit-breaker and segment D auto-reclosing of the circuit-breaker and system restore (Ukil & Zivanovic 2006).
Figure 11. Wavelet transform-based analysis of voltage recording during a phase-to-ground fault.
7.5. Other Uses of Wavelet

The wavelet transform is particularly suitable for the power system disturbance and fault signals which may not be periodic and may contain both sinusoidal and impulse components. In addition, time-frequency resolution is needed for the power system fault analysis. This is another reason for using the wavelet transform because it provides a local representation (both in time and frequency) of a given signal, unlike the Fourier transform which provides a global representation of a signal. In particular, the ability of the wavelets to focus on short intervals for high-frequency components and long intervals for low-frequency components improves the decomposition of the fault signals into finer and detailed scales, facilitating further effective signal processing and analysis.
249
In the past ten years, the application of the wavelet transform in the field of power systems has received particular attention. In power systems signal analysis, time-frequency resolution is needed, which states the reason for using the wavelet transform because it provides a local representation (both in time and frequency) of a given signal. It is different from the Fourier transform which provides a global representation of a signal. Examples of the use of the wavelet transform in power systems can be found in the works of Robertson, Camps and Mayer (Robertson et al. 1994), Robertson, Camps, Mayer et al. (Robertson et al. 1996), Wilkinson and Cox (Wilkinson & Cox 1996). Santoso, Powers and Grady (1994) and Santoso, Powers, Grady et al. (1996) used the wavelet transform for power quality assessment. Ribeiro (1994) proposed the use of the wavelet transform for analyzing the harmonics in power systems. Pillay and Bhattacharjee (1996) used the wavelet transform to model short-term power system disturbances.
8. CONCLUSION
Disturbance analysis plays a crucial role in a secure and reliable electrical power supply. With the advent of new digital technology, the quantity and quality of fault and disturbance data have increased considerably. We have encountered in this chapter basics about power systems disturbance analysis: its importance, relevance and procedures. Relevant digital equipments used in modern power distribution systems for disturbance data acquisition and analysis, are discussed in details. This is followed by discussion of different types of power systems disturbance like faults, oscillation, transients, power quality, etc. With the huge amount of available disturbance data, automated disturbance analysis is inevitable and of great importance for utilities worldwide. Idea and requirements of the automated disturbance analysis has been introduced in this chapter. This is followed by two important signal processing techniques commonly used in power systems disturbance analysis: Fourier transform and the wavelet transform.
9. REFERENCES
Anderson PM (1999) Power system protection. New York, McGraw-Hill. Barth P, Ludwig A, Schegner P (2003) Development of the fault and disturbance data andling in the German HV-network. In Proc. CIGRE SC-34 Colloquium, Sydney. CEA Technologies Inc.(CEATI), Electromagnetic Transient Program EMTP-RV, 1998. Available: http://www.emtp.com Daubechies I (1992) Ten Lectures on Wavelets. Philadelphia: Society for Industrial and Applied Mathematics. Grigsby LL (editor) (2007) The electric power engineering handbook, 2nd Ed., five vol. Florida: CRC Publication. IEEE (1991) IEEE Standard Common Format for Transient Data Exchange. Standard 37.1111991, version 1.8. Kezunovic M, Chen-Ching L, McDonald JR, Smith L (2001) IEEE tutorial on automated fault analysis.
250
Abhisek Ukil
Kundur P (1994) Power system stability and control. New York: McGraw-Hill. Mallat S (1989) A theory for multiresolution signal decomposition: the wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11 (7) 674 693. Mallat S (1998) A wavelet tour of signal processing. New York: Academic Press. Nagrath J, Kothari DP (1998) Power system engineering. New Delhi: Tata McGraw Hill Publications. Pillay P, Bhattacharjee A (1996) Application of wavelets to model short-term power system disturbances. IEEE Transactions on Power Systems 11 (4) 2031 2037. Rabiner LR, Gold B (1975) Theory and Application of Digital Signal Processing. Englewood Cliffs, NJ: Prentice Hall. Ribeiro PF (1994) Wavelet transform: an advanced tool for analysing non-stationary harmonic distortions in power systems. In Proc. of IEEE Int. Conf. on Harmonics in Power Systems, Bologna, Italy. Robertson D, Camps OI, Mayer JS (1994) Wavelets and power system transients: Feature detection and classification. In Proc. SPIE Int. Symp. on Optical Engg. in Aerospace Sensing, pp. 474 487. Robertson D, Camps OI, Mayer JS, Gish WB (1996) Wavelets and electromagnetic power system transients. IEEE Transactions on Power Delivery 11 (2) 1050 1056. Santoso S, Powers EJ, Grady WM (1994) Electric power quality disturbance detection using wavelet transform analysis. In Proc. IEEE-SP Int. Symp. on Time-Frequency and TimeScale Analysis, Philadelphia. pp. 166 169. Santoso S, Powers EJ, Grady WM, Hoffman P (1996) Power quality assessment via wavelet transform analysis. IEEE Transactions on Power Delivery 11 (2) 924 930. Spasojevic P (2004) Analysis of digital fault recording. M.Sc. dissertation, Dept. Electrical Engineering, Tshwane University Technoloy, Pretoria, South Africa. Strang G, Nguyen T (1996) Wavelets and filter banks. Wellesley, MA: Wellesley-Cambridge Press. Taylor CW (1993) Power system voltage stability. McGraw-Hill, New York. Ukil A (2005) Abrupt change detection in automatic disturbance recognition in electrical power systems. Ph.D. dissertation, Dept. Math. Tech., Tshwane University Technoloy, Pretoria, South Africa. Ukil A, Zivanovic R (2006) Abrupt change detection in power system fault analysis using adaptive whitening filter and wavelet transform. Electric Power Systems Research, 76 (910) 815823. Ukil A, Zivanovic R (2007) Application of abrupt change detection in power systems disturbance analysis and relay performance monitoring. IEEE Transactions on Power Delivery 22 (1) 5966. Wilkinson WA, Cox MD (1996) Discrete wavelet analysis of power system transients. IEEE Transactions on Power Systems 11 (4) 2038 2044.
Chapter 8
ARCING AND FUSION PROCESS OF EXPLODING WIRES IN INDUSTRIAL APPLICATIONS

C. S. Psomopoulos1,*and C. G. Karagiannopoulos2,
1
Technological and Education Institute of Piraeus, Dep. of Electrical Engineering, High Voltage Laboratory, 250 Thivon & P. Rali str., GR-12244 Athens, Greece 2 National Technical University of Athens, School of Electrical and Computer Engineering, 9, Iroon Polytechniou St., GR-157-80 Athens, Greece
ABSTRACT
The fusible wires and electrically exploded conductors are frequently used in fuses that protect the electrical installations of the low and medium voltage distribution network and in a wide variety of power applications. They generally operate either under a minus value of their nominal current or momentarily under excess or sort-circuit currents. Despite a wide variety of experimental work, there remains much that is not understood about the electrical explosion of conductors. Furthermore the surrounding medium has a significant role in the fusion process. Most of the experimental work is focused in air and some in SiO2 due to use as the most common surrounding material in industrial fuses. Recently the fusion process in capillaries presented also interest among researchers. The power supplies used in these experimental works were pulses mostly, produced by typical L-C circuits. This chapter will investigate the fusion process of exploding wires using common industrial power supply directly feed from the network of 50Hz. The measured magnitudes were the voltage drop and the current waveforms during the current interruption process and the total duration of the fusion. Experiments performed for several diameters and lengths and different current densities. The resistance and the I=f(V) waveforms during fusion process were directly calculated from the experimental curves.
E-mail: c-psomop@ath.forthnet.gr E-mail: ckarag@central.ntua.gr
252
C. S. Psomopoulos and C. G. Karagiannopoulos

In this chapter, measurements better than before through fast digital oscilloscopes connected with PC are presented and the observed phenomena are investigated. Measurements of the total time duration of the fusion process are presented and analysed. The presented measurements are compared with existing ones (taken with the use of A/D converters) and the differences are evaluated. In the related literature the experiments are carried out usually with pulse voltages or currents, as it is mentioned above. The measurements presented here were performed in random position of the supply voltage curve. These measurements give the possibility to investigate the physicochemical process during the fusion process further, and especially the arcing behavior. The obtained experimental results are investigated and evaluated with the use of analytical and statistical methods, and are compared with other relevant ones in literature. The analysis of the results is focused on the fusion duration and the arcs formation for a variety of parameters (current density, geometric characteristics, etc) because of their importance in the protection of networks during short circuit. The digital measurements with PCs, can provide a more accurate approximation and understanding of the phenomena during the fusion of exploding wire. Another parameter that contributes to the above is the use of common industrial voltage source and random sinus values in the experiments, because provide a more accurate simulation of the actual conditions.
INTRODUCTION
The fusible wires and electrically exploded conductors are frequently used in fuses that protect the electrical installations of the low and medium voltage distribution network and in a wide variety of power applications. They generally operate either under a minus value of their nominal current or momentarily under excess or sort-circuit currents. During the operation under nominal current or less the Joule heating produced on fuse element dissipates to the surrounding area of the element and thermal equilibrium is attained after a time period [1,2,3,4]. For the case of the operation under excess or heavy fault currents the design of fuses is based on the well known fundamental principle, that they must interrupt those currents in a very sort time period. The phenomena developed are in brief as follows. The increase of the fuse elements resistance causes the temperature to rise rapidly, until the melting point is reached. The latent heat of fusion is produced gradually by the current during the melting time until the material is completely vaporized [2,3,5]. When the material of the element vaporizes, electric arcs struck between the remaining solid parts of the element. This dynamic process gives rise to a rapid temperature increase, while the current decreases rapidly until its flow is interrupted. The fundamental fuse operation has been described extensively in the relevant literature [113]. The understanding of this operation relies mainly on work on exploding wires and sand fuses mostly performed from 1960s to 1990s [313]. Several theories have been developed to explain the disruption of fuse elements, but no conclusive proof has been found in favour of one of them. Hence new ideas on the exploding wires phenomenon and the sand fuse operation often appear [13, 1317]. Present research activities are focused mostly in the arcing period and a number of models have been developed to simulate the fundamental operation. This operation is well
Arcing and Fusion Process of Exploding Wires in Industrial Applications
253
understood in general, but all the physicochemical processes during the arcing period seem not to be completely known [1215]. However some models have been developed which can simulate fuses operation under fault current during the arcing period [1,6,1215, 18]. During the electrical explosion of a wire under certain conditions, the wire fragments and current is diverted around condensed fragments, through a plasma sheath surrounding the fragments, due to arcing process there. Initially, the plasma is seen as bright spots along the wire length, in many experimental results. These plasma spots develop along the wire, expanding axially into other spots until the entire wire is seen to become covered in plasma. Experimental results using X-radiography reveals significant, continuous lengths of condensed (liquid) wire beneath the plasma [19,20,21]. The reasons for the current diversion are a matter of debate, and various models have been put forward to explain these phenomena [1922]. Another interesting question regarding the wire fragmentation during fusion process is whether the wire is in the solid or liquid phase during fragmentation. Experiments and modelling have been conducted for many decades showing that solid phase fragmentation of exploding wires can occur [2326]. In this work, measurements of the voltage drop across fuse elements and the current, using fast digital oscilloscopes connected with PC are presented. The fusion total duration is measured for a variety of thin wire elements lengths and different current densities. To our knowledge there are no similar measurements in the related bibliography especially considering the method used here were the fusion process was measured using actual industrial voltage (230V, 50Hz), which is simulating actual conditions for the circuit and the exposed fusible wire. In the related literature the experiments are carried out usually with pulse voltages or currents. The measurements in this work were performed in random position of the supply voltage curve. These measurements give the possibility to investigate the physicochemical process during the pre-arcing and arcing periods. An identification of the existing knowledge in the measurements performed, and a further interpretation of the phenomena during sort-circuit is attempted.
EXPERIMENTAL SET-UP AND MEASUREMENTS RESULTS

Figure 1 shows the simplified schematic diagram of the experimental set-up, for the estimation of the voltage drop across a fuse element under sort-circuit conditions. Thin wires simulating the fuse elements, were connected to the circuit using appropriate terminals. A typical ohmic load of 0,5kW, connected in parallel with an open switch was used to simulate the nominal current operation and sort-circuit conditions respectively. A measuring resistance Rm was used to estimate the current through the fuse element. The material used for the measuring resistance had zero value coefficient of variation of resistivity with temperature ( 0). The variable resistance R was of a similar material as Rm and used to determine the peak current value during sort-circuit.
254
Figure 1. Simplified schematic diagram of the used experimental set-up. 1) Thin-wire element, 2) R: variable resistance for determine the current peak value, 3) Digital oscilloscope with an RS-232 connection to the PC, RU1, RU2 voltage divider for the voltage across the element, R1, R2 and Rm voltage divider and measuring resistance for the current through the element.
The measurements were performed through a digital oscilloscope with high sampling frequency up to 100MHz per channel and maximum voltage of 40V pick to pick (5V/div). The sampling frequency was 0,1MHz. The oscilloscope was directly connected to PC, via an RS232 port, for direct transfer of the measurements. The two voltage dividers were used in order to reduce the voltage peak value provided to the oscilloscope. The software of the interconnection could transfer the values and the curves in excel sheets (generally a spreadsheet software) directly with spreadsheet software format, and the measurement on the oscilloscope screen, as image independently inserted in the spreadsheet with common bmp format. The maximum absolute measuring error was 2.5mV as the measurements were performed in 2.5V scale. Any possible capacitance of the experimental setup (coaxial cables, etc.) could not affect the measurements accuracy due to the high sampling frequency used. Also, all the resistances were checked in order to verify that their impedance was zero. Measurements were performed on conductors stretched in air. The conductors were thin cylindrical copper wires (99,9% Cu), with several diameters. Figures 2 and 3 show the voltage drop, the current and the resistance waveforms during the current interruption process in fuse elements with diameter of 0,14mm and 0,16mm and length which was varied between 35mm to 100mm with a step of 5mm. The current density was changed through the variable resistance (R) of the circuit.
255
Figure 2a. Typical measurements of the fusion process in thin copper wire elements. Channel 1 is the voltage drop across the element (the sinus waveform), channel 2 is the current during fusion process (rapidly decreases to zero while voltage remains). The calculated maximum current density was 4543A/mm2. The thin wire element had length 60mm, diameter 0.14mm.
Figure 2b. Typical measurements of the fusion process in thin copper wire elements. The calculated maximum current density was 3303A/mm2. The thin wire element had length 90mm, diameter 0.14mm.
256
400 350 300 250 200 150 u(V) J (x10A/mm2) 100 50 0 -50 -100 -150 -200 -250 -300 -350 -400 0 2 4 6 8 10
j=f(t) , u=f(t)
12
14
16
18
20
22
24
26
28
V (Volt) J (x10A/mm2)
t (ms)
Figure 3a. Typical measurements and the respective actual values calculated from experimental set up characteristics, of the fusion process in thin copper wire elements. The waveform of the of the fusion process R=f(t) is also presented. Channel 1 is the voltage drop across the element (the sinus waveform), channel 2 is the current during fusion process (rapidly decreases to zero while voltage remains). The calculated maximum current density is 4543A/mm2. The thin wire element has a length 40mm, diameter 0.14mm.
257
INTERPRETATIONS REGARDING THE ARCING AND FUSION PROCESS IN EXPLODING WIRES

In figures 2, 3 and 4, which present actual measurements from the oscilloscopes screen, it can be clearly seen the sort-circuits duration, the voltage drop across the element, the current value (impulse and continuous sort-circuit currents). From the waveforms presented in this figures it can be notice that they follow the form of the ones in literature, and similar characteristics can be observed. Here the difference that exists is that the measurements presented in this work were taken in random positions of the sinus curve of the voltage source. Another difference concerned the voltage source, which was a common industrial sinus sapped waveform exactly the same one with the one that the equipment it is supplied, and not a laboratory produced voltage. With these two mentioned here basic conditions the fusible wire is exposed to the actual conditions of a network and thus the phenomena observed are closer to the actual ones in practical applications. General, the waveforms observed in the literature can be divided in two basic areas known as pre-arcing period and arcing period. In the measurements presented here the following can be observed: In the beginning of the pre-arcing period it can be seen the peak value of the sort-circuit current (impulse current). After few sec, the current is almost stabilized (continuous sort-circuit current). The voltage across the elements edges, after a very fast rising period was also almost stabilized. This value was lower than the source peak value (310V).
34 32 30 28 26 24 22
R ()
R= f(t)
20 18 16 14 12 10 8 6 4 2 0 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
t (ms)
Figure 3b. The waveform of the of the fusion process resistance R=f(t) of the measurement 3a.
258
450 400 350 300 250 200 150 V(V)-I(A) 100 50 0 -50 0 -100 -150 -200 -250 -300 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 t (ms) I (ampere) V (volt)
450 425 400 375 350 325 300 275 250 225 200 175 150 125 100 75 50 25 0 -25 0
R=f(t)
R ()
8 9 t (ms)
10
11
12 13
14
15 16
17
Figure 4. The calculated by the measuring system characteristics actual time resolved spectrum of the voltage across the element and the current flow through the exploding wire, for the measurement presented in Figure 2b. The waveform of the of the fusion process resistance R=f(t) of the measurement is also presented.
259
During this period and after the first 0,51msec the waveforms presented the following behavior: The voltage drop was almost stabilized with increasing tense while the current although it was almost stabilized showed decreasing tense. These can be explained by the increasing of the specific resistance due to overheating. The graduated increase of the wires specific resistance follows the well known relation = 0 (1 + a + b 2 ) and considering the sort-circuit conditions and the above mentioned, seems to give an explanation to the measured high voltage drop in the elements edges [8,13]. At the end of the prearcing period it can be observed a sinking of the voltage drop and the current for a sort time duration (0,25-0,75ms). This sinking may be explained from the accelerated temperature rise until the melting point of the fuse wires material, and the changes from solid to liquid conditions (the wire starts to liquefy in the inner parts). After reaching the melting point, the wire starts to liquefy. In general the inner parts of the cross section are hotter than the outer parts and fuse first (Figure 5). Then the melting zone spreads out until it reaches the periphery. At melting temperature the resistivity of the fused material is much higher than that of solid material, and increases with progressive melting. (Some data for the specific resistivity of the copper are: when solid at melting point 10,210-6 Ohm-cm, when liquid at melting point 21,310-6 Ohm-cm and in melting state 13,810-6 Ohm-cm) [3,6,10,13,17]. During the sort times here in question and the surface tension, the gravitational forces do not play any role according to literature and thus the liquid wire material will remain in place and be heated further to the temperature of vaporization. At some instant before the material is completely vaporized, the metallic coherence within the fuse wire is interrupted and an arc is formed. [3,7,13]. However, it is doubtful whether this phenomenon will occur in a regular way or whether the liquid cylinder will explode under the mechanical action of the vapor pressure. [3,7,8,13]. According to the literature there is another action that could lead to arcs formation in fusible wires. A cylindrical wire completely melted by an electric current flowing through it, experiences a pressure in connection with the surface energy and a magnetic pressure created by the electric current. If the diameter of the melted wire shows small variations the magnetic pressure is largest at the smallest diameter. As a consequence of the surface tension and the magnetic pressure, a cylindrical current-carrying liquid conductor stretched in air will deform into a number of globules (undoloids). If the source voltage of the circuit is sufficiently high, small arcs arise between these globules. This phenomenon is referred as multiple arcing and it has been observed [1,2,3,7,17].
Liquid metal
Solid metal
Figure 5. The cross-section of a fusible element during melting.
With the volatilization and interruption of the metallic coherence, electric arcs are formed (arcing period) between the remaining solid parts of the conducting material, in an
260
environment of metal vapors. In this complex environment, as the arcs expand produce both elevated temperature and pressure within the plasma and hence an active environment. In these conditions thermal ionization can be expected. There is a rapid change in arc dimensions, which is a function of the rate of burn-back of the metallic parts (arc elongation). This dynamic process gives rise to rapid increase in the electric resistance of the arc and its eventual extinction [1,2,3,6,13]. The metal vapor column, which arises during the evaporation process of a fusing wire surrounded by air, expands very rapidly due to the high pressure in this vapor, as has been made visible with cylindrical wires in the past [7]. The theoretical and experimental investigation concerning the exploding wires showed that fuse elements, after reaching the evaporation temperature under atmospheric conditions, could become highly superheated in the liquid state [9,11,12,18]. At the actual evaporation temperatures the metal vapor column possesses a resistance which is much larger than that of the fuse element at the same temperature, but in liquid state. The fuse resistance in these circumstances is determined by the cross-section of the metal vapor column, the length of this column and the specific resistance of the ionised metal vapor. Experiments showed that the cross-section of the metal vapor column depends on the current [1,2,7,18]. In this metal vapor column the arcs (which are characterized as a self-sustained electric discharge in which the plasma occurs at high temperature, high pressure and the surrounding is in local thermodynamic equilibrium) are formed. The most important feature of the arc seems to be its temperature [1,2,3,7,18]. In these conditions, to sustain a high electron density, a small electric field strength is needed and there is a small electrode fall (~1050V) [13,18]. Figure 6 illustrates the main regions of the arc, which are the electrodes with the fall regions and the main plasma column. The main column is known as the thermal plasma or thermal layer in which most of the energy transfer takes place due to collisional process involving free electrons and heavy ionized particles. Under these conditions the collisions can be considered as elastic and most energy is used to heat the colliding ions. The rapid changes in arcs dimensions and the expansion of the metal vapor column, which increase rapidly the arc resistance, as well as the decreasing of the voltage drop across the fuse edges (caused by the source voltage that is in the falling part of the positive semiperiod) cause the current interruption.
Figure 6. The main regions of an arc.
261
Figure 7. Plasma spot formation prior to voltage peak. Image taken by M.J. Taylor and J. Dunnet [21].
The development of the wire fragmentation in exploding wires and the plasma formation in experiments above mentioned, have been observed, and described also in M.J. Taylors recent research works [19,20,21]. During his experiments he used mainly capacitor based pulsed power supply and he also performed time-resolved measurements of the voltage across the wire and the current flowing through the wire. The wire material he used was also high purity (99,9%) copper with higher geometric characteristics (1,0mm diameter, length varied between 150-180mm). These experimental results presented the resistance change and the plasma formation and expansion through x-radiological image, photographic study and measurement of the current and voltage waveforms of the used pulses [19,20,21]. In these experiments the development of the plasma spots has been closely studied using photographic images. According to them the development in the radial direction of the spots was characterized by a sudden expansion to between 5 and 12 mm, followed often by a pause and then a slower expansion to higher lengths [19,20,21]. A typical photographic image of M.J Taylors experiments regarding plasma spot formation is presented below in Figure 7. According also to these experiments, it is feasible to be produced local boiling by introduced local thermal gradients as the forming vapor bubble is free to expand rapidly but constrained longitudinally, due to the existence of the remaining solid parts of the wire [1,2,3,6,7,13,21]. The axial development of the plasma spots follows the typical procedure during the abovementioned multiple arcing. However, discharge conditions are typically longer (relatively steady state) with lower current densities than those employed in similar research experiments regarding with electrothermal chemical (ETC) gun [19,20,21,22]. During their experimental results Molokov and Allen [24] reported that, solid phase fragmentation due to thermal stress waves is likely to occur during the current loading, for current densities in the range 108 109 A m-2. Results of similar to those of Molovcov et al [24] proposed that thermally induced stress or other wave phenomena cause the fragmentation, possibly while the wire is still in the solid phase [24,26,27]. Graneau [23] showed that it is possible for the wire to fragment whilst still in the solid phase. Microscopic examination of the wire in these experiments [23,24] showed little evidence of melting in, but in all cases the wire length was rather high and wit greater diameter than those used in this work. Lukyanov and Molokov [25] proceeded further the experimental and theoretical research as to predict the number of fragments expected by this process. Current densities in the work presented here are somewhat higher at peak current, but may pass through this range while
262
the wire is still solid, especially during low current densities experiments. Whether there is sufficient time for solid phase fragmentation to occur is debatable and especially when the wire length is rather small compared to the one used in the above experiments. [19,20, 23,24] Considering the findings and proposal of researchers regarding the flexural vibrations and fragmentation of wire elements stressed under high current densities [23,24,25] the theory of multiple arcing seem to be supported further. Assuming that the heat stress of the wire is increased and results expansion of the wire element longitudinally the element shaped changes and presents an S-formation instead of the initially straight line. This was photographically proved in recent literature [19,20], and it was expected theoretically [8,10,13,14]. The S-type form presents higher elasticity and lower mechanical strength due to higher mobility in the molecules inside the metallic bonds [13,14,28]. The proven flexural vibrations [23,24,25] stresses further the wire and enhanced the explosion especially in the boiling spots reported above, were globules are formed. This stress increases the mobility of the volatilized wire and enhances the interruption process of the metallic coherence, and consequently the electric arcs formation between the remaining solid parts of the conducting material, in an environment of metal vapors. Experimental work in literature where the multiple fragmentation and arcing was proven [19,20,21] support these assumptions. The same experimental work proved that during the multiple arcing that the solid parts remain in place until the complete evaporation as was stated before. This also supports the assumption that the electric field developed during arcing between the electrodes even though is rather small supports and enhances the procedures. Focusing in a typical part with small time duration the I=f(V) plots can provide more information for these physicochemical process. Figure 8 represents the I=f(V) characteristic of the exploding wires measurement of Figure 2b and 4 for the time intervals between 3.25ms up to 4.75ms and between 13.0ms up to 16.15ms, respectively. These intervals correspond to a pick rise in the voltage drop (with the respective drop in current) along the metal vapour column and to the ending of the fusion process until the arcs extinguished and the current flow reduced rapidly to zero. In the first interval it can be clearly seen the continuous formation of N-type negative differential resistance (NDR) while in the second interval and especially during arcs extinction, the S-type NDR is formed. The observation of NDR existence during the fusion process supports the abovementioned assumptions. Figures 9 and 10 represent the total time duration of the fusion process in thin wire elements exposed in open air. It can be easily observed that the smallest is the length of the wire the higher is the time duration of the fusion process. This seems to explained by the high volume of the metal structure that supports the element which operates as a heat dissipater and since the wire has small length the supporting absorbs a significant amount of the produced heat in the element [3,4,10,13,14]. Further more to this explanation, as it is mentioned above, during the interruption of the metallic coherence of the wire is followed by an electric arc. Thus the ionised metal vapor column has smaller length so its elongation is very small considering higher lengths, and at the same time the current density presents high values (see Figure 3). During this process, the heavily ionised surrounding air is considered that is operating in the smaller gap (smaller length) [1,5,7,10,11,16,17].
263
I = f (V) t [3.25, 4.75]ms -250 -245 -240 -235 -230 -225 -220 -215 -210 -205 -200 -195 -190 -185 -36 V (volt) -38 -40 -42 -44 -46 -48 I (A)
65 60 55 50 45 40 35 30 25 20 15 10 5 V (volt) 0 230 240 250 260 270 280 290 300 310 320 330 340 350 360 370 380 390 400 410 I (A) I = f(V) t [13,16.15] ms
-50
Figure 9. Fusibles wire I=f(V) plots during the interruption process. Focus on two small intervals of Figure 2b and 4, in the time periods between 3.25ms up to 4.75ms and between 13.0ms up to 16.15ms, measuring from the short circuit starting.
264
Figure 9. Waveforms of the average mean value of the fusion process duration in relation with the wire length. The parameter of the current density is also presented. The copper fusible wire has a diameter of 0,14mm.
Considering Figs. 3b and 4, we can clarify how the resistance of a fuse element during short circuit comes into existence and by which factors it is determined. The pre-arcing and arcing periods can be clearly seen, and short circuit duration can be determined with considerable accuracy. The time period where impulse short circuit current is in progress is represented in Figs. 3b and 4 by an area where the resistance is not stabilized. This time period, according to these plots, seems depend from the current density and instant in the sinus power supply voltage in which the short-circuit occurred. After this period the resistance is almost stabilized, with an increasing tense. This may be explained by the increase of the specific resistance due to overheating of the thin fusible element, as it has been explained above in detail. The end of the pre-arcing period is marked by instabilities of the resistance presented to a time instant of few ms after short circuits initiation. Instabilities result from the sinking of the voltage drop, which is a consequence of the progressive and finally complete liquefaction of the element. The phenomena presented could be explained adequately from the above mentioned. Regarding Figs. 3b and 4, it can be notice that resistance during arcing period seems to present two distinctive behaviours (areas). In the first area, the resistance has a rather moderate increasing tense. In the second area, which lasts until the interruption of the current occurs (respectively marked by infinite resistance), fast oscillations groups with corresponding peaks are presented. Those oscillations present a duration (period) of approximately 230 s each and cannot be explained by classical arc theory, as established to
265
date. According to this theory, a metal vapour column sustains electric discharges in which plasma occurs at high pressure and temperature in local thermodynamic equilibrium. It would lead one to expect, as the burnback procedure is in progress, that the vapour column extends lengthwise (arcs elongation) with respective reduction of its cross-section. This normally results in a progressive rapid increase of the arcs resistance. The rapid changes in the arcs dimensions and the expansion of the metal vapour column (which rapidly increases the arc resistance), as well as the possible decrease of the voltage drop across the fuse edges (caused by the source sinusoidal voltage), lead to the arcs eventual extinction. Interpretations regarding the fast oscillations of the resistance presented in Figs. 3b and 4 seem to involve fluctuation of the arcs current density due to the collisional process involving heavy ionized particles and free electrons. It is unlikely that the phenomenon of multiple arcing, caused by globule formation, contributes to the oscillations of the resistance value. A possible explanation of the phenomena could be as follows. The inductance presented in any circuit in practice generally dictates a variation of current. The time constant = L/R of the circuit and the rise time of the fuse resistance during current interruption precise the voltage variation to the end-cups of the fuse. At the moment of the current interruption, overvoltage appears between the still-remaining solid parts of the fuse element. This is reasonable, as the current interruption process is fast.
Figure 10. Waveforms of the average mean value of the fusion process duration in relation with the wire length. The parameter of the current density is also presented. The copper fusible wire has a diameter of 0,16mm.
266
Therefore, it can be assumed that there is ignition of a new arc, which cannot be sustained for long, as the overvoltage to the solid parts of the fuse element no longer exists. This procedure, if repeated, leads to ignition and extinguishment of arcs. Figures 3b and 4support the aforementioned. The final extinguishment of arcs could be owed to an incidental event or to the elongated distance of the two remaining solid parts of the fuse element. The resistance waveform in figure 3b and 4 present significant similarities with the ones in literature and especially with the ones of M.J. Taylor [19,20] even though the experimental procedures were completely different. The differences exist in the waveforms can be easily attribute to the different geometric characteristics and different waveforms. But the similarities provide strong evidence for the plasma formation and disruptions of the electrical coherence occur due to similar physicochemical procedures in both experiments. In figures 8 and 9 also it is observed that the higher the current density is the lower the fusion duration is. This could be easily explained from the above mentioned, since the produced heating power is analog to the term j2, where j is the current density (Joule relation) [3,4,10,17,28]. In these figures it can be easily observed also that in lower current densities the smaller the length is (lower than a threshold value) the higher is the duration of the fusion process. This seems that can be explained from the following assumptions: The heat from the produced Joule energy is lower per time interval. At the same time the energy that is transferred from the thin wire through the end tags is almost stable per time interval and depends from the materials and the volumes. Thus the energy used for melting and arc ignition is lower per time interval, and as a consequence the required time is rising [8,13,18,28]. Furthermore the lower current densities produced lower magnetic pressure values and thus the thin wire seems not to be exploded. In this smaller gap between the electrodes (smaller distance between the end tags, when the wire has smaller length) the metal vapour expands [7]. Here it must be notice that in small lengths of the gap between the end tags, the electric field is higher as it can be easily concluded from the abovementioned [13,18,28]. Thus it seems that the metal vapours which are ionized expand not as rapidly as it has been observed in cylindrical wires with higher lengths where the field is lower and the metal vapours do not remain constrained [7]. These assumptions explain also the differences up 25% that can be observed in the total short circuit duration, in higher lengths of the wire elements and low current densities. Observations concerning the short circuit duration and the measured resistance values in relation to the different geometric characteristics of the fusible elements seem to present a normality, as the different wire diameters imply different resistance and consequently current densities. The corresponding energy to melt is obviously analogous to the diameter of the fusible element. Important role in the fusion duration seems to have the time instant in which this occurs regarding the value of the source voltage. The lower was the source voltage in the starting of the sort-circuit the longer was the duration. This can be explained from the resulted small current density in the beginning of the phenomenon which results a greater pre-arcing period, according to the aforementioned. Of course, differences existed because of the different geometric characteristics of the fusible elements. Those differences, as it can be easily presumed are the consequences of [8,12,14,28]:
Different wires volumes which require different energy values to melt.
267
Different wires resistance and thus different current densities in the elements. The higher current densities have as a consequence smaller affect of the length in the fusion duration.
CONCLUSION
The measurements through fast digital oscilloscopes combined with PCs, can provide a more accurate approximation and understanding of the phenomena during the operation of the fuse elements. Another parameter that contributes to the above is the use of common industrial voltage source and random sinus values in the experiments. The resistance measurements combined with the voltage drop and current densities measurements can provide a more accurate approximation and understanding of the phenomena involved. Overvoltages during current interruption process, due to circuits inductance, provide a possible explanation to the phenomena. The I=f(V) plots supports these assumptions, with the observed continuous NDR formation. The observed fast oscillations for different wire lengths and current densities provided useful information concerning the fusion process duration and thus for short-circuit, which fuse the thin wire. The fusible elements geometric characteristics affect only the fusions process duration and the values of the measured magnitudes, for small current densities. The results could be useful in understanding the fusion process.
REFERENCES
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Barrow, D.R.; Howe A.F.; Cook N. IEE Proc. A. 1991, 138(1), 83-88. Cheim L.A.V.; Howe A.F. IEE Proc. Sci. Meas. Technol. 1994, 141(2), 123-128. Wright A.; Newbery P.G. Electric fuses (2nd Edition); IEE-Peter Peregrinus Ltd: London, 1995; pp 1-52, pp 112-117 Leach J.G.; Newbery P.G.; Wright A. IEE Proc. 1973, 120(9), 987-993. Narancic V.N.; Fecteau G. Ibid., 236-251. Chicata T.; Ueda Y.; Murai Y. ; Miyamoto T. Proc. of Inter. Conf. on Electric Fuses and their Applications 1976, 114-121. Vermij L. IEEE Trans. Plasma Sc. 1980, PS-8(4), 460-468. Morgan V.T. Proc. IEE 1971, 118(3/4), 555-570. Lebedev D.V. Sov. Phys. JETP, 1957, 5(2), 243-252. Psomopoulos C.S.; Karagiannopoulos C.G. Electric Power Systems Research, 2002, 61, 161-167. Vitkovitsky I.M.; Scherrer V.E. J. Appl. Phys 1981, 52(4), 3012-3015. Berger T.L. IEEE Trans. Plasma Scien. 1980, PS-8(3), 213-216. Rdenberg R. Transient Performance of Electric Power Systems, M.I.T. Press: 1970, p.p.431-453. Chapman A.J. Heat Transfer, 4th Edition, Maxwell Macmillan International: New York, NY, 1989. Saqib M.A.; Stokes A.D. Thin Solid Films 1999, 345, 151-155.
268 [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28]
C. S. Psomopoulos and C. G. Karagiannopoulos Bussire W.; Bezborodko P. J. Phys. D: Appl. Phys. 1999, 32, 1693-1701. Psomopoulos C.S.; Karagiannopoulos C.G. Measurement Journal of the International Measurement Confederation 2002, 32, 15-22. Raizer Y.P. Gas Discharge Physics, Springer-Verlang: Heidelberg, Germany, 1991; pp 128-161, pp 245-288, pp288-291. Taylor M.J. J. Phys. D.: Appl. Phys. 2002, 35, 700-709. Taylor M.J. Proc. of 7nth Inter. Conf. on Electric Fuses and their Applications 2003, 1-9. Taylor M.J.; Dunnet J. IEEE Trans. Megnetics 2003, 39(1), 269-274. Powell J.D. Thornhill L.D. IEEE Trans. Mag. 2001, 37(1), 183-187 Graneau P. Phys. Lett. 1987, 120A, 77-79. Molokov S.; Allen J.E. J. Phys. D: Appl. Phys. 1997, 30, 3131-3141. Lukyanov A.; Molokov S. J. Phys. D: Appl. Phys. 2001, 34, 1543-1552. Me-Bar Y.; Harel R. J. App. Phys.1996, 79(4), 18641868. Coffman M.L. Exploding Wires; W.G. Chace and H.K. Moore: New York, NY, 1964, vol 3, pp 89102 Karagiannopoulos C.G. International Journal of Power and Energy Systems 2002, 22(3), 142-145.
Chapter 9
WIDE-AREA MONITORING OF POWER SYSTEM OSCILLATIONS IN THAILAND POWER NETWORK USING SYNCHRONIZED PHASOR MEASUREMENT UNITS VIA DEMAND SIDE OUTLETS
Issarachai Ngamroo1,* and Yasunori Mitani2,
Department of Electrical Engineering, Faculty of Engineering, King Mongkuts Institute of Technology Ladkrabang, Bangkok 10520, Thailand 2 Department of Electrical Engineering, Graduate School of Engineering Kyushu Institute of Technology, Fukuoka 804-8550, Japan
1
ABSTRACT
The longitudinal structure of the tie-line interconnection between central and southern areas of Thailand power system causes the dynamic stability problem due to the inter-area oscillation with poor damping. To detect the dominant inter-area oscillation mode, a sophisticated wide-area dynamic stability monitoring system based on the phasor measurement units (PMUs) is proposed. The main feature of the presented system is the convenient installation of PMU at 220 V level. PMUs are located at two universities representing both areas. The single-phase phasor voltage data measured by PMUs at both sites are synchronized by global positioning system. The phase voltage difference data are used to analyze the dynamic characteristic of the inter-area oscillation based on signal processing techniques. The proposed monitoring system provides vital information of wide-area dynamic stability.
Keywords: Synchronized phasor measurement units, Wide-area monitoring, Power system dynamic stability, Inter-area oscillation, Signal processing
Email: knissara@kmitl.ac.th Email: mitani@ele.kyutech.ac.jp
270
Issarachai Ngamroo and Yasunori Mitani
INTRODUCTION
The interconnection between central and southern areas by 230 kV tie-line in Thailand 50 Hz electric power system establishes a longitudinal structure. This configuration causes the problem of inter-area oscillations. Especially, the dominant inter-area oscillations with poor damping can deteriorate the dynamic stability of the system. However, such inter-area oscillations between both areas have never been observed and analyzed before. The practical monitoring system as a sophisticated energy management system is highly expected. Recently, the phasor measurement units (PMUs) which are synchronized based on the time stamp of the global positioning system (GPS) [1,2] have been applied to power systems such as monitoring of power system dynamics [3,4], analysis of power system events [5,6], detection of transient power swing [7], wide-area stabilizing control [8] etc. In these works, synchronized phasor measurement systems have been established at a high-voltage transmission level. To increase the feasibility of the wide-area monitoring of power system oscillations, a new monitoring system with convenient installation at a distribution level, high accuracy of the measured data, Internet based data transmission, and cost-effective is highly expected. To achieve these requirements, the Internet based GPS synchronized wide area frequency monitoring network (FNET) developed at the 110-V distribution level has been established in US system [9]. The objective of FNET is to create an extremely low cost and quickly deployable wide-area frequency measurement system. Moreover, the power system frequency can be accurately measured and synchronized at 110-V typical demand side outlet. Many applications of FNET are described in [9]. In this paper, a wide-area dynamic stability monitoring system developed in Thailand power network is proposed. The monitoring system is established by GPS-synchronized PMUs. The salient feature of the monitoring system is the convenient installation of PMUs at 220 V domestic outlets. Besides, the initial installation cost is very low. In our system, PMUs are located at two universities which represent central and southern areas of Thailand power system. The PMU data transmission between both universities is performed via the Internet. The phase difference between two universities can be calculated from phasor voltages collected by PMUs. By using the phase difference data, the physical characteristics of power oscillations can be analyzed by Fast Fourier Transform (FFT) and discrete wavelet decomposition. In addition, the 2nd order model of the equivalent two-machine system is employed to establish the mathematical model of the interconnection areas embedded with the dominant oscillation mode. Then, the least-squares regression is applied for system identification. Consequently, the eigenvalue corresponding to the oscillation mode can be determined. Monitoring of variation of eigenvalue not merely clarifies the oscillatory behavior, but also provides applicable information of the wide-area dynamic stability. The organization of this paper is as follows. First, the proposed monitoring system is overviewed. Next, the noise elimination, the analyzed results using FFT and discrete wavelet decomposition are carried out. Subsequently, system identification studies by least-square error technique are provided. Lastly, a conclusion is given.
Wide-Area Monitoring of Power System Oscillations in Thailand Power Network 271
PMU-BASED MONITORING SYSTEM

An overview of the Thailand power system is delineated in Figure 1. The Electricity Generation Authority of Thailand (EGAT) is responsible for generation and transmission grids of high voltage levels. Central and southern areas are connected by 230 kV tie-line with 800 km long. This interconnection establishes a longitudinal structure between both areas which may cause an inter-area oscillation. To monitor the inter-area mode between both areas, PMUs are located at King Mongkuts Institute of Technology Ladkrabang (KMITL), Bangkok and Prince of Songkla University (PSU), Songkla, which represent central and southern areas, respectively.
Figure 1. Location of PMUs.
272
PM U
220 V Outlet
Figure 2. PMU based monitoring system via 220 V outlet.
Figure 2 shows the proposed PMU-based monitoring system via conventional power outlets. The phasor measurement system employs a manufactured PMU as a commercial product, which is a Network Computing Terminal Type-A, NCT2000 of Toshiba [10,11]. The PMU can measure a single-phase instantaneous voltage at 220 V outlet, with correcting its clock based on the time stamps of GPS. As the PMU uses the time stamps of GPS for its clock correction, the time synchronization between two PMUs located at distant places is easily accomplished by keeping their internal clocks synchronous with the time stamps of GPS. As given in [10,11], the phasor voltage is computed by
V =
N 2 N Vk sin k + j Vk cos k N k =1 k =1
(1)
& where, V is the phasor voltage, Vk is a sequential data of the instantaneous voltage of 220 V
outlet sampled by an A/D converter, N is the sampling number of the voltage and is the sampling angle. For this apparatus, N = 96 and = 2 N = 3.75deg. Based on (1), the phase angle can be calculated by
& & = tan 1 Im (V ) Re (V )
(2)
& & & where, Re(V ) and Im(V ) are the real part and the imaginary part of the phasor voltage V ,
respectively. The phase angle is accumulated in the PMU as the time sequential data. Calculation in (2) provides phase angle referred to the GPS time. The PMU records the calculated phasor voltages every 40 ms (2 cycles) and measures at the domestic outlets for 20 minutes twice an hour, for example, 0.20-0.40 a.m. and 0.50-1.10 a.m. etc. Hence, there are 30,000 data of phasor voltages for 20 minutes. The measured phasor voltages at PSU are transmitted via the Internet to a server at KMITL.
EVALUATION OF PMU DATA

Generally, the phase difference between two interconnected areas reflects the active power flow in tie-lines between both areas. In this part, the variations of phase difference between KMITL and PSU in two periods are considered. The average value of phase differences for 20 minutes twice an hour in each day is calculated. Thus, there are 48 average phase difference data per day. Figure 3 shows the variations of the average phase difference between KMITL and PSU from June 1 to 7, 2005. The variation of the average phase difference on each day is significantly different. There are both positive and negative values. The positive average phase difference implies that the active power flows from the central area to the southern area and vice versa. To verify the validity of the measured data of PMU during steady-state conditions, the average phase difference is compared to the actual power flow data of 230 kV tie-lines between central and southern areas of EGAT from June 1 to 7, 2005, as demonstrated in Figure 4. Because the power flow data of EGAT are measured at every half hour, there are also 48 data per day same as a number of average phase difference data of PMU. Obviously, the average phase difference changes in the same fashion as the actual power flow. Besides, the direction of power flow is consistent with the sign of the average phase difference. Figures 5 and 6 depict the average phase difference between KMITL and PSU measured by PMU and the actual tie-line power flow of EGAT from July 4 to 10, 2005, respectively. The same variations of the average phase difference and the actual power flow are also obtained in this period. These comparison results signify that the variation of the actual tie-line power flow can be conveniently observed by the variation of the average phase difference between KMITL and PSU measured by PMUs via demand side outlets.
Figure 3. Average phase differences between KMITL and PSU, June 1 7, 2005.
274
Figure 4. Actual tie-line power flow from EGAT, June 1 - 7, 2005.
Figure 5. Average phase differences between KMITL and PSU, July 4 - 10, 2005.
Figure 6. Actual tie-line power flow from EGAT, July 4 - 10, 2005.
Figure 7. Least-squares linear equation, June 1 7, 2005.
276
Figure 8. Least-squares linear equation, July 4 10, 2005.
The correlation between the actual power flow data of EGAT ( P ) and the average phase 12 difference data between KMITL and PSU measured by PMU ( ) is evaluated by a least-
squares regression. The linear equation P = K12 + c with two variables is used as a 12
regression model. The resulted least-square linear equations of the actual power flow and the average phase difference from June 1 to 7, 2005 and from July 4 to 10, 2005 can be delineated in Figures 7 and 8, respectively. To evaluate the fitness of data, the correlation coefficient between two variable is calculated by
R2
n ( xy ) ( x )( y ) = n x 2 ( x )2 n y 2 ( y )2
(3)
where x is average phase difference data, y is actual power flow data, and n is the number of data. The correlation coefficient R 2 indicates the fitness of data. If R 2 is close to unity, then the power flow strongly correlates with the average phase difference. If not, the correlation is weak. As given in Figs 7 and 8, the correlation coefficients of two variables in both periods are nearly equal to one. The phase difference data measured by PMUs strongly correlates with the actual power flow data of EGAT during steady-state conditions. The correlation equations of Figures 7 and 8 have very different parameters K12 and c . This is
Wide-Area Monitoring of Power System Oscillations in Thailand Power Network 277 because the measurement is taken at 220 V distribution level. The phase shifts of distribution transformers significantly affect the correlation. Additionally, the correlation is also a function of the local loads through the distribution network.
(a) 2.20 a.m. - 2.40 a.m.
(b) 8.20 a.m. - 8.40 a.m. Figure 9. (Continued)
278
(c) 2.20 p.m. - 2.40 p.m.
(d) 8.20 p.m. - 8.40 p.m. Figure 9. Phase differences between Bangkok and Songkla.
SIGNAL PROCESSING FOR DETECTION OF OSCILLATION MODES

In the proposed monitoring system, the phase angles of single-phase voltages at both locations are measured simultaneously. Because, the time derivative of phase angle corresponds to the deviations of system frequency, the frequency deviation data at any location can be generated. After applying Fast Fourier Transform (FFT) [12] to phase angle
Wide-Area Monitoring of Power System Oscillations in Thailand Power Network 279 or frequency deviation at the single location as well as phase difference between two locations, it is found that using phase difference is much easier to observe the dominant power oscillation mode. Accordingly, the phase difference is used to analyze the power oscillation mode.
Figure 10. FFT results of phase difference before noise elimination.
Figure 9 depicts waveforms of phase difference between Bangkok and Songkla every six hours from 2.20 a.m. to 2.40 a.m., 8.20 a.m. to 8.40 a.m., 2.20 p.m. to 2.40 p.m. and 8.20 p.m. to 8.40 p.m. on June 4, Saturday, 2005. Positive and negative phase differences imply the opposite direction of the tie-line power flow. It can be seen that the phase difference signals are composed of many oscillatory components with different frequencies. However, in the phase difference signals, it is possible to contain noises caused by the switching of different loads. In general, noises due to such loads occur at high frequencies which are beyond the frequency of power system oscillations. Thus, the noises elimination technique is carried out prior to the detection of power oscillation modes. Figure 10 shows the FFT result of phase difference signal from 8.20 p.m. to 8.40 p.m. It can be observed that the phase difference signal contains noises at high frequencies. To eliminate noises, the discrete wavelet decomposition [13,14] is employed to decompose the original phase difference signal into signals with different frequencies. Figure 11 illustrates wavelet decomposition results of phase difference from 8.20 p.m. to 8.40 p.m. by Symlet wavelet functions with 12 levels. By applying FFT to each wavelet element (d1-d12) in Figure 11, the frequency range of each element can be determined as in Table 1. It is well known that power system oscillations such as inter-area oscillations and local generator oscillations occur at frequencies below 2 Hz [15]. As shown in Table 1, the d1, d2, and d3 signals are at high frequencies. These signals are noises which do not contain power oscillations. Accordingly, d1, d2, and d3 signals are isolated from the original phase difference signal, in order to eliminate any effects of the
280
switching loads. As a result, the original phase difference signal can be decomposed into signals containing mainly power oscillations. Figure 12 demonstrates the FFT result of the phase difference signal after noises elimination. Explicitly, the noises at high frequency are absolutely eliminated. Figure 13 shows the FFT results of phase differences between Bangkok and Songkla after noises elimination every six hours from 2.20 a.m. to 2.40 a.m., 8.20 a.m. to 8.40 a.m., 2.20 p.m. to 2.40 p.m. and 8.20 p.m. to 8.40 p.m. on June 4, Saturday, 2005. The FFT results signify that the dominant power oscillations at each elapsed time are observed at a frequency around 0.4 - 0.6 Hz, which is in the vicinity of the inter-area oscillation mode (0.2 0.8 Hz) [15]. In addition, the oscillation frequency of the inter-area mode varies with the elapsed time. This is due to the variation of power demands in each elapsed time.
Table 1. Frequency range of each wavelet element
Wavelet level d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11 d12 Frequency (Hz) >4 1.5 10.2 1.3 4.52 0.52 2.32 0.2 1.23 0.11 0.52 0.05 0. 19 0.03 0.15 0.01 0.08 0.01 0.04 0.005 0.03 0.001 0.01
Figure 11. (Continued)
282
Figure 11. (Continued)
284
Figure 11. Wavelet decomposition results of phase differences.
Figure 12. FFT results of phase difference after noise elimination.
(a) 2.20 a.m. - 2.40 a.m. Figure 13. (Continued)
286
(b) 8.20 a.m. - 8.40 a.m.
(c) 2.20 p.m. - 2.40 p.m.
(d) 8.20 p.m. - 8.40 p.m. Figure 13. FFT results of phase differences.
IDENTIFICATION OF INTER-AREA OSCILLATION MODE
Figure 14. Enlarged view of the amplitude of the combination of d4, d5 and d6 elements.
288
Figure 15. FFT of the combination of d4, d5 and d6 elements.
As provided in Table 1, frequencies of d1-d3 and d7-d12 elements are not in the vicinity of the inter-area oscillation mode. Frequencies of the remaining d4, d5 and d6 elements are in the range of interest. Accordingly, the combination of d4, d5 and d6 elements (d4+d5+d6) is justified to be the signal containing the inter-area oscillation mode. Figure 14 outlines the enlarged view of the amplitude in degree of the combination of d4, d5 and d6 elements of phase difference between 310 and 330 seconds from 8.20 to 8.40 p.m. Clearly, the oscillation period is about 2.0 second. Figure 15 shows the FFT result of the combination of d4, d5 and d6 elements. The dominant frequency occurs at a frequency around 0.4 - 0.6 Hz. To identify power oscillation mode from phase difference data, the multi-variable regression method is applied. It is assumed that the central and southern areas are represented by an equivalent two-machine system. The swing equation is given by
&& && & & M 1 2 + D 1 2 + K ( 1 2 ) = 0
(4)
where M is an inertia constant, D is a damping coefficient, K is a synchronizing power coefficient, 1 and 2 are phase angles at Bangkok and Songkla, respectively, &1 and &2 are && && the first derivative of phase angles at Bangkok and Songkla, respectively and and are
1 2
the second derivative of phase angles at Bangkok and Songkla, respectively. Rewriting (4) as
&& && 1 2 =
D & & K 1 2 ( 1 2 ) M M
(5)

&& && By applying the wavelet decomposition to measured data 1 , 2 , &1 , &2 , 1 and 2 ,
extracting the combination of d4, d5 and d6 elements of each variable and substituting in (5), the coefficients D M and K M can be estimated by a multi-variable regression method. Based on this identification technique, the steady state phasor fluctuations can be used to model the system dynamics. In other words, a large disturbance like a line fault is not necessary. As an example, the phase difference from 8.20 p.m. to 8.40 p.m. on June 4, 2005 is used to identify the two-machine system. As a result, the estimated values of D M and K M are equal to 0.3617 and 8.9480, respectively. To evaluate the regression results, the combination && && of d4, d5 and d6 elements of the 1 and 2 are substituted in the left side of (5), while the estimated D M , K M and the combination of d4, d5 and d6 elements of , , & , & are
1 2 1 2
replaced in the right side of (4). Figure 16 shows the comparison results of measured values (left side of (5)) and estimated values (right side of (5)). Clearly, both measured and estimated values are almost the same. Since (4) is a second-order system, the characteristic equation can be represented by
2 s 2 + 2n s + n = 0
(6)
Accordingly, the undamped natural frequency ( n ) and the damping ratio ( ) are determined by
n = =
K M D 2 M n
(7)
(8)
The undamped natural frequency and the damping ratio lead to the calculation of the eigenvalues. Suppose that the eigenvalues corresponding to the oscillation mode are jd , the real part ( ) and the imaginary part (d ) of the eigenvalues can be determined by
= n
d = n 1 2
(9) (10)
As a result, the eigenvalues corresponding to the oscillation mode from 8.20 p.m. to 8.40 p.m. on June 4, 2005 are 0.1808 j 2.9858 . The oscillation frequency is 2.9858 rad/sec (0.4752 Hz) and the damping ratio is 0.0605.
290
Figure 16. Comparison of measured and estimated values.
Figure 17. Variation of real parts of eigenvalues, June 1 7, 2005.
Figure 18. Variation of imaginary parts of eigenvalues, June 1 7, 2005.
Figure 19. Variation of real part of eigenvalue, July 4 - 10, 2005.
292
Figure 20. Variation of imaginary part of eigenvalue, July 4 - 10, 2005.
Figures 17 and 18 indicate the variations of real part and imaginary part of eigenvalues from June 1 to 7, 2005, respectively. The variation of the real part in each day has almost the same pattern. In the nighttime, the oscillation mode is more stable than that in the daytime. This is due to large electric power demand in the daytime which deteriorates the system stability. For the imaginary part which is the oscillation frequency, it also varies with the power demand in each day. The more the power demand increases, the lower the oscillation frequency is. In the daytime, the participation of synchronous generator in the power system increases. This is equivalent to a larger system inertia which leads to the longer oscillation period and the lower oscillation frequency. On the other hand, the oscillation frequency becomes higher due to low power demand in the nighttime. Figures 19 and 20 depict the variations of real part and imaginary part of eigenvalues from July 4 to 10, 2005, respectively. The oscillatory behavior can be described as in the former period. These results exhibit that the dynamic stability of the power system based on eigenvalue analysis can be conveniently monitored via home power outlets.
CONCLUSION
This paper proposes a wide-area dynamic stability monitoring based on synchronized PMUs. The aim of the presented system is to monitor the inter-area oscillation in the 230 kV tie-line interconnection between central and southern areas of Thailand power system. The proposed system has many advantages such as a convenient installation at a 220 V home outlet level, an Internet based data transmission, an easy access for university researchers and
Wide-Area Monitoring of Power System Oscillations in Thailand Power Network 293 a low capital cost. Signal processing techniques based on Fast Fourier Transform and Discrete Wavelet Decomposition have been applied to eliminate noises in the measured signal and analyze physical behavior of inter-area oscillation. System identification using least-square error has been applied to calculate the eigenvalue corresponding to the dominant inter-area mode. Consequently, the investigation of system dynamic stability can be easily achieved by monitoring the variation of the dominant inter-area mode in any time period. The proposed monitoring system directly provides applicable information of wide-area power oscillations for power system researchers, energy managers, system operators, and university students. It leads to the high feasibility of wide-area monitoring and stability analysis for the future electrical power systems.
ACKNOWLEDGEMENTS
This work was supported by the Thailand Research Fund under Research Career Development Grant No. RSA4780015.
REFEERENCES
Phadke AG. Synchronized phasor measurements in power systems. IEEE Computer Application in Power. 1993;6(2):10-15. [2] Martin KE et al. IEEE standard for synchrophasors for power systems. IEEE Trans. Power Delivery. 1998;13(1):73-77. [3] Hauer JF, et al. Keeping an eye on power system dynamics. IEEE Comp. App.in Power. 1997;10(4):50-54. [4] Hauer JF, Bhatt NB, Shah K, and Kolluri S. Performance of WAMS East in providing dynamic information for the North East blackout of August 14, 2003. Proc. IEEE PES General Meeting 2004; 1685 1690. [5] Rasmussen J. and Jorgensen P. Synchronized phasor measurements of a power system event in Eastern Denmark. IEEE Trans. Power Syst. 2006;21(1): 278-283. [6] Burnett RO Jr. Butts MM, Cease TW, Centeno V, Michel G, Murphy RJ, and Phadke AG. Synchronized phasor measurements of a power system event. IEEE Trans. Power Syst. 1994; 9(3):1643-1650. [7] Liu CW, Thorp JS, Lu J, Thomas RJ, and Chiang HD. Detection of transiently chaotic swings in power systems using real-time phasor measurements. IEEE Trans. Power Syst.1994;9(3):1285-1292. [8] Kanwa I, Grondin R, and Hebert Y. Wide-area measurement based stabilizing control of large power systems a decentralized/hierarchical approach. IEEE Trans. Power Syst.2001;16(1):136-153. [9] Zhong Z, Xu C, Billian BJ, Zhang L, Tsai SJS, Conners RW, Centeno VA, Phadke AG, and Liu Y. Power system frequency monitoring network (FNET) implementation. IEEE Trans. Power Syst. 2004;20(4):1914-1921. [10] Tsukui R, Beaumont P, Tanaka T and Sekiguchi K, Power system protection and control utilizing Intranet technology, Power Eng Journal. 2001;15:249-255. [1]
294
[11] TOSHIBA Corporation. PMU Document. 2001. [12] MATLAB Signal Processing Toolbox, Users Guide. The MathWorks. 2005. [13] Wilkinson WA and Cox MD, Discrete wavelet analysis of power system, IEEE Trans. Power Syst. 1996;11(4):2034-2044. [14] MATLAB Wavelet Toolbox, Users Guide. The MathWorks. 2005. [15] Rogers G. Power System Oscillations. Kluwer Acaemic. 2000.
INDEX
aspiration, 25 assessment, 37, 44, 99, 152, 211, 212, 213, 218, 238, 249, 250 assignment, 135 assumptions, 121, 123, 124, 130, 131, 133, 168, 262, 266, 267 Athens, 152, 153, 155, 174, 251 atmospheric pressure, 106, 157 atoms, 68 attention, 21, 51, 215, 249 automobiles, 75 availability, vii, x, 1, 2, 4, 6, 18, 19, 21, 26, 28, 29, 32, 37, 38, 39, 41, 56, 94, 106, 120, 209, 211, 212, 216, 217, 227, 239 avoidance, 12
A
absorbents, 84, 93, 94 abundance, vii, 4 AC, 64, 67, 156, 169, 172, 175 access, x, 209, 211, 229, 230, 292 accuracy, 105, 106, 108, 109, 112, 113, 116, 119, 124, 126, 150, 157, 215, 231, 237, 239, 254, 264, 270 achievement, 12 acid, 2, 64, 65, 66, 67, 72, 73, 75 activation, 66 adjustment, 213 adsorption, 94, 96 AEA, 43 aerospace, 224 Africa, 43 ageing, 162, 163 agent, 13 air pollution, 179 air quality, 54 algorithm, 107, 142, 223, 224, 240, 242, 246 alternative, 4, 6, 19, 26, 28, 29, 37, 40, 42, 51, 115, 135, 157, 179, 188, 191 alternatives, 22, 40, 139, 142 aluminium, 28 ambient air, 84, 86 ammonia, 86, 87, 89, 188, 190 amortization, 140 amplitude, 160, 161, 162, 167, 168, 235, 287, 288 AN, 172 antimony, 67 arsenic, 67 artificial intelligence, 215 ash, viii, 1, 4, 12, 13, 14, 22, 25, 26, 28, 29, 37, 38, 39, 40, 44 aspect ratio, 11
B
banks, 247, 250 barriers, viii, 1, 20, 40, 69 batteries, 56, 64 behavior, ix, x, xi, 10, 12, 14, 22, 29, 59, 155, 209, 211, 227, 228, 252, 259, 270, 292, 293 Belgium, 21 benefits, 2, 8, 18, 20, 96, 98, 119, 120, 121, 138, 199, 219, 220 bias, 111, 122, 125, 133 biodegradable, 20 biofuel, 43 biofuels, 40, 43, 53, 56, 58 bio-fuels, viii, 1, 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, 14, 15, 18, 19, 20, 21, 22, 29, 32, 34, 35, 38 bio-fuels, 41 bio-fuels, 42 bio-fuels, 42 bio-fuels, 42 biological activity, 11
296
Index
chloride, 93, 94 chlorine, 13, 25, 26, 27 classes, 166, 167 classification, ix, 31, 38, 79, 138, 155, 166, 167, 168, 173, 188, 191, 192, 233, 250 cleaning, 31, 173 closure, 121, 124 clusters, viii, 56, 103 CO2, 2, 13, 15, 16, 21, 26, 29, 32, 34, 35, 38, 41, 42, 67, 74, 179 coal, vii, 1, 2, 3, 4, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 31, 32, 34, 35, 36, 37, 38, 39, 40, 41, 43, 44, 49 coal particle, 22, 32 coefficient of variation, 253 coherence, 259, 262, 266 cohesion, 10 collaboration, ix, 25, 155 collisions, 260 combustion, 3, 10, 12, 14, 15, 16, 21, 22, 25, 26, 27, 28, 29, 31, 37, 38, 39, 42, 43, 44, 53, 54, 55, 56, 58, 59, 62, 71, 77, 78, 79, 82, 95 combustion chamber, 39 communication, 230 compatibility, 179 compensation, 138, 143, 214 competition, 210 complexity, 98, 210, 213 components, viii, 4, 14, 39, 41, 47, 48, 50, 52, 53, 58, 65, 66, 67, 69, 71, 78, 79, 83, 84, 86, 96, 98, 158, 180, 181, 241, 242, 247, 248, 279 composition, 5, 14, 27, 42 compost, 34 compounds, 13, 14, 188 computation, 106, 108, 114, 119, 150, 213 computing, 211, 219, 221 concentration, 14, 22, 26, 39, 42, 66, 67, 87, 94, 164 concrete, 15, 25 condensation, 55, 79, 84, 86, 157 conditioning, 48, 71, 79, 86, 87, 91, 94 conduction, 164, 174 conductivity, 69, 156, 157, 159, 161, 162, 163, 164, 169, 174, 187 conductor, 259 configuration, ix, 14, 25, 58, 89, 94, 98, 177, 180, 185, 199, 202, 231, 232, 238, 270 Congress, 207 connectivity, 99 conservation, 8 constraints, vii, 1, 3, 13, 30, 37, 78, 113, 116, 135, 138, 139, 142, 144, 213, 214, 218, 219, 223 construction, 3, 8, 12, 15, 28, 38, 63, 67, 71, 74, 223, 229
biomass, vii, 1, 2, 3, 4, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 24, 25, 26, 28, 29, 32, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 179 biomass materials, 2, 3, 6, 7, 9, 10, 11, 12, 13, 14, 15, 18, 19, 21, 30, 35, 40 blackouts, 51, 97, 210, 218, 222, 234 boilers, 2, 3, 13, 15, 21, 22, 27, 28, 32, 38, 39, 40, 41, 43, 44, 83 boils, 86 bonds, 164, 262 branching, 164 breakdown, 173, 174 Britain, 17, 41 Bromide, 88 burn, 22, 31, 42, 44, 260 burning, 12, 13, 15, 59, 62 burnout, 13, 25 by-products, 13, 14, 40
C
C++, 175 cables, 231, 254 calcium, 13, 39, 94 calibration, 12 California, 99 campaigns, 20 Canada, 210, 222 candidates, ix, 67, 178, 179, 188 capacitance, 164, 170, 236, 254 capital cost, 20, 36, 42, 53, 55, 58, 59, 62, 90, 98, 99, 190, 293 capital expenditure, 4, 9, 20, 35, 37 carbon, 8, 12, 16, 17, 19, 22, 26, 28, 64, 70, 71, 72, 73, 74, 75 carbon dioxide, 8, 64, 70, 71, 74, 75 carbon monoxide, 70, 71 carbonyl groups, 164 carrier, 75, 163 case study, 35, 36, 37, 43, 99 catalyst, 14, 53, 64, 67, 71, 72, 77 catalysts, 74 categorization, 167 category b, 161 cell, 64, 65, 66, 67, 69, 71, 72, 73, 75, 76, 78 Central Europe, 21 ceramic, 70, 71, 73, 173, 175 cereals, 2 certainty, 210 certificate, 42 channels, 94, 157, 232 chemical composition, 13 chemical energy, 78
Index
consulting, 43 consumers, vii, 119 consumption, 12, 13, 29, 35, 51, 179 contaminants, 94, 156, 162 contamination, 156, 163, 164, 166, 171, 172, 173, 215 contingency, x, 134, 209, 211, 218, 224 continuity, 10 control, viii, 9, 10, 11, 12, 24, 31, 37, 40, 43, 69, 71, 84, 92, 103, 104, 105, 134, 137, 138, 139, 140, 142, 143, 146, 150, 151, 181, 182, 210, 211, 212, 213, 214, 215, 219, 220, 222, 223, 225, 229, 230, 250, 270, 293 convergence, 142, 167 conversion, 11, 13, 21, 30, 31, 40, 49, 53, 62, 63, 64, 66, 69, 78, 158, 207, 232 conversion rate, 13, 158 cooling, viii, 47, 48, 50, 52, 53, 54, 55, 57, 58, 62, 74, 76, 78, 84, 91, 92, 93, 96, 98, 179, 208, 215, 239 cooling process, 91, 96 Copenhagen, 151 copper, 254, 255, 256, 259, 261, 264, 265 corona discharge, 165 correlation, 27, 108, 156, 276 correlation coefficient, 276 corrosion, 8, 12, 13, 25, 26, 27, 29, 30, 40, 41, 62, 75, 90, 179 cost benefit analysis, 220 cost of power, vii cost saving, 96 cost-benefit analysis, 212 costs, 6, 8, 10, 18, 19, 20, 30, 32, 37, 40, 41, 42, 51, 54, 55, 57, 58, 62, 64, 65, 67, 69, 71, 76, 78, 89, 96, 97, 99, 104, 123, 124, 129, 139, 140 coupling, 41, 89, 232 covering, 40, 207 creep, 169 Crete, 175 crops, 2, 4, 21, 36 customers, 49, 105, 120, 210 cycles, 15, 179, 180, 207, 272
297
D
damping, xi, 269, 270, 288, 289 data availability, 239 data base, 230 data set, 233 data transfer, 158 database, 230 dating, 51
DD, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 146, 147, 148, 149, 150 decay, 164 decision making, 8, 100, 229 decisions, 116 decomposition, 31, 246, 247, 248, 250, 270, 279, 284, 289 defects, 164 deficiency, 244 deficit, 119 definition, 228 degradation, 90, 234 delivery, vii, 12, 14, 20, 32, 49 demand, 10, 26, 44, 56, 98, 99, 116, 146, 151, 214, 270, 273, 292 Denmark, 3, 21, 24, 41, 116, 118, 151, 152, 293 density, xi, 12, 25, 28, 126, 156, 159, 164, 174, 252, 254, 255, 256, 262, 264, 265, 266 Department of Defense, 100 Department of Energy, 100, 210 deposition, viii, 1, 25, 44, 159 deposits, 11, 13, 26, 28, 39, 161 deregulation, viii, 47 designers, 206 desire, 76 destruction, 28 detection, 28, 250, 270, 279 deviation, 111, 112, 121, 211, 217, 278 devolatilization, 11 DFT, 240, 241, 242 diamonds, 131 dielectrics, 163, 164, 173, 174 diesel engines, 179, 207 diesel fuel, 53 dilation, 246 diodes, 174 dioxin, 38 discharges, ix, 155, 156, 159, 161, 162, 163, 165, 168, 169, 170, 171, 173, 174, 265 discrete variable, 142 discrimination, 159 discs, 232 distortions, 234, 250 distribution, vii, x, 32, 49, 51, 112, 114, 126, 128, 134, 137, 166, 219, 223, 229, 249, 251, 252, 270, 277 district heating, 21, 44, 83 division, 158 domestic markets, 101 drying, 9, 11, 31, 34, 39 duplication, 12 durability, 63
298
Index
energy density, 6, 10, 11, 19, 41, 77 energy management system, 270 energy markets, 210 energy recovery, 207 energy supply, 42 energy transfer, 260 England, 60, 152 enlargement, 40 entanglements, 164 entropy, 66, 181, 185, 187, 188, 206 environment, x, 2, 15, 31, 32, 71, 84, 86, 156, 157, 159, 179, 209, 210, 211, 214, 218, 225, 260, 262 environmental impact, 37 environmental issues, vii, 1, 2, 13, 14, 20, 38, 51 environmental regulations, 134 environmental threats, 15 epoxy, 165 equality, 141 equilibrium, 170, 234, 252 equipment, 3, 4, 8, 9, 10, 12, 20, 24, 31, 37, 38, 42, 52, 53, 54, 55, 56, 65, 78, 92, 94, 96, 179, 188, 210, 211, 214, 221, 228, 229, 230, 231, 234, 237, 257 erosion, 13, 168, 215 error detection, 218 estimating, 105 ethanol, 188, 190 ethylene, 94 ethylene glycol, 94 EU, 15, 16, 30, 38, 40, 43 eucalyptus, 4 Europe, vii, 1, 3, 4, 40, 41, 42, 60, 101, 119, 222 European Commission, 42, 43, 45, 71 European Community, 2 evacuation, 116 evaporation, 79, 170, 260, 262 evidence, 13, 261, 266 excitation, 174 exothermic, 65 expenditures, 42, 48 expertise, 210 exploitation, 37, 40 exposure, 156 extinction, 165, 174, 260, 262, 265
duration, x, xi, 159, 162, 168, 170, 171, 234, 239, 240, 244, 251, 252, 253, 257, 259, 262, 264, 266, 267
E
economic losses, viii, 103, 104, 137 economics, vii, 2, 6, 19, 20, 37, 42, 43, 44, 69 Education, 251 EIA, 98 eigenvalue, 270, 291, 292, 293 elasticity, 10, 262 electric arc, 252, 259, 262 electric circuit, 66 electric current, 259 electric energy, 84, 104 electric field, 161, 163, 164, 174, 260, 262, 266 electric power, vii, viii, ix, 49, 50, 52, 54, 61, 62, 78, 103, 177, 179, 180, 183, 186, 190, 196, 197, 198, 199, 201, 202, 203, 205, 210, 216, 249, 270, 292 electric power production, vii, ix, 177 electric utilities, viii, 47 electrical conductivity, 14 electrical power, vii, viii, x, 37, 47, 48, 49, 51, 53, 64, 78, 98, 99, 180, 210, 222, 227, 234, 237, 249, 250, 293 electrical system, 134, 228, 238 electricity, vii, viii, 2, 10, 15, 16, 17, 20, 21, 35, 37, 40, 41, 43, 44, 47, 48, 49, 50, 51, 65, 67, 69, 71, 73, 74, 76, 78, 97, 99, 103, 104, 105, 119, 120, 121, 128, 129, 150, 153, 179, 210, 228 electricity system, 40 electrocatalyst, 72 electrodes, 69, 164, 260, 262, 266 electrolyte, 65, 66, 67, 68, 69, 70, 71, 72, 73, 75 electromagnetic, 158, 163, 250 electron, 164, 174, 260 electron density, 260 electrons, ix, 65, 66, 68, 70, 73, 74, 155, 163, 164, 260, 265 elongation, 260, 262, 265 emission, ix, 13, 14, 16, 21, 30, 31, 32, 35, 155, 174, 179 endurance, 73 end-users, 228 energy, viii, ix, 2, 4, 6, 8, 10, 11, 15, 18, 19, 20, 21, 24, 34, 35, 37, 40, 41, 42, 43, 44, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 60, 64, 65, 69, 75, 77, 78, 79, 84, 88, 92, 95, 96, 97, 99, 103, 104, 105, 119, 120, 121, 125, 128, 129, 131, 150, 151, 153, 155, 163, 164, 165, 179, 181, 182, 184, 207, 210, 212, 217, 218, 222, 260, 266, 270, 293 energy consumption, 48, 49, 96, 210
F
fabrication, 71, 73 failure, 210, 217, 228, 238, 239 family, 108, 246 farm efficiency, 107 farms, 104, 105, 106, 107, 108, 113, 114, 116, 119, 134, 136, 137, 138, 139, 141, 142, 150
Index
fatigue, 153 fault analysis, x, 175, 227, 233, 247, 248, 249, 250 feet, 94 FFT, 270, 278, 279, 285, 287, 288 fibers, 30 filters, 94, 118, 232 finite impulse response filters, 247 Finland, 21, 100 fire hazard, 35 fire suppression, 10 fires, 16, 32 first generation, 72 fitness, 276 flame, 11, 14, 22, 28, 38, 39, 41 flammability, 179 flexibility, 3, 9, 57, 60, 65, 66, 97 float, 83 flood, 69 fluctuations, 104, 161, 163, 165, 167, 234, 289 flue gas, 13, 14, 24, 27, 29, 42, 83 fluid, ix, 62, 63, 79, 80, 83, 89, 177, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 190, 191, 192, 194, 198, 199, 203, 205, 206, 207 fluidized bed, 83 focusing, viii, 1, 24 food, 4 food production, 4 forecasting, 119, 123, 165, 170, 175, 215 fossil, 2, 3, 4, 8, 14, 15, 18, 20, 21, 26, 29, 34, 40, 41, 42, 43, 44, 57, 179 fossil fuels, 3, 4, 14, 15, 21, 26, 40, 41, 42, 43, 179 fouling, viii, 1, 8, 12, 13, 22, 24, 26, 30, 39, 43, 179 Fourier, x, 227, 240, 244, 248, 249, 270, 278, 293 Fourier analysis, 240, 244 fragmentation, 253, 261, 262 France, 17, 43, 152, 153, 224 free energy, 66 freezing, 87 frequency distribution, 126, 128 frequency resolution, 246, 248, 249 frost, 170 fruits, 5 fuel, vii, 1, 2, 3, 4, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 44, 49, 51, 52, 55, 56, 57, 59, 60, 62, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 95, 97, 99, 179, 210, 212, 215, 217 fuel cell, 64, 65, 66, 67, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 95, 97, 99 fuel type, 2, 10, 12, 21, 55 funding, 20 furan, 38
299
fusion, x, xi, 159, 251, 252, 253, 255, 256, 257, 258, 262, 264, 265, 266, 267
G
gas diffusion, 69 gas turbine, 3, 59, 76 gases, 2, 3, 53, 54, 57, 58, 61, 62, 64, 68, 78, 79, 83, 89 gasification, 3 gasoline, 53 gel, 93 generation, vii, viii, 10, 14, 18, 38, 40, 41, 42, 43, 44, 47, 48, 49, 51, 52, 54, 56, 60, 61, 75, 76, 78, 83, 88, 95, 103, 104, 105, 116, 123, 133, 134, 135, 137, 139, 140, 142, 144, 146, 147, 149, 150, 151, 156, 180, 186, 212, 213, 214, 217, 224, 271 Geneva, 175 Georgia, 99, 101 Germany, 21, 26, 30, 118, 119, 153, 268 Gibbs free energy, 66, 71 glass, 91, 159 goals, 42 government, 21, 29, 38, 76 GPS, 270, 272 grains, 5, 90, 91 gravitational force, 259 Great Britain, 101 Greece, 1, 4, 6, 19, 37, 38, 44, 152, 153, 155, 173, 175, 251 greenhouse gas, 8, 16, 31, 41 greenhouse gases, 8 Grid connectivity, 99 grids, 51, 99, 103, 104, 105, 119, 150, 210, 271 grounding, 12 groups, 113, 116, 138, 234, 264 growth, 90, 92, 113 guidelines, 43
H
halogen, 94 harvesting, 37, 49 hazards, 73 HD, 293 health, 90, 92 heat, viii, ix, 2, 3, 4, 9, 10, 13, 21, 22, 24, 25, 26, 27, 29, 30, 32, 34, 35, 36, 39, 47, 48, 49, 50, 51, 52, 53, 54, 55, 57, 58, 59, 61, 62, 63, 64, 67, 69, 71, 72, 74, 76, 77, 78, 79, 80, 82, 83, 84, 86, 87, 88, 89, 91, 94, 95, 96, 168, 177, 179, 180, 181, 182,
300
Index
induction, 138, 151, 158, 235 industrial application, 75, 83 industrialized countries, 52 industry, vii, viii, ix, 1, 15, 18, 35, 38, 42, 47, 51, 177, 179, 210, 212 inertia, 69, 114, 288, 292 infinite, 179, 188, 240, 264 information sharing, x, 209, 211 infrastructure, 20, 21, 29, 38, 54, 99, 230 initiation, 162, 239, 264 injections, 134 innovation, 101 insight, x, 113, 133, 227, 228 inspections, 55 institutions, 7, 41 instruments, 157, 229, 230, 231 insulators, ix, 155, 156, 157, 158, 159, 162, 164, 165, 166, 168, 170, 171, 172, 173, 175 integrated waste management, 42 integration, 105, 150, 153, 229 integrity, 15 intensity, ix, 20, 146, 155, 159, 164, 170 interaction, 164, 175, 234, 236 interface, 230, 231 interference, 9, 32, 156, 158, 172 internal clock, 272 internal combustion, 54, 55, 59 Internet, 270, 272, 292 interpretation, 175, 237, 253 interval, 126, 129, 142, 144, 146, 149, 157, 168, 262, 266 intrinsic value, 18 investment, vii, 1, 2, 8, 10, 30, 32, 34, 37 ionization, 164, 260 ions, 65, 66, 67, 68, 70, 73, 74, 75, 161, 260 isobutane, 198, 199, 208 Italy, 38, 151, 153, 222, 250
183, 185, 187, 190, 192, 194, 199, 201, 202, 203, 205, 206, 207, 215, 252, 260, 262, 266 heat exchangers, viii, 47, 48, 59, 78, 79, 80, 82, 83 heat loss, 13, 62 heat release, 168 heat transfer, 13, 22, 26, 30, 79, 80, 82, 83, 86, 181, 182, 185, 201, 202 heating, viii, 11, 13, 18, 25, 29, 38, 47, 48, 50, 52, 53, 54, 56, 57, 58, 71, 72, 74, 78, 79, 83, 84, 215, 252, 266 heating rate, 11 heavy metals, 14, 30 high power density, 78 higher quality, 48, 97 histogram, 165 homogeneity, 39, 162, 163 hospitals, 76, 90, 98 hot water, viii, 47, 48, 53, 54, 55, 62, 71, 74, 75, 76, 78, 88, 89 hotels, 57, 90 households, 48, 51 housing, 11 hub, 49 human resources, x, 227, 228 humidity, 10, 11, 39, 69, 84, 90, 91, 92, 94, 96, 156, 157, 159, 163, 170 hybrid, 57 hydrocarbons, 67 hydrogen, 10, 56, 64, 65, 66, 67, 68, 69, 70, 71, 73, 74, 75, 188, 190 hydrogen bonds, 10, 188 hydrogen gas, 65 hysteresis, 164
I
identification, 165, 224, 232, 253, 270, 289, 293 IEDs, 229, 230 images, 119, 165, 261 imbalances, viii, 103, 105, 122, 125, 128, 131, 150 implementation, x, 3, 4, 8, 14, 18, 19, 20, 21, 35, 37, 40, 42, 209, 216, 220, 221, 229, 239, 293 imports, 2 in situ, 134 incentives, 16, 18, 41, 139, 149 incidence, 164 inclusion, 98, 99 income, 16, 18, 35, 51, 105, 123, 125 incompatibility, 235 independent variable, 123 indication, 66, 228 indices, 211, 218, 220 indigenous, 19, 21, 35
J
Japan, 269 jobs, 38 judgment, 217
L
labor, 6, 20 laminar, 94 landfill gas, 54, 57, 75 lanthanum, 70 laws, 181, 205 layering, 12
Index
leaching, 29 leakage, ix, 155, 156, 157, 158, 159, 160, 161, 162, 163, 166, 167, 168, 170, 172, 173, 175 learning, 166, 167 legislation, 15, 16, 28, 40, 42, 135 life cycle, 2 lifespan, 8 lifetime, 164 limitation, 16, 116, 213 linear programming, 213, 214, 223, 225 liquefaction, 264 liquid phase, 253 liquids, 78, 79, 84, 93 literature, xi, 13, 107, 252, 253, 257, 259, 262, 266 lithium, 73, 86, 88, 89, 94 loans, 20 location, 20, 35, 40, 54, 213, 228, 230, 231, 238, 278 location information, 238 logic programming, 152 logistics, 6, 40 long distance, vii, 18, 26 low cost, vii, 21, 41, 43, 270 low temperatures, 67 lubricating oil, 58, 187 lying, 19
301
M
magnetic field, 158 maize, 25 Malaysia, 206 management, ix, x, 69, 71, 104, 105, 106, 139, 150, 151, 209, 211, 216, 218, 219, 220, 222 manufacturer, 62, 67 manufacturing, 94 mapping, 166 market, viii, ix, 2, 8, 15, 16, 18, 19, 40, 41, 43, 47, 48, 49, 51, 52, 58, 69, 71, 73, 76, 97, 98, 99, 103, 105, 119, 120, 121, 122, 123, 124, 125, 128, 129, 130, 134, 138, 139, 149, 150, 209, 210, 211, 212, 217, 219 market prices, 122, 123, 124, 130, 139 market trends, viii, 47 market value, 18, 19, 69 marketing, 54 markets, viii, 2, 42, 43, 51, 75, 97, 104, 105, 116, 119, 120, 121, 122, 123, 124, 125, 128, 129, 130, 131, 133, 150, 153 Marx, 189, 207 Maryland, 100 matrix, 73
measurement, ix, 155, 156, 157, 158, 159, 168, 211, 221, 229, 234, 254, 257, 258, 261, 262, 270, 277, 293 measures, 9, 109, 115, 157, 272 media, 86, 188 Mediterranean, 4, 19, 37, 44 Mediterranean countries, 4, 37 melt, 73, 266 melting, 13, 39, 252, 259, 261, 266 melting temperature, 13, 259 memory, 98, 157, 214, 242 mercury, 27 metals, 13, 14, 39 methane, 67, 70 methanol, 67 microwave, 24, 238 microwaves, ix, 155, 165 minerals, 13 mixing, 3, 9, 11, 12, 22, 26, 29, 35, 80, 86 mobility, 73, 161, 262 model reduction, 217 modeling, 36, 165, 171, 175, 211, 222 models, 54, 60, 106, 107, 108, 114, 115, 116, 118, 119, 150, 153, 169, 170, 175, 180, 188, 230, 252 modernization, 8 modules, 108 moisture, 6, 10, 11, 12, 13, 14, 24, 25, 26, 31, 35, 41, 90, 91, 93, 94, 96 moisture content, 6, 10, 11, 13, 14, 24, 25, 26, 31, 35, 41, 91 mole, 74, 189 molecular oxygen, 164 molecular structure, 162 molecules, 70, 93, 188, 190, 262 motels, 57 motion, 59 motivation, 49 moulding, 11 multiples, 234
N
NaCl, 161 natural gas, 29, 49, 53, 56, 69, 71, 73, 74, 76, 89 natural resources, 37 neglect, 56 Netherlands, 21, 29, 44, 69, 71 network, vii, x, 137, 138, 142, 151, 166, 167, 210, 214, 217, 218, 219, 225, 233, 234, 247, 249, 251, 252, 257, 270, 277, 293 neural network, ix, 108, 155, 165, 166, 167, 168, 173, 174, 175, 223, 224 neural networks, 108, 165, 166, 175, 224
302
neurons, 166, 167 New Jersey, 175 New York, 99, 100, 101, 153, 174, 206, 207, 222, 223, 249, 250, 267, 268 next generation, 188 nickel, 70, 71, 74 nitrogen, 5, 14, 41 nitrogen compounds, 41 nodes, 134, 135, 214 noise, 54, 57, 59, 62, 78, 97, 242, 270, 279, 285 North America, 210, 222, 224 nuts, 36
Index
permit, 16, 20 phase shifts, 277 phasor measurement, xi, 269, 270, 272, 293 phosphorous, 67 phosphorus, 39 planning, 159, 210, 211, 212, 213, 214, 215, 222 plants, vii, 1, 2, 7, 14, 15, 16, 19, 20, 21, 26, 29, 35, 37, 39, 40, 42, 49, 51, 54, 62, 79, 179 plasma, 253, 260, 261, 265, 266 platinum, 67, 72 PM, 249 polarity, 234 polarization, 66 pollutants, 6, 15, 16, 37, 42 pollution, viii, ix, 15, 47, 49, 78, 156, 157, 159, 161, 163, 169, 170, 172, 173, 174, 175, 177, 179 polymer, 67, 69 poor, xi, 10, 12, 98, 269, 270 portfolio, 112, 114, 116 Portugal, 40 potassium, 13, 26, 39, 73 poultry, 34 power generation, viii, 4, 20, 41, 43, 44, 47, 48, 49, 51, 52, 53, 54, 56, 59, 60, 62, 65, 69, 75, 76, 78, 79, 97, 137, 138, 139, 142, 146, 149, 179, 188, 214 power plant, vii, 1, 2, 3, 4, 6, 7, 8, 10, 12, 18, 22, 34, 35, 37, 40, 41, 49, 51 power plants, vii, 1, 2, 3, 4, 7, 8, 9, 12, 15, 17, 18, 19, 20, 21, 24, 26, 27, 29, 30, 35, 36, 37, 38, 40, 41, 42, 43, 44, 49, 51, 61, 63, 207, 210 precipitation, 156 prediction, viii, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 118, 119, 120, 121, 122, 123, 124, 125, 126, 128, 129, 130, 131, 133, 136, 150, 152, 153 prediction models, 108, 152 predictors, 115, 116 pressure, 12, 35, 54, 55, 56, 62, 74, 76, 84, 86, 88, 93, 180, 182, 187, 191, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 210, 239, 259, 260, 265, 266 Pretoria, 250 prevention, x, 227 price taker, 123 prices, 19, 123, 129, 130, 131, 133, 138, 139, 140, 143, 145, 215, 217 probability, 114, 125, 126, 128, 129, 156, 164, 233 probability density function, 125, 126, 129 probability distribution, 114, 128 process duration, 264, 265, 267
O
observations, 162, 167 oil, 54, 55, 58, 75, 83 operator, 4, 6, 7, 8, 9, 116, 151, 217, 220, 228, 240 Operators, viii, 103, 105, 128 optimal performance, 108 optimization, viii, 39, 103, 142, 144, 146, 149, 150, 207, 213, 215, 216, 217, 219, 223, 225 optimization method, 216, 219 organic compounds, 207 organization, 107, 120, 270 organizations, 210 oscillation, xi, 233, 234, 249, 269, 270, 271, 279, 288, 289, 292 overhead power transmission lines, vii oxides, 28 oxygen, 28, 37, 41, 64, 65, 67, 68, 70, 73, 74 ozone, 179, 188
P
parameter, xi, 41, 115, 123, 124, 194, 236, 252, 264, 265, 267 Pareto, 212 Paris, 153 particles, 10, 11, 12, 22, 26, 30, 32, 34, 35, 38, 39, 72, 83, 260, 265 partnership, 76 payback period, 69, 89 penalties, 122, 125 perception, 20 percolation, 174 performance, ix, x, 3, 8, 9, 10, 12, 14, 15, 20, 22, 24, 25, 32, 36, 37, 42, 63, 64, 66, 72, 73, 75, 82, 87, 94, 96, 98, 104, 106, 107, 108, 115, 133, 152, 171, 172, 177, 179, 180, 187, 188, 191, 194, 198, 205, 206, 211, 215, 227, 228, 229, 230, 239, 250 Periodic Table, 67
Index
producers, viii, 47, 103, 104, 129, 134, 135, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 149, 150 production, 4, 15, 21, 26, 29, 37, 39, 40, 42, 43, 49, 54, 60, 63, 73, 76, 79, 103, 104, 105, 106, 107, 108, 109, 113, 114, 116, 118, 119, 120, 123, 125, 126, 128, 129, 134, 135, 136, 137, 138, 139, 141, 142, 143, 144, 145, 146, 149, 150, 153, 157, 163, 164, 210, 214 production costs, 60, 63, 76 profit, 107, 225 profits, 18 program, 24, 28, 76, 104, 105, 106, 107, 108, 113, 116, 121, 123, 125, 128, 150, 219, 229, 231, 232 programming, 142, 144, 151, 214, 225 promote, 15, 80 propagation, 166, 169 propane, 53, 56, 75, 191, 192, 194, 197, 208 Proteins, 5 protocol, 16 protocols, 230 proton exchange membrane, 64, 67 prototype, 89 pruning, 5 PTFE, 67 pulse, xi, 157, 167, 170, 252, 253 pulses, x, 157, 158, 159, 160, 161, 162, 166, 167, 168, 171, 173, 251, 261 pumps, 53, 84 purification, 27
303
Q
quality control, 25 quantization, 166 quantum mechanics, 164
R
radar, 119 radiation, ix, 13, 54, 82, 155, 157, 163, 164, 165 Radiation, 162, 173 radio, ix, 155, 156, 165, 172 radiography, 253 radius, 18, 40 rain, 157 rainfall, 156, 157, 159 range, 4, 9, 10, 12, 19, 21, 40, 42, 53, 55, 56, 58, 59, 61, 62, 70, 71, 75, 76, 79, 87, 88, 125, 126, 128, 129, 165, 166, 168, 171, 191, 194, 195, 199, 203, 206, 261, 279, 280, 288 ratings, 232
reactants, 66 reactivity, 26 reading, 234 real numbers, 245 real time, ix, x, 105, 106, 108, 114, 115, 116, 126, 209, 210, 211, 212, 217, 218, 219, 220 reception, 11, 20, 31, 35 recognition, 250 recombination, 164 recovery, 40, 53, 55, 57, 58, 59, 63, 78, 79, 82, 83, 179, 190, 207 recycling, 28 reduction, 2, 8, 9, 14, 16, 21, 26, 29, 31, 32, 34, 35, 38, 39, 40, 84, 121, 123, 124, 125, 134, 135, 136, 137, 138, 139, 140, 141, 143, 146, 149, 150, 162, 220, 234, 265 redundancy, 11, 246 regenerate, 96 regeneration, 91, 93, 94, 95, 96 regional, 107, 119, 120, 134, 135, 212 regression, 270, 276, 288, 289 regression method, 288, 289 regulation, 19, 116, 134, 139, 215 regulations, viii, 30, 38, 47 regulatory framework, 13, 120 rejection, 180, 182 relationship, 240 relaxation, 142 relevance, 15, 249 reliability, x, 12, 53, 57, 58, 62, 69, 77, 78, 89, 104, 150, 151, 209, 210, 211, 212, 214, 216, 218, 219, 220 renewable energy, 2, 15, 21, 29, 43, 49, 134 representative samples, 171 reserves, 120 residential buildings, 90 residuals, 40 residues, 2, 4, 18, 19, 21, 28, 29, 32, 34, 36, 37, 40, 42, 43, 44 resistance, ix, x, 143, 155, 157, 158, 169, 170, 171, 231, 251, 252, 253, 254, 257, 258, 259, 260, 261, 262, 264, 265, 266, 267 resolution, 106, 107, 114, 116, 119, 247 resources, 2, 4, 6, 8, 19, 21, 26, 40, 42, 54, 99, 212 response time, 75 restaurants, 57 restructuring, viii, ix, 47, 209, 211, 212 retail, 57, 90 revenue, 8, 20, 121, 122, 124, 125, 129, 133, 139, 151 rewards, 150 risk, 2, 8, 12, 20, 37, 41, 175, 220 rolling, 123, 124
304
roughness, 106, 107, 112, 118 routing, 9 rural areas, 101
Index
sites, xi, 31, 164, 214, 235, 269 skewness, 112 slag, 26 sludge, 2, 27, 28, 29, 30, 32, 34, 36 smoke, 22 smoothing, 247 society, 151, 214 sodium, 39, 73 software, x, 58, 119, 140, 158, 188, 209, 211, 254 solar energy, 59 solid phase, 253, 261, 262 solid state, 70 solid waste, 28 sorption, 94 sorption process, 94 sorting, 28 South Africa, 247, 250 Spain, 45, 103, 113, 114, 116, 118, 119, 134, 135, 151, 152, 173 species, 8 specific heat, 206 spectrum, 161, 163, 213, 258 speed, 24, 106, 107, 114, 116, 151, 156, 157, 159, 213, 214, 242 spot market, 121 stability, vii, xi, 22, 39, 72, 94, 104, 134, 162, 179, 211, 214, 218, 219, 220, 250, 269, 270, 292 stages, 11, 28, 88, 89 standard deviation, 111 standardization, 7, 38 standards, 26, 38, 54, 71, 99, 100 statistics, 108, 118 steel, 74 storage, 6, 8, 10, 12, 18, 20, 29, 31, 32, 35, 37, 57, 58, 71, 99, 157, 158 strain, 162 strategies, viii, 103, 104, 138, 146, 150, 211, 212 strength, 163, 210, 260, 262 stress, 170, 261, 262 strikes, 234 subsidy, 104 substation, vii, 49, 229, 230 substitution, 24, 30, 32, 34, 36, 37, 38 successive approximations, 247 suffering, 173 sulfur, 13, 67 sulfuric acid, 94 sulphur, 14, 22, 37 supervision, 166 suppliers, 3, 6, 17, 19, 20, 26, 37, 76 supply, viii, x, xi, 1, 7, 17, 19, 20, 25, 26, 30, 35, 38, 40, 41, 44, 62, 66, 75, 78, 89, 210, 214, 219, 227, 234, 235, 249, 251, 252, 253, 261, 264
S
SA, 153, 206, 207, 222 safety, 6, 98, 179, 187, 216 sales, 58 salinity, 156, 159 salt, ix, 93, 155, 156, 159, 163, 171, 172 salts, 73 sample, 131, 229, 233, 239 sampling, ix, 155, 157, 158, 168, 229, 230, 240, 242, 247, 254, 272 satellite, 119 saturation, 94, 179, 188, 191, 203 savings, 20, 26, 220, 229 sawdust, 4, 10, 21, 22, 35, 36 scaling, 242, 246 scheduling, 105, 113, 116, 212, 224 search, 151, 188 seasonal variations, 115 security, viii, x, 35, 51, 97, 103, 135, 136, 209, 210, 211, 212, 213, 214, 217, 218, 219, 220, 223, 224, 225 seed, 25 selecting, 67 selenium, 67 semiconductors, 164 sensitivity, 135, 220, 222, 229 sensors, 156, 157 separation, 11, 31, 234 series, 8, 9, 39, 79, 99, 100, 131, 132, 143, 156, 158, 232, 235 set theory, 214 severity, 10, 234 sewage, 2, 32 shape, 11, 79, 126, 161 shares, viii, 1, 28, 40, 41, 42 sharing, 137 shortage, 19, 214 side effects, 22 sign, 163, 273 signals, ix, 209, 211, 229, 230, 234, 236, 237, 244, 246, 247, 248, 279 silicon, 32, 39 simulation, ix, xi, 118, 155, 169, 170, 171, 175, 231, 252 sine wave, 234, 236, 244 Singapore, 223 sinus, xi, 252, 255, 256, 257, 264, 267 SiO2, x, 251
Index
supply chain, 7 suppression, 10, 12, 224 surface area, 32, 94 surface energy, 259 surface tension, 259 surfactants, 15 surplus, 16 Sweden, 21, 151, 152 switching, 15, 29, 104, 138, 230, 234, 279 Switzerland, 43, 44, 152, 227 synchronization, 272 synthesis, 150 system analysis, 150, 239 systems, viii, ix, x, 8, 10, 12, 14, 20, 21, 31, 37, 39, 41, 43, 47, 48, 51, 52, 53, 54, 55, 57, 58, 61, 62, 69, 71, 78, 79, 82, 83, 84, 86, 88, 89, 91, 93, 94, 95, 96, 97, 98, 99, 100, 103, 104, 105, 131, 156, 157, 158, 159, 165, 175, 180, 188, 190, 207, 209, 210, 211, 212, 213, 214, 215, 218, 220, 221, 223, 225, 227, 228, 229, 230, 231, 233, 234, 237, 239, 240, 242, 243, 247, 249, 250, 270, 293
305
T
targets, 2, 16, 21, 42, 239 tariff, 120 tax incentive, 21 TBP, 206 technology, vii, viii, x, 1, 8, 15, 20, 22, 25, 36, 42, 43, 44, 47, 48, 51, 53, 54, 55, 56, 57, 58, 60, 62, 65, 69, 71, 72, 78, 97, 98, 134, 137, 138, 207, 217, 227, 229, 237, 249, 293 tellurium, 67 temperature, ix, 11, 12, 14, 25, 30, 32, 39, 52, 54, 55, 58, 59, 66, 67, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 83, 84, 86, 87, 88, 89, 91, 92, 93, 94, 95, 96, 106, 156, 157, 159, 162, 163, 164, 170, 174, 177, 179, 180, 181, 182, 183, 185, 186, 187, 188, 190, 191, 192, 193, 194, 195, 196, 197, 198, 203, 204, 205, 206, 239, 252, 253, 259, 260, 265 tension, 187 terminals, 253 Thailand, xi, 269, 270, 271, 292, 293 theory, 162, 163, 213, 215, 218, 250, 262, 264 thermal energy, 26, 51, 53, 63, 79, 83, 84, 86, 95, 96 thermal stability, 187 thermodynamic cycle, 62 thermodynamic equilibrium, 260, 265 thermodynamic properties, 188, 207, 208 thermodynamics, 205 thinking, viii, 47, 214 threat, x, 11, 227 threshold, ix, 107, 155, 157, 163, 164, 170, 247, 266
thresholds, 55 TID, 125 time, vii, ix, x, xi, 1, 4, 9, 10, 11, 12, 29, 35, 48, 49, 59, 69, 72, 75, 98, 99, 105, 106, 107, 108, 109, 111, 113, 114, 115, 116, 119, 120, 121, 122, 123, 124, 125, 126, 129, 155, 156, 157, 159, 160, 161, 163, 166, 168, 170, 171, 209, 210, 211, 214, 215, 216, 217, 219, 220, 223, 224, 227, 228, 230, 232, 233, 237, 239, 240, 244, 245, 246, 247, 248, 249, 252, 258, 259, 261, 262, 263, 264, 265, 266, 270, 272, 278, 280, 293 time factors, 9 time frame, ix, 155 time lags, 126 time periods, 29, 263 time series, 107, 108, 114, 115, 116 topology, 218, 224 total energy, 29, 119 total product, 136, 137 toxicity, 86, 87, 179, 188 trace elements, 14, 27 tracking, 165 trade, 17, 119 tradeable permits, 17 trading, 42, 119, 124, 150, 212 training, 58, 166, 167 transactions, 212, 219 transducer, 157 transition, 51, 218 translation, 246 transmission, vii, 49, 51, 134, 135, 136, 137, 139, 141, 142, 143, 144, 145, 146, 148, 163, 172, 210, 211, 212, 217, 218, 219, 222, 223, 231, 247, 270, 271, 292 transmission path, vii transport, 11, 18, 19, 24, 26, 67, 163 transport costs, 18, 19, 26 transportation, 6, 18, 28, 29, 41, 49, 57, 76, 224 trend, vii, 1, 111, 119, 120, 133, 191, 192, 195, 197, 198, 203 trial, 37, 64 triggers, 229 tunneling, 164
U
UK, 3, 4, 6, 12, 16, 17, 18, 19, 35, 36, 43, 44, 119 uncertainty, viii, 51, 103, 104, 125, 126, 128, 129, 150, 153, 210, 211, 212, 213, 214, 215, 216, 217, 218, 222 uniform, 181 unit cost, 54, 90 United States, 51, 98, 99, 100, 217, 222
306
universities, xi, 76, 269, 270 university students, 293 updating, 105, 115, 121, 130 users, 26, 41, 116, 119, 151
Index
Warsaw, 207 waste treatment, 26 water absorption, 87 water heater, 83, 180, 184 water vapor, 56, 71, 87, 90, 93, 94 wavelet, x, 227, 244, 245, 246, 247, 248, 249, 250, 270, 279, 280, 289, 294 wavelet analysis, 244, 250, 294 wear, 39 wells, 49 wetting, 96, 156 wind, viii, 99, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 118, 119, 120, 121, 123, 125, 126, 128, 129, 130, 131, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 148, 149, 150, 151, 152, 153, 156, 157, 159 wind farm, viii, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 114, 115, 116, 118, 119, 120, 121, 123, 125, 126, 128, 129, 131, 134, 136, 137, 138, 139, 141, 142, 143, 146, 149, 150 wind generators, viii, 103, 104, 119, 120 wind turbines, 106, 107, 120, 140, 151 windows, 123, 124 winning, 166 wires, x, 251, 252, 253, 254, 259, 260, 261, 266 wood, 2, 4, 5, 11, 13, 14, 22, 24, 25, 27, 28, 30, 31, 32, 34, 36, 37, 38, 42, 44 wood species, 44 working conditions, 180
V
vacuum, 164, 174 validity, 273 values, viii, ix, xi, 11, 16, 22, 29, 38, 56, 96, 103, 104, 114, 122, 123, 124, 126, 128, 130, 131, 133, 135, 141, 142, 143, 155, 157, 158, 160, 164, 165, 166, 167, 171, 195, 212, 213, 215, 217, 219, 232, 233, 236, 242, 243, 244, 252, 254, 256, 262, 266, 267, 273, 289, 290 vapor, 53, 56, 71, 84, 86, 87, 88, 89, 93, 179, 180, 187, 188, 259, 260, 261, 262 variability, 42, 104, 120 variable, 20, 125, 134, 141, 166, 244, 253, 254, 276, 288, 289 variables, ix, 117, 164, 209, 211, 212, 213, 214, 219, 240, 276 variance, 109, 126, 128, 213 variation, 41, 146, 159, 162, 168, 191, 192, 194, 197, 198, 199, 201, 203, 205, 265, 270, 273, 280, 292, 293 vector, 166 vehicles, 6, 57, 60 velocity, 25, 83 ventilation, 71, 87 vibration, 62 Virginia, 100 viscosity, 187 visualization, 211, 212, 220 volatility, x, 72, 209, 211 volatilization, 259 voltage management, vii
Y
yield, 39, 89, 128, 194
Z
zeolites, 94 zirconia, 70
W
warm air, viii, 47, 48, 83

Electric Power Research Trends

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Electric Power Research Trends

Uploaded by

Copyright:

Available Formats

ELECTRIC POWER RESEARCH TRENDS

ELECTRIC POWER RESEARCH TRENDS

Nova Science Publishers, Inc.

Copyright 2007 by Nova Science Publishers, Inc.

Published by Nova Science Publishers, Inc.

ISBN: 978-60021-978-8 2007 Nova Science Publishers, Inc.

Panagiotis Grammelis and Emmanuel Kakaras

Keywords: secondary fuels, coal, retrofitting, SRF

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

2. CURRENT CO-FIRING TECHNIQUES

Panagiotis Grammelis and Emmanuel Kakaras

3. PRACTICAL CONSIDERATIONS FOR RETROFITTING A COAL-FIRED UNIT TO CO-FIRING BIOMASS

3.1. Fuel Availability

Table 1. Main characteristics of solid bio-fuels [33]

No nitrogen Extracts (%) 40-50 40-50 40-45 45-50 70-80 10 - 20 35 - 40 38 - 54 39 51 53 - 63 34 - 36

Barley Oak Rice Maize

Grains Grains Grains Grains

Straw Straw Straw Straw Cob stalks Pruning Pruning Pruning

sarmenta (pruning) (w) wood, (b) branches pruned

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

3.2. Plant Modifications

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

3.3. Legislative Framework on Environmental Issues

Solid 200 200 30 6%

Valid until 31 December 2015

A ROC is equivalent to 1MWh of electricity generated of renewable sources

3.4. Financial Evaluation

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

Figure 1. Final costs of biomass fuels in UK, Source DTI-2006.

3.5. Societal Issues

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

4. REVIEW ON CO-FIRING EXPERIENCE AT EUROPEAN LEVEL

4.2. Demonstration of Co-firing at Commercial Level in Finland

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

Figure 3. Direct co-firing of straw with coal in Studstrup Power Plant.

Panagiotis Grammelis Table 7. Danish power plants experienced co-firing

Plant name Studstrupvaerket 1 Studstrupvaerket 4 Amager 3 Esbjergvrket

Location Aarhus Aarhus Amager Esjberg

Output (MWe) 150 350 250 150

% substitution 20% heat 20% heat Up to 20% heat Up to 25%

Status Commercial (1996-1998) Commercial (since 2001) n.a n.a

Suppl. Fuels straw straw straw / miscanthus straw /wood

4.3. Demonstration of Co-firing at Commercial Level in Denmark

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

4.4. Demonstration of Co-firing at Commercial Level in Germany

Up to n.a 10% n.a n.a

n.a 235 1933 n.a n.a

Tests with approx.8t/h 8% n.a n.a n.a n.a 7%

n.a 1 to 13% n.a

n.a 1995 tests

Sewage sludge Straw pellets Wood chips SRF /paper pulp

Operation since 6/1999 Cont.operati on starting 2000

Co-firing Solid Bio-fuels or Waste Recovered Fuels with Coal

4.5. Demonstration of Co-firing at Commercial Level in the Netherlands

Table 9. Dutch power plants experienced co-firing

Amer 8/ Geertruidenberg Massvlakte1/2,Rotterdam

EPZ (=Essent) EZH (Germany)

0.8% heat 5% heat 4% heat

commercial commercial tested