You are on page 1of 7
Available online at www.sciencedirect.com ScienceDirect Procedia Computer Science 55 ( 2015 ) 388 – 394 Information

Available online at www.sciencedirect.com

ScienceDirect

Procedia Computer Science 55 ( 2015 ) 388 – 394

Available online at www.sciencedirect.com ScienceDirect Procedia Computer Science 55 ( 2015 ) 388 – 394 Information

Information Technology and Quantitative Management (ITQM 2015)

Bottom-up long-term forecasting of Brazilian commercial class electricity consumption: First results

Bruno Q. Bastos a *, Reinaldo C. Souza a , Fernando L. Cyrino Oliveira b

a Electrical Engineering Department, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rua Marquês de São Vicente 255, Gávea, 22453-900, Rio de Janeiro, RJ, Brazil b Industrial Engineering Department, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rua Marquês de São Vicente 255, Gávea, 22453-900, Rio de Janeiro, RJ, Brazil

Abstract

In Brazil, the electricity consumption of the commercial class has been growing more than the consumption of the other classes, e.g. residential, industrial, and others. Understanding why this is happening and how it would progress is essential for policy makers and for agents of the electrical sector. Bottom-up models consider a detailed and disaggregated representation of a region’s economy, and allow the incorporation of technological changes and policy impacts in its forecasts. In this context, the paper presents the first results of the long-term bottom-up modelling of Brazilian commercial class electricity consumption. The bottom-up model used in this work is the FORECAST model adapted for Brazil. It differentiates the five regions of the country, 8 subsectors of the tertiary sector, and 14 building and end user related energy services, such as lighting in buildings, street lighting, ventilation and air conditioning, and others. Despite the lack of consolidated information at the required level of disaggregation in Brazil, the first results show proximity to the official long-term forecasts. The results are analyzed and discussed.

© 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license

© 2015 The Authors. Published by Elsevier B.V.

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Selection and/or peer-review under responsibility of the organizers of ITQM 2015

Peer-review under responsibility of the Organizing Committee of ITQM 2015

Keywords: Bottom-up model; electricity demand; commercial; Brazil

1. Introduction

Long term forecasting of the electricity consumption is essential for the electric power sector. It can provide a basis for capacity expansion, business planning, policy making, and others. For example, a distribution

* Bruno Q. Bastos. Tel.: +55 21 3527-1406. E-mail address: brunoq.b@ele.puc-rio.br.

1877-0509 © 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the Organizing Committee of ITQM 2015 doi: 10.1016/j.procs.2015.07.088

Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

389

company could estimate future investment in its concession area based on demand forecasts of the region, and a power generation company could plan the purchasing of generation units. In the literature, there are applications of many different models providing long-term electricity demand or consumption forecasts, most of which are top-down models: Bodger and Tay [1] use logistic curve to predict New Zealand’s electricity consumption 15 years ahead; Atakhanova and Howie [2] calculate sectorial and aggregate electricity consumption of Kazakhstan via econometric model; Bianco, Manca, and Nardini [3] use regression models to forecast Italy’s electricity consumption from 2008 to 2030; Kermanshahi and Iwamiya [4] apply neural networks to forecast Japanese peak load up to year 2020 with data until 2000; and many others studies. In contrast with top-down models, which consider the aggregated electricity consumption of a country, region, utility, etc., bottom-up models consider the consumption in a disaggregated manner, detailing the energy end-uses like lighting, information and communication equipment, air conditioning, etc. [5]. Bottom-up models, therefore, have the ability to model technological changes, and are a strong tool for policy makers. This paper applies a bottom-up model to calculate the long-term electricity consumption of the Brazilian commercial class. This class was selected because its consumption is growing faster than that of the other classes, i.e. “industrial”, “residential”, and “others”, as presented in Fig.1. With bottom-up modelling one may analyze why this is happening. The objective of the paper is to show the results of the long-term bottom-up forecast for Brazil, obtained with FORECAST -Tertiary model, and to compare them with the country’s official forecasts. The paper will present the results of total consumption, and will not show the disaggregation at end-use level nor at regional level. The paper is structured as follows: the first section introduces the paper; the second section describes the model; the third section presents the results, and the fourth section provides the conclusion of the paper.

2. Model description

The model used in this work is the FORECAST-Tertiary, which has been developed by Fraunhofer Institute, T EP Energy and IREES over the years. The detailed methodology can be found in the study by Jakob et al. (2013) [6], where the model is applied to estimate the impact of an eco-design policy in the electricity demand of the European tertiary sector. The study extends the methodology of preceding papers [7] [8].

2.1. The model

T he FORECAST -Tertiary bottom-up model allows the calculation of electricity consumption, for a target year, disaggregated by region, by subsectors of the terti ary sector, and by energy services (e.g. lighting,

CAGR 2009 - 2013 Others Industrial Residential Commercial CAGR 1995 - 2013 Others : 3,8% a.a.
CAGR 2009 - 2013
Others
Industrial
Residential
Commercial
CAGR 1995 - 2013
Others : 3,8% a.a.
Commercial: 5,4 % a.a.
Industrial: 2,8% a.a.
2004
Others : 5,6% a.a.
Commercial: 6,4 % a.a.
Industrial: 3,4% a.a.
Residential: 5,5% a.a.
-
500
400
300
200
100
Electricity consumption
(TWh)
2012
1999
1997
1998
1996
2013
2003
2005
2002
1995
2011
2001
2009
2007
2008
2006
2010
2000

Fig. 1. Brazilian annual electricity consumption (1995 – 2013)

  • 390 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

refrigeration, ventilation and air-conditioning). FORECAST -Tertiary can model energy policies and, thus, can estimate their impacts on the electricity demand: the model uses a concept of energy efficiency measures (EEMs), which represent options that improve energy efficiency related to specific energy services, subsectors, or regions, and which are defined according to the policy being modelled. The forecasting approach consists, basicall y, of calculating the electricity consumption via sum and product of global drivers (e.g. number of employees per square meter), specific energy service drivers (e.g. number of computer per employee), and of specific energy consumption indicators (e.g. installed power per square meter, utilization hours) [6]: in order to obtain the total consumption for a year, the consumption per energy service, subsector, and region, of that year, is aggregated. When taking policies into account, EEMs are also considered in the aggregation, and reduce the total electricity consumption. In order to create the forecast of a country, exogenous forecasts of different variables of the model have to be made. It is possible to perceive that the model is very data intensive, once it requires detailed data (e.g. installed power, utilization hours, maintenance cost) of several energy services per subsector of the tertiary sector and per region. The good representation of a country is, therefore, directly related to the database that will be used in the forecast. In order to represent Brazilian commercial class, Brazil was divided into five regions, seven subsectors of the tertiary sector, and fourteen energy services, as presented in Table 1. After listing the information demanded by the model (input data required), a major research was made (contact with qualified consultancy, research in governmental databases and studies, etc.) with the purpose of constructing the database. When data was not available or was not found, information from Europe was used.

Table 1. Regions, subsectors and energy services considered in the model

Regions

Subsectors

Energy Services

South

Wholesale and retail trade

Lighting

Southeast

Hotels, cafes, restaurants

Lighting street

Midwest

Traffic and data transmission

ICT office

Northeast

Finance

ICT data centers

North

Health Education Other services

Ventilation and air-conditioning Circulation pumps and others Electric heating Heat pumps Hot water Elevators Misc. building technologies Cooking Laundry Refrigeration

2.2. Assumptions for Brazil

Assumptions regarding the future of Brazil were made for the variables that had to be exogenously forecasted. The global drivers of the model, e.g. number of employees per subsector per region, and floor area per employee, greatly influences the forecast results; the forecast for these variables are, thus, of great importance and will be shown. The floor area data was obtained via calculation based on European data and on reports of the National Electrical Energy Conservation Program (PROCEL); Table 2 shows the values obtained from the calculation. As the historic data was not found for Brazil, the forecasts were made considering the annual growth rate of European countries. The forecasts of the number of employees were made considering

Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

391

the official historic and extrapolating the trends using annual growth rate year by year. Additionally, a consultancy was contacted in order to understand the expectancy of future employment of the service sector, what gave a basis to the reduction of the growth rate on the long-run. The regional forecasts of the employment are presented in Figures 2 to 6.

Table 2. Specific floor area (m²/employee) per subsector

Subsector

m²/employee

Wholesale & Retail trade

41,33

Hotels, restaurants, etc.

44,91

T&D trans.

24,60

Finance

19,80

Health

10,11

Education

47,07

Other services

21,54

North Region

1 1.5 0.5 0 2012 2002 2032 2022 2008 2018 2028 2006 2016 2026 2010 2030
1
1.5
0.5
0
2012
2002
2032
2022
2008
2018
2028
2006
2016
2026
2010
2030
2020
2014
2034
2004
2024
Million employees

Fig. 2. Number of employees forecast for the North Region of Brazil

Northeast Region

2030 Million employees 0.0 2.0 4.0 6.0 2024 2014 2004 2034 2020 2012 2010 2026 2016
2030
Million employees
0.0
2.0
4.0
6.0
2024
2014
2004
2034
2020
2012
2010
2026
2016
2006
2028
2018
2008
2022
2032
2002
Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394
Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

Fig. 3. Number of employees forecast for the Northeast Region of Brazil

  • 392 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

Midwest Region

2016 Million employees 2024 2034 2014 2004 2030 2020 2010 2026 2006 1 2028 2018 2008
2016
Million employees
2024
2034
2014
2004
2030
2020
2010
2026
2006
1
2028
2018
2008
2032
2022
2012
2002
0
0.5
1.5
392 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 –

Fig. 4. Number of employees forecast for the Midwest Region of Brazil

Southeast Region

2030 2012 2022 2032 2008 2018 2028 2016 2006 2026 2010 2002 2020 2034 2004 2014
2030
2012
2022
2032
2008
2018
2028
2016
2006
2026
2010
2002
2020
2034
2004
2014
2024
8
6
4
2
0
Million employees
392 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 –

Fig. 5. Number of employees forecast for the Southeast Region of Brazil

South Region

2002 2012 2022 2032 2018 2008 2028 2016 2006 2026 2010 2020 2030 2014 2034 2004
2002
2012
2022
2032
2018
2008
2028
2016
2006
2026
2010
2020
2030
2014
2034
2004
2024
3
2
1
0
Million employees

Fig. 6. Number of employees forecast for the South Region of Brazil

392 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 –

Some variables are made constant throughout the forecasting horizon due to the dynamics of the model (for example, installed power per energy service is kept constant, and would change over time depending on the diffusion of EEMs; in the case of this paper, we do not model the diffusion of EEMs).

3. Results

After consolidating the database for Brazil with information found and adjusted, and with the required forecasts for the variables, the model was used. The results and their comparison with the official demand

Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

393

forecasts [9] are shown in Fig. 7. In the figure, black color is realized value, and hachures are estimated values. It is possible to see that the forecast from the proposed bottom-up model has a lower growth rate (3,7% a.a. from 2013 to 2023) compared to the official forecast (5,5% a.a. from 2013 to 2023), which is made by the Energetic Research Company (EPE): this indicates that the model provides more conservative forecasts than the official source. The short-term results are nearly the same because the FORECAST-Tertiary adjusts the bottom-up (aggregated/calculated) result with a factor (i.e. the model calibrates the bottom-up calculation in order to provide more accurate forecasts). The long term result from the bottom-up model indicates a scenario of low electricity consumption: it estimates that the annual growth rate on the period of 2018 to 2023 to be 2,3% a.a. This appointment is in accordance with the assumptions made for employment for the Brazilian service sector, which are assumed to retract from 2018 on. In contrast to the bottom-up model, EPE’s forecast indicate relative high consumption growth on the long run – 5,7% a.a. from 2018 to 2023. The bottom-up model also allows one to analyze the electricity consumption: it is possible to disaggregate the results per subsector or per region or per energy services. This feature is positive to policy makers and to companies that would like to plan investments in the energy efficiency business or in electrical power sector based on consumption scenarios. This type of result will not be provided by this work because it must still be consolidated.

2016 EPE Commercial FORECAST Commercial 2023 2013 2015 2012 2022 2021 2011 2009 2019 2017 2018
2016
EPE Commercial
FORECAST Commercial
2023
2013
2015
2012
2022
2021
2011
2009
2019
2017
2018
100
2010
2020
2014
Electricity consumption
(TWh)
120
140
160
20
40
60
80
0

Fig. 7. Electricity consumption forecasts for Brazil with FORECAST-Tertiary

4. Conclusion

The bottom-up model is an important tool for evaluating scenarios of electricity consumption and for explaining the occurrence of those scenarios, as it allo ws one to disaggregate the results per energy service, subsectors or regions. In contrast, top-down models use aggregated information and forecast total electricity consumption based on exogenous variables (or not). In this context, the present work proposed a first application of a bottom-up model, the FORECAST- Tertiary, to long-term forecast the total electricity consumption of the Brazilian commercial class. It is shown that the results of the model are much more conservative than that of the official forecast, and that the model is very sensitive to the assumptions of the global drivers, such as “number of employees per subsector per region” and “specific floor area”, specific drivers, such as “utilization hours of energy services”, and others. In order to model Brazil, the consolidation of the database had to be done. The data was based on research and on contact with consultancy. In spite of this, there was data that could not be found, because it does not

  • 394 Bruno Q. Bastos et al. / Procedia Computer Science 55 ( 2015 ) 388 – 394

exist for Brazil; in this case, data from Europe was used. The idea is to consolidate, in the future, the database only with information of Brazil, so that disaggregated evaluation can be accurate about the country. The proposed future studies are: model existing Brazilian policies with the bottom-up model, and evaluate their impacts on the electricity consumption.

Acknowledgements

T he authors would like to thank the colleagues from PUC-Rio for their valuable comments and suggestions, which improved the paper, and the R&D program of the Brazilian Electricity Regulatory Agency (ANEEL) for the financial support (R&D project PD-7625-0003/2014). Finally, the authors would like to thank CNPq and CAPES, which are governmental agencies that have also given support to this research.

References

[1] P. S. Bodger and H. S. Tay. Trend extrapolation in long-term forecasting: An investigation using New Zealand electricity consumption data. Technological Forecasting and Social Change 1986; 30: 167-188.

[2] Z. Atakhanova and P. Howie. Electricity demand in Kazakhstan. Energy Policy 2007; 35: 3729-3743.

[3] V. Bianco, O. Manca and S. Nardini. Electricity consumption forecasting in Italy using linear regression models. Energy 2009; 34:

1413-1421.

[4] B. Kermanshahi and H. Iwamiya. Up to year 2020 load forecasting using neural nets. International Journal of Electrical Power & Energy Systems 2002. 24: 789-797.

[5] T. Fleiter, E. Worrell and W. Eichhammer. Barriers to energy efficiency in industrial bottom-up energy demand models - A review. Renewable and Sustainable Energy Reviews 2011; 15: 3099-3111.

[6] M. Jakob, G. Catenazzi and T. Fleiter, 2013. "Ex-ante estimation of the EU Ecodesign Directive's impact on the long-term electricity demand of the tertiary sector," eceee 2013 Summer Study. Toulon/Hyères, France.

[7] T. Fleiter, S. Hirzel, M. Jakob, J. Barth, L. Quandt, F. Reitze, F. Toro and M. Wietschel, 2010. "Electricity demand in the European service sector: A detailed bottom-up estimate by sector and by end-use," 6 th International Conference on Energy Efficiency in Commercial Buildings. Frankfurt, Germany.

[8] M. Jakob, T. Fleiter, G. Catenazzi, S. Hirzel, F. Reitze and F. Toro, 2012. "The impact of policy measures on the electricity demand of the tertiary sector of the European countries: An analysis with the bottom-up model FORECAST," 7 th International Conference on Energy Efficiency in Commercial Buildings. Frankfurt, Germany.

[9] EPE, 2013. "Projeção da demanda de energia elétrica para os próximos 10 anos (2014 -2023)," Rio de Janeiro, Brazil.