90%(10)90% found this document useful (10 votes) 3K views546 pagesPower Plant Performance Monitoring
Power Plant Performance Monitoring
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content,
claim it here.
Available Formats
Download as PDF or read online on Scribd
Power Plant Performance
Monitoring
Rodney R. Gay
With
Carl A. Palmer
Ne LIBS ae nat
TECHNIZ BOOKS INTERNATIONALPower Plant Performance Monitoring
Rodney R.Gay
Carl A. Palmer
Michael R. Erbes
TECH BOOKS INTERNATIONAL
Publishers & Distributors
New Delhi-110019, India© 2006 Tech Books International, New Delhi-110019, India,
First Indian Edition
All rights reserved. No Part of this book may be reprinted or reproduced or utilised in any form or by any
electronic, mechanical, or other means, now known or hereafter invented, including photocopying and
recording, or in any information storage or retrieval system, without permission in writing from the
Publishers.
Published by .
‘TECH BOOKS INTERNATIONAL
Publishers & Distributors
4/12, Kalkaji Extn., Opp. Nehru Place, Kalkaji, New Delhi -110 019, India
Phone : (011) 26284791, 26284790, 26414791 Fax: 91-11-26473611, 26231799
E-mail : abitechbooks@ vsnl.net; booksales@bol.net.in Visit us at : wwww.technobooks.com,
Library of Congress Cataloging-in-Publication Data
Gay, Rodney R,
Power plant performance monitoring / by Rodney R.
Gay, Carl A. Palmer, Michael R. Erbes, - Ist ed.
pem.
Includes bibliographical references and index.
LCCN 2004096640
ISBN 81-88305-83-9
1. Electric power-plants--Efficiency. 2. Plant
performance--Monitoring. I. Palmer, Carl A.
IL Erbes, Michael R. I]. Title.
‘TK1005.G39 2006 621.3121
(QB 104-200607
ISBN : 81-88305-83-9Dedicated to the most important people in my life:
Henry, my father who taught me to be curious
Joan, my mother responsible for my success in life
Wendy, my wife, companion and friend for 27 years and for life
Christopher, my son who is nearly perfect
Richard, my twin brother and good friend
David, my brother, tennis partner and hiking buddy
Timothy, my brother who gives me medical and political advice
William, my grandson who makes me smileAuthors
Rodney R. Gay received his PhD in mechanical engineering from Stanford
University in 1975. He served as founder and president of Enter Software,
Inc. from 1988 until the company was sold to General Electric in 1999. He
remained as president of GE Enter Software for two years, then left GE to
become a writer and engineering consultant.
Carl A. Palmer earned his PhD in mechanical engineering from the
University of Wisconsin in 1991, after which he became an employee with
Enter Software, Inc. Carl is currently an engineering manager working on
sensor development for the power industry.
Michael R. Erbes received his PhD in mechanical engineering from
Stanford University in 1987. He co-founded Enter Software, Inc. in 1988
where he served as vice president and director of engineering. Mike is now
president of Enginomix, LLC (www.enginomix.net), a consulting and
software development company focusing on integrated engineering and
economic modeling solutions for power plant design and operations.Table of Contents
Foreword 5
1. Overview of Performance Monitoring B
1.1 Concept of Performance Monitoring 13
1.1.1 “Where You Are” Vs “Where You Should Be” 13
1.1.2 Performance Calculation Procedure 7
1.1.3 Expected Performance: “Where You Should Be” 18
1.1.4 Equipment Ratings 20
1.1.5 Corrected Performance 26
1.1.6 What is My Degradation? 29
1.1.7 How Much is Degradation Costing Me? 30
1.1.8 Optimization: “Where You Could Be” 32
1.1.9 Controllable Loss Displays 33
1.2 ASME Test Codes 35
1.3. Performance Testing versus Online Monitoring 37
1.4 Curve Based Methods 38
1.4.1. Performance Curves 38
1.4.2 Expected Performance from Curves. 44
1.4.3 Additive Performance Factors 46
1.4.4 Expected Performance from Curves 47
1.4.5 Correction Factors 50
1.4.6 Percent Change Correction Factors 53
LS Model-Based Performance Analysis 54
2. Heat Balance Analysis 63
21 Local Heat Balances 64
2.2 Combined-Cycle Overall Plant Heat Balance 9
23 Combined-Cycle Balence Using Commercial Software 76
2.4 — Rankine-Cycle Overall Plant Heat Balance 9
2.5 _ Rankine-Cycle Balance Using Commercial Software 85
3. Data Validation 89
3.1 Definition of Data Validation 89
3.2 Range Checking 90
3.2.1 Static Ranges 90
3.2.2 Dynamic Ranges 90
3.2.3 Rejected Values 92
3.3 Averaging Sensor Data 93
3.4 Time Averaging 94
be Heat Balances for Data Validation 9o7
4. Accuracy of Calculated Results 105
4.1 Instrument Error 1054.1.1 Measurement Error
4.1.2. Random Uncertainty
4.13. Systematic Uncertainty
4.2 Uncertainty of a Calculated Test Result
4.3 Monte Carlo Method
4.3.1 Definition of the Monte Carlo Method
4.3.2 Probability Distributions
4.3.3 Sampling from Probability Distributions
43.4 Running the Monte Carlo Simulation
4.3.5 Results of the Monte Carlo Simulation
Overall Power Plant Performance
5.1 Equipment Performance versus Plant Performance
5.2 Specification of Overall Power Plant Performance
5.3 Overall Plant Expected Performance Models
5.3.1 Curve-Based Method Expected Plant Perf
5.3.2. Model-Based Method Expected Plant Perf
5.3.3 Impact Method Jor Expected Plant Performance
5.4 Degradation of the Overall Power Plant
Impacts of Degradation on Overall Plant Performance
6.1 Definitions of Plant Impacts
6.2 Gas Turbine Impacts
6.3 Heat Recovery Steam Generator Impacts
6.4 Steam Turbine Impacts
65 Boiler Impacts
6.6 Feedwater Heater Impacts
6.7 Condenser Impacts
6.8 Cooling Tower Impacts
6.9 Inlet Air Filter Impacts
6.10 Exhaust Pressure Loss Impacts.
Gas Turbine Performance
7.1 Overview
7.2 Power Generation
7.3 Airflow, Firing Temperature and Pressure Ratio
7.4 Control Algorithms
7.5 Correction Curves (Baseload Performance)
7.5.1" Effect of Inlet Temperature
7.5.2 Effect of Inlet Humidity
7.5.3 Effect of Atmospheric Pressure or Altitude
7.5.4 Effect of Inlet Pressure Loss
7.5.5 Effect of Exit Pressure Loss
7.5.6 Effect of Steam or Water Injection
7.6 Part-Load Performance (Industrial Engines)
105
107
108
109
110
110
MW
113
115
115
7
117
119
122
122
124
126
128
129
129
131
135
136
137
139
140
141
142
144
147
147
147
149
153
156
159
159
160
162
163
164
16617
78
19
7.10
7
7.12
7.13
114
715
Part-Load Correction Curves
7.7.1 Under-Firing Correction
7.7.2. Inlet Guide Vane Correction
7.7.3 Part-Load Expected Heat Rate
Aeroderivative Engine Performance
Overall Gas Turbine Heat Balance
7.9.1 Determination of Exhaust Gas Specific Heat
7.9.2. Detailed Gas Turbine Heat Balance
7.9.3. Tuning Detailed Gas Turbine Heat Balance
7.9.4 Step-by-Step Solution of the Equations
7.9.5 Simultaneous Solution of the Equations
7.9.6 Combustion Mass Balance Analysis
7.9.7 Specific Heat of a Mixture
Model-Based Gas Turbine Heat Balance
Physically-Based Models of Expected GT Performance
Gas Turbine Performance Evaluation
A theoretical degradation curve versus time is in
Experience with Measured Data from Operating GT
Performance Degradation and Engine Life
Heat Recovery Steam Generator Performance
8&1
8.2
83
8.4
85
8.6
8.7
88
8.9
8.10
8.11
8.12
Overview
8.1.1 Economizers
8.1.2 Evaporators
8.1.3 Blowdown
8.1.4 Superheaters
Duet Burner
HRSG Efficiency and Effectiveness
Expected HRSG Performance
8.4.1 Effect of Duct Burner Firing
8.4.2 Effect of Exhaust Gas Temperature
8.4.3 Effect of Exhaust Gas Flow
8.4.4 Effect of Steam Pressure
HRSG Heat Balance Analysis
Model-Based HRSG Heat Balance Analysis
Expected Section-by-Section Performance
Impact of Fouling on HRSG Performance
HRSG Performance Evaluation
Example Performance Analysis Fouled HP Evaporator
Example of Section-by-Section Expected HRSG Perf
Conclusions and Recommendations
Steam Turbine Performance
9.1
Overview
168
168
169
172
174
177
179
185
195
197
198,
201
209
210
214
216
225
226
227
231
231
232
235
245
246
248
251
257
260
263
264
265
266
273
280
284
289
294
297
299
301
3019.2
93
94
9.5
9.6
97
98
99
Steam Turbine Configurations
9.2.1 Inlet Section
9.2.2. Condensing Section
9.2.3 Back-Pressure Steam Turbines
9.2.4 Extractions
9.2.5 Controlled (‘Automatic’) Extraction
9.2.6 Uncontrolled Extraction
9.2.7 Admission
9.28 Reheat
Seals and Leaks
Steam Turbine Thermal Performance
9.4.1 Steam Turbine Efficiency and Heat Rate
9.4.2 Pressure, Temperature and Flow Relationships
Steam Turbine Heat Balance Analysis
9.5.1 Combined-Cycle ST Heat Balance Analysis
9.5.2 Rankine Cycle ST Heat Balance Analysis
Curve-Based Expected Performance
9.6.1 Rankine Cycle ST Correction Curves
9.6.2 Combined Cycle ST Performance Curves
Model-Based Expected Steam Turbine Performance
9.7.1 Expected Performance of Overall ST
9.7.2 Section-by-Section Expected ST Performance
Building ST Expected Performance Models
Steam Turbine Degradation
Boiler Performance
10.1
10.2
10.3
10.4
10.5
10.6
10.7
10.8
10.9
Boiler Efficiency
Theoretical Air
Boiler Losses
Flue Gas Loss
10.4.1 Generalized Chemical Balance Method
10.4.2 Products of Combustion Method
10.4.3 Loss Due to Moisture
Loss Due to Ash
Loss Due to Radiation
Credits for Heat Addition to Boiler
Boiler Heat Balance Analysis
10.8.1 Furnace Heat Balance Analysis
10.8.2 Analysis of Boiler Convective Heat Exchangers
10.8.3 Desuperheater Heat Balance
10.8.4 Air Heater Heat Balance
10.8.5 Simultaneous Solution of the Equations
Model-Based Boiler Heat Balance Analysis
302
302
308
314
315
315
316
320
321
323
325
325
328
329
329
335
341
342
345
348
348
352
355
362
367
367
369
372
373
373
374
375
379
379
380
381
385
395.
396
397
399
4082
13.
14.
15.
16.
17.
10.10. Expected Boiler Performance
10.10.1 Curve-Based Method for Exp Boiler Perf
10.10.2 Model-Based Expected Boiler Performance
10.11 Boiler Degradation
10.12 Sootblowing Analysis
Air Heater Performance
11.1 Overview
11.2 Air Heater Heat-Balance Analysis
11.3 Air Heater Expected Performance
11.4 Air Heater Degradation
Feedwater Heater Performance
12.1 Overview
12.2 Feedwater Heater Heat-Balance Analysis
12.3. Expected Feedwater Heater Performance
12.4 Feedwater Heater Degradation
Deaerators, Drums and Open Heaters
Condenser Performance
14.1 Overview
14.1.1 ASME Method for Condenser Heat Transfer
14.1.2 The HEI Method for Condenser Heat Transfer
14.2 Condenser Heat Balance Analysis
14.2.1 Overall Plant Energy Balance for Cond Duty
14.2.2 Steam Turbine Expansion Line Analysis
14.2.3 Condenser Heat Balance Equations
14.2.4 Condenser Cleanliness from Measured Data
14.2.5 Validation of Condenser Heat Balance Data
14.3 Condenser Expected Performance
14.3.1 Predicting Expected Condenser Performance
14.4 Condenser Degradation
14.5 Diagnosing Condenser Performance Problems
Cooling Tower Performance
15.1 Overview
15.2 Cooling Tower Performance Curves
15.3 Cooling Tower Heat Balance Analysis
15.4 Expected Cooling Tower Performance
15.5 Cooling Tower Degradation
Inlet and Exhaust Pressure Losses
16.1 Overview
16.2 Fitting the Pressure Loss Equation to Data
16.3 Pressure Loss Degradation
Pump Performance
17.1 Overview
410
412
416
426
430
433
433
440
443,
44s
447
447
449
453
437
461
461
461
464
466
467
467
468
469
47
474
415
475
477
480
487
487
490
493
500
502
503
503
504
505
507
50717.2
173
174
17.5
17.6
17.7
17.8
Extended Bernoulli Equation
Pump Curves
Affinity Laws
Corrected Pump Performance
Pump Flow Control
Model-Based Pump Performance
Pump Degradation
References and Links
Nomenclature
APPENDIX Definition of Terms
507
512
514
sis
518
519
521
523
527
531Foreword
I developed an interest in performance monitoring in 1983 when I worked as
a consultant to Pacific Gas and Electric (PG&E). I was asked to review
measured data from the steam cycle of a nuclear power plant. The task was,
to evaluate feedwater heater performance by comparing plant measured data
toa PEPSE™ computer model (built by someone else at PG&E) of the
steam cycle, and identify any discrepancies which might indicate a
performance problem.
I compared the measured feedwater heater TTD’s (terminal temperature
differences) to the predictions of the PEPSE™ model. The computer code
predicted integer numbers for each of the feedwater heater TTD’s. The
measured data was within 43 F of the predicted values from the computer
code for all of the feedwater heaters, but did not match any of them.
The fact that the predicted TTD’s were integers indicated to me that the
computer model was a design prediction of the plant performance, and that,
the TTD’s were inputs to the analysis. Did the predicted results mean that
some of the feedwater heaters were better than expected, and that some of
them were fouled? Or was this computer prediction just a theoretical design
model that did not necessarily represent reality at the plant. [ did not know,
so I decided to run some alternate computer calculations to see if I could
learn more.
Tlooked at the measured drain flows from each feedwater heater. I knew
that the TTD’s and the drain flows were related to each other by mass and
energy balances. The beauty of a computer code like PEPSE™ is that all
these mass and energy balances are automatically calculated. I just needed
to input plant measured data (TTD’s), and PEPSE™ would calculate heat
balance data that is consistent with my inputs.
Unfortunately, the measured drain flows bore little resemblance to the
computer calculated values for these flows. Typically the calculated flow
values were different from the measured values by 30% or more. I came to
the conclusion that the measured drain flows were of little value. A PG&E
engineer confirmed that the drain flows were known to be inaccurate.
I decided to see what effect a TTD has on the overall steam-cycle
performance. I chose a feedwater heater that appeared to be performing
poorly (TTD measured three degrees F higher than predicted) and entered
this measured TTD into the PEPSE™ input field for the chosen feedwater
heater. I ran the PEPSE™ prediction and was astonished to see that the only
predicted result than changed very much was the TTD for the feedwater
Forewordheater that I had input to the analysis. The remained of the steam cycle was
almost unchanged. The feedwater temperature at the exit of every feedwater
heater, except for the one feedwater heater that I changed, was exactly the
same as in the original computer prediction.
This exercise taught me the difference between design analysis, where
equipment operating data (such as a TTD) is input, and an off-design
analysis (also called predictive analysis) where the equipment performance
capabilities (such as surface area and heat transfer coefficient at the design
point) are input and the operating data is predicted. In my case, the
computer analysis increased the size (heat transfer coefficient times surface
area) of the feedwater heater following the degraded feedwater heater such
that the fall-off in feedwater temperature caused by the degradation in one
feedwater heater was exactly made-up by the improved performance in the
following feedwater heater. I refer to design analysis as running the
“tuber” power plant because the heat balance code changes the size of
plant equipment to meet the specifications (temperatures, flows, power
levels and pressures) of the software user.
‘The concept of design analysis versus off-design analysis was not developed
{for application to performance monitoring, but it has turned-out to be one of
the key characteristics that makes heat balance codes useful as the
calculation engines for on-line performance monitoting systems.
Performance monitoring involves the comparison of current performance to
expected performance. Design analysis can be used to mach a computer
heat-balance analysis to the current plant operating conditions (current
performance), and the off-design analysis can be used to predict the
expected equipment performance (expected performance) given those
current operating conditions.
Later in my career, when my company developed the GateCycle™ heat
balance code, the concept of design versus off-design analysis was built
directly into code structure. We wanted it to be easy for a user to establish
the design-point performance of a power plant in one calculation, and then
switch to the off-design analysis to predict the power plant performance
over a range of postulated operating conditions.
Several vendors of commercial heat balance products have built the concept
of design versus off-design analysis into their produets. The GT PRO™ and
STEAM PRO™ products from Thermoflow, Inc. perform design analysis,
and allow the user to transfer the results of the design analysis to the
predictive (off-design) analysis that is done by the GT MASTER™ and
STEAM MASTER™ computer codes.
ForewordThe GateCycle™ user can establish the design-point performance of a
power plant in one calculation, and then switch to off-design analysis (by
clicking a check box on the user interface) to predict the power plant
performance over a range of postulated operating conditions.
PEPSE™ allows the user to perform design or off-design analysis through
the option to either input a desired equipment output parameter (such as
feedwater heater TTD), or detailed equipment characteristics (such as the
number of tubes, the tube sizes, the tube material and surface type of a heat
exchanger). The choice of inputs must be selected individually for each icon
in the PEPSE™ model.
‘The engineer that I reported to at PG&E just wanted me to compare the
predicted temperatures to the measured temperatures, and identify any
potential problems. Unfortunately there were too many discrepancies, and I
didn’t know how to resolve them. I soon realized that the running of a
design computer model of the plant and comparing to measured data was
not an adequate process to identify equipment degradation.
I describe this example to illustrate the point that it is not obvious how to
use commercial heat balance codes to monitor performance, even when you
are familiar with all of the inputs and outputs of the heat balance code. 1
spent next two decades developing and implementing procedures for using
heat balance analysis and commercial computer codes to monitor the
performance of power plants and their equipment. This book documents
what I learned.
Returning to the discussion of the performance monitoring evaluation at the
PG&E nuclear power plant, I can now say what I should have done to
evaluate the plant data back in 1983.
First, I would assume that the PEPSE™ computer model, given to me by
PG&E, was a design analysis model that matched a vendor heat balance or
guarantee prediction of the power plant performance at full load. The TTD’s
in this model probably came from a heat-balance diagram delivered to
PG&E by the plant vendor. This model would become the basis of
predictive models for the expected performance of the plant equipment.
The equipment performance characteristics from the design model, such as
the UA (heat transfer coefficient multiplied by surface area) of each heat
exchanger, and the design isentropic efficiency of each steam turbine
section, would be the inputs to the predictive (off-design) models. I would
copy and rename the design PEPSE™ model, and then manually edit the
inputs from design to off-design for each piece of plant equipment in the
ForewordPEPSE™ model (these equipment inputs are changed for you automatically
in GateCycle™ by selecting the off-design analysis mode),
Fora feedwater heater, the input would be changed from TTD to surface
area plus the design value of heat transfer coefficient. If I did not know the
actual surface area, I would choose a correlation for heat transfer coefficient
and then iterate on the input value of surface area until the feedwater heater
TTD matches the value used in the design analysis. The only requirement is
that the product of surface area and heat transfer coefficient result in the
desired TTD. I would then run this off-design PEPSE™ model to confirm
that its prediction matches the results of the design model when the plant is
running at the design operating conditions.
I would now have two PEPSE™ models, the original design model and an
expected plant performance model that matches the design model at the
design plant operating conditions, and will correctly predict plant
performance at various plant operating conditions (such as changes in load
or environmental conditions). The original design model would not correctly
predict plant performance over a range of plant operating conditions because
the TTD’s (and other parameters) are held constant,
I will also need separate PEPSE™ expected performance models of each
feedwater heater, each steam turbine section and the condenser. These
models will be used to predict the expected performance (TTD, condenser
pressure, steam turbine power) of each piece of plant equipment. Each of
these PEPSE™ models would contain a single piece of plant equipment plus
sources representing the flows into the equipment from other locations in
the steam cycle, and sinks to receive the flows going out of the equipment.
I would use the following procedure to evaluate the performance of the
feedwater heaters.
1, Perform Plant Heat Balance Analysis
Run the design PEPSE™ model of the overall plant with plant
| measured data (such as the TTD of each feedwater heater) as the
input Values to each piece of equipment in the plant. This is called
the “heat balance analysis” of the power plant. The heat balance
analysis yields a set of current plant operating data that matches the
plant measured data, but is more complete than the measured data.
This heat balance data will include all the steam turbine extraction
flow rates and their enthalpies, and it will also include feedwater
heater drain flow rates that are consistent with the extraction flows
and the feedwater heater TTD’.
Foreword2. Calculate Expected Equipment Performance
Predict the expected performance of each feedwater heater using a
separate PEPSE™ model for each feedwater heater. The inputs to
each feedwater heater model are the mass flow rates and enthalpies
of all three streams that flow into the feedwater heater: the extraction
steam, the feedwater and the incoming drain water from the higher
pressure heaters, The design heat transfer characteristics (UA) of
each feedwater heater are part of the input data for each feedwater
heater. The expected feedwater heater outlet temperatures (from
which TTD and DCA can be calculated) are outputs of this
calculation.
3. Evaluate Degradation
Degradation is based upon the difference between the current
performance and expected performance. The current performance of
‘a feedwater heater is the measured TTD. The expected performance
is the expected TTD, calculated for each feedwater heater. The
difference between the measured TTD and expected TTD is an
evaluation of degradation in the feedwater heater.
The expected TTD’s from the above performance monitoring procedure will
be very different from the design point TTD’s that were used as inputs in the
original PEPSE™ model. Visualize what would happen if one feedwater
heater is degraded such that its TTD is three degrees higher than the design
TTD. The plant heat balance analysis (step 1. above) will calculate a lower
extraction flow for this degraded feedwater heater. This lower extraction
flow will result in higher steam turbine pressure at the extraction location
and at all the extraction locations downstream in the steam-turbine steam.
path. These higher extraction pressures result in higher steam saturation
temperatures, which will change the TTD’s of all the feedwater heaters
receiving this higher pressure steam.
The higher TTD from the degraded feedwater heater will result in a lower
feedwater temperature at the inlet to the next (higher pressure) feedwater
heater, and this will change both the TTD and the extraction steam flow to
that higher pressure feedwater heater. The changes in extraction flows will
change the steam turbine power and the exhaust energy from the steam
turbine, which changes the condenser pressure.
These complex interrelationships between the feedwater heaters, the steam
turbine and the condenser indicate that degradation at one point in the steam
cycle will cause changes in the measured (and expected) performance all
around the steam eycle. The expected TTD of any feedwater heater is
Foreworddependent upon the performance of the remainder of the steam cycle
equipment. It is necessary to perform a heat balance analysis (step | above)
to quantify these interrelationships. Then an expected equipment
performance calculation (step 2 above) can predict the expected output from
each feedwater heater given the performance of other equipment in the
plant
In the case of the degraded feedwater heater, the expected TTD of the higher
pressure feedwater heater will be higher when the lower pressure feedwater
heater is degraded than it would be if the lower pressure feedwater heater is
not degraded. If the monitoring system assumes a “target” TTD for a
feedwater heater that doesn’t depend upon the performance of the other
feedwater heaters, it is assuming that a higher pressure feedwater heater will
make-up any degradation in the lower pressure feedwater heaters. This is
equivalent to assuming that the “rubber” (design analysis) model of the
plant is an accurate prediction of the expected equipment performance.
The one remaining problem with the performance analysis described above
is that the expected performance prediction is based on a design model that
might not represent the actual performance expected for plant equipment.
One way to resolve this is to obtain data from early in the plant operational
history when degradation can be consicered to be zero. Use this plant data
as the design data in the plant design aralysis model instead of vendor
design or guarantee data. This “tuning” or “base-lining” process can be
repeated at any time during the life of the plant: use measured plant data as
the plant design data such that the plant degradation is zero at the time of the
measured data,
An on-line performance monitoring system should be thought of as a
relative evaluation instead of an absolute evaluation. Because plant
measured data does not normally come from calibrated, precision-test
instruments, the absolute magnitude of the results may not be accurate.
However, on-line performance monitor:ng systems can be very good at
detecting changes in performance.
Because an on-line performance monitcring system produces relative
results, the degradation of plant equipment must be tracked over time to
identify changes that have occurred. For this reason, itis important to install
a performance monitoring system early in the operational history of the
plant. Any performance changes that occurred before the monitoring system
‘was installed at the plant may be missed by the monitoring system.
The only way I would be able to tell my PG&E manager about degradation
in the feedwater heaters would be to run earlier measured data through the
performance monitoring calculation described above, and then compare the
Foreworddegradation calculated from the earlier time to the degradation calculated
from the current plant measured data. The changes in calculated degradation
are an accurate indication of changes in equipment performance, but the
absolute values of the degradation may not be accurate.
Rodney R. Gay
ForewordForeword1. Overview of Performance Monitoring
1.1 Concept of Performance Monitoring
1.1.1. “Where You Are” Versus “Where You Should Be”
Performance monitoring is the process of continuously evaluating the
production capability and efficiency of a power plant and its equipment over
time using measured plant data. Performance monitoring evaluations are
repeated at regular intervals using data readily available ftom on-line
instrumentation. This differs from a performance test, a one-time event that
relies on precision instrumentation installed specifically for that test.
The objective of performance monitoring is to continuously evaluate the
degradation (decrease in performance) of the plant and its equipment in
order to provide plant operators additional information to help them identify
problems, improve performance, and make economic decisions about
scheduling maintenance and optimizing plant operation. A successful
performance monitoring system can tell plant operators how much the plant
performance has changed and how much each piece of equipment in the
plant contributed to that change. This information enables operators to
localize performance problems within the plant and to estimate the
operational cost incurred because of the performance deficits.
While it is expected that performance monitoring will help operators
diagnose and repair faults in plant equipment, the diagnostic procedures to
accomplish this are beyond the scope of this book.
To answer the question “How good is my performance?” one must compare
the current capability of the power plant and its equipment to its expected
capability. Thus, performance monitoring is a comparison of the current
capability, “Where You Are”, to the exnected capability, “Where You
Should Be”.
Production capability is a measure of tke ability of equipment to produce the
output that the equipment is designed to produce; it is not the current
production. In other words, a plant that is designed to generate (produce)
600 MW, might only be able to generate $50 MW on a hot day, but still be
capable of generating 600 MW when operating at its design conditions. The
objective of performance monitoring is to continuously evaluate this
capability and monitor its change over time.
Degradation is defined as the shortfall in equipment performance caused by
‘mechanical problems in the equipment (such as wear, fouling, and
page 13 1. Concept of Performance Monitoringoxidation), but not by changes to set points under the control of the plant
operators. For example, if plant operators increase the excess oxygen on a
coal-fired boiler to reduce CO emissions when burning low quality fuel, the
boiler efficiency will decrease. The boiler capability has not changed: if the
fuel and excess oxygen level were returned to their original value, the boiler
efficiency would also improve to its original value. Thus, the observed
efficiency decrease in this example is not degradation, but is instead an
opportunity for economic optimization.
A second example is a gas turbine whose water-to-fuel injection ratio must
be increased to meet more restrictive NOX emissions requirements. The
engine power would increase and the heat rate would get worse (increase).
These changes in performance do not represent degradation, just a change in
operating conditions.
Economic optimization is concemed with finding the plant operating mode
and control set points that meet all constraints on plant operation (such as
equipment protection and emissions limits) and maximize plant profits. The
current degradation of plant equipment is an important input to optimization
analysis and the current plant control set points are important inputs to
degradation analysis, but the two are separate evaluations.
Performance monitoring involves two calculations: current production and
expected production. The evaluation of performance degradation is a
comparison between these two values. For example, a plant designed to
produce 600 MW on a 59 F day may be expected to produce 550 MW on a
100 F day. If the plant meets its expected production of 550 MW on the 100
F day, then its performance is as expected (zero degradation) even though it
did not perform at its design production level of 600 MW.
Table 1-1 lists the plant equipment types discussed in this book, the
production objective(s) of each equipment type, and an output parameter
that is a measure of each production objective. Any performance evaluation
of the equipment listed in the table must relate the current production
capability of the equipment to the expected production capability. Notice
that the equipment types that consume fuel have two production objectives,
and hence two measurements of performance. This is because output and
efficiency are independent parameters for these equipment types. For fuel-
consuming equipment, efficiency needs to be evaluated along with output
production capability because it may be possible to achieve higher output by
simply consuming more fuel.
For other equipment types (non-fuel-consuming types such as heat
exchangers and steam turbines), the input source of energy is fixed (that is,
not determined by the performance of the equipment type being monitored)
1.Concept of Performance Monitoring page 14and therefore higher efficiency causes higher output. Thus, for these
equipment types output and efficiency are not independent performance
parameters,
Table 1-1 List of equipment types and their production objectives
Equipment Production Objective
Measured Output
Electricity
Power Plant
Efficiency
Efficiency
Net Power (MW)
Net Heat Rate
Heat Rate
Steam Generation
Steam Flow(s),
Steam Turbine
Electricity
Boiler Temperature(s) and
Pressure(s)
Efficiency Boiler Efficiency
Heat Recovery Steam | Steam Generation Steam Flow(s),
Generator ‘Temperature(s) and
Pressure(s)
Condenser Vacuum Condenser Shell Pressure
Cooling Tower Energy Rejection Cooling Water Temperature
to Condenser
Feedwater Heater Feedwater Heating Feedwater Outlet
Temperature
‘The performance of a power plant has two measures: power and heat rate.
‘They are independent measures of peiformance in that the highest power is
not necessarily achieved at the best (lowest) heat
rate. A plant operator
generally has the option to control the plant for maximum power output or
to control for maximum efficiency. A performance evaluation of a power
plant must include evaluations of both the power generation capability and
the heat rate capability.
page 1 1. Concept of Performance MonitoringA gas turbine is like a power plant; in ‘act, a simple-cycle gas turbine is a
power plant. Thus, both power and heat rate are independent performance
parameters that must be evaluated when monitoring a gas turbine.
A boiler consumes fuel to generate steam. Both the’steam generation
capability and the boiler efficiency are important parameters of boiler
performance, and both must be evaluated.
The job of a heat recovery steam generator (HRSG) is to convert the
available exhaust gas energy into as much steam as possible. When the plant
is operating at full load, the temperature and pressure of the steam are
controlled by the plant, and therefore are not independent parameters of
HRSG performance. They represent requirements that the HRSG must meet.
Improved HRSG effectiveness or efficiency results in increased steam
generation. There is no opportunity to increase steam generation capability
without increasing HRSG efficiency; thus, efficiency and steam generation
are not independent parameters of performance. A performance monitoring
system must compare the current value of HRSG steam generation or
efficiency to its expected value.
‘A condenser’s job is to condense all of the steam exhausted from the steam
turbine at a pressure as low as possible. The need to condense alll of the
steam is a requirement that must be met. Condenser pressure is the measure
of condenser performance: the lower the pressure the better the
performance. A performance monitoring system must compare the current
value of this pressure to its expected value. Other parameters of condenser
performance, such as cleanliness, are only important because they are an
indication of the ability of a condenser to reduce steam turbine exhaust
(condenser) pressure to its expected value.
A cooling tower must reject all of the steam condensation energy (condenser
duty) to the cooling media (air or water). The quantity of energy to reject is
a requirement that the cooling tower must meet. The measure of
performance of a cooling tower is the cooling water temperature at the exit
of the cooling tower (or at the inlet to the condenser). A lower value of thi
temperature indicates better performance. A performance monitoring system
must compare the current value of this temperature to its expected value,
1.1.2 Performance Calculation Procedure
Performance monitoring involves a calculational procedure that is repeated
at regular time intervals. The details of the calculation vary greatly from
plant to plant, depending upon the measured data that is available, the plant
type, and the degree of sophistication cf the calculations. However, a
1. Concept of Performance Monit page 16performance monitoring calculational procedure always involves some or all
of the following steps:
1. Acquire measured data.
2. Review, check and/or validate the raw measured data to find errors
and omissions.
3. If possible, fix errors or omissions identified in the measured data.
4. Improve precision of the measured data by averaging and/or other
techniques.
5. Compute fluid thermal properties, such as enthalpy and entropy,
from measurements.
6. Use mass, energy and/or chemical balances to calculate data that is
not measured, but can be computed from the measurements that do
exist.
7. Compute current values for equipment parameters of performance
such as heat rates, efficiencies, effectiveness’s, temperature
differences, and cleanliness.
8. Predict expected values for equipment parameters of performance.
9. Compute the corrected performance of the plant equipment
10. Calculate the shortfall in performance (degradation), based upon the
difference between the expected and current values of the
performance parameters.
11, Estimate the effect (impact) that the equipment degradation has on
plant performance and plant operating cost.
12, Perform plant optimization calculations to predict the most cost-
effective way to run the degraded plant equipment.
A given performance monitoring system often will not perform all of these
calculational steps, but the list is a fairly complete compilation of the
calculations that can be and probably should be done in a comprehensive
and successful performance monitoring system.
1.1.3. Expected Performance: “Where You Should Be”
For performance monitoring to be meaningful, one must compare current
performance to expected performance, and track that comparison over time.
This process is equivalent to tracking degradation (the difference between
expected and current performance) over time. Since performance
monitoring is a continuous process, as opposed to a one-time event like a
page17 1. Concept of Performance Monitoringperformance test, the performance evaluation will be performed over a
variety of plant operating conditions. This makes the evaluation of expected
performance the most challenging aspect of performance monitoring.
Figure 1-1 illustrates the concept of expected versus actual (degraded)
power for a gas turbine engine. The baseload power of a gas turbine engine
varies with inlet air temperature, as illustrated by the expected power line in
this figure. It is assumed for the purposes of this discussion that ambient
temperature is the only environmental parameter that is changing, Gas
turbine vendors typically provide performance curves which show how
performance will change with environmental conditions. A vendor
performance curve can be used to compute the expected power line. Notice
that one point on the expected power line is the rated power, which occurs
at only one air inlet temperature (shown as Trefeence in Figure 1-1).
When a degraded gas turbine is operated over a range of inlet air
temperatures, the measured gas turbine power levels will likely be along a
line below the expected power line, as illustrated in Figure 1-1.
The corrected power is the power that the actual (degraded) engine would
produce if operated at the reference temperature. The difference between the
tated and corrected power is the degradstion of the engine from rated.
The procedure to calculate expected performance is to start with the
expected performance at the reference operating conditions (the rated
performance), and then use a model or models of equipment performance to
predict the change in equipment performance when the equipment is
operated at conditions different from the reference operating conditions. The
‘model(s) of equipment performance can be very simple, such as table look-
ups, or very complex, such as a physically based computer code.
1. Concept of Performance Monitoring page 18Performance Evaluation Terms
GT Power
Expected (ne degradation)
Expected
Performance
‘Actual
Performance
rae Trotoane Inlet Temperature
dooF
Figure 1-1, Comparison of rated, expected, measured and corrected power for a
gas turbine
The expected performance line in Figure 1-1 is actually a simple example of
a performance model of a gas turbine. This line shows how gas turbine
power will change as the gas turbine inlet temperature changes. This line
could be converted into a table look-up as part of a computerized
performance model.
Of course, any gas turbine performance monitoring system must also
account for changes in other reference operating conditions such as inlet
pressure loss, exhaust pressure loss, fuel properties, inlet pressure, inlet
relative humidity, steam/water injection, inlet guide vane angle and firing
temperature. Since these are independent parameters of gas turbine
performance, separate models can be used for each condition, and the total
power change is the product of the power changes predicted from the
changes in each reference operating condition. Using curves to evaluate
equipment performance is discussed later in this chapter under “Curve-
Based Methods”.
page19 1, Concept of Performance MonitoringAn alternative model of gas turbine performance is a computer code that
includes physically based mathematical models of the compressor,
combustor and expander. Such a code would take the operating conditions
as inputs and predict the gas turbine power and heat rate at those operating
conditions. It would be necessary to adjust (tune) such a computer code so
that it accurately predicts the gas turbine rated performance at the reference
operating conditions, Then the performance monitoring system could input
‘measured data into the computer code to predict the expected performance
at the current measured operating conditions. Using physically based
computer models to evaluate equipment performance is described later in
this chapter under “Model-Based Performance Analysis”.
1.1.4 Equipment Ratings
The rated performance of plant equipment must include a specification of all
the external conditions and control settings that change equipment
performance but are not part of the equipment itself. Table 1-1 lists all of the
specifications that are required to state the rating of a gas turbine.
‘There are several ways to obtain the rating data for a plant and its
equipment. For performance monitoring purposes, the choice is somewhat
arbitrary since a monitoring system tracks changes in performance or
degradation over time. If the monitoring system defines degradation as the
fall-off in performance over time, the absolute value of the rating cancels
‘out. Several ways to define equipment ratings are:
‘© Use vendor guarantees
© Use acceptance test (as-built) data for the plant and equipment
‘* Use plant measured data at the time the monitoring system is
installed
© Baseline (tune) the ratings on a regular basis using plant measured
data
A gas turbine will produce its rated power and heat rate only at the reference
operating conditions listed. The values of the reference operating conditions
are called the reference data. All of the data in Table 1-2 are related. Change
any of the operating conditions (from their reference values), and the power
and heat rate of the engine will change (from rated),
1. Concept of Performance Monitoring page 20Table 1-2 Typical rating specifications for a gas turbine engine
Gas Turbine Rating Specifications Example Data
RATING:
Gross Power 170oMw
Gross Heat Rate 9400 BewKW-hr
REFERENCE OPERATING CONDITIONS:
Ambient Temperature 59 deg-F
Ambient Pressure 14.65 psia
Ambient Specific Humidity 0.0065
Ibm H,0/lbm air
Inlet Pressure Loss 4 in H,0
| Exhaust Pressure Loss 12 in HO
‘Steam/ Water Injection none
Fuel Type Natural Gas
| Fuel Lower Heating Value 20200 Brw/lbm
Inlet Guide Vane Angle 86 deg,
Firing Temperature 7300 deg-F
Inlet Cooling or Heating none
The expected performance prediction for a gas turbine, or for any equipment
type, requires both a set of rating specifications (which includes both the
rated performance and the reference operating conditions), plus a model of
performance that predicts how performance changes as the operating,
conditions change.
Table 1-3 lists the rating specifications for a typical heat recovery steam
generator.
page2l 1. Concept of Performance MonitoringTable 1-3 Typical rating specifications for a heat recovery steam generator (HRSG)
Heat Recovery Steam Generator Example Data
Rating Specification
RATING:
HP Steam Flow 415,000 lb/hr
IP Steam Flow 70,000 liar
REFERENCE OPERATING CONDITIONS:
Exhaust Gas Flow 3,250,000 Ilr
Exhaust Gas Temperature 1138 F
Exhaust Gas Composition 3% HO
HP Drum Pressure 1900 psia
IP Drum Pressure 400 psia
LP Drum Pressure 100 psia
Inlet Feedwater Temperature 140F
HP Steam Temperature 1100 F
Duct Bumer Fuel Flow none
Steam Extraction to Process 20,000 Ilar
Water Extraction to Process 30,000 tar
Once again, the rating specifications for an HRSG indicate that the HRSG
will produce the rated steam flows only if it is operating at the reference
‘operating conditions. To predict HRSG expected performance, a monitoring
system must be able to predict the change in HRSG performance as
operating conditions change from their reference values.
Table 1-4 gives typical rating specifications for a steam turbine.
1. Concept of Performance Monitoring page 22Table 1-4 Typical rating specificaiions for a steam turbine/generator
Steam Turbine Rating Specification
T
Example Data
RATING:
Gross Power 190 Mw
REFERENCE OPERATING CONDITIONS:
Throttle Steam Flow £930,000 Ibrar
Throttle Steam Temperature 137F
Throttle Steam Pressure 1800 psia
Condenser Back Pressure 0.8 psia
Reheat Steam Temperature 1000 F
UP Extraction Flow none
LP Admission Flow 160,000 tbrhr
A steam turbine will generate its rated power only at the rated steam flow
conditions and condenser pressure. Any change in these flow conditions or
pressures will cause the steam turbine power to change.
page 2! 1. Concept of Perfarmance Monitoring‘Table 1-5 Typical rating specifications for a condenser
Condenser Rating Specification Example Data
RATING:
Shel (Steam) Pressure 0.8 psia
REFERENCE OPERATING CONDITIONS:
Inlet Steam Flow 930,000 tbrhr
Inlet Steam Enthalpy 1000 Brullb
Cooling Water Flow 6,000,000 Ibe
Cooling Water Inlet Temperature 80F
‘A condenser is required to condense all of the incoming steam and transfer
the energy released from condensation to the cooling water. The condenser
duty, the cooling water flow and the cooling water inlet temperature are
‘imposed upon the condenser by the performance of other equipment in the
plant (external to the condenser). The condenser is designed to achieve its
rated pressure at a given (reference) set of inlet flow conditions. Any change
in the inlet steam or water flows will be expected to change the condenser
pressure.
1. Concept of Performance Monitoring page 24Table 1-6 Typical rating specifications for a coal-fired boiler
Boiler Rating Specification
Example Data
RATING:
Main Steam Generation 2,560,000 tbvhr
Boiler Efficiency 89.59%
REFERENCE OPERATING
CONDITIONS:
Fuel Input Energy 3374 mmBtwhr
‘Steam Drum Pressure 2800 psig
Steam Temperature 1005 F
Reheat Steam Temperature 1005 F
Reheat Steam Flow 2,275,000 lb/hr
Reheat Steam Inlet Pressure 592 psig.
Fuel Higher Heating Value 11,495 Bru
Fuel Composition (C, H, N, $, HO, Ash)
(64.2,4.1,25,4.4,08,4.1,19.9)
Inlet Feedwater Temperature 475 F
Inlet Air Temperature 80F
Inlet Air Relative Humidity 60%
page 2S 1. Concept of Performance MonitoringTable 1-7 Typical rating specications for a feedwater heater
Feedwater Heater Rating Specification Example Data
RATING:
Outlet Feedwater Temperature 420F
Outlet Drain Water Temperature 380F
REFERENCE OPERATING CONDITIONS:
Inlet Steam Flow 120,000 lbvhr
Inlet Steam Temperature 890 F
Inlet Sieam Pressure 320 psia
Inlet Feedwater Flow 2,600,000 lb/hr
Feedwater Inlet Temperature 370F
Inlet Drain Water Flow 150,000 tbr
Inlet Drain Water Temperature 460 F
1.1.5 Corrected Performance: The Indicator of
Degradation
For combined-cycle power plants, the expected performance varies greatly
over time. This makes it difficult to track changes in performance, as the
measured values of most performance parameters vary due to changes in
plant operating conditions. One methodology to make the identification of
performance changes over time easier is to “correct” the current
performance to a standard operating condition, usually the reference
operating conditions. To correct the performance means to account for the
performance variations that would be expected due to the changes in
environmental conditions and control set points. The corrected performance
is the performance that would be expected if the current (degraded) engine
were operating at the reference operating conditions. The virtue of corrected
performance is that its expected value zemains constant and equal to the
rated value. Thus, any change in a corrected value represents a change in
equipment performance capability
1.Concept of Performance Monitoring page 26Corrected power is a barometer of engine performance. It goes down when
degradation increases and it goes up when degradation decreases. In fact,
the degradation in performance from one point in time to another is equal to
the change in corrected performance over that time range.
a
||
ef —___ i
Figure 1-2 Measured and corrected gas turbine power over a nine-month time
period. An overhaul was performed on the gas turbine during October 2002; this
time period is evident on the plot as the time during which there is no measured
data
Corrected gas turbine power accounts for changes in engine operating
conditions and predicts the equipment performance if the equipment were to
operate at the reference operating conditions (including inlet filter delta-P,
ambient conditions, load level, water/steam injection, fuel heating value,
and exhaust delta-P). If changes in the engine operating conditions were to
‘cause changes in gas turbine power, the corrected power would not change.
Figure 1-2 is an actual trend of gas turbine measured and corrected power.
The measured power is shown on the plot only when the engine was
operating at or above 99% percent of baseload power. Notice that measured
page27 1. Concept of Performance Monitoringbaseload power varies during each day, and is higher in the winter months
than in the summer months.The trend display of measured power is a history
of operation, but gives the viewer little information about degradation,
Each corrected power point shown in Figure 1-2 is an average of calculated
corrected power over a time period of approximately two hours. Corrected
power is a prediction of the power that the engine would generate if
operating at reference operating conditions. Itis essentially the current,
rating of the engine, or it is a prediction of the power the engine would
achieve in a performance test at reference operating conditions.
The corrected power is a convenient plotting parameter because it shows
degradation in the engine. Notice that the engine corrected power started at
over 161 MW in July and degraded to approximately 158 MW by October, a
loss of approximately 3 MW over a three-month period. The engine
‘overhaul in October improved the corrected power back up to approximately
162 MW. In other words, this plot shows that the overhaul improved the
engine's power capability by 3 MW to 4 MW.
Figure 1-3 Shows corrected condenser pressure over an eight-month period
Figure 1-3 shows corrected condenser pressure at a combined-cycle power
plant in the United Kingdom. Notice the slow increase in corrected pressure
‘over 150 days, indicating fouling of the condenser tubes and/or blockage in
the waterboxes. Cooling water flow through the tubes also decreased about
4% during this time period (not shown on the figure). When the tubes and
1. Concept of Performance Monitoring page 28waterboxes were cleaned during a plant outage, the corrected condenser
pressure improved back to approximately the same level as the beginning of
the trend, and the cooling water flow rate also recovered (not shown on the
figure)
4.1.6 What is My Degradation?
Degradation is the reduction in equipment performance capability that has
occurred over time. It is a relative parameter; it compares equipment
capability at one point in time to that at another time. Since the corrected
gas turbine power is a prediction of the current rating of the engine, the
difference in corrected power from one point in time to another is the
degradation that has occurred over the time period. Thus, degradation may
be defined as the change in corrected performance over time.
For the value of degradation to be meaningful, the start time and end time of
the degradation must be stated. If no time range is stated, it is usually
assumed that the degradation is over the operational lifetime of the
equipment, which is from the time the equipment was put into service to the
present.
Often when historical data is not available, degradation may be stated as the
difference between the current equipment capability and its rated capability.
This is equal to the difference between the rated performance and the
corrected performance. Since degradation is defined as a change in
performance over time, this definition of degradation is only true if the
equipment actually achieved its rated performance at some point in time.
Rated performance is often set equal to the vendor guarantee as opposed to a
performance test at the beginning of equipment life. Thus, the equipment
‘may not have ever operated at its rated performance.
Degradation will be defined throughout this book as the difference between
corrected and rated performance. Ideally, the rated performance should be
defined as the actual performance at some given point in time, but if
sufficient plant data is not available it may be set equal to the vendor
guarantee.
The definition of degradation as a change over time instead of the change
from vendor guarantee is significant to the concept of performance
monitoring because a change over time is an aid in identifying changes in
equipment performance, while a change from guarantee may be misleading.
If the degradation is defined as the change from guarantee (a level of
performance that the equipment may never have actually operated at), some
plant equipment may show may show negative degradation, indicating that
the equipment is performing better than the guarantee level.
page29 1, Concept of Performance Monitoring1.1.7 How Much is Degradation Costing Me?
Knowing the amount of degradation is important, but it’s not the full story. '
In order to make decisions about which maintenance to perform, plant
operators need to know how much the degradation is costing plant
operation.
For example, a reduction in gas turbire performance (power and heat rate)
has an effect on overall combined-cycle plant performance, which can be
calculated using an overall plant model. The power reduction in the gas
turbine reduces plant power because both the gas turbine and the steam
turbine power levels will change. The steam turbine power changes because
the gas turbine exhaust flow and temperature normally change as a result of
the gas turbine degradation. The heat rate increase of the gas turbine will
cause the plant to consume more fuel per MW-hr of power produced. These
effects on plant power and heat rate can then be converted to operating costs
by applying a fuel cost to the extra fuel being burned, and/or a MW-hr cost
to the power which is not being sold because of the degradation.
Here isa situation where the definition of degradation as a change in
performance over time, as opposed toa change from vendor guarantee, is
particularly important. Once the plant is accepted and goes into commercial
operation it is too late to worry about equipment guarantees. The best that
the operators can be expected to do is to maintain plant performance at a
level that the plant actually operated in the past. Therefore, degradation is an
estimate of the performance improvernent that is possible, and any existing
degradation can be looked upon as the source of an operational cost that is
potentially avoidable.
Degradation normally is evaluated in different engineering units for each
‘equipment type: gas turbine degradation is in MW while condenser
degradation is in either psia or percent cleanliness. This makes it difficult to
compare degradations calculated for cifferent parts of the plant or for
different equipment types. One way to make a meaningful comparison is to
calculate the impacts of the degradation on overall plant power, heat rate
and operating cost. The definition of z plant impact is the change in plant
performance that would be realized if the degradation were to be returned to
zero by some maintenance action.
For example, a condenser may have a degradation of 0.1 psia (6.9 mbar).
This means that the condenser is operating at a pressure 0.1 psi (0.69 mbar)
higher than it would operate if the degradation were zero. This degradation
causes a reduction in steam turbine power, which is also a reduction in plant
power. The impact of the condenser degradation on plant power is equal to
the change in plant power caused by the degradation of the condenser.
1. Concept of Performance Monitoring page 30The reduction in plant power due to the condenser degradation increases
plant heat rate since fuel flow is not changed, Actually, fuel flow ina
Rankine cycle plant with condenser degradation may decrease slightly
because the increased condenser pressure will lead to higher feedwater
temperature entering the boiler, which will reduce boiler fuel consumption.
Even so, the plant power always decreases and heat rate always increases
when condenser pressure increases. The change in plant heat rate caused by
the degradation in the condenser is called the impact of condenser
degradation on plant heat rate.
‘These changes in plant performance reduce electric sales revenues and
increase fuel costs, resulting in a net operating cost to the plant. The change
in plant revenues minus fuel costs is called the impact of condenser
degradation on plant cos
‘The idea behind the overall plant impacis is to convert all of the
degradations in the plant to their respective costs on plant performance.
Then these degradations can be compared and evaluated on a consistent
(apples to apples) basis. Table 1-7 belov’ illustrates the concept for a
combined-cycle plant.
Table 1-8 Example of plant equipment degradations and their impacts on plant
performance
Impact on Plant Performance
Equipment | Degradation
Power | Heat Rate | Operating
(MW) | Btwkw-h) | Cost
(S/hr)
Inlet Air Filter | 1.1 in-tt,0 03 16 46
Gas Turbine 19MWw 22 15 294
HRSG 12,000 tbe 4 169
Steam Turbine | 0.8 MW 08 31 84
Condenser O41 psi oa 25 56
Cooling Tower 21F 02 B 2
Total Plant 31 191 on
page 31 1, Concept of Performance Monitoring‘The inlet air filter in Table 1-8 has a pressure-loss degradation equal to 1.1
in-H,0. If this degradation were eliminated by replacing the air filters, the
‘gas turbine inlet pressure would increase, resulting in a gas turbine power
increase, The steam turbine power would also increase because of the
increase in gas turbine exhaust energy. The total plant power would increase
by 0.3 MW, which is defined as the impact of the air filter on plant power.
This plant power increase would cause a plant heat rate decrease equal to 16
BtwkW-hr. Overall these changes in plant power and heat rate would yield a
net increase of 46 $/hr in plant operating profits (electric sales revenues
minus fuel costs). Methods to calculate these impacts are reviewed in the
chapter 6, “Impacts of Degradation on Overall Plant Performance”.
The total plant power degradation is equal to the sum of the equipment
impacts on plant power. In other words, the total of the equipment impacts
on plant power is equal to the degradetion in plant power, which is equal to
the rated plant power minus the corrected plant power, when the degradation
is calculated from rated. For example, if the plant were rated at 400 MW,
and the total power degradation from rated is 5.1 MW. Then, the plant
‘would be expected to now produce orly 394.9 MW if operated at the plant
reference operating conditions. This power (394.9 MW) is called the
corrected plant power.
Ina similar manner, the corrected plant heat rate is equal to the rated plant
heat rate plus the total of the equipment degradations in plant heat rate (191
Btu/kW-hr in Table 1-8). The current plant operating costs (electric sales
revenues minus fuel costs at the reference operating conditions) are S671/hr
higher than they would be if the plant was performing as rated and was
operating at the reference operating conditions.
1.1.8 Optimization: “Where You Could Be”
Once the degradation of the plant and its equipment is known, the plant
operator is prepared to answer the question, “What is the best way to operate
the plant so as to maximize plant profits?” The idea is to adjust the plant set-
points that are under the control of the operator to make as much money as
possible for the plant. The equipment degradation listed in Table 1-8
summarizes maintenance issues, but optimization is concemed with actions
the operator can take to improve performance without maintenance.
‘An example of calculated optimization outputs for a combined-cycle power
plant with two gas turbines is illustrated in Table 1-9.
1. Concept of Performance Monitoring page 32Table 1-9 Example optimization outputs for a combined cycle power plant
Controllable Current Value | Optimal Cost Savings
Set-point Value (Shr)
GTI Power 170 MW limw | 90
GT2 Power 150 MW 159 MW 88
Inlet Chiller #1 On oft 2
Inlet Chiller #2 on on u
Duct Burner # on on o
Duct Bumer #2 om on 0
Number of Cooling, 7 6 au
‘Tower Fans On
‘Total Savings Possible 22
The Current Value column shows current plant operating data, and the
Optimal Value column shows where the plant could operate if the operator
took the appropriate control actions. Finally the Cost Savings column
‘estimates the increase in plant operational profit that would be achieved if
the operator took the suggested actions. This screen is different from the
degradation screen in Table 1-8 in that no maintenance actions are required,
and the optimal operating conditions are achievable by operator action. No
one knows if the degradation in Table 1-8 is fully recoverable, but the
control actions suggested in Table 1-9 can be taken (assuming no
environment or other operational limit on plant operation is violated), and
the cost savings achieved.
4.1.9 Controllable Loss Displays
Controllable loss displays are an alternate way to present the degradation
and optimization data of tables 1-7 and 1-8. These displays are most often
used for Rankine cycle plants where the expected or target values of plant
performance parameters do not vary widely with plant operating conditions.
Controllable loss displays show the current value of selected plant
performance parameters, their target values, and the cost incurred by not
operating the plant at these target values.
pige33.— 1, Concept of Performance Monitoringanes opzosebn (ilemarowasumtossisomunyitinf omoes Loto vm son?
oeey oe sess on i226
we 4 4P AP ae Se Se Se
207] pe ET
om eE JE IE 4
anor wom | [Cz] [Cae sere] [Caene
‘wey ""(oronw
Figure 1-4 Example controllable loss display for a fossil (Rankine cycle) plant
The advantage of controllable loss displays is that they are readily
understandable summary of the plant performance status. If there is no
degradation in plant equipment, the controllable loss display will show
small losses and vice versa. The disadvantage is that they give little
information as to the location of plant performance problems. Controllable
loss displays are a very useful way to summarize plant status; they inform
the operator if there is a plant performance problem.
The target values for controllable loss displays are generally based upon
expected overall plant performance with no equipment degradation
anywhere in the plant. Due to the regenerative nature of a Rankine cycle,
degradation in one area of the plant will likely show up as deviations in
several controllable loss parameters calculated from measured data in other
areas of the plant. Thus, controllable loss parameters do not report
degradation specific to individual plant equipment, but instead report a
departure in overall plant performance from the values that the performance
parameters would have if the entire plant were “new and clean”,
For example, in order to achieve the target main steam temperature in a
boiler, the economizers, the air preheater, and the feedwater heaters must all
1.Concept of Performance Monitoring page 34operate with their target performance. Degradation in any of these may
cause the steam temperature to change. A change in the steam temperature
may change the throttle pressure, which might change the steam turbine
efficiency and the condenser pressure. Thus, many of the controllable loss
parameters are related to each other, and several will likely change when
one of them changes.
The target values used in controllable loss displays are a very different
concept from the equipment degradation calculations described above where
the expected performance of each equipment type depends upon the
operational conditions that the equipment is exposed to and is independent
of the degradation of other equipment in the plant.
1.2 ASME Test Codes
ASME Performance Test Codes provide test procedures that yield results of
the highest level of accuracy consistent with the best engineering knowledge
and practice currently available, The test procedures were developed by
balanced committees of professional individuals representing all concerned
interests. The test codes specify procedures, instrumentation, equipment
operating requirements, calculation methods, and uncertainty analysis,
When tests are run in accordance with an ASME code, the test results will
be of the highest quality and the lowest uncertainty available,
The focus of the ASME test codes is to provide test specifications
appropriate for verification of compliance with guarantee or warranty
performance. As such, the absolute eccuracy of measured performance is,
stressed as opposed to ease of testing. In general it is very difficult to
implement the AMSE test code procedures as the basis of performance
monitoring at an operating power plant.
The following table lists the test codes that are most closely related to power
plant performance monitoring.
page 35 1. Concept of Performance MonitoringTable 1-10 ASME Performance Test Codes closely related to pefformance
monitoring
ASME Test Code Description
PTC 1 - 1999 General Instructions
PTC 2- 1980 (R197)
Code on Definitions and Values
PIC 4.3 ~ 1968 (R191) Air Heaters
PTC 4.4 1981 (R2003) Gas Turbine Heat Recovery Steam
Generators
PIC6- 1996 Steam Turbines
PTC 6A - 2000 Appendix to PTC 6
PTC 6 Report 1985 (R1997)
Evaluation of Measurement Uncertainty in
Performance Tests of Steam Turbines
PTC 6S ~ 1988 (R195)
Procedures for Routine Performance Test of
‘Steam Turbines
PTC 82-1990
Centrifugal Pumps
PTC 11 ~ 1984 (R198)
Fans
PTC 12.1 -2000 Closed Feedwater Heaters
PIC 12.2- 1998 Steem Surface Condensers
PTC 123 - 1997 Deaerators
PTC 19.1 - 1998 ‘Test Uncertainty
PTC 22-1997 Performance Test Code on Gas Turbines
PTC 23 ~ 1986 (R197)
Atmospheric Water Cooling Equipment
PTC 46~ 1997
Overall Plant Performance
PTC PM~ 1993
Performance Monitoring Guidelines for
‘Steam Power Plants
1, Concept of Performance Monitoring
page 361.3. Performance Testing versus Online Monitoring
A performance test is a one-time evaluation of equipment performance that
relies on precision instrumentation installed specifically for that test. The
equipment being tested is operated at conditions as close to design and/or
guarantee as possible. The objective of a performance test is to measure the
absolute capability of the equipment. The tests are often done to verify
vendor guarantees on new or upgraded equipment.
The objective of performance monitoring is to detect changes in equipment
performance (degradation) so that proper corrective action can be taken. The
absolute value of performance is not necessarily important to performance
monitoring; instead, repeatability of results is most important, so that
changes over time can be evaluated.
The principal differences between testing and monitoring are summarized in
Table 1-11 below.
Table 1-11 Comparison of performance testing and online monitoring
Performance Test | Online Monitoring
Objective ‘Absolute Performance Detect Degradation
Insiumentaion Type | Precision Test insiruments | Whatever Is Available
Measurement Requirement | Accuracy Repeatability
‘Test Interval One Time Event Repeated Often.
‘Test Conditions Equipment kolated and at | Normal Plant Operation
Full Load
The basic difference between performance monitoring and performance
testing is that monitoring uses whatever instrumentation is continuously
available at the plant to give the operators an indication of plant
performance status. As such, monitoring data is usually not adequate for
vendor guarantee testing, but is usually acceptable for tracking changes in
equipment degradation. The fact that monitoring evaluations are repeated
many times gives the engineer the opportunity to reject results that are not
consistent with long-term trends.
page37 1. Concept of Performance MonitoringThe uncertainty of a measurement is considered to be the sum of two
components called the bias and the random uncertainties. Accuracy is,
achieved only if both the bias and random uncertainties are small. However,
repeatability is the long-term variation in bias error. Although the relative
contributions of random and bias errors are unknown for most instruments,
the ASME Performance Test Code Committee has estimated the
repeatability as one-half the overall instrument uncertainty.
The conclusion is that even though installed plant instrumentation may not
be adequate for precision tests, the repeatability of performance monitoring
results often approaches the accuracy of precision tests. This means that
degradation (change in performance) can be measured more accurately than
absolute performance.
1.4 Curve Based Methods
1.4.1. Performance Curves
Performance monitoring involves a comparison of the expected (new and
clean) equipment performance to its current (measured) performance. The
current performance is usually directly measured or is calculated from
measured data. The prediction of expected equipment performance requires,
both a measurement of equipment operating conditions and a method or
‘model to use to predict how the equipment performance changes as
operating conditions change.
Curve based methods are a simple and reliable method to predict equipment
performance changes as long as the operating conditions have not changed
too much from the reference conditions. The basic concept behind curve
based methods is to assemble a set of performance or correction curves that
plot the variation in a specific equipment performance parameter (such as
power, heat rate or efficiency) when one of the operating conditions
changes. The total equipment performance fractional change is then
computed by multiplying together the fractional changes for each operating
condition, where cach multiplying factor is generated using a separate
correetion curve.
Two equipment characteristics must be known in order to predict the
expected performance of any plant equipment:
1, Rating specification for the equipment that includes both the
rated performance and the reference operating conditions at
which the rating applies.
1.Concept of Performance Monitoring page 382. A method or model, which could be in the form of performance
curves, of equipment performance that can predict how the
performance changes when any of the reference operating
conditions change.
Table 1-12 is an example of the rating specifications for a heat recovery
steam generator (HRSG), and Figures 1-5 through 1-8 are example
performance curves for that same heat recovery steam generator. These
curves may come from vendor performance guarantee tables, a computer
model of the HRSG, or from measured data. Each curve shows how
equipment performance will change if only one of the equipment operating
conditions changes. When generating a performance curve it is assumed that
all other equipment-operating conditions remain constant and equal to theit
reference values, Thus, figure 1-5 shcws the variation of HP steam flow and
HRSG effectiveness as the gas turbine exhaust temperature varies, but only
if the other operating conditions (exhaust gas flow, exhaust gas composition,
drum pressures, inlet feedwater temperature, HP steam temperature, and
duct burner fuel flow) remain equal to their reference values as stated in
Table 1-12
page39 1, Concept of Performance MonitoringTable 1-12 Ratings specification for the example heat recovery steam generator
Heat Recovery Steam Generator | Data
Rating Specification
RATING:
HP Steam Flow 511,700 Tovar
LP Steam Flow 88,300 Ibvar
ifectiveness 934
REFERENCE OPERATING
CONDITIONS:
Exhaust Gas Flow. 3,200,000 Ib/ar
Exhaust Gas Temperature 1135 F
Exhaust Gas Composition 10% H.0
HP Drum Pressure 1900 psia
LP Drum Pressure 100 psia
Inlet Feedwater Temperature 1B6F
HP Steam Temperature 1000 F
Duct Bumer Fuel Flow 0.00
1, Concept of Performance Monitoring page 40RSG Perfomance vs Ehaust Gas Temperature
Figure 1-5 Example HRSG performance (HP steam flow and HRSG effectiveness)
versus changes in gas turbine exhaust gas temperature
page4l 4. Concept of Performance MonitoringHRSG Perormance vs Exhaust Gas Flow
Figure 1-6 Example HRSG performance (HP steam flow and HRSG effectiveness)
‘versus changes in gas turbine exhaust gas flow rate
1, Concept of Performance Monitoring page 42HRSG Pertormance ve HP Drum Pressure
0 nm Pree
Figure 1-7 Example HRSG performance (HP steam fiow and HRSG Effectiveness)
versus changes in high-pressure steam drum pressure
page 43/1. Concept of Performance MonitoringHRSG Performance ve Burner Ful Enersy
He amFlow ie
Figure 1-8 Example HRSG performance (HP steam flow and HRSG effectiveness)
versus changes in duct burner fuel energy input
1.4.2 Expected Performance from Curves
The basic assumption behind the curve-based performanee-prediction
methodology is that the individual operating conditions impact equipment
performance independently. When this assumption is true, the total impact
in performance can be computed by combining the impacts of the individual
parameters.
‘The methodology used to combine the individual impacts into a net impact
on performance is to convert all the individual impacts into a fractional or
percentage change in the performance parameter. The fractional change in
HP steam flow when the exhaust temperature changes fiom the reference
value, Try, to some value T;, is,
Fractional Change from T; to Trot:
re = £4 Tr)
lec col an
where
1. Concept of Performance Monitoring page 44Curver(T;) is the look-up value of the HP steam flow from the
HRSG exhaust temperature performance curve, Figure 1-5, at
‘temperature T;
Curver(Tr) is the steam flow at the reference exhaust temperature
from the same performance curve
The expected HP steam flow is the combination of all the fractional changes
from all of the parameters that affect the HP steam flow.
Expected HP Steam Flow from Performance Curves:
cy Lure dT) Curve) Curven(R) Currey» Won)
© Curve Tog) CHIE g Wen) CHV p (Pry CHIME pg py)
(1.2)
Ware
where
Wares is the expected value of the HP steam flow, at the exhaust,
temperature 7}, exhaust flow waxy, drum pressure P;, and duct
burner fuel flow wpg).
Wrra is the rated value of the HP steam flow, which occurs at the
reference exhaust temperature Tuy
Curver(T;) is the value read from the exhaust temperature
performance curve at temperature 7
Curver(Tre) is the value read from the exhaust temperature
performance curve at temperature Trey
Curvew(Wexhi) is the value from the exhaust flow rate performance
curve at temperature Weshs
Curvey(Wees rep is the value from the exhaust flow rate performance
curve at temperature Werk rey
Curvep(P.) is the value from the drum pressure performance curve at
temperature P,
Curvep(Prep is the value from the drum pressure performance curve
at temperature Prey
Curveps(Woarep is the value from the duct burner fuel flow
performance curve at temperatute Wes
Curveps(wpay) is the value from the duct burner fuel flow
performance curve at temperature wpa,
page 451. Concept of Performance MonitorAs an example, consider what happens to the HRSG performance when the
gas turbine exhaust conditions change, such as when the exhaust
temperature into the sample HRSG changes from its reference value of 1135
F to 1100 F, and the exhaust flow reduces from 3200 kib/hr to 2800 kib/hr.
‘The exhaust temperature performance curve (Figure 1-5) gives HP steam
flow values of 511.7 klb/hr at the reference temperature (1135 F), and 480.6
lb/hr at exhaust temperature equal to 1100 F. The exhaust flow
performance curve (Figure 1-6) gives HP steam flow values of 511.7 klb/hr
at the reference flow (3200 kibjhr), and 448.5 klb/hr at exhaust flow equal to
2800 kib/hr. Thus, the expected HP steam flow at the new exhaust
conditions is equal to:
weg ew, Lurvey (1100) Curve, (2800)
HPexy Priel’ Curve, (1135) Curve,,(3200)
480.6 448.5 _ yg hlb
S117 511.7 hr
Notice, only two terms out of four possible change factors are included in
the calculation because the drum pressure and duct bumer fuel flow did not
change, and their contributions to the calculation would equal unity (1.0).
(13)
=SIL7
The expected HRSG effectiveness (7Aey) at the new gas turbine exhaust
conditions would be calculated in the same manner, except that the
calculation must use curve look-up values for the effectiveness instead of
the steam flow.
Curve, (1100) Curve, (2800)
4 Curve, (1135) Curve, (3200)
93 (28/24) =92.8%
93.4 A934
1.4.3. Additive Performance Factors
Some operational parameters are not best represented by fractional changes
in performance, but instead by incremental changes in performance. An
addition of a quantity of energy to a system will likely cause the outputs of
the system to increase by an additive amount that is proportional to the
quantity of energy added, For example, a given amount of water or steam
injected into a gas turbine will increase the gas turbine power by an
increment that is proportional to the amount of steamy/water injected, but is
not closely related to the power level of the gas turbine without the
steanvwater injection. It would make little sense in this case to use a
multiplier on the reference gas turbine power. In this situation, it would be
Nesy
(a)
1. Coneept of Performance Monitoring page 46non-intuitive to express this impact as a multiplier on the reference gas
turbine power. Instead, this impact is better presented by adding an
increment of power proportional to the amount of steam/water that is
injected. This argument also applies to other situations such as adding duct
burner fuel energy added to an HRSG, steam, and to admission or extraction
from a steam turbine where the flow rate is not directly related to the throttle
flow.
Note also that if the impact of water injection on gas turbine power is
expressed using the water-to-fuel ratio as the independent parameter instead
of a specified water injection flow rate, then the effect on performance is
better represented as a multiplicative factor. This is because the water
injection has been normalized back to rated conditions by dividing the
quantity of water injection by the quantity of fuel flow. Additive correction
factors generally are only used to represent discrete quantities being added
to or taken from the equipment or system. Thus, gas turbine vendors have
the option of expressing the effect of water injection on gas turbine
performance as an additive factor (when the amount of water injection is
plotted versus gas turbine power) or asa multiplicative factor (when water
to fuel ratio is plotted versus gas turbine power).
HISG vendors have the same option cn duct bumer fuel energy. If the duet
bumer fuel energy were expressed as ¢ fraction of the input exhaust gas,
energy, the performance effect would be multiplicative.
Additive changes in performance are computed by adding the increment in
performance calculated from the performance curves to the reference
performance value, For example, the performance increment when the duct
bumer fires at level equal to some value, wr, instead of firing at the
reference duct burner firing level is:
Performance Increment = (Curveps(wr) — Curvepa(0)} (1.5)
Where Curvepa(wr) is the value from the HRSG performance versus duct
bummer firing curve (Figure 1-8) at the x-axis value equal to wp. The rated
HRSG performance occurs at a duct burner firing level equal to the
reference value, which equals zero. Thus, Curvepg(0) is the HRSG
performance at the reference duct burner firing level.
If the duct burner in the example HRSG fires at a level equal to 200
mmBTU/hr when all other operating conditions remain at their reference
values, the expected HP steam flow would-be:
page47 1. Concept of Performance MonitoringW(20D)em = = Wray + {Performance Increment}
Wray + {Curvepa(200) - Curvens(0)}
SUL.7 + 704 - 511.7
704 klb/he
1.4.4 Expected Performance from Curves
In summary, the expected equipment performance at actual operating
conditions can be calculated from a set of performance curves of the
equipment performance versus equipment operating conditions by the
following formula,
Expected Performance from Performance Curves:
CurveValudactualconditions)
snurptetacers Curve Valudreferenceconditions)
+L [Curvevatudactual)—CurveValudreferencd]
bsvePciore
erfOr MANE eg
Performaneg,
(1.6)
where
Performance. is the expected equipment performance at the actual
operating conditions, if the equipment performs with rated capability
Performanceraea is the expected or rated equipment performance at
the reference operating conditions (expected equals rated at the
reference operating conditions)
is a mathematical operator indicating that all the terms in the
following parenthesis are to be multiplied together, one term for
each performance curve until terms from all the performance curves
are included in the final product.
_isamathematical operator indicating that all the following
steer
terms are to be added together, one term for each additive
performance increment until terms from all the additive performance
increments are included in the firal sum,
CurveValue(i) is the value off the performance curve at operating
condition i
Note that the above formula can be used to predict the performance at any
set of operating conditions when the performance is known at any other set
1. Concept of Performance Monitoring page 48of operating conditions. Simply define the reference conditions to be equal
to the operating conditions where the performance is known, and the rated
performance to be equal to that known performance value.
Predicted Performance at (1) Given Test Performance at (2):
Curve Valuéconditions)
satiptcFucens Curve Valugconditons 2)
+ — Y[CurveValuéconditions|)—CurveValuéconditions2)]
(7)
Performane(1) = Performane(2)
where
Performance(1) is the predicted equipment performance at the
operating conditions (1) if the equipment performs with the same
capability as the known or test performance at conditions (2)
Performance(2) is the known or test equipment performance at the
operating conditions (2)
[] isa mathematical operator indicating that all the terms in the
following parenthesis are to be multiplied together, one term for
‘each performance curve until terms from all the performance curves
are included in the final product.
2. __ isamathematical operator indicating that all the following
terms are to be added together, one term for each additive
performance increment until terms from all the additive performance
increments are included in the final sum.
CurveValue(i is the value off the performance curve at operating
condition i
If the example HRSG is operated at inlet exhaust gas temperature of 1100 F,
and inlet gas flow rate of 2800 Klb/hr, and with the duct burner consuming
200 mmBtw/hr of fuel, the expected HP steam flow rate would be:
Curve, (1100) Curve,(2800)
(Curve -a1
reap = Wired Cryer (1135) Curve. 700) * WUrVe pg(200) ~ Curveng(0)}
=suif 82) AB) p04—si1.7 = 61282
S17 \ S117 hr
(1.8)
page49 1. Concept of Performance Monitoring1.4.5 Correction Factors
The traditional method to account for operational and environmental effects
on equipment performance is the correction factor method. This method was
developed by equipment vendors to enable their customers to predict the
performance of the vendor’s equipment at various operating conditions, and
to avoid the need to provide the physically based computer models of
equipment performance from which the curves are usually derived.
‘The methodology is to apply independent correction factors for each
operational and environmental effect. For example, gas turbine base-load
power is known to be dependent upon inlet air temperature, inlet air
pressure, inlet air humidity, inlet pressure loss, exhaust pressure loss, steam
injection rate, water injection rate and fuel type. The gas turbine vendor
would rate the engine power at a given set of these conditions, and provide
correction curves for each of these operational and environmental effects.
Each correction curve would quantify the change or percent change in
engine performance that would result when the given operational or
environmental condition changes.
The basic assumption of this curve-based methodology is that the individual
operating conditions have independent impacts on equipment performance.
This means that the total impact on performance can be computed by
combining the individual parameter impacts.
‘A correction curve is simply a normalized performance curve. The
equipment output parameter (Y axis on the performance curve) value is
divided by its rated value. This forces the Y-axis value to equal unity (1.0)
at the X-axis value equal to the reference value. The advantage of correction
curves is that the value read directly from the plot is equal to the correction
factor needed to predict performance at the reference conditions given
performance at some other operating condition. There is no need for the user
to divide by the rated performance value to obtain a correction factor. The
disadvantage is that the absolute value of performance is not available from
the curve.
1. Concept of Performance Monitoring page 50Figure 1-9 Correction factor curve for the effect of exhaust gas temperature on HP
steam flow, this curve is equal to the curve in Figure 1-5 divided by the rated HP
steam flow
Correction factors are defined as the fractional change in performance from
rated when an operational condition changes from the reference conditions.
They are often used in performance testing to predict the equipment
performance at the reference conditions when the performance was
‘measured at conditions other than the reference conditions. This predicted
performance at reference conditions is called the corrected performance.
Expected Performance at Test Condtions from Correction Factors:
Performance. = Performance. [| CorrectionFactors
Aion
(19)
+) AdditiveCorrections
akadave
where
is a mathematical operator indicating the product of
ier
all the following terms (each term multiplied by the
next term)
pageS1 i, Concept of Performance Monitoringisa mathematical operator indicating the sum of all
suave
the following terms (each term added to the next
term)
Performanceryea is the expected or rated performance at the
reference operating conditions
Performanceep is the expected performance at test operating
conditions if the equipment performs with rated
capability
CorrectionFactors are the values from the correction curves
at the test operating conditions
AdditiveCorrections are the values off the additive correction
curves at the test operating conditions
The correction factor curves, just like performance curves, can be used to
predict equipment performance at any operating condition given the
performance at one other operating condition. The formula for the predicted
equipment performance at operating condition (1), given known
performance at operating condition (2) is below.
Predicted Performance at Operating Conditions (1):
Performan) = Performan [1 omectionFaciorstt) - [| CorrectionFactors
+ Performance(2)+ Y” AdditiveCorrections(\)~ ¥° AdditveCorrections(2)
(1.10)
where
Performance (1) is the predicted performance at operating conditions
(1), ifthe equipment performs with the same capability as the known
or test performance at operating conditions (2)
Performance (2) is the known or test performance at operating,
conditions (2)
Performanceraie is the tated performance at the reference operating
conditions
CorrectionFacotrs(1) are the values off the correction curves at
‘operating conditions (1)
CorrectionFacotrs(2) are the values off the correction curves at
operating conditions (2)
1. Concept of Performance Monitoring page 52AdditiveFacotrs(1) are the values off the additive correction curves
at operating condition (1)
AdditiveFacotrs(2) are the values off the additive correction curves
at operating condition (2)
Because the correction factor curves are based upon a rated performance
value at reference operating conditions, the prediction of performance at
some operating conditions (1) requires the knowledge of both the rated
performance as well as the performance at operating conditions (2).
1.4.6 Percent Change Correction Factors
Sometimes the variations in equipment performance with operating
conditions are presented as a percentage change in performance versus the
percentage change in the operating condition. These curves are fully
normalized performance curves where the y-axis is equal to the change in
equipment performance (equipment performance minus the rated
performance) divided by the rated performance, and the x-axis is equal to
the change in reference condition (current operating condition minus the
reference operating condition) divided by the reference condition. An
example of such a performance curve is shown in Figure 1-10.
ageS3- 1. Concept of Performance Monitoring‘% Change in Steam Flow Versus % Change in GT Exheust Temperatre
Figure 1-10 An HRSG percent change correction curve for the HP steam flow
versus exhaust gas temperature
‘The use of these percentage change correction curves is essentially the same
as that for correction curves, except that the correction factor must be
calculated from the curve look-up value in the following manner,
ValueOnPercentageCorrectionCurve
100
CorrectionFactor
a.)
1.5 Model-Based Performance Analysis
The correction curves that equipment vendors supply to customers are based
upon physically based computer models of the equipment performance, and
are a convenient way for vendors to transmit the results of complex
computer analysis to their customers. However, in this time of powerful
‘computers on every desk, it is not necessary to simplify the analysis into a
few curves. Why not use the computer codes directly to calculate expected
and corrected performance?
Computer software programs like GateCycle™, Pepse™ and GTMaster™,
contain complex, physically based models of equipment performance that
1. Concept of Performance Monitoring page 54can be used in place of correction curves. In fact, equipment vendors often
use these computer codes to create the correction curves.
Some advantages of using the computer codes (model-based analysis)
instead of performance or correction curves are listed below:
* Interaction of varying operating conditions can be modeled,
* Physically based models can allow wide variations (far from
reference) in operating conditions,
* Physically based models can compute impacts of parameters for
which no curves are available.
* Physically based models give detailed information about the
expected performance, not available from curves. This additional
information may help the engineer diagnose problems,
The individual equipment operating conditions may not have independent
effects on equipment performance, wiich is an assumption of the curve-
based method. In other words, the assumption that the overall effect of
changes in all the operating conditions can be computed by multiplying the
correction factors together may is not valid over a wide range of operating
conditions. As long as each correction factor is near unity, the method words
very well; but when correction factors get far from 1.0, their product may
not represent the true performance change in the equipment.
Computer models can handle wide variations in environmental parameters
and operational modes for which curves do not exist or do not accurately
‘model. In particular, as conditions change over a broad range, the
interactions between environmental parameters become more and more
important, and computer codes are often built specifically to handle these
interactions.
Computer models can compute corrections for parameters that the vendor
may not have supplied correction curves for. For example, if the gas turbine
uses varying amounts of water injection or switches from natural gas to oil
fuel, the exhaust gas compositions will change. These changes are handled
directly by computer models, but seldom accounted for in correction curves.
age $5 1. Concept of Performance Monitoring‘The methodology for model-based performance analysis is,
1.
Build a computer model of the equipment being monitored. The
procedure to build such a model is specific to the software used.
Test the model versus vendor guarantee data and/or plant-measured
data over a wide range of operating conditions. Correct the model
where necessary.
‘At each performance monitoring calculation interval input the
‘measured equipment operating conditions into the model
Run the model and obtain the expected equipment performance as a
model output.
. Evaluate degradation by comparing the expected performance from
the model to the measured performance.
The following figures illustrate the use of physically based computer code
analysis for the example heat recovery steam generator used in the chapter
on performance curves.
The computer model in Figure 1-11 was constructed to replicate the actual
steam/water flow path in an existing HRSG. The actual surface areas of the
tube bundles were obtained from vendor information and input into the
computer model. Then the design-point heat transfer coefficient in each tube
bank was adjusted so that the model prediction matches the rating
specification for the HRSG. This resulted in a design point model of the
HRSG.
1. Concept of Performance Monitoring page 56meen a
|
sonal Y : ms | re
Figure 1-11 Computer model (screen from GateCycle™ computer code) of
example HRSG at reference operating conditions
Next, the predictions (off-design mode in GateCycle™) of the HRSG model
‘were compared to vendor warrantee data over a range of operating
conditions to verify model accuracy. If necessary, corrections were made to
the design-point model, and the verification process repeated until the
predictions of the HRSG model matched the vendor data to within one
percent over the entire operating range of the HRSG. The resulting model
can then be used to give predictions of the expected performance of the
HRSG as operating conditions change.
Figure 1-11 shows the output of the model when the reference operating
conditions from Table 1-12 are input to the computer model. Notice that the
model predicts the rated steam flows to three digits of accuracy or better. In
addition to predicting the rated steam flows, the model also computes the
complete temperature distributions within the HRSG. Figure 1-12 is a plot
of these temperature distributions.
pageS7 1, Concept of Performance Monitoring1100
gages
Temperature, F
400
Heat Transfer, BTU/hr
Figure 1-12 Temperature profile from GateCycle™ for the example HRSG at
reference conditions
‘The upper straight line in Figure 1-12 is the exhaust gas temperature as the
gas goes from the HRSG inlet to the stack. The lower set of straight lines is,
the corresponding steam/water temperature distribution. There is one
steam/water straight line for each tube bundle modeled in the HRSG. The
‘computer code only predicts the inlet and outlet conditions for each tube
bundle, and does not predict temperature distributions within a tube bundle.
‘A straight line is drawn from one predicted point to another.
Notice on the left hand side of the plot (at zero on the x-axis), the gas enters
at a temperature of 1135 F, where the corresponding steam temperature is
1000 F. On the right hand side of the plot the gas exit (stack) temperature is
206 F, while the inlet feedwater temperature is 136 F. The closer the gas exit
temperature is to the feedwater inlet temperature the higher the effectiveness
of the HRSG.
1. Concept of Performance Ma
page $8r+
Sts
Figure 1-13 Model-based prediction (from GateCyclo™) of the HRSG performance
at exhaust gas temperature equal to 1100 F, and exhaust gas flow equal to 2800
Kibmhr
Figure 1-13 shows the predicted HRSG performance when the exhaust gas
inlet temperature and flow rate are changed to 1100 F and 2800 mmBtu/hr
respectively. Notice that when using model-based analysis, all operating
conditions are input to the model and the output accounts for changes in all
the operating conditions at once. Interactions between the inputs can be
predicted only if all changes in operating conditions are input to the model.
Figure 1-13 shows the model-based prediction of HP steam flow and
effectiveness. How do these compare to the curve-based method? Table 1-
13 below compares the model-based results to the curve-based results for
the situation where the exhaust gas temperature and flow change from 1135
F and 3200 kib/hr to 1100 F and 2800 klb/hr respectively.
Table 1-13 Comparison of results between curve-based and model-based methods
for a change in HRSG inlet conditions
Predicted Performance
after exhaust gas flow & Curve-Based Model Based Method
temperature change Method
HP Steam Flow (Klb/hr) 420 421
HRSG Effectiveness (%) 928 92.9
page S¥ 1: Concept of Performance MonitoringThus, the curve-based and the model-based methods yield approximately
equal predicted performance values when exhaust gas temperature and flow
change over a relatively narrow range. Changes in exhaust conditions of this,
size could be expected to occur as a result of ambient temperature changes
on the order of 40 F.
asa crecroeness
ross SESPEEATERS——
i
oral
Figure 1-14 Predicted HRSG performance (from GateCycle™) when exhaust gas
temperature, exhaust gas flow, and duct burner fuel flow all change from reference
Now let’s add a significant change in duct bumer firing level, from zero at
the reference conditions to the maximum possible for this HRSG (200
mmBtwhr), and once again compare the predictions of the curve-based
method to the model based method.
Table 1-14 Comparison of results between curve-based and model-based methods
for high duct-burner firing situation
Predicted Performance
after exhaust gas flow & | Curve-Based Method | Model Based Method
temperature & duct
firing level all change
HP Steam Flow (kltvhr) on 614
HRSG Effectiveness (%) 945 95.1
1. Concept of Performance Monitoring page 60Notice that differences between the curve-based method and the model-
based method begin to become important, at least for effectiveness, as the
changes in operating conditions get larger. Since, the curves for this
example were calculated from the model, all of the differences in calculated
results are due to the simplifying assumptions inherent in the curve-based
method. In other words, the curve-based method is based upon the model,
and is a simplification to the model that makes it possible to predict
performance without needing to run the computer code.
page61 1. Concept of Performance Monitoring1. Concept of Performance Manito rage 622. Heat Balance Analysis
2.1 Overview
The term heat balance analysis in the context of performance monitoring
describes the application of mass and energy balance equations to model
power plant systems with the objective of determining detailed
thermodynamic properties of the operating system. The heat balance
analysis process takes measured data as input and outputs a complete set of
thermodynamic data for each flow stream in the model, including both the
measured data and data that were not measured. Heat balance analysis, in
this context, is not a prediction of plant or equipment performance: itis
instead a process of matching @ mass and energy balance model of the
system as closely as possible to measured data.
The objective of heat balance analysis for performance monitoring is to
obtain a complete set of actual operating data for the system. This operating
data will be used in subsequent equipment evaluations to determine the
degradation in the plant equipment. Therefore, the heat balance analysis,
should not make assumptions about the performance capability of the
equipment being modeled. Mass and energy balances are true no matter
what the performance capability of the system, and therefore do not require
any assumptions about performance capability. Thus, mass and energy
balance calculations are appropriate for heat balance analyses when applied
to performance monitoring, but assumptions about the efficiency of a
turbine or the effectiveness of a heat exchanger are not.
One often hears the term heat balance code to refer to software programs
used to predict power plant performance. Heat balance codes typically
perform mass and energy balance calculations; in addition, they often
incorporate models of the physical characteristics of the equipment and
apply them to predict the equipment performance over a range of operating
conditions.
A commercial heat balance code may or may not be useful for the type of
heat balance analyses used in performance monitoring that is described in
this chapter. In order for a commercial heat balance code to be appropriate,
it must allow the user to input measured plant data as opposed to parameters
of equipment performance. When matching a heat balance code to measured
data itis generally correct and necessary to input values such as
temperatures, flow rates, pressures and power levels. If the heat balance
code requires the user to input values for isentropic efficiency or heat
page 63 2. Heat Balance Analysis,exchanger effectiveness, it may be making implicit assumptions about
equipment degradation that negate the purpose of the heat balance analysis.
2.2 Local Heat Balances
The application of mass and energy balance calculations (also called simply
“heat balances” in this book) are a method to infer data that is not or cannot
be measured, and can also be used to improve the accuracy of existing
power plant measured data. Heat balances are an essential component of any
on-line performance-monitoring system, because plants rarely have
sufficient instrumentation to adequately assess equipment performance. It is
often necessary to infer data at locations where measurements are not
normally made. Suich locations include the extraction steam flow rates to
feedwater heaters, the power generated in each steam turbine stage group,
the exhaust flow from a gas turbine, the cooling water flow in the
condenser, and the boiler flue gas properties. Heat balance analysis can
often provide accurate estimates of data values at these locations.
Some plant locations have more instrumentation than needed to determine a
data value (redundant measured data). Heat balance analysis can verify and
possibly improve the accuracy of redundant measured data,
Performance monitoring is a process where operating information
(measured data) from the plant is processed so as to make judgments or
Teach conclusions about the current capabilities of the plant. Mass and
energy balances are a way to add information about the plant. These balance
equations should be based on as few assumptions as possible about the
performance capability of the plant, but instead establish relationships
between the measurements that are always true. For example, heat balance
equations should not contain assumptions about the heat transfer
coefficients in heat exchangers or the efficiency of rotating equipment.
Of course, some assumptions about the operational capability of the power
plant must always be made. A mass balance equation for a pipe will
generally assume that all of the water entering a pipe flows out the other end
of the pipe. This is equivalent to assuming that there is no leak between the
inlet and the outlet of the pipe, and that the pipe is operating in steady state.
The basic mass and energy balance equations for a system are stated below:
Rate of Storage of Mass = Mass Inflow Rate - Mass Outflow Rate (2.1)
Rate of Storage of Energy = Energy Inflow Rate - Energy Outflow Rate(2.2)
Power plant performance monitoring is almost always done at steady-state
conditions, where the system is not changing with time. At steady state, the
2, Heat Balance Analysis page 64storage terms in the above equations are equal to zero and the balance
equations reduce to setting the inflow rates equal to the outflow rates.
Stack
Ems |
ee
z nas
PH2
| Power pump
cro] [Hea] [rw] [PH
Figure 2-1 Control volume for performing heat balance analysis
To perform a power plant heat balance analysis, an engineer first selects the
area of the power plant around which the mass and energy balances are to be
performed. The selected area is called the control volume. The balance
equations include only the mass and energy flows into and out of the control
volume: that is, flows which cross the control volume boundary. Flows that
stay inside of or outside of the control volume are not part of the analysis.
This makes the selection of the contrcl volume very important. Ifa flow
value is not measured or is measured inaccurately, it may be possible to
select a control volume that completely surrounds or excludes the unknown
flow and thereby eliminates the flow from the analysis. Any flow or energy
transfer that crosses the boundary of the control volume is included in the
balance equations; no other flows appear in the balance equations.
In Figure 2-1 a sample control volume, which looks something like a
HRSG, is shown with arrows indicating the flows of mass and energy into
page 65 2. Heat Balance Analysisand out of the control volume. The labels on the arrows indicate the type of
flow at the location of the arrow:
2. Heat Balance Analysis page 66Table 2-1 Types of flow associated with she arrows on sample control volume in
Figure 2-1
Label ‘Type of Flow
INSTACK Exhaust Gas In
Exhaust Gas Out
PHI Preheat Water In
PH2 Preheat Water Out
FW Feedwater
HP HP Steam,
1p IP Steam
LP LP Steam
RH Cold Reheat Steam
HRH Hot Reheat Steam
Powetnnp | Power to Pump(s)
Eras Heat Energy Lost
‘The mass balance equation for the sample HRSG control volume is as
follows:
Mass Inflow Rate= Mass Outflow Rate
wiv + Worn + Wew + Weai= Wsrack + Warn + Wei + Wp + wip + WLP
(2.3)
Where w is the mass flow rate at the location specified in the subscript.
Notice that the power (Powerjump) and heat transfer energy terms (Eig) do
not appear in the mass balance because they do not involve the movement of
mass across the boundary of the control volume. The energy balance
equation for the HRSG control volume is given below.
Energy Inflow Rate = Energy Outflow Rate
win*hin + worn'*hcan + Wew*hew + wens *hens + Powerpump
=wsrack*hsrice Ware * hei wens *hpret wre hue + wip hipt wip hip + Etoss
(2.4)
Where h refers to the enthalpy of the fluid at the location specified in the
subscript. The term for power input to the system (Powerump) only appears
in the energy balance equations for an HRSG if a pump or some other
power-consuming device is located within the boundary of the control
volume and the electric current to power the pump crosses the boundary of
page 67 2. Heat Balance Analysthe control volume. To choose whether or not to include a pump within the
HRSG control volume, consider where the feedwater temperature is
‘measured. Ifthe feedwater temperature is measured after the pump, the
control volume can exclude the pump so that the temperature into the
control volume is the temperature downstream of the pump. If the feedwater
temperature is only measured before the pump, it is better to establish the
control volume so that it includes the pump so that the feedwater
temperature into the control volume is the temperature upstream of the
pump.
‘The energy transferred out (Eion) is due to heat transfer to the environment.
If the energy lost to the environment were due to a leak, an additional mass
flow term would be required to account for the leakage flow. The remaining
terms in the energy balance equation are the same as the terms in the mass
balance equation, except that each flow rate is multiplied by an enthalpy (/),
which is the energy content per unit mass.
‘The mass balance equation (Equation 2.1) includes both gas and
steam/water flows. It may be convenient to develop separate relationships
for the gas and steam/water flows. To obtain a relationship including only
the gas flows, set up a control volume which includes only the gas (shell)
side of the HRSG. The boundary of this control volume would be along the
outside of the pressurized tubes and drums in which the steam/water flows
and along the inside of the metal walls that contain the gas flow within the
HRSG. For this gas-side control volume the mass balance is:
Mass Inflow Rate =Mass Outflow Rate
Wow =Wsrack (2.5)
The energy balance for the gas-side control volume is:
Energy Inflow Rate=Energy Outflow Rate
win*hiv=wsrack*hstac + Eloss + Qurse (2.6)
Where Qnasc is the heat transfer rate from the gas-side to the steam/water
side of the HRSG, also called the duty of the HRSG.
These equations illustrate that changing the location of the control volume
will change the balance equations, and that can yield additional information
from the analyses. In this case, the additional information is that the gas
flow into the HRSG must equal the gas flow out of the HRSG, and that the
difference in energy of the gas from the inlet to the stack is equal to the duty
of the HRSG plus heat transferred to the environment.
There are two ways to generate the equations for the steam/water mass and
energy balances, One is to mathematically subtract the gas-side balance
2, Heat Balance Analysis page 68equations from the overall HRSG balance equations. The second is to use a
control volume consisting of the steam/water flow paths through the HRSG.
The mass balance equation for this steanvwater control volume is:
Mass Inflow Rate = Mass Outflow Rate
Wer + Wew + wrx! = Wurutwenst wet wit wip 27
The energy balance equation for the steam/water control volume is,
Energy Inflow Rate= Energy Outflow Rate
Wern*hcrn + Wew*hew + wey; "hens + Powerpunp + Oursc
= wart*huen + Wex2*hpx + wup*hup + wre*hip + wip *hip (2.8)
‘These heat balances around an HRSG illustrate how an engineer can use
heat balances to add additional information about the system. For example,
if all the steam/water flows and temperatures are measured and the gas inlet
and outlet temperatures are also measured, the HRSG heat balance
equations can be used to calculate both the duty of the HRSG and the gas
turbine exhaust gas flow rate. To perform these caleulations, first calculate
the duty using the steam/water and pump power measurements:
Qiesc = Powerpump + Wrn*hrinn - Wern*hcan + Wen*hpr -
wer*hen* wue*hup + wir*h + wip*hip- wew*hew (2.9)
‘Then, use the gas-side energy balance (Equation 2.6) to get the exhaust gas.
flow rate:
win = (Eas + Quase) / Ohov- hisrace) (2.10)
‘The energy loss to the environment, Eioss is difficult to measure, and
therefore is frequently estimated (assumed) to equal approximately 2% of
the HRSG duty (heat transfer rate). In this case the gas turbine exhaust gas
flow rate is:
Wor = 1.02 * Quase/ (haa hsrack) 4)
Local heat balance analysis, when applied as in this example for an HRSG,
adds useful information for performance analysis. The input/output methods
presented in several of the ASME Performance Test Codes are developed
from local mass and energy balances.
One of the primary difficulties with the application of local heat balance
analysis to performance monitoring is the number of measurements required
and the resulting accuracy of those measurements when obtained from
available plant measured data. The calculation of the duty (Qyrsc) of an
page 69 2. Heat Balance AnalysisHRSG from a complete energy balance requires the measurement of every
inlet and outlet steam/water flow, temperature and pressure. Unfortunately,
steam flow measurements at power planis are often accurate to only 2% to
5% (or worse). This means that the calculated HRSG duty will only be
accurate to 2% to 5%. Ifan engineer wants to track performance so as to be
able to detect 1% degradation, he has to avoid dependency on steam flow
measurements.
2.3, Combined-Cycle Overall Plant Heat Balance
The application of overall plant heat-balance analysis has the potential to
improve the precision of performance monitoring results by allowing the
performance engineer to use the most accurate instruments in the plant and
avoid less accurate measurements. If the degradation of plant equipment can
be computed from temperature, pressure and power measurements instead
of steam flow measurements, the uncertainty of the results will be reduced.
Overall heat balance analysis cannot completely eliminate the need to use
inaccurate steam flow measurements during the performance evaluation, but
it can greatly reduce the impact of these measurements on the overall
determination of equipment degradation.
‘An overall plant heat balance does not mean simply setting up a control
volume that encompasses all of the plant. Instead, overall plant heat balance
analysis involves using control volumes at many locations within the power
plant system and formulating mass and energy balance equations for each of
those control volumes. It is not necessary to generate the balance equations
for an overall plant control volume because that is equivalent to the
mathematical sum of all the balance equations of all the control volumes
used throughout the plant system, and its use therefore adds no additional
information.
Overall plant heat balances are considered a separately because the large
number of equations arising from the required control volumes comprise a
set of simultaneous algebraic equations. The solution of this set of equations
requires a level of mathematical sophistication beyond the scope of one-line
expressions used for calculated tags in a control system or a plant data
historian, Instead, the solution of an overall heat balance analysis usually
requires the sophistication of a mathematical equation solver or a
commercial heat-balance computer code.
Figure 2-2 illustrates the concept of using multiple control volumes to
model a combined-cycle plant system. The plant has been simplified for the
purposes of this discussion, but enough detail is included to illustrate the
concepts of overall plant heat balance analysis.
page 70Looking at the outermost control volume (the set of dotted lines that
surround the plant) and applying conservation of energy, one can see that:
Condenser Duty=Fuel Energy -GT Power Stack Energy -ST Power (2.12)
Thus, the condenser duty can be computed from the fuel energy, plant
power and the stack energy loss. However, the calculation of the stack
energy loss requires knowledge of the stack temperature and the gas turbine
exhaust gas flow rate. The gas turbine exhaust flow could be computed from
the HRSG energy balance described above (Equations 2.9 and 2.11), but it
can also be computed without the use of measured steamv/water flows by
applying a gas turbine energy balance:
Fuel Energy = Powergr + GT Exhaust Gas Energy
= Powercr + Wes * heh
and solving for the exhaust gas flow rate,
Wen = (Fuel Energy - Powercr) / hea (2.13)
This expression enables one to compute the exhaust gas flow rate from the
measured gas turbine fuel flow, gas turbine power and gas turbine exhaust
temperature if the specific heat or enthalpy of the exhaust gas is known as a
function of temperature. The beauty of these relationships is that the
condenser duty and exhaust gas flow rate (two quantities that cannot be
directly measured) can be calculated from measurements that are normally
acceptably accurate in a combined-cycle power plant; namely fuel flow,
power, and gas temperatures,
page 71 2. Heat Balance AnalysisStack
Power
Figure 2-2 Control volumes for Combined-Cycle Overall Plant Heat Balance
Analysis
Those familiar with gas turbine heat balances may be saying, “This method
looks easy enough, but I know that a gas turbine heat balance is
complicated. You need to guess the air flow rate in order to get the exhaust
gas properties.”
‘The gas turbine heat balance is an iterative process that is outlined in Table
2-2: Several methods to solve the gas turbine heat balance are detailed in
Chapter 7 “Gas Turbine Performance”.
2. Heat Balance Analysis page 72Table 2-2 Overall Gas Turbine Heat Balance Analysis Procedure
Gas Turbine Heat Balance Iteration To Get Exhaust Flow
1. | Guess the inlet airflow.
2. | Compute the air/fuel ratio, or use a chemical mass
balance to get the exhaust gas composition.
3. | Get the exhaust gas specific heat or enthalpy from
table look-up or gas properties routine.
4. | Compute the exhaust gas flow rate from Equation
212,
5. _ | Compare the exhaust gas flow minus the fuel flow
to the guessed inlet air flow.
If the two air flows (from steps 1. and 5.) are not
equal go back to step 1
Such an iterative solution process is probably too complex to implement as a
calculated tag in a control system or daia historian, and can even be very
challenging to implement in a spreadsheet environment. Note, too, that this
is just considering the gas turbine heat balance. Overall plant heat balances
involve many more equations and unknowns, and these unknowns must be
determined by some iterative process. The rewards of overall plant heat
balance are available only to those willing and able to make the leap to the
iterative mathematical analyses provided by the application of heat balance
computer programs and/or equation solvers.
Once the exhaust gas flow is known, the condenser duty can be calculated
from the overall plant energy balance, Equation 2.12. The cooling water
flow that runs through the tube-side of the condenser can then be computed
from an energy balance on the tube-side of the condenser.
Condenser Cooling Water Flow Rate:
CondenserDusy “8
CoolingWaterFlowRate =—__CondenserPuty_
[SpecificHeat * TemperatureRiselcosingwaur
The HRSG duty can be calculated from the gas-side energy change from the
HRSG inlet to the stack (Equation 2.6). The energy change of the exhaust
page 73 2. Heat Balance Analysis,gas is accurately indicated by the temperature measurements at the gas
turbine discharge and at the HRSG stack.
Equation 2.8 is the steam/water energy balance for the HRSG. This equation
includes all of the steam and water flows into and out of the HRSG, but the
high pressure steam flow is the most important of these flows. From an
energy perspective, the HP steam energy represents 70% to 90% of the total
HRSG heat transfer. The reheat steam flow mostly derives from this same
HP steam flow, so that the determination of reheat steam flow is improved if
the HP steam flow is known more accurately.
Assuming that HP steam is at least three-fourths of the total steam energy,
the accurate measurement of this flow is three times more important to an
accurate energy balance on the HRSG than are all the other steam flows
combined. In other words, a 1% error in the measurement of this HP steam
flow will cause the same size HRSG energy balance error as a 3% or 4%
‘measurement error in either the IP or LP steam flows. Therefore, a modeling
technique that improves the precision of the HP steam flow measurement is,
valuable, even if the IP and LP steam flow rates still must come from
measurements.
‘The HRSG steam/water side energy balance, Equation 2.8, can be solved for
the HP steam flow, resulting in:
wap = {Wean*hcan + Wew*hew + Went *hpar + Powerpump + Queso -
Waat*hunu - Wora*hpna- Wie*hip - Wiehe} | hup
(2.15)
Al the terms on the right hand side of this equation must be measured
except the HRSG duty (Quasa), which comes from the overall plant energy
balance plus the HRSG gas-side energy balance. The beauty of the method
is that the heat balances used to get the HRSG duty do not involve measured
flows (except plant fuel flow). If measured power levels, temperatures and
pressures are accurate to 2%, the HRSG duty will be accurate to 2%. Since
the energy contribution of the other steam flows (IP and LP) is small
relative to Quinsc, even a 4% measurement error in these other steam flows
will not add much to the uncertainty in the HP steam flow, wie, and the
uncertainty of the HP steam flow will still be on the order of 2%.
‘The HP steam flow that is calculated from the heat balance is called the heat
balance value of HP steam flow. This heat balance value can be compared
to the measured value of HP steam flow. This comparison can serve either
as a confirmation of the heat balance analysis or, more likely, as a
diagnostic on the accuracy of the measured HP steam flow.
2. Heat Balance Analysis page 74Now that the steam flow into the steam turbine is known, we can get the
steam turbine discharge steam enthalpy by using heat balances on the
control volume surrounding the steam turbine. We can then estimate the
overall steam turbine efficiency by using the enthalpy-drop method
In summary, overall plant heat balances for combined-cycle power plants
‘can add data thiat were not measured and improve the accuracy of selected
measured data. The additional data added by the heat balance analysis are
essential for the expected equipment performance evaluations that follow
the heat balance analysis.
age 75 2, Heat Balance Analysis,Table 2-3 Summary of Overall Plant Heat Balance Inputs and Outputs for
‘Combined-Cycle Power Plant
Overall Plant Heat Balance Analysis for Combined-Cycle Plants
Required Measurements
Calculated from Heat Balance
Ambient Conditions: Temperature, Pressure
and Humidity
Condenser Duty
Steam Turbine Power
Condenser Cooling Water Flow
Fuel Input Energy
Gas Turbine Exhaust Gas Flow
Fuel Chemical Composition and Heating Value
Gas Turbine Exhaust Enthalpy
Gas Turbine Exhaust Temperature
Firing Temperature of Gas Turbine
Compressor discharge temperature and
pressure
First Stage Nozzle Flow Area of GT
Steam/Water Injection Rate to Gas Turbine | Compressor Efficiency
Gas Turbine Power Expander Efficiency
HIRSG Stack Temperature HRSG Duty
IP Steam Flow HP Steam Flow fiom HRSG
LP Steam Flow Steam Turbine Discharge Enthalpy
Steam/Water Extraction Flows from HRSG (to
process or deaerator heating)
‘Steam Turbine Isentropic Efficiency
Preheater Water Flow
Preheater Outlet Water Temperature
All Steanv Water Temperatures & Pressures
Condenser Cooling Water Inlet Temperature
Condenser Cooling Water Outlet Temperature
2. Heat Balance Analysis
page 762.4 Combined-Cycle Heat Balance Using Commercial
Software
GateCycle™ is one of several commercial “heat balance” codes that can
perform a heat balance analysis of a combined-cycle power plant. Figure 2-3
shows a screen from a GateCycle™ model of a combined-cycle power plant
consisting of two industrial gas turbines with HRSG’s and a single steam
turbine.
Figure 2-3 Model of Combined-Cycle Power Plant in GateCycle™ Software
GateCycle™ and most other heat balance codes do not allow the user to
input the power generation directly, but instead require inputs such as the
steam turbine efficiency or exit enthalpies, which are then used to calculate
the power. Since power is an important measured parameter that the heat
balance analysis must match, when using GateCycle™ for heat balance
analyses the user must set-up what that program calls a macro (a user-
defined control calculation) that iterates upon a selected plant variable, such
as LP steam turbine efficiency, until the measured power is matched.
The HRSG models built using heat balance codes are typically more
detailed than the measured data can support. That is, a complete heat
page 77 2. Heat Balance Analysisbalance analysis around the HRSG would require measured steam/water
temperatures at the outlet of each economizer and superheater icon in order
to have a complete set of inputs for the analysis without the need to make
assumptions about the performance of any of the icons (tube banks in the
HRSG). If some the economizer or superheater outlet steam/water
temperatures are not measured, then the heat balance analysis must make
assumptions such as the effectiveness of the superheater or the outlet
subcooling of the economizer.
‘The incorporation of assumptions such as these into the configuration of the
heat balance analysis will mean that any degradation that has actually
occurred in the icons with assumed performance will, according to the
results of the heat balance analysis, be shifted to other equipment in the
plant. For example, if the HP economizer tubes in the operating plant were
to foul and reduce the effectiveness of the economizer, but the heat balance
analysis was configured assuming no fouling in the economizer; then in the
results from the heat balance analysis, the degradation would shift from the
economizer to the evaporator. The calculated overall (or sectional) HRSG
degradation would still be correct; and, as long as the performance
monitoring system reports overall HRSG degradation and not the
degradation for each icon in the HRSG, there will be no error in the reports
from the performance monitoring system.
In GateCycle™, the HRSG model can be configured so that steam generation
in the HP evaporator is an input to the calculation. However, there is only
cone HP steam generation that would be consistent with the measured values
of IP steam flow, LP steam flow and the stack temperature. An iteration can
be used to vary the input steam generation in the HP evaporator until the
measured stack temperature is achieved (the IP and LP steam flows,
temperatures and pressures are fixed inputs). A GateCycle™ macro is used
to accomplish this iteration.
The third and most complex iteration required in the application of a
GateCycle™ heat balance analysis for a combined-cycle power plant is the
‘gas turbine heat balance. In the gas turbine heat balance, the measured fuel
flow, compressor discharge conditions, exhaust temperature, and power are
input to an iteration to determine the compressor efficiency, expander
efficiency, firing temperature, exhaust gas flow rate, first stage nozzle flow
area, and exhaust enthalpy. This detailed gas turbine heat balance analysis
uses three GateCycle™ icons and is preferred to the overall gas turbine heat
balance described as described in Table 2-2, which uses a single control
volume for mass and energy balances. The detailed gas turbine heat balance
yields more detailed output information than does the single control volume
gas turbine heat balance analysis.
2, Heat Balance Analysis, page 78A GateCycle™ detailed gas turbine heat analysis model (for a typical
industrial or frame gas turbine configuration) is illustrated in Figure 2-4,
This model includes a macro to iterate on air inlet flow rate until the user-
input power is matched by the results from the heat balance model,
case: ore
POWER: 168.08
Gas Turbine Heat Balance Analysis
Figure 2-4 Detailed Gas Turbine Heat Ealance Analysis Model in GateCycle™
Since the gas turbine heat balance is an iterative calculation and the overall
plant heat balance s also iterative, itis efficient and safer (meaning there is a
higher probability of convergence) to solve the two problems in separate
GateCycle™ models. To do this, first run the gas turbine heat balance and
then transfer the results as input data into the gas turbine (type GTDATA}
icon in the overall plant heat balance model. The overall plant heat balance
can then be executed without fear of the gas turbine iteration causing
convergence problems in the overall plant heat balance.
page 79 2. Heat Balance AnalysisTable 2-4 Macros Required in GateCycle™ Heat Balance Analysis of Combined-
Cycle Plant
Macros Required
in GateCycle™ Heat Balance Analysis for Combined-Cycle Plant
Measured Value to Match Iteration Implemented in a Macro
Overall Plant Power Heme on LP Turbine Efficiency
Stack Temperature Ienite on HP Steam Generation
Gas Turbine Power Iterate on Gas Turbine Airflow
First Stage Inlet Flow Area (optional) | Iterate on Gas Turbine Fuel Flow
2.5 Rankine-Cycle Overall Plant Heat Balance
The application of heat balance analyses to Rankine cycle power plants is
often more mathematically complex than that for combined-cycle power
plants because the flow streams in a Rankine plant are more interconnected
than in a combined-cycle plant. In addition, there is generally little
‘measured data available inside the boiler of a Rankine cycle plant. This
forces the heat balance model to make assumptions about the performance
of some of the equipment in the boiler.
One useful and attainable objective of Rankine cycle heat balance analysis,
is to generate a complete set of thermoéynamic data about the boiler such
that the boiler efficiency from the ASME loss method is equal to the boiler
efficieney from the input/output method when heat balance data is used in
the calculations. The result is a set of operating boiler data that is more
complete than the measured data and that is consistent with the available
‘measurements.
{A second objective is to improve the accuracy of the boiler fuel flow
‘measurement. The heat balance analysis requires measurement of the
feedwater flow and/or the main steam flow; the fuel energy input to the
system is then determined by the heat balance. The result is a fuel flow
estimate that is at approximately the same accuracy as the feedwater flow
measurement.
A third objective of the Rankine cycle heat balance analysis is to generate
input and output flow information about each piece of equipment in the
plant that can be used to evaluate the degradation for each piece of
2, Heat Balance Analysis page 80equipment. The boiler heat balance data will enable the performance of the
walls, superheaters, and economizer in the boiler to be estimated. The
additional information about the steam cycle (such as the flow rates and
enthalpies of all flows leaving the steam turbine) will enable the degradation
in the steam turbine and feedwater heaters to be evaluated.
The heat balance analysis does not determine the degradation: the heat
balance analysis provides a more complete set of current plant operating
data so that degradation can be determined in subsequent calculations.
Degradation is the difference between current or actual performance and the
expected or rated performance. The heat balance analysis provides
additional information about the current performance of plant equipment,
and it provides input data for the expected performance prediction. The
difference between the current and expected performance determines
degradation.
Ash
Figure 2-5 Control Volumes for Heat Balence Analysis of Rankine Cycle Plant
Figure 2-5 illustrates the control volumes used for heat balance analysis of
Rankine cycle power plants. The diagram is simplified in that it does not
show all of the flow streams, and the number of feedwater heaters has been
reduced to two closed heaters and one open heater. In a complete power
page 81 2. Heat Balance Analysisplant analysis, a separate control volume would be used for each feedwater
heater in the power plant, regardless of the number of feedwater heaters.
In this example, the main steam superheating has been modeled as two tube
banks (SH and SH2). The primary superheater (SH) is in the radiative
section of the boiler and the secondary superheater (SH2) is in the
convective section of the boiler. A similar modeling convention has been
used for the reheat tube banks (RH and RH2).
In actual practice, it may be difficult to determine performance for each of
these tube banks because of limited measured data. If the steam
temperatures at the exit of each tube bank are measured, the heat balance
can determine the effectiveness or cleanliness of each tube bank. If fewer
measurements exist, the heat balance analysis must assume the performance
of tube banks where the exit steam temperature is not known. This makes it
impossible to determine degradation in those tube banks. Any degradation
that actually occurs in tube banks where performarice is assumed to be as
rated will appear instead as degradation in other (downstream) tube banks.
Boiler heat balance analysis is discussed in more detail in Chapter 10,
“Boiler Performance”.
The objective of a heat balance analysis is to determine the thermodynamic
properties of the mass and energy flows into and out of the control volumes
‘used to model the power plant. A set of measured data is input to the heat
balance analysis and a complete set of heat balance data for the system is
output. Table 2-5 summarizes the inputs and outputs for a Rankine cycle
power plant heat balance.
Often the number and location of the measurements available for use are
such that the heat balance is mathematically over-specified in some areas of
the plant. This means that there is more measured data than needed by the
heat balance analysis. An example is that if the feedwater flow into the
boiler (minus the desuperheating spray flows) is measured along with the
main steam flow. If boiler blowdown is zero, then the heat balance
equations would state that the feedwater flow must equal the main steam
flow. If the measured values of the feedwater flow and the steam flow are
not equal to each other, then there is no solution to the heat balance
equations that matches both measurements. In this ease, the performance
engineer is forced to use only one of these measurements as an input to the
heat balance analysis or, alternatively, to use both measurements to improve
the accuracy of the calculated results through the heat balance optimization
procedure described in Chapter 10.
2, Heat Balance Analysis page 82‘Table 2-5 Inputs and Outputs for Heat Balance Analysis of a Rankine Cycle Power
Plant
Overall Plant Heat Balance Analysis for Rankine-Cycle Plants
Required Measurements
Calculated from Heat Balance
Ambient Air Temperature, Pressure &
Humidity
Boiler Efficiency
Steam Turbine Power
Boiler Fuel and Air Flows
Plant Auxiliary Power
Reheat Steam Flow
Main Steam Flow to Steam Turbine
‘Steam Turbine Extraction and Discharge
Flows
Main Steam Desuperheat Spray Flow
‘Steam Turbine Extraction and Discharge
Enthalpies
Reheat Steam Desuperheat Spray Flow
LP Steam Turbine Inlet Flow
Main Steam Temperature and Pressure
HP Turbine Power and Efficiency
Hot Reheat Steam Temperature and Pressure
IP turbine Power and Efficiency
Cold Reheat Steam Temperature and Pressure
LP turbine Power and Efficiency
‘Steam Turbine Extraction and Discharge
Pressures and Temperatures
Feedwater Heater Effectiveness
IP Steam Turbine Exit Temperature
Main Steam Superheat Effectiveness
Feedwater Heater Outlet Feedwater Water
‘Temperatures
Reheat Steam Superheat Effectiveness
Feedwater Heater Outlet Drain Water
‘Temperatures
Water Wall Effectiveness
Economizer Outlet Water Temperature
Economizer Effectiveness
Fuel Composition
Air Heater Effectiveness
Fuel Heating Value
‘Air Heater Leakage
Unburned Carbon in Ash
‘Condenser Cleanliness
page 83
2. Heat Balance Analysis,‘Oxygen and/or CO; at Air Heater Gas Inlet
Condenser Duty
‘Oxygen and CO, at Air Heater Gas Outlet
Condenser Cooling Water Flow
Gas Temperature at Air Heater Inlet
Flue Gas Flow and Composition
Gas Temperature at Air Heater Outlet
Air Temperature at Air Heater Outlet
Power of Fans in Boiler
Condenser Pressure
Condenser Cooling Water Inlet Temperature
Condenser Cooling Water Outlet Temperature
2. Heat Balance Analysis age 84Table 2-6 Optional Measurements and Modeling Assumptions for a Rankine Cycle
Power Plant Heat Balance
Overall Plant Heat Balance Analysis for
Rankine-Cyele Plants
Optional Measurements | Modeling
Assumption
Required If
Parameter Is Not
Measured
Economizer Exit Water Correlation for Economizer
Temperature Heat Transfer Coefficient
Secondary Main Steam Correlation for Superheater
Superheater Exit Temperature | Heat Transfer Coefficient
Secondary Reheat Steam Correlation for Superheater
‘Superheater Exit Temperature | Heat Transfer Coefficient
Steam Turbine Extraction Use Overall Steam Turbine
Temperatures Efficiency to Get
Extraction Temperatures
Leakage Steam Flow from HP | Seal Leakage Model
to IP Turbine
Radiation Loss from Boiler to | Assume a Percent Loss or
Environment use Table Look-up Versus
Lead
Pump Inlet and Outlet ‘Assume a Pump Efficiency
Temperatures & Pressures
Water Wall Exit Steam Quality | Assume a Wall Exit Steam
or Boiler Recirculation Ratio | Quality
page 8 Heat Balance Analysis2.6 Rankine-Cycle Heat Balance Using Commercial
Software
TR
Figure 2-6 Model of Rankine-cycle power plant in GateCycle™ software
‘Commercial heat balance codes such as GateCycle™, PEPSE™,
VirtualPlant™, and SteamPro™ can perform heat balance analysis of
Rankine cycle power plants at the level of detail needed for performance
monitoring applications. Figure 2-6 shows a screen from GateCycle™
displaying a sample heat balance mode! of a Rankine cycle power plant.
This GateCycle™ model replicates closely the control volume schematic
illustrated on Figure 2-5.
It is important to configure the models built using commercial heat balance
codes in a manner such that the inputs and outputs (see Table 2-5) are
appropriate for heat balance analysis as used here. Heat balance analysis as
described in this chapter is not a prediction of plant performance; it is
instead a mass and energy balance of the power plant that matches the
measured data.
2. Heat Balance Analysis, page 86The use of commercial heat balance codes to perform heat balance analysis
normally involves an iterative procedure where selected inputs to the heat
balance code are changed until a desired measured result is achieved. For
example, the steam turbine power is normally measured, but itis usually not
an input to a commercial heat balance code. Thus, the user must iterate on
(guess values for) an input to the code (such as LP steam turbine efficiency)
until the desired result (measured value for steam turbine power) is output
by the heat balance code. Most commercial heat balance codes enable the
user to set up such iterations so that they are automatically executed as part
of the plant model. GateCycle™ calls these macros, and PEPSE™ calls
these controls: in both cases, user-specified iterations can be configured as
part of the plant model such that the iterations run automatically when the
heat balance code is executed..
page $7 2. Heat Balance Analysis,2. Heat Balance Analysis page §83. Data Validation
3.1 Definition of Data Validation
Data validation is the process by which the quality of raw input data is
evaluated and improved if possible. Data validation, when applied to power
plant performance monitoring, has the following objectives
© Identify data that is not or cannot be correct
‘© Reject data that is in error
* Provide replacement values for data that is determined to be in error
Improve the accuracy of input data
‘* Determine the accuracy of the resulting input data
Data validation can be viewed as a decision-making process where tests are
applied to the raw input data, and individual data values are either accepted
or rejected. If the raw input data is rejected, replacement values may be
supplied. In either case, the quality of the resulting input data may be
improved if possible. Finally, the ideal data validation system will estimate
the accuracy of the resulting input data.
3.2 Range Checking
3.2.1 Static Ranges
The most often-used data validation technique is range checking. Range
checking is simply testing the raw data value against minimum and
maximum values and rejecting the data value if itis lower than the
minimum or higher than the maximum. If the data value is rejected, it may
be replaced with a default value or removed from the set of data that is input
to the analysis.
The minimum and maximum values (ranges) used in range-checking
algorithms have a direct impact on the ability of the rangé-checking
algorithm to identify incorrect data. The first step in establishing the
minimum and maximum values is to determine the objective of the range
checking. If the objective is simply to detect and reject data from failed
sensors, a minimum value lower than ary reasonable measured value and a
‘maximum value higher than any reasonable measured value can be used.
This is because failed sensors usually report data values that are far from the
expected measured result. That is, the failed sensor data usually goes to zero
page 89 3. Duta Validationor negative, or it goes to a very high value, beyond any physically
reasonable value of the quantity being measured.
3.2.2 Dynamic Ranges
If the objective of a data validation process is to detect sensor drift or partial
failure, then the ranges (minimum and maximum values) should be derived
from a model which produces a prediction of the expected value of the
sensor. In this case, the ranges result from a prediction, and therefore
usually change over time. Ranges that change over time are called dynatnic,
ranges.
The optimal technology to predict the expected data value from a sensor is a
computer-based model of the overall power plant expected performance,
implemented in a commercial heat-balance computer code. The
methodology to establish dynamic ranges using an overall plant predictive
model is to first acquire measured values for plant operating conditions such
as ambient conditions, load, fuel type, and process steam requirements.
Then, predict the overall plant expected performance using the plant model.
This model should also predict the expected values for all the remaining
data sensors in the plant.
‘An interval of acceptable values around the expected sensor value must be
established for each sensor. This interval may be expressed as either a
percentage of the expected value or as an absolute change in value (delta)
from the expected value. Then the minimum and maximum values for the
ranges used in range checking will be at the extremes of the interval of
acceptable values. Ifa percentage is used, the minimum value will equal the
expected value minus the percentage of the expected value, and the
maximum value will equal the expected value plus the percentage of the
expected value. If an absolute change in value is used, then the minimum,
value is the expected minus the change and the maximum value is the
‘expected value plus the change.
‘The interval of acceptable values must be chosen carefully. If the interval is
too small the algorithm will reject data representative of the degradation that
the monitoring system is trying to detect. If the interval is too large the
algorithm will fail to detect sensor drift.
A simpler alternative to using an overall plant model is to establish
relationships among sensors that are local to one another. Many different
techniques have been used to establish these relationships. The most
frequently used techniques are based on the application of the laws of
3. Data Validation page 90thermodynamics and the development cf correlations between related data
values.
Examples of the relationships among sensor values derived ftom application
of the laws of thermodynamics are as follows. The cooling water outlet
temperature from a heat exchanger should be lower that the inlet hot fluid
temperature and higher than the inlet cooling water temperature. The water
temperature out of an air-cooled cooling tower must be greater than the
ambient wet-bulb temperature and less than the hot cooling water from the
condenser. The gas temperatures down the path of an HRSG should
decrease after each heat exchanger and increase after a fired duct burner.
The measured steam temperatures should be equal to or above the saturation
temperature. The steam temperature in a steam drum or condenser should
equal the saturation temperature, unless there are non-condensable gases
(air) in the system,
Often the relationship between sensor values is too complex to model or is
not known. In this case a relationship can be developed by correlating the
expected value of a sensor with the histcric values of related sensors. The
engineer who develops such a correlation needs to identify which sensors a
given sensor is related to and dependent upon, and then acquire the
appropriate amount of historic data for all the related sensors. Various
methodologies, including neural networks, may be used to create the
correlations. The resulting relationship (correlation) can be used as a
predictive model of the expected data value from the sensor.
3.2.3 Rejected Values
When a value is rejected by the range check, some action must be taken.
Some actions that may be taken are:
* Do nothing, but record a warning about data accuracy
© Replace the data with a static default value
* Replace the data with the expected value of the sensor calculated
from a model
* Halt execution of the monitoring system calculation, wait for the
next time step
‘The action to be taken may depend upon the importance of the measured
data value to the performance-monitoring calculation, Some input data
values are vital to the calculated results, but others have little impact. For
example, the inlet air temperature to a gas turbine and the gas turbine power
are critical parameters, and there is little use in running the performance-
page 91 3. Data Validationmonitoring calculations if these data values are incorrect. However, the inlet
air humidity to the compressor has almost no effect on the performance-
‘monitoring calculations, and so a static default value can be used to replace
an incorrect inlet air humidity data value.
Generally, for a robust implementation of performance-monitoring
calculations, it is recommended that all rejected data values be replaced with
reasonable default values appropriate for the sensor so that mathematical
problems in the ensuing calculations can be minimized. Some data values,
even if they are not normally important to the calculated results, can lead to
mathematical or programmatic problems such as divide by zero, failed table
look-ups, negative flow rates, or negative heat transfer rates.
Since a monitoring system has no plant control or equipment protection role
during normal operation of the power plant, it is usually of litle
consequence if the calculations are not executed at a particular time step.
Thus, the procedure of choosing the action to not run the calculations, when
data vital to the calculated results fails a range check, can be an effective
way to avoid errors in the calculations. It is usually better to not report a
degradation value at a given time step than to report a value that is incorrect.
Similar logic can also be used to stop the monitoring calculations during
plant operational modes where the monitoring calculations are not
appropriate, such as during start-up or at low plant loads.
3.3 Averaging Sensor Data
Often there is more than one sensor available to measure a single quantity
such as the stack temperature, the gas turbine exhaust temperature, or the
compressor inlet temperature, but when only one data value is needed or
used in the performance calculations to represent the measurement. In this
situation, the use of multiple sensors can improve both the reliability and the
accuracy of the measured data.
‘There are two ways that multiple sensors can be used to improve data
quality. The first is to allow a measurement to survive the failure of a single
sensor without losing validity. The second reason is to improve the
accuracy, Multiple readings at various locations are often more accurate
than a single reading, especially if the average of a flow stream is desired.
‘The exhaust temperature of a gas turbine varies across the exhaust duct, and
temperature measurements are taken at multiple points, and it is the average
exhaust temperature that is most appropriate for use in calculating the
overall energy balance. Multiple exhaust gas temperature sensors provide
enough data to calculate a meaningful average temperature.
3. Data Validation page 92There are many schemes used to identify and account for failed sensors in a
multiple-sensor situation, and then compute an average of the accepted
sensor values. The following algorithm has been used successfully in the
EfficiencyMap™ performance monitoring system, and is called the “Smart
Average”.
TM
EfficiencyMap™ Smart Average Procedure:
Acquire data values from all-sensors that measure the same plant
characteristic
2. Range check each of the data values against minimum and maximum.
values
Discard any data values that fail the range check
Ifall data values fail the range check, provide a default data value
Calculate the median (midway point) of the remaining data values
awh we
Discard any data values for which the absolute value of the data
value minus the median is greater than the required precision, where
the required precision is an input parameter for each measurement
7. Compute the average (mean) of the remaining values
This procedure will use all data values which pass the test criteria. This
procedure rejects obviously incorrect values via the minimum and
maximum values used for range checking, and also rejects data values that
are inside the range-check limits but disagree with the other data values by
‘more than the value of the required precision.
The required-precision value should be determined for each set of sensors
based upon the estimated accuracy of the sensors and the expected variation
of the quantity being measured across the field of measurement. The
required precision should be small enough so that sensors that are badly out
of calibration are rejected, but large enough so that calibrated sensors that
are measuring different locations in the flow field are not rejected. A
reasonable estimate for the required precision is to double the expected
variation (sensor accuracy plus variations due to location) in sensor readings
when multiple sensors are in the flow field.
As an example, the average exhaust temperature from a gas turbine is a
number typically accurate to within a few degrees Fahrenheit; however, the
required precision used to average these exhaust temperatures should be at
least 50 F (28 C) because the actual gas temperatures normally vary by as
much as 15 F to 25 F around the annulus of the exhaust duct.
page 93 3. Data Validation,In another example, the appropriate required precision for the inlet air
temperature to a gas turbine should be only 3 to 5 degrees Fahrenheit
because the inlet air temperature normally does not vary much across the
inlet flow area, and the measured temperature is typically accurate to within
approximately one degree Fahrenheit (0.55 C).
3.4 Time Averaging
Averaging both measured (input) data and calculated (output) data over time
can be a very effective method to improve both the accuracy and the
reliability of a performance-monitoring system. An online performance-
monitoring system has the advantage that many data sets can be evaluated
over long periods of time. Reviewing trends of measured and calculated data
versus time helps engineers to evaluate when and if performance changes
actually occurred.
The standard deviation of a measured data value due to the random
fluctuations can be reduced by taking multiple measured data values
(readings) over a period of time and then averaging the data values. The
standard deviation of the time-averaged data value is related to the standard
deviation of a single measured data value by the following formula.
Standard Deviation of a Measured Value
Se
3.1
Un GA)
where
‘Syis the standard deviation of ¢ single measured data value
Sz is the standard deviation of the average (mean) of a set of data
values taken over time and then averaged
Nis the number of data values used to determine the time-averaged
value.
Time averaging of data values from a data acquisition system can
dramatically reduce the uncertainty in a measurement due to random error.
However, time-averaging does not reduce the measurement error due to the
systematic uncertainty (bias) in the messurement. The systematic error is
constant over the time interval of the time-averaged data values, and is not
affected by the averaging.
Performance-monitoring calculations are based upon steady-state
predictions of plant and equipment performance. Time-averaging the
3. Data Validation page %4measured data can smooth out transients and improve the accuracy of the
calculations. ASME test procedures typically call for time-averaged data as
input to the calculations.
‘Some measurements vary in such a manner that they must be averaged in
order to get a value that is representative of steady-state conditions. An
example is the feedwater flow rate into an HRSG, which is controlled by
water-level sensors in the steam drums of the HRSG. The water levels in
these drums change slowly relative to other parameters in the system,
causing the feedwater flows to lag the corresponding steam flows by several
minutes.
The feedwater flows into some HRSG’s have been observed to behave ina
cyclic manner with the length of a cycle on the order of ten to twenty
minutes. The amplitude of the cycle is ten to twenty percent of the value of
the average feedwater flow rate. Thus, even when a combined-cycle power
plant is operating at base load, a measured feedwater flow value at one point
in time can vary by ten percent from the sum of the steam flows at that same
point in time. In such a case, it may be necessary to average the feedwater
flow measurements over a time interval as long as twenty minutes.
‘The author has found that time averaging of measured data is typically not
used in most on-line performance monitoring implementations even though
it is generally agreed that it can improve accuracy. One reason for not using
time averaging in an on-line performance monitoring system is the response
time of the monitoring system. Some power plant operators ask for less than
one minute update rates on the performance calculations. Such a
requirement makes it impossible to do much time averaging.
A second reason is that time-averaging increases the complexity of an
already complex system, and many engineers do not think that the small
improvement in results is worth the effort, especially when it is difficult to
quantify the improvement in results.
Note also that the time averaging of the outputs from a performance
monitoring system can always be used as an aid to interpreting the results,
and may alleviate much of the need for time-averaging of the input data.
Figure 3-1 displays calculated gas turbine corrected power values calculated
from data taken from an operating gas turbine over a six-month period.
These corrected power values were generated by an on-line performance
monitoring system (EfficiencyMap™) that calculates the gas turbine
corrected power at approximately five-minute intervals, No time-averaging
was used in the calculation of gas turbine corrected power; however, the
data was filtered such that the plot shows corrected power values only when
page 95 3. Data Validationthe gas turbine was operating at baseload, In this plot, the darker black data
Points on the plot are the corrected power values when the engine was
‘operating at baseload, and the lighter grey data points on the plot were
calculated using EfficiencyMap™'s smart-averaging process over the last
24 corrected power values, It is obvious from the plot that the time-averaged
corrected power values have less scatter than do the corrected power values
that were not time-averaged.
Figure 3-1 Gas Turbine Corrected Power calculated from data taken only when the
engine was at baseload. Darker points are the corrected power values; lighter
Points are the EfficiencyMap™ “smart average" of the last 24 corrected power
values.
3.5 Heat Balances for Data Validation
Often there is redundancy in the measured plant data and a performance
monitoring system can use that redundancy to improve the accuracy of the
measured data values. Redundancy exists when more measurements are
available than are necessary to determine the quantity being measured, One
example of a redundancy is when several sensors are used to measure the
same physical quantity, such as multiple thermocouples used to measure the
3. Data Validationinlet air temperature. In this situation, the sensor-averaging techniques
described above can be used to improve the measurement accuracy.
A less obvious example of redundancy is when mass and energy balance
relationships can be applied to measured data in order to calculate a
parameter that is measured. For example, the measured HP steam flows
from each of the HRSG’s in a combined-cycle power plant are redundant
with the measured throttle steam flow into the steam turbine if all the HP
steam is going to the steam turbine. That is because the sum of the measured
HP steam flows should equal the measured steam turbine throttle flow. If all
these values are measured, the redundancy can be used to improve the
accuracy of the measurements, and to estimate the uncertainty in the
‘measurements.
Consider the case where there are two measured HP steam flows from two
HRSG's, and a measured steam turbine throttle flow, as illustrated on figure
3-2.
HRSG #1 HP Steam
106 KPPH, 1000 F, 1000 psia
‘Throttle Steam
220 KPPH, 1020 F, 1000 psia
99 KPPH, 1040 F, 1000 psia,
TIRSG #2 HP Steam
Figure 3-2 Example of Redundant HP Steam Flow Measurements
In this example, it is easy to see that all of the measured data cannot be
correct. According to the flow measurements, the total measured inflow
from the HRSG is 205 KPPH (thousands of pounds per hour) of steam while
the measured outflow to the steam turbine is 220 KPPH. The measurements
violate the principle of conservation of mass which states that the inflows
page 97 3. Data Validationmust sum to equal the outflow. The measurements also violate conservation
of energy since the total energy into the system is greater than the outflow of
energy. Some of the measured data values must be incorrect.
Conservation of Mass for Figure 3-2:
Waser + Warscr = Wrote (3.2)
Conservation of Energy for Figure 3-2:
Waascrunser * Wuesoa aesc2 = arate Mirae (33)
where
w is the steam mass flow rate
nis the enthalpy of the steam
‘One way to calculate a set of validated data that satisfies the conservation of
mass and energy is to “believe” the HRSG measurements, and use mass and
energy balances to calculate the throttle flow conditions. In fact, in this,
example, there are three different sets of measured data that could be
‘generated by believing the data at two of the measurement locations, and
calculating the data at the third location. Any one of these sets of data might
be correct if the only test for correctness is to satisfy mass and energy
balances. It would then difficult to know which of these possible sets of
validated measurements to use in the monitoring system. Some way to test
and rank them would be desirable.
‘The first three rows of results on Table 3-1 show the mass flow rates
calculated when two of the three measurement sets are believed and the
third is calculated from mass and energy balances. The problem with all
three of these sets of validated data is that in each case only two of three
possible measurements were used. and the third measurement was ignored.
Ifall three sensors are of equal quality it might be better to find a set of
validated data that uses all three of the sets of measured data values as
inputs, using them to generate a set of validated data values that are some
sort of weighted average of the sensor values while still satisfying the mass
and energy balances.
3. Data Validation page 98Table 3-1 Possible Sets of Validated Measurements that Satisfy Conservation of,
Mass and Energy (uncertainties set equal to 1.0 for all sensors)
Validation Assumption | Valideted | Validated | Validated | Least-
HRSG #1 | HRSG #2 | Throttle | Squares
Flow | Flow | Flow | Error
1. Believe HRSG #1 and 1060 99.0 | 205.0 | 225,
HRSG #2 Measurements
2. Believe HRSG #1 and 1060 | 1140 | 2200 | 227.
Throttle Measurements
3. Believe HRSG #2 and 121.0 990 | 2200 | 238.
Throttle Measurements
4, Believe All 1075 | 1075 | 2150 | 995
‘Temperatures, Equal
Uncertainties on Flow
Sensors
5. Optimal with Equal 110 | 1040 | 215.0 | 75.2
Uncertainties
‘One way to accomplish this “weighted averaging” of the sensor values is to
find the set of validated data that satisfies mass and energy balances and is
as close as possible to all the measuremer
nts using the concept of “least-
squares error”. The set of validated data values that minimizes the difference
between the measured data values and th
ie validated data values, based on
the calculated least-squares error, is the preferred solution. This criterion of
“least-squares error” is calculated using the following expression.
x,
LSE= >
where
vated — X mensrd
S sersor
2
)
LSE is the least-squares error for the set of measurements and
validated data
Xvpeanees 18 the measured data value for a given sensor
X saves i8 the validated data value for a given sensor that satisfies
conservation of mass and energy
page 99
3. Data Validation.yuu iS the estimated uncertainty for measured data from a given
sensor
& isa mathematical symbol indicating that the following terms
are to be summed for all the sensors (measurements) being validated,
This formulation of the least-squares errcr uses the sensor uncertainty to
ht the importance of each sensor in finding the validated results. Ifa
particular sensor is known to be more accurate than the other sensors, a
small value of the uncertainty for that sensor will cause the validated results
to be relatively closer to the measured value for that sensor than to the
values measured by other sensors which have larger uncertainties.
Another factor to take into consideration when determining the optimal set
of validated data from a performance-monitoring system is that certain
classes of sensors are known to be more accurate than other sensors. In this
case the data validation procedure can be constrained to “believe” these
sensors and then adjust the remaining sensors so as to satisfy mass and
energy balances. For example, the tempe-ature measurements are typically
more accurate than are the flow measurements. The data validation
procedure can be forced to “believe” a more accurate measurement by
adding the constraint expressed in Equation 3.5 to the set of equations to be
solved.
For measurements to be believed by the least-squares validation:
AG x,
waned = X mused
G.5)
To force the least-squares validation procedure to believe all three of the
temperature measurements (Assumption 4 in Table 3-1 above). three
equations of the form of Equation 3.4 were added to the mass and energy
balance equations. These three equations are:
Tpnsetcatito = Fronsitmeasiod Ce)
Tausei2vunkacs = Tes a 3.7)
Tirentevaitnst = Tirana G8)
The fourth set of results on Table 3-1 shows the validated flow rates that
result from believing all of the temperature measurements and then finding
the three flow rates that minimize the least-squares error. Notice that with
this assumption (believing all three temperatures), the two HRSG inlet
steam flows are forced by the mass and energy balances to be equal because
the outlet (throttle) steam temperature is midway between the two inlet
3. Data Validation page 100,steam temperatures. The only way that this can happen, according to
conservation of mass and energy, is if the inlet flows are equal.
The optimal solution to the data validation problem is the solution which
finds the set of validated data that is as close as possible to all of the
measurements (minimizes the least-squares error), and still satisfies the
mass and energy balances. The validated flows for the optimal solution are
shown on the last row of validated results on Table 3-1. If all of the sensors
are of equal accuracy, this set of validated is better than any of the other
possible sets of validated results because it makes use of all of the sensors,
and weights all of the sensors equally (absolute uncertainties set equal to 1.0
for all sensors) in the least-squares e:ror.
This method of least-squares data validation can be applied anywhere in the
power plant where redundant sensors exist. If the gas turbine exhaust gas
flow is known from a measurement or from a heat-balance analysis of the
gas turbine, then there are typically multiple redundancies in the HRSG
data. The inlet feedwater flows are usually measured, and the outlet steam
flows are also measured. These flows are redundant. In addition, there is
redundancy because the gas-side energy loss must equal the energy gain of
the steam/water side. Least-squares data validation can be used to determine
more accurate values for the HRSG steam flows based upon measurements
of feedwater flow and exhaust-gas inlet and outlet temperatures.
Ina Rankine-cycle power plant, the measured feedwater drain flows are
typically redundant with the measured feedwater flow rate and the outlet
feedwater temperature from each feedwater heater, when mass and energy
balances are applied to each of the feedwater heaters.
As an example of the methodology required to implement the least-squares
data-validation procedure described above, the two shaded boxes (Figures 3-
3 and 3-4) which follow present a mathematical model containing Equations
3.2 through 3.8, and show a set of results that are a solution to the
mathematical model.
The first shaded box (Figure 3-3) contains the complete set of equations
required to implement the least-squares data-validation procedure that
generated the results on Table 3-1. The equations are in the format required
by the LINGO™ optimization equation solver. LINGO™ is described in
more detail in Chapter 7, “Gas Turbine Performance”, and in Chapter 10,
“Boiler Performance”. The second shaded box (Figure 3-4) contains a
results report generated by LINGO™ as the solution to the set of equations
presented in the first shaded box.
page 101, 3. Data ValidationILINGO Model £6 Apply Least-Squares Mass-and-Energy-Balance Data-
Validation to Flow Redundancy Data!
MODEL:
IMINIMIZE THE LEAST-SQUARES ERROR OBJECTIVE FUNCTION;
min = ((wi-wi_mes) /Wi_UNC)"2 + ((w2-w2_mes)/W2_UNC)~2 + ((w3-
w3_més) /W3_0NG)*2 + ((T1-T1_mes)/T1_UNG)*2 # ((12-
32 mes) (B2)UNC) *2 +. ((T3-T3/mes}/73_UNC) “2;
\Conservation of Mass;
wt 2 = 37
'Consezvation of Energy:
wit (HEgsCp* (M-Tref)) + w2* (HEg#Cp* (72-Tret) }=w3" lHEg#Cp* (73
Tref));
lUse these equations to force belief in a given measurement;
't2=t2_mes;
't3=t3_mes;
twi=wismes;
! CONSTANTS;
cp=0.53
Tref=550.;
Hfg=1200.7
!UNCERTAINTIES;
wiune = 1.
w2-une = 1.
w3-une
Tloune
r2_une
T3-une
te
re
Li
Le
IMBASUREMENTS ;
w1_mes
w2Lmes
w3omes
Times
72/mes
73mes
END
Figure 3-3 Model used to solve the redundant steam flow data validation analysis
using the LINGO™ equation solver
3. Data Validation page 102LINGO RESULTS REPORT
Rows= 3 Vars= 6 No. integer vare= 0
Nonlinear rows= 2 Nonlinear vars= 6 Nonlinear constraints= 1
Nenzeros= 16 Constraint non2= 9 Density=0.762
No. <: No. =: 2No. >: 0, Obj=MIN Single cole=
optimal solution found at step: 22
Objective value! 75.27935,
variable Value Reduced Cost
a 110.9605, =0.14156938-04
Wi_MES 106-0000 9.000000
wove 3.000000 60000000
we 104.0403, 60000000
W2_MES 93.0000 9.000000
1.000000 0.000000
215.0008, -0.22308098-05
220.0000, 0.000000
1.000000 00000000
1039.79 0, 0000000
1040.000 0.000000
1.000000 00000000
999.7924 0.7487616E-05
1000..000 ‘0.000000
1.000000 0,0090000
1020.429 -0.35073878-05
1020000, ‘0000000
1.000000, 0.000000
0.S00a000 0. 0000000
550.0000 00000000
1200.00 0.000000
Figure 3-4 Output results from the LINGO™ model in Figure 3-3,
page 103 3. Data Validation,3. Data Validation page 1044. Accuracy of Calculated Results
4.1 Instrument Error
4.1.1 Measurement Error
‘The terms and methods used in this book to describe the uncertainty in
‘measurements and in the results calculated from those measurements are
consistent with the terms and methods described in “ASME PTC 19.1-1998,
Test Uncertainty”, which is a supplement to the ASME performance test codes.
Every measurement may be considered to equal the true value of the quantity
being measured plus an error. The accuracy of the measurement is the
closeness of the measured value to the true value. Normally that accuracy is
expressed mathematically in terms of the measurement uncertainty.
Uncertainty is the interval around the measured value that contains the trie
value for a given confidence level
Measurement uncertainty with 95% Confidence Level
TrueValue =X £U yy 1
where
TrueValue is the true value of the quantity being measured
Xis the measured value
Ups is the uncertainty in the measurement with 95% probability that the
true value lies within an interval equal to the measurement plus or
minus the uncertainty
It is convenient to divide the error in the measurement into two components
called the random error (or precision) and the systematic error (or bias). Every
‘measurement will fluctuate about a mean; the error in the measurement due to
these fluctuations is called random error. Another tetm for the random error is
precision. The mean of the measurements will be different from the true value
by an error amount called the systematic error. Another term for the systematic
error is bias.
Measurement Error:
b=fre (42)
where
6's the total measurement error
page 105 4. Accuracy of Calculated ResultsBis the systematic measurement error (bias)
is the random measurement error (precision)
41.2 Random Uncertainty
One reason to divide the error into random and systematic is that the
uncertainty due to random error may be estimated by inspection of the scatter
in the measurements. The standard deviation of a set of measurements can be
used to estimate the uncertainty in the measurement due to random error. The
standard deviation of a data sample can be calculated by the following formula.
‘Standard Deviation of Measured Data Values:
spo [peat 43)
N
where
'Syis the standard deviation of the set of measured values
X,, is a measured value with index k
% is the mean (average) of the measured values
Nis the total number of measured values
‘kis an index to keep track of the measurements; the first measurement
has k equal to one, the second measurement has k equal to two, and so
forth,
w
> isa symbol indicating that the expression which follows is to be
summed for all values of k from one to N.
‘The mean or average of the measurements is calculated as follows.
Mean of Measurements:
— 12
x 7 2 x, (44)
In typical on-line monitoring situations a computer-based data acquisition is
used and the measurement of a variable may be only a single data value or a
single average of data values taken over a short period of time (milliseconds).
If the period of time is short relative to the time period of the random variations
4, Accuracy of Calenlated Results. page 106in the process, then this single average of data values should be treated as if it
were single data value.
The random uncertainty of this single data value from a computer based data
acquisition system can be estimated by taking a set of data values over a longer
time frame (over which the true value is expected to remain constant) and
calculating the standard deviation of the set of data values
Sometimes computer-based data acquisition systems will take a data value at a
fixed time interval over a medium length of time (minutes), and then average
these data values to get a time-averaged data value to use as current input to a
performance monitoring calculation. If the standard deviation of a single data
value is known from prior testing to be Sx, then the standard deviation of the
mean of the time-averaged data values is,
Standard Deviation of a Mean of Data Values’
(4.5)
where
‘Sys the standard deviation of a test set of measured data values, taken
prior to the performance test
Ss
data values
the standard deviation of the mean (average) of the time-averaged
Nis the number of current data values used to determine the time-
averaged value.
4.1.3. Systematic Uncertainty
Systematic error is constant as measurements are repeated and therefore cannot
be quantified by observation of a set of measurements. Thus, the systematic
uncertainty, # , must be estimated based on engineering judgment and
knowledge of the equipment being used. Often the instrument manufactures
provide an estimate of the likely systematic uncertainty to be expected from a
given sensor.
The total measurement uncertainty can be calculated from knowledge of the
random and systematic measurement uncertainties as follows.
Us =2 (4) +(S¢P 46)
page 107 4, Accuracy of Calculated Resultswhere
Uisis the total uncertainty in the measurement with 95% probability
that the true value lies within an interval equal to the measurement plus
or minus the uncertainty
Bis the systematic uncertainty (bias) in the measurement
Sy is the random uncertainty (precision) in the measurement
4.2 Uncertainty of a Calculated Test Result
To understand how the measurement uncertainty associated with each sensor
used in a performance test affects the uncertainty of a result calculated from the
‘measurements, one must know how much a given change in a measurement
will change the calculated result. The ratio of the change in a result to a unit
change in a measured value used in the calculation of the result is called the
sensitivity.
The sensitivity of a calculated result toa change in a measured value can be
found by the following procedure. First calculate the result value, R(1), using
the measurement value, X(1). Then, change the measurement by a small
amount to the new measured value, X(2), and calculate a new result, R(2),
based upon the changed measured value. The sensitivity is the change in the
result divided by the change in the measurement.
Sensitivity of a Calculated Result to a Change in a Measured Value:
R(2)- R()
9=x0@)-x0) mo
where
@ is the sensitivity of the result to the measurement
X is a measured value used as input to the calculation of the result
Ris the calculated result associated with the given value of the
measurement
Equation 4,7 is strictly valid if the calculated result depends only upon a single
‘measurement or if all the measurements are independent of each other in the
calculation of the result. Neither of these cases is typically true in a
performance monitoring calculation where the calculated result is usually a
non-linear combination of many measurement variables. Fortunately the error
in using this first order approximation is usually small.
4, Accuracy of Calculated Results page 108Using this first order approximation, the calculated result from a set of
‘measurements is the result that would be obtained from a given set of
measurements plus the change in the result that is caused by the change in any
of the measurements.
Result of Multivariable Calculation:
RIX, Xy Xr)
= Rey, Hay Hee vo) + Ox, (X= He, ) + Ox, Ka ~My, ) + Ox, (X5 — Hoe
(48)
where
R is the test result calculated at given values of the measurements
4, is the sensitivity of the result 1o the measurement i
H,, is the given value of measurement i where the result is known
‘is the value of the measurement / where the value of the result is
desired
Ifall the variables, X;, are independent in the calculation of the result, then the
standard deviation of the result depends upon the standard deviation of the
‘measurements in the following way.
Standard Deviation of Result:
28%
H (4.9)
where
Sq is the standard deviation of a calculated result that depends upon the
measured values of n parameters
,, is the sensitivity of the calculated result to measured parameter i
S,, is the standard deviation the measured values of parameter i
The total uncertainty of the calculated result depends upon both the random
and the systematic uncertainties of the measured parameters. If the systematic
uncertainty of each of the measurements is known then the systematic
uncertainty of the result can be estimated as follows.
Page 109 4. Accuracy of Calculated ResultsSystematic Uncertainty of Result:
Bro {Xe Br, (4.10)
By is the systematic uncertainty of a calculated result that depends
upon the measured values of n parameters
where
4,, is the sensitivity of the calculated result to measured parameter 7
Bj, is the systematic uncertainty the measured values of parameter i
The total uncertainty of a calculated result depends upon the standard deviation
and the systematic uncertainty of that result.
Uncertainty of a Calculated Result:
EZ ) +(Sp) (4.11)
Uyo, is the total uncertainty in the result with 95% probability that the
true value lies within an interval equal to the result plus or minus the
uncertainty
By, is the systematic uncertainty (bias) in the result
S, is the random uncertainty (precision) in the result
4.3, Monte Carlo Method
4.3.1 Definition of the Monte Carlo Method
The first order (linear) method, described above, to estimate the uncertainty in
a calculated result given the uncertainty in the measurements is an
approximation. This approximation may be inaccurate when applied to the
situation where a complex, non-linear heat balance computer code is used to
process the measurements and calculate the results.
There is a way to correctly propagate the measurement uncertainties through a
computer-based model, and it can be done to any desired level of accuracy. The
method is called the Monte Carlo method, and it involves running many sets of
simulated measured data through the computer model and compiling the
4, Accuracy of Calculated Results page 110calculated results. Each input measured data value is varied according to its
own probability distribution using a random number generator. After many
calculations, each using a different set of input (simulated measured) data, the
probability distribution of the results will be an accurate representation of the
uncertainty in the calculated result given the probability distributions of the
inputs. See Figure 4-1.
It’s called the Monte Carlo method because it’s a lot like gambling (Monte
Carlo is a world famous gambling location) in that random numbers generate
the input to each calculation, Like in gambling, the results from many trials are
determined by the probability of occurrence of each possible outcome. The
more trials run, the more accurate the probability distributions of the outputs.
Monte Carlo Method
Pick Input Values Rua Otwene
‘According to. Calculations Distibaont
Their Probability Ech Sof the
Distributions eps Clie
X,4U, ———
RU,
XU, Performance
Caleiaton
X4U; Module RAU,,
XaU; ReU,,
INPUTS OUTPUTS
Figure 4-1 Overview of the Monte Carlo Method
4.3.2 Probability Distributions
Each measured data value from a power plant has its own probability
distribution. The Monte Carlo method involves randomly sampling (picking)
simulated measured values from their probability distributions and inputting
sets of these measured values to the performance monitoring calculation.
page 11 4. Accuracy of Caleulated ResultsThe probabitity distributions of the measured data are normally not known; but,
for purposes of Monte Carlo simulation, the distributions are often assumed to
be cither uniform or normal. The uniform distribution is such that all values
between the minimum possible value, a, and the maximum possible value, b,
have equal probabilities of occurrence. The mean of the uniform distribution is
the average of a and b. The variance (or standard deviation of a set of values) is
the interval between a and b divided by the square root of twelve (3.464). The
interval from a to b contains 100% of the data sampled from the uniform
probability distribution. Figure 4-2 is a plot of the uniform probability
distribution function. While the uniform distribution does not appear in nature,
it is easy to implement in a Monte Carlo simulation, and the results are often
easy to interpret.
£0)
Figure 4-2 Plot of the Uniform Probability Distribution Function
The normal distribution is the bell-shaped curve that we are all familiar with,
Figure 4-3 is a plot of the normal probability distribution function. The mean,
t, and the variance, o, of the normal probability distribution are such that the
interval 2c contains 95% of all values sampled from the distribution. This
level of uncertainty is often used in the statement of uncertainty for a
measurement or for a calculated result.
4. Accuracy of Calculated Results. page 112‘The normal probability distribution function:
1 eat
£) pr a @.12)
where
JAés) is the probability of sampling data over the interval xtdx
nis the mean of the distribution
iis the variance of the distribution
f(x)
HW x
Figure 4-3 Plot of the Normal Probability Distribution Function
4.3.3 Sampling from Probability Distributions
Sampling simulated measured data from a probability distribution requires the
use ofa random number generator. A random number generator is a comptiter
algorithm that outputs a number with a uniform probability distribution
between the values of zero and one. Each time the random number generator is
called, it outputs a new number.
Ifa value equal to ¢ is retrieved from a random number generator, then a value,
X, can be computed from the following formula, where X has a uniform
probability distribution between the values a and b.
Page 134. Accuracy of Calculated ResultsSampled Value with Uniform Probability Distribution:
Xsat(b-ag (4.13)
where
Xis a sampled value with a uniform probability distribution between a
and b
isa random number between 0 and I.
‘Sampling from a normal probability distribution can be accomplished using the
Box-Muller transformation, The Box-Muller transformation states that if ¢, and
@ are two uniformly and independently distributed numbers between zero and
one, then X; and X2 have a normal distribution with mean equal to zero and a
variance equal to one, where X; and X» are defined by equations 4.14 and 4.15,
Values with Normal Probability Distributions with means equal to 0.0, and
variance 1.0:
X, =J-2inG,) cos(2x¢,) (4.14)
and
X, = y-2in(G,) sin(27¢;) (4.15)
To obtain a simulated measured value with a mean equal to jt and a variance
equal to 0, one needs to generate two random values, ¢; and 2, and then use
either of the following two formulas (equation 4.16 or 4.17) to calculate the
simulated measured value.
Values with Normal Probability Distributions with means equal to y and
variance 0:
X, = t+a-2InG,) cos(2a¢,) (4.16)
and/or
X, =M+oy-2in(G,) sin(27¢,) (4.17)
where
X isa sampled value with a normal probability distribution with mean
and variance o
(is a random number uniformly distributed between 0 and 1.
Gis a second random number between 0 and 1
4, Accuracy of Caleulated Results page 1144.3.4
Running the Monte Carlo Simulation
Each run of the Monte Carlo simulation involves the following steps:
Generate a set of random numbers, one or two random numbers for
each simulated measured value to be input to the performance
calculation module
Calculate a set of input values to the performance calculation using the
sampling formula (Equation 4.13 or Equation 4.16) appropriate for the
probability distribution chosen for each input value.
Execute the performance calculation using the input values from step 2.
4. Record (store) the results of the performance calculation for future
4.3.5
processing
Retum to step 1. and generate a new set of calculated results;
enough executions have been completed, go to step 6.
Calculate the mean and standard deviation of each calculated result
variable.
Results of the Monte Carlo Simulation
The uncertainty of a calculated result can be quantified by calculating the
standard deviation of the calculated result from the result values stored from all
of the Monte Carlo simulations performed. First calculate the mean using
Equation 4.18; then calculate the standard deviation of each result variable
using Equation 4.19.
Mean of Result Variable:
where
= ie
R we (4.18)
R, is a calculated result value form the Ath simulation
Ris the mean (average) of the result values
Nis the total number of simulations executed
Kis an index to keep track of the simulations; the first simulation has k
equal to one, the second simulation has k equal to two, and so forth.
Page IIS 4. Accuracy of Calculated Results> isa symbol indicating that the expression which follows is to be
ma
summed for all values of k from one to N.
Standard Deviation of Calculated Result Variable:
y (4.19)
If the result variable has a normal distribution (which it probably does not),
then the uncertainty of the result with a 95% confidence level is plus or minus
two times the standard deviation. This is because the standard deviation of a
sample of data values is an estimate of the variance of the probability
distribution; the more data values in the sample the more accurate the estimate
of the variance. The uncertainty of the calculated standard deviation of the
calculated result variable can be estimated from the number of simulations
performed.
Uncertainty of the Standard Deviation of Calculated Result Variable:
(4.19)
If only two Monte Carlo simulations are executed then there is 100%
uncertainty in the calculated standard deviation of the result. But if one
hundred and one simulations are executed, the standard deviation of the result
is known to within 10%.
4. Accuracy of Caleulated Results page 1165. Overall Power Plant Performance
5.1 Equipment Performance versus Plant Performance
Overall plant performance evaluation is a comparison of the expected plant
performance to the measured plant performance. A unique characteristic of
overall plant performance is that many pieces of equipment in the power plant
must all operate with rated performance capability in order for the plant to
perform as expected. That is, the expected overall plant performance is based
upon the assumption that all equipment in the plant are operating at their rated
capabilities.
Itis important to understand that the expected equipment performance
calculated in the performance evaluation of each piece of plant equipment is
not the same as the performance of that same piece of equipment calculated in
the evaluation of overall expected plant performance. The operating conditions
imposed on plant equipment at the current (actual) plant operation (taking into
account the degradation of the surrounding equipment) are not the same as the
expected operating conditions if all plant equipment performs as rated. Since
expected equipment performance is evaluated at the current (actual) operating
conditions of each piece of equipment, the expected equipment performance
will not match the equipment performance when the entire power plant is
operating as expected (with rated capability).
The sum of the expected powers generated in each piece of equipment in the
power plant is not equal to the expected power of the overall power plant. The
example below illustrates this situation for a steam turbine that is operating in a
degraded power plant system.
Consider a steam turbine whose throttle steam flow comes from a steam
generator (boiler for a Rankine cycle, or HRSG for a combined-cycle) located
upstream of the steam turbine. At the reference plant operating conditions, the
rated steam flow is 700,000 Ib/hr, and the rated steam turbine power is 100
MW. If the steam generation degrades such that the throttle steam flow
becomes 600,000 lb/hr at the reference operating conditions, the expected
steam turbine power would be evaluated at the actual steam flow of 600,000
Ib/hr instead of the rated steam flow 700,000 Ibvhr, and would equal
approximately 85 MW.
Thus, at the plant reference operating conditions, the expected steam turbine
power used to evaluate the steam turbine performance would be 85 MW, but
the steam turbine power calculated in the expected overall power plant
performance evaluation would be 100 MW. If under these degraded steam flow
page 117 5. Overall Plant Performanceconditions, the steam turbine actually produced 85 MW at the plant reference
operating conditions, the degradation of the steam turbine would be zero
because the expected steam turbine power equals the actual steam turbine
power. The degradation of the steam generator would be 100,000 Ib/hr, and the
power degradation of the overall plant due to the degradation of the steam
generator would be 15 MW.
If the steam turbine in this example was the only power generating equipment
in the power plant, the expected power of the steam turbine (85 MW) would
not equal the expected power of the overall power plant (100 MW). This
illustrates the concept stated above that the sum of the expected powers
generated in each piece of equipment in the power plant is not equal to the
expected power of the overall power plant,
5.2 Specification of Overall Power Plant Performance
A power plant is typically warranted to produce a guaranteed amount of
electric power and a guaranteed plant heat rate if operated at the guarantee
operating conditions. In this book, the operating conditions where the plant
performance guarantee is specified are called the reference operating
conditions, and the plant performance at these conditions is called the rated
performance. Table 5-1 illustrates typical parameters used to rate the
performance of a power plant. A power plant can be expected to produce rated
power and rated heat rate only at the reference operating conditions.
5. Overall Plant Performance page 118,Table 5-1 Rating Specification for a Power Plant
Power Plant Rating Specification Example
Data
RATIN¢
Net Power 500 Mw
Net Heat Rate 8800 BlukW-hr
REFERENCE OPERATING CONDITIONS:
Ambient Air Temperature OF
Ambient Air Pressure 146 psia
Ambient Relative Humidity 60%
Fuel Type Natural Gas.
Fuel Lower Heating Value 20800 Buu/tbm
Load Level or Operational Mode Base
Process Steam Requirement: 100.000 ib/hr 200
Flow, Pressure and Temperature pia
300F
Process Return Water Temperature 140F
River or Ocean Temperature NA
‘The rated power and rated heat rate are the expected performance values for the
power plant at the reference operating conditions. If any of the operating
conditions change, the expected power and the expected heat rate will change.
The plant vendor will often guarantee plant performance at a set of selected
operating conditions other than the reference operating conditions. This set of
selected plant performance specificatiors is typically part of the plant “thermal
kit”. For a combined-cycle power plant the plant performance may be
guaranteed on a hot day, on a cold day, with various process steam
requirements, with an alternate fuel (oil instead of gas), with one gas turbine
‘out of service, and/or with duct bumers in operation.
page 119 5. Overall Plant Performance‘A performance monitoring system must predict the expected plant power and
heat rate at any possible operating concition and then compare that expected
performance to the actual performance. The plant guarantee points form the
basis for the validation of an expected performance model of the power plant.
Any model used to predict the expected performance of the plant must match
the guaranteed plant performance at all operating conditions where those
guarantees are stated.
Vartation of Combined: cyet
lant Performance with Ambiont Temperature
10
° = ” © ” 00 By 0
able: Temperature)
Figure 5-1 Variation of Combined-Cycle Plant Performance with Ambient
Temperature
Figure 5-1 illustrates the variation in overall plant performance that can be
expected for a combined-cycle power plant as ambient temperature changes.
The large change in plant power as the ambient air temperature changes is due
primarily to the increase in gas turbine power at lower inlet temperature. The
variation of condenser pressure with ambient temperature also contributes to
the plant power change, but its effect is much smaller than the effect of the
change in gas turbine power.
The primary driving force for the variation of power with ambient in a
combined-cycle power plant is the increase in air mass flow rate into the gas,
5. Overall Plant Performance page 120turbines with lower ambient temperatures. To a good approximation, the air
flow into a gas turbine is at constant velocity. The density of air, since it acts
like an ideal gas, varies directly with pressure and inversely with absolute
temperature. As the air temperature goes down, the air density goes up, and the
inlet mass flow rate to the gas turbine goes up.
Rankine Power Plant Performance Versus Ambient Temperature
Temperature)
Figure 5-2 Variation of Rankine-Cycle Plant Performance with Ambient Temperature
The performance changes of a Rankine cycle power plant with changes in
ambient temperature are relatively less pronounced and are sometimes ignored
in Rankine cycle performance-monitoring systems. Itis difficult to generalize
the performance changes of Rankine cycle plants with ambient temperature.
The data in Figure 5-2 illustrates performance changes that might be expected
for a Rankine cycle power plant as ambient conditions change, but the precise
performance changes will be plant specific. The change in condenser pressure
and cooling tower performance are important parameters in the Rankine cycle
plant performance changes with ambient temperature. One reason these effects
are so plant specific is that boiler forced-draft fans and induced-draft fans may
page 121 5. Overall Plant Performancereach limits at either low or at high ambient temperatures, depending upon the
boiler design.
5.3. Overall Plant Expected Performance Models
There are three methods to predict expected overall plant performance given
the plant operating conditions.
* Curve-Based Method: use performance or correction curves,
© Model-Based Method: use a computer-based model of the power plant
© Impact Method: use the total of the equipment impacts on plant power
5.3.1 Curve-Based Method for Expected Plant Performance
Performance or correction curves may be used to predict expected overall plant
power and heat rate as long as the parameters that impact plant performance are
independent of each other. Figures 5-1 and 5-2 show performance curves
illustrating the change in plant performance as ambient temperature changes
for two different types of power plants.
‘The basic idea behind curve-based performance methods is to assemble a set of
performance or correction curves that plot the variation in a plant performance
parameter (power or heat rate) when one of the operating conditions changes.
‘The total power plant performance fractional change is then computed by
multiplying together the fractional changes for each operating condition, where
each multiplying factor is generated from a separate correction curve.
Curve-Based Method for Expected Plant Performance:
Powergen = Poets T] CurveValue(actual conditions)
sitspes\ CurveValue(reference conditons) } ser cares
(6.1)
HeatRateyayg = HEAIROE yay TT] ( CurveValue(actual conditions) |
llduries\ Curve Value(reference conditons) )ysecunss
(5.2)
where
is a mathematical operator indicating that all the terms in the
silos
following parenthesis are to be multiplied together, one term for each
plant performance curve, until terms from all the plant performance
curves are included in the final product
5, Overall Plant Performance page 122,A curve-based performance-prediction method becomes inaccurate when
multiple parameters (operating conditions) change the plant performance and
the performance change is large (greater than five percent). This happens
because the operating conditions are not truly independent of each other, but
this independence is assumed when a separate curve is used for each operating
condition.
Consider the effects that ambient temperature and relative humidity (or wet-
bulb temperature) have on overall plant performance. In a combined-cycle
power plant the ambient temperature strongly affects gas turbine power, steam
turbine power and condenser duty, which changes with plant load. The wet-
bulb temperature (for a wet cooling-tower) has a strong impact on condenser
pressure, but the magnitude of this impact depends upon condenser duty, which
in turn depends upon ambient temperature.
The result is that the change in plant performance caused by a change in wet-
bulb temperature is not independent of ambient temperature. [fa curve of plant
power versus wet-bulb temperature is used, the curve will not account for
changes in ambient temperature. It may be possible to produce a family of wet-
bulb temperature curves, one for each ambient temperature. This two-
dimensional curve would surely be better than using a single curve for wet-
bulb temperature.
Unfortunately, even two-dimensional curves do not always solve the problem.
‘The ambient air pressure will also affect plant load in a manner that is not
independent of ambient temperature or wet-bulb temperature. Maybe a set of
ambient temperature curves is required for each ambient pressure, and then a
set of wet-bulb temperature curves for each ambient temperature.
The fuel type, process steam loads, return water fraction, and return water
temperature will all impact the plant performance in ways that are not
independent of ambient conditions. The basic conclusion is that a curve-based
method will generally work well for small changes in overall plant
performance, when the expected plant performance is within five percent of the
tated plant performance. For application in situations resulting in larger
changes in plant performance need, a curve-based method is probably not the
best choice.
5.3.2. Model-Based Method for Expected Plant Performance
The model-based method overcomes the problems associated with the eurve-
based method because the effects of each operating condition can be combined
with the effects of the other operating conditions in a single calculation. The
disadvantages of the model-based method are the high degree of plant
page 123, 5, Overall Plant Performanceinformation and knowledge required to create an accurate model, and the level,
of engineering skill and effort required to implement and maintain the model.
Figure 5-3 Overall Plant Model for a Combined Cycle Power Plant
When applying the model-based method, the power plant mode! must be
“tuned” so that it predicts the rated plant performance when run at the reference
operating conditions. The model should be validated versus guarantees at those
operating points other than the reference operating conditions where the plant
performance is guaranteed.
Note that to generate an accurate plant model for model-based methods of
performance prediction; itis usually not enough to simply generate a physical
model of the key plant equipment (turbines, pumps, heat exchangers, piping,
etc.). It will probably also be necessary to recognize and understand the
operational limits on the plant equipment and how the plant is controlled to
respond to those limits, and implement chese limits and controls when building
5, Overall Plant Performance age 124