Thermal Power Plant & Simulation

Thermal Power Plant Simulation and Control

Edited by Damian Flynn

The Institution of Electrical Engineers

Published by: The Institution of Electrical Engineers, London, United Kingdom

2003: The Institution of Electrical Engineers

This publication is copyright under the Berne Convention 2003 and the Universal Copyright Convention. All rights reserved. Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, this publication may be reproduced, stored or transmitted, in any forms or by any means, only with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Inquiries concerning reproduction outside those terms should be sent to the publishers at the undermentioned address:

The Institution of Electrical Engineers, Michael Faraday House, Six Hills Way, Stevenage, Herts., SG1 2AY, United Kingdom

While the authors and the publishers believe that the information and guidance given in this work are correct, all parties must rely upon their own skill and judgment when making use of them. Neither the authors nor the publishers assume any liability to anyone for any loss or damage caused by any error or omission in the work, whether such error or omission is the result of negligence or any other cause. Any and all such liability is disclaimed.

The moral rights of the authors to be identified as authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

British Library Cataloguing in Publication Data

Thermal power plant simulation and control. - (lEE power & energy series ; 43) 1. Electric power-plants - Management 2. Electric power systems - Control 3. Electric power systems - Computer simulation I. Flynn, D. II. Institution of Electrical Engineers 621.311210113

ISBN 0 85296 419 6

Typeset in India by Newgen Imaging Systems Printed in the UK by MPG Books Limited, Bodmin, Cornwall

List of contributors

A. Alessandri Institute for the Studies of Intelligent Systems for Automation National Research Council of Italy Genova, Italy

A.E Armor Electric Power Research Institute Palo Alto, California, USA

M.D. Brown Atkins Aviation and Defence Systems Bristol, England

A. Cipriano Electrical Engineering Department Pontificia Universidad Cat61ica de Chile Santiago, Chile

P. Coletta Institute for the Studies of Intelligent Systems for Automation National Research Council of Italy Genova, Italy

M. Cregan School of Electrical and Electronic Engineering The Queen's University of Belfast Belfast, Northern Ireland

G.Q. Fan Veritas Software Sydney, Australia

x List of contributors

D. Flynn School of Electrical and Electronic Engineering The Queen's University of Belfast Belfast, Northern Ireland

A. Fricker Innogy plc Swindon, England

R. Garduno-Ramirez Electrical Research Institute Cuernavaca, Morelos, Mexico

G.W. Irwin School of Electrical and Electronic Engineering The Queen's University of Belfast Belfast, Northern Ireland

K.Y. Lee Department of Electrical Engineering Pennsylvania State University Pennsylvania, USA

A. Leva Department of Electronic Engineering and Information Sciences Politecnico di Milano Milan, Italy

K. Li School of Mechanical and Manufacturing Engineering The Queen's University of Belfast Belfast, Northern Ireland

C. Maffezzoni Department of Electronic Engineering and Information Sciences Politecnico di Milano Milan, Italy

T. Moelbak Elsam A/S Fredericia, Denmark

List of contributors xi

J.H. Mortensen Tech-wise A/S Fredericia, Denmark

G. Oluwande Innogy plc Swindon, England

T. Parisini Department of Electrical, Electronic and Computer Engineering University of Trieste Trieste, Italy

G. Poncia United Technologies Research Center East Hartford, Connecticut, USA

G. Prasad School of Computing and Intelligent Systems University of Ulster Londonderry, Northern Ireland

N.W. Rees School of Electrical and Telecommunication Engineering The University of New South Wales Sydney, Australia

J.A. Ritchie School of Electrical and Electronic Engineering The Queen's University of Belfast Belfast, Northern Ireland

D. S~iez Electrical Engineering Department Universidad de Chile Santiago, Chile

S. Thompson School of Mechanical and Manufacturing Engineering The Queen's University of Belfast Belfast. Northern Ireland

Preface

During the past decade power generation has undergone several extremely significant changes. These include deregulation of the electricity industry in many parts of the world, with a greater focus on economic and financial concerns instead of purely engineering issues. In conjunction with this, environmental matters are of increasing interest, leading to an assessment of existing greenhouse gas emissions and the exploitation of renewable energy sources. Additionally, combined cycle gas turbines (CCGTs) have emerged as an extremely economic and efficient means of electricity generation. Finally, many power plants have been retro-fitted with modern and sophisticated, plant-wide instrumentation and control equipment. These computer- based distribution control systems (DCSs) are intended to enhance regulation control performance and more importantly provide a means for implementing supervisory control/monitoring schemes.

These various considerations have led to significant changes in the philosophy of how power stations are operated, while at the same time affording engineers the opportunity to introduce monitoring and plant-wide control schemes which were previously infeasible. However, a distinction has largely arisen between those working in the power and control oriented research communities, with centres of excellence in scattered locations, and engineers engaged in power plant design, operation, con- sultancy, etc. The objective of this book is to address this issue, through a number of case studies, which illustrate how various methodologies can be applied to various subsystems of power plant operation, or indeed introduced into the overall control hierarchy. The case studies presented focus on what can feasibly be achieved with an indication of the subsequent benefits of doing so, using results from live plant where possible.

The level of the book makes it suitable for engineers working in the power generation industry who wish to gain an appreciation of the advances which have taken place in this field within the research community. It should also provide a very useful overview for new and experienced researchers working in this area. A number of the contributions to this book arise from work carried out at, or in collaboration with, universities and research institutions, while others benefit from the experience of practitioners in the industry. A natural consequence of this is that a mixture of viewpoints is offered, with a contrast between the use of academic and industrial

xiv Preface

terminology. The mathematical content of the book is sufficient to give an indication of the underlying technologies, and the deficiencies of more traditional techniques, with the reader directed to related work for further detail.

The text is split into three main parts covering, respectively, power plant simulation, specific control applications and optimisation/monitoring of plant operations. Chapter I provides a brief introduction to power plant fundamentals, outlining different plant configurations, the control requirements of various loops, and the hardware and instrumentation on which these systems are based. An essential aspect of investi- gating and developing novel control and monitoring schemes is a detailed simulation of the system in question. Chapter 2 illustrates how a complex power plant model can be constructed using an object-oriented approach. The reader is introduced to the Modelica modelling language, and issues such as testing and validation are discussed.

Part 2 (Control) comprises five contributions and forms a major part of the book. A number of diverse applications are considered, and differing control strategies are proposed and implemented. Chapter 3 investigates the highly complex problem of both modelling and controlling pulverised fuel coal mills. Linear quadratic and predictive control techniques are investigated, with a supervisory operator support system introduced. Chapter 4 tackles the problem of excitation control of a synchronous machine. Local model network and adaptive control-based approaches are exam- ined in detail. Chapter 5 then examines steam temperature control of a once-through boiler for both the evaporator and superheaters. Linear quadratic Gaussian, fuzzy logic and predictive control schemes are applied, with the benefits of feedforward action using suitable instrumentation strongly highlighted. Chapter 6 examines the problem of controlling combined cycle plant. An objective function is defined based on operational costs, and alternative hierarchical control configurations are exam- ined. Finally, in this section, Chapter 7 explores the development of a multi-input multi-output (MIMO) predictive controller sitting on top of the plant's conventional control systems to improve the overall plant's capabilities.

Part 3 (Monitoring, optimisation and supervision) again comprises five contributions, and demonstrates how the ability of distributed control systems to gather plant-wide, real-time data can be constructively employed in a range of applications. Chapter 8 introduces a sophisticated plant-wide, neurofuzzy control scheme with feedback and feedforward actions to provide improved unit manoeuvrability and an improved distribution of control tasks. Chapter 9 then focuses on the task of modelling NOx emissions from a coal-fired power station. A grey-box modelling approach is proposed, taking advantage of a priori knowledge of NOx formation mechanisms. Chapter 10 introduces model-based approaches for fault detection of a high-pressure heater line. Again grey-box identification, coupled with non-linear state estimation techniques are considered, to aid fault diagnostics. Chapter 11 continues with an examination of how the data stores which distributed control systems now offer can be exploited for both fault identification and process monitoring activities. The part concludes in Chapter 12 with an overview of a number of performance support and monitoring applications that have been successfully applied to real plant, largely based around a real-time expert system.

Preface xv

The final part of the book highlights some possibilities and issues for the future. Chapter 13 demonstrates how a physical model of a power plant can be integrated into a predictive control strategy to provide enhanced unit control by recognising the true system characteristics. Finally, Chapter 14 discusses some topics of concern including the impact of age and maintenance requirements on existing units in an increasingly competitive environment, and how technology is expanding the capabilities of modern power plant.

The editor would like to take this opportunity to thank all the authors for their contributions, and for their assistance in bringing together the final text. The support and guidance from Roland Harwood and Wendy Hiles of the IEE has also been most welcome. The editor also wishes to acknowledge the significant role played in the creation of this work by Brian Hogg and Edwin Swidenbank in establishing the Control of Power Systems research group at The Queen's University of Belfast. Finally, the advice and encouragement offered by Brendan Fox and Nataga Marta6 from Queen's has been greatly appreciated.

Damian Flynn April 2003

Contents

List of contributors ix

Preface xiii

List of abbreviations xvii

1 Advances in power plant technology M. Cregan and D. Flynn

1.1 Power plant historical development 1.2 Plant configuration and design 1.3 Control and instrumentation 1.4 External influences 1.5 Plant technology developments 1.6 References

1 2 5 9

13 13

Part 1: Modelling and simulation

2 Modelling of power plants A. Leva and C Maffezzoni

2.1 Introduction 2.2 Model structuring by the object-oriented approach 2.3 Basic component models 2.4 Modelling of distributed control systems 2.5 Application of dynamic decoupling to power plant models 2.6 Testing and validation of developed models 2.7 Concluding remarks and open problems 2.8 References

17

17 18 27 50 52 53 56 57

vi Contents

Part 2: Control

3 Modelling and control of pulverised fuel coal mills N.W. Rees and G.Q. Fan

3.1 Introduction 3.2 Modelling of coal mills 3.3 Plant tests, results and fitting model parameters 3.4 Mill control 3.5 Intelligent control and operator advisory systems 3.6 Conclusions 3.7 Acknowledgements 3.8 References

4 Generator excitation control using local model networks M.D. Brown, D. Flynn and G. W. Irwin

4.1 Introduction 4.2 Local model networks 4.3 Controller design 4.4 Micromachine test facility 4.5 Results 4.6 Conclusions 4.7 References

5 Steam temperature control T. Moelbak and J.H. Mortensen

5.1 Introduction 5.2 Plant and control description 5.3 Advanced evaporator control 5.4 Advanced superheater control 5.5 Conclusions 5.6 References

6 Supervisory predictive control of a combined cycle thermal power plant D. Sdez and A. Cipriano

6. l Introduction 6.2 A combined cycle thermal power plant 6.3 Design of supervisory control strategies for a combined cycle

thermal power plant 6.4 Application to the thermal power plant simulator 6.5 Discussion and conclusions 6.6 Acknowledgements 6.7 References

63

63 64 71 80 92 97 97 97

101

101 102 108 113 117 124 127

131

131 133 137 147 159 159

161

161 162

168 171 176 177 177

Contents vii

7 Multivariable power plant control

G. Poncia 7.1 Introduction 7.2 Classical control Of thermal power plants 7.3 Multivariable control strategies 7.4 An application: MBPC control of a 320 MW oil-fired plant 7.5 Conclusions 7.6 Acknowledgements 7.7 References

Part 3: Monitor ing, optimisation and supervis ion

8 Extending plant load-following capabilities

R. Garduno-Ramirez and If. Y Lee 8.1 Introduction 8.2 Power unit requirements for wide-range operation 8.3 Conventional power unit control 8.4 Feedforward/feedback control strategy 8.5 Knowledge-based feedforward control 8.6 Design of neurofuzzy controllers 8.7 Wide-range load-following 8.8 Summary and conclusions 8.9 Acknowledgements 8.10 References

9 Modelling of NOx emissions in coal-fired plant

S. Thompson and K. Li 9.1 Emissions from coal-fired power stations 9.2 An overview of NOx formation mechanisms 9.3 NOx emission models for a 500 MW power generation unit 9.4 Conclusions 9.5 Acknowledgements 9.6 References

10 Model-based fault detection in a high-pressure heater line

A. Alessandri, P Coletta and T. Parisini 10.1 Introduction 10.2 Description of power plant application 10.3 Grey-box modelling and identification of a power plant 10.4 A general approach to receding-horizon estimation for

non-linear systems 10.5 Conclusions 10.6 References

179

179 181 184 189 200 200 201

205

205 207 209 213 221 224 228 238 239 239

243

243 248 253 263 267 267

269

269 271 287

295 307 307

viii Contents

11 Data mining for performance monitoring and optimisation J.A. Ritchie and D. Flynn

11.1 Introduction 11.2 Outline of data mining applications 11.3 Identification of process and sensor faults 11.4 Process monitoring and optimisation 11.5 Non-linear PLS modelling 11.6 Discussion and conclusions 11.7 Acknowledgements 11.8 References

12 Advanced plant management systems A. Fricker and G. Oluwande

12.1 Plant management in a deregulated electricity market 12.2 Supervisory control 12.3 System integration and HMI issues 12.4 Performance monitoring 12.5 Added value applications 12.6 Conclusions 12.7 References

Part4: The future 13 Physical model-based coordinated power plant control

G. Prasad 13.1 Introduction 13.2 A review of physical model-based thermal

power plant control approaches 13.3 Control problems of a thermal power plant 13.4 Applying a physical model-based predictive control strategy 13.5 Simulation results 13.6 Discussion and conclusions 13.7 Acknowledgements 13.8 References

14 Management and integration of power plant operations A.E Armor

14.1 Introduction 14.2 Age and reliability of plants 14.3 Improving asset management 14.4 The impacts of cycling on power plant performance 14.5 Improving maintenance approaches 14.6 Power plant networks: redefining information flow 14.7 Conclusions 14.8 References 14.9 Bibliography

Index

309

309 310 311 325 334 338 341 341

345

345 346 350 351 354 360 361

365

365

366 368 375 381 389 391 391

395

395 396 401 405 407 410 413 414 415

417

List of abbreviations

AF ANN API APMS ARMAX ARX ASME AVA AVR BETTA BMS CARIMA CBR CCGT CCR CEGB CFD COL DCDAS DCS DMA EAF EC EDL EKF EPRI FB FERC FF FFPU FGD GHG

availability factor artificial neural network application program interface advanced plant management system AutoRegressive Moving Average model with eXogenous input AutoRegressive model with eXogenous input American Society of Mechanical Engineers added value application automatic voltage regulator British-wide Electricity Trading and Transmission Arrangements burner management system controlled auto-regressive integrating moving-average case-based reasoning combined cycle gas turbine central control room Central Electricity Generating Board computational fluid dynamics cost of losses distributed control and data acquisition system distributed control system direct memory access equivalent availability factor European Commission electronic dispatch and logging extended Kalman filter Electric Power Research Institute feedback Federal Energy Regulatory Commission feedforward fossil fuel power unit flue gas desulphurisation greenhouse gas

xviii List of abbreviations

GMV GPC HMI HP HRSG HSC IAF ICOAS IGCC ILC ILM IOAS IPCC IPP ISA KBOSS LMN LP LPC LQ LQG LQR LS MBPC MCR MIMO MISO MLP MLR MVC NARMAX

NARX NDE NETA NIPALS NPMPC OIS OOM OSC PCA pf PFBC PLC

generalised minimum variance generalised predictive control human-machine interface high-pressure heat recovery steam generator hierarchical supervisory control integrated application framework intelligent control and advisory system integrated gasification combined cycle integrated load control integrated load management intelligent operator advisory system Intergovernmental Panel on Climate Change independent power producer Instrumentation, Systems and Automation Society knowledge-based operator support system local model network low-pressure lumped parameter components linear quadratic linear quadratic Gaussian linear quadratic regulator least squares model-based predictive controller maximum continuous rating multi-input multi-output multi-input single-output multilayer perceptron multiple linear regression multivariable steam control Non-linear AutoRegressive Moving Average model with eXogenous input Non-linear AutoRegressive model with eXogenous input non-destructive evaluation New Electricity Trading Arrangements non-linear iterative partial least squares non-linear physical model-based predictive control operational information system object-oriented modelling one-side components principal component analysis pulverised fuel pressurised fluidised bed combustion programmable logic controller

List of abbreviations xix

PLS PRBS PRESS RBF RLS RMS RSME SCADA SEGPC SISO SMS TSC UV VOC

projection to latent structures pseudo-random binary sequence predicted residual sum of squares radial basis function recursive least squares root mean square root squared mean error supervisory control and data acquisition state estimation-based generalised predictive control single-input single-output startup management system two-side components ultraviolet volatile organic compound

Chapter 1

Advances in power plant technology

M. Cregan and D. Flynn

I.I Power plant historical development

Fossil fuelled power plants have been supplying electricity for industrial use since the late 1880s. At first, simple d.c. generators were coupled to coal-fired, reciprocating piston steam engines. Electricity was delivered over relatively short distances, and was primarily used for district lighting. The first central generating station was opened by Thomas Edison in September 1882 at Pearl Street, Lower Manhattan, New York City. Lighting alone, however, could not provide an economical market for successful commercial generation, so new applications for electricity needed to be found. The popularity of urban electric tramways, and the adoption of electric traction on subway systems, coincided with the widespread construction of generating equipment in the late 1880s and 1890s.

Initial power plant boiler designs generated steam in a simple water tube boiler, from a coal or coal gas supply. They typically operated at 0.9 MPa (8.6 bar) and 150 C (300 F), and would have been connected to a 30 kW generator. Since then the topography of the typical power plant has evolved into a highly complex system. Today, advanced turbine and boiler designs, utilising new metal alloys, can operate at supercritical conditions of 28.5 MPa (285 bar) and 600 C (1112 F), generating 1300 MW of electricity (Smith, 1998; DTI, 2000).

In a search for reduced operating costs, plant design has moved on from generating units based on the Rankine cycle, which typically achieved thermal efficiencies in the range 30--40 per cent. Now combined cycle gas turbine (CCGT) units utilising the latest heat recovery steam generator (HRSG) plant can achieve efficiencies of 50-60 per cent. The removal of European Community restrictions on burning gas for power generation, coupled with other factors, has resulted in increased deployment of CCGT units. However different current plant may now appear, the underlying principles of generation and distribution had been mastered by the end of the nineteenth century. Since then the evolution of power plant design has been largely incremental, driven mainly by new technology. The past three decades

2 Thermal power plant simulation and control

have witnessed the integration of microprocessor equipment into every aspect of generation and distribution. The next 20 years should see this technology develop further, bringing with it pseudo-intelligent applications which truly harness the rapidly expanding computational power available. New computer-based systems will increase plant automation, improve unit control and permit more flexible plant operation, while at the same time maximising unit efficiency and reducing harmful emissions.

New developments in plant design are continually being sought and investigated to improve unit performance. Currently integrated gasification combined cycle (IGCC) and advanced pressurised fluidised bed combustion (PFBC) are emerging technologies that are showing great potential for yielding high efficiency and low emissions. The short-to-medium term targets that have been mapped out by Vision 21 (US Depart- ment of Energy) for new plant designs are 60 per cent thermal efficiency for coal/solid fuels and 75 per cent efficiency for natural gas units, combined with zero or very low environmental emissions (DOE, 1999).

1.2 Plant configuration and design

Although there are many variations in power plant configuration and design, at the most basic of levels, a fossil fuel is combusted to raise steam, which then rotates a turbine that drives an alternator, to provide three-phase a.c. electricity at 50/60 Hz. Illustrated in Figure 1.1 is the turbine hall of a modern power station. At the heart

Figure 1.1 Premier Power turbine hall

Fu~

Advances in power plant technology

Turbines

3

Boiler

Air ~ Feedwater Boiler feed pump

Figure 1.2 Simplified power plant

of a conventional power plant is the boiler, which operates by following the thermodynamic Rankine steam cycle, a practical implementation of the ideal Carnot cycle. Figure 1.2 provides a simplified illustration of the steam flow path in a fossil fuel power plant.

1.2.1 Subcritical plant

In subcritical boilers steam temperatures and pressures never exceed the 'critical point' of steam which occurs at 373.9 C (705.1 F) and 22.1 MPa (220.6 bar). The steam cycle is conveniently analysed by beginning with the feedwater flow from the condenser. The condenser's hotwell maintains a large reservoir from which boiler feed pumps draw their supply. The temperature and pressure of the feedwater is raised by a series of low- and high-pressure feed heaters which draw heat from steam, bled from the turbines, thereby improving unit efficiency. The economiser is the last stage of heating prior to entering the drum.

The dual role of the drum is to supply feedwater to the walls of the furnace and to separate the resulting steam from incoming water. The temperature of the superheated steam leaving the drum is constrained only by the metallurgical limits of the pipework. In the turbine, normally consisting of multiple stages, the kinetic energy of the steam is converted to mechanical torque as the steam expands across the turbines. After the initial high-pressure stage the steam returns to the boiler for reheating. On exiting the low-pressure turbine the now saturated steam is condensed back into liquid by the cooling water in the condenser.

One of the advantages of operating in the subcritical region is that the differential density of water and steam, before and after the drum, permits 'natural circulation' of the feedwater around the boiler. If the flow becomes unbalanced, as a result of operating at elevated temperatures and pressures, then boiler feed pumps are required to provide the extra driving force. Hence, in unbalanced 'forced circulation' boilers the rating of the boiler feed pumps is significantly increased.


1.2.2 Supercritical plant

The world's first supercritical power plant began operating in 1957 and was com- mercially operated until 1979. This 125 MW installation at the Philo Plant operated at 31 MPa (310 bar) and 621 C or 1150 F (Smith, 1998). When operating a boiler in the supercritical region improvements can be made to both unit efficiency and heat-rate, due to the elevated temperatures and pressures. Currently, operating at state-of-the-art steam conditions a 3 per cent improvement in unit efficiency can be achieved, as compared with subcritical plant (Goidich, 2001). Supercritical boilers operate at pressures greater than 22 MPa (220 bar) and are also referred to as 'once- through' boilers, since the feedwater circulates only once through the boiler in each steam cycle. When operating at pressures above the critical point of steam there is no clear distinction between the vapour and liquid states, rather a fluid results, whose density can range from vapour-like to liquid-like. Consequently, a drum, normally required to separate the steam from the water, is eliminated, as shown in Figure 1.3.

Control in a supercritical boiler is somewhat different from that in a drum boiler. While for a supercritical boiler it is the boiler feed pump that determines the steam flow rate, this requirement is met by the fuel-firing rate for a drum boiler. Consequently, to control superheat steam temperatures a once-through boiler first adjusts the fuel-firing rate, as opposed to using spray water attemperation for a drum boiler (Goidich, 2001).

While supercritical plant should be more efficient than conventional drum plant, their development and deployment has been slow. Despite the reduction in operating costs resulting from higher unit efficiency, their increased installation cost can not often be justified over the life of the plant.

To turbine [[ , To turbine

~ P 'Drum' boiler

erfee .7. 'Once-through' boiler

Figure 1.3 Boiler steam flow paths

Advances in power plant technology 5

Exhaust, gases

Turbine Gas

ill

HRSG Flue

Air

Turbine

Alternator

tl Alternator

I I Steam

Boiler feed pump

L Condenser

.......................... 1~_.~

Cooling water

I Feed water

Figure 1.4 Simplified CCGT plant

1.2.3 Combined cycle gas turbines

Combined cycle gas turbines, as their name suggests, combine existing gas and steam technologies into one unit. They bring together the Rankine cycle from conventional steam plant and the Brayton cycle from gas turbine generators, yielding significant improvements in thermal efficiency over conventional steam plant. In both types of plant it is the inherent energy losses in the plant design that constrain their thermal efficiency. However, in a CCGT plant the thermal efficiency is extended to approximately 50-60 per cent, by piping the exhaust gas from the gas turbine into a heat recovery steam generator, operating on the Rankine cycle. In general, the heat recov- ered in this process is sufficient to drive a steam turbine with an electrical output of approximately 50 per cent of the gas turbine generator. A simplified multishaft CCGT plant is illustrated in Figure 1.4.

Alternatively, for single-shaft systems, the gas turbine and steam turbine are coupled to a single generator, in tandem. For startup, or 'open cycle' operation of the gas turbine alone, the steam turbine can be disconnected using a hydraulic clutch. In terms of overall investment a single-shaft system is typically about 5 per cent lower in cost, with its operating simplicity typically leading to higher reliability. Benefits of multishaft arrangements are shorter shafts, fewer stability problems and more degrees of freedom in the mechanical design.

1.3 Control and instrumentation

Modern power plant is a complex arrangement of pipework and machinery with a myriad of interacting control loops and support systems. However, it is the boiler


control system that is central in determining the overall behaviour of the generating unit. All the main control loops must respond to a central command structure, which sets their individual setpoints and controls the behaviour of the plant. It is the demand for steam that resides at the top of this control hierarchy. From this all other individual loop controllers receive their demand or setpoint signal. Due to its importance, the steam demand signal is often known as the master control signal.

The strategic behaviour of the unit is governed by various boiler control configurations, and the behaviour of the master control signal within these arrangements is now discussed.

Boiler following mode Boiler following or 'constant pressure' mode utilises the main steam governor as a fast-acting load controller, since opening the governor valves, and releasing the stored energy in the boiler, meets short-term increases in electrical demand. Conversely, closing the governor valves reduces the generated output. These actions alter the main steam pressure, so it is the role of the master pressure controller to suitably adjust the fuel-firing rate. Operating a unit in this mode does, however, contain inefficiencies as throttling of the governor valves reduces the available steam flow, creating energy losses.

Turbine following mode A generating unit may alternatively be configured to operate in turbine following mode, whereby the combustion controls of the boiler are set to achieve a fixed output. The position of the main steam governor valve is controlled by the valve outlet pressure, not the input as in boiler following. Consequently, such units can be operated with their governor valves remaining fully open.

Turbine following mode is preferred for thermal base load and nuclear plant, since it allows the generating units to operate continuously at their maximum capacity rating. However, such units do not respond to frequency deviations and so cannot assist in a network frequency support role. For nuclear plant, there are also safety benefits in providing continuous steady state operating conditions.

Sliding pressure mode Although boiler following mode is commonly used, sliding pressure mode is an 'instructive' development, where the constant steam pressure is replaced by a variable steam pressure mode. The reduced throttling- back action by the governor control valves, at lower outputs, leads to improved unit efficiency. Variable pressure operation also provides faster unit loading, and enables operation of the turbines at lower temperatures and pressures. However, the ability to use the stored energy of the boiler to meet short-term changes in demand is restricted. For safety reasons, fast-responding, electrically operated safety valves are essential for variable pressure operation to protect against sud- den, dangerous increases in steam pressure that may occur while the pressure setpoint is low.

1.3.1 Combustion control

Burning a fossil fuel releases energy in the form of heat, which is absorbed by the feedwater through convection and radiation mechanisms. Controlling the volume of heat released when burning large quantities of fossil fuel is a demanding and


potentially dangerous problem, which is very much dependent on the fuel being burned- a coal-fired boiler being significantly different from that of an oil- or gas-fired boiler.

The fundamental problem of combustion control is to adjust the fuel and air flow rates to match the energy demand of the steam leaving the boiler. The ideal or 'stoichiometric' ratio for complete combustion of the fuel is impracti- cal and results in incomplete combustion due to unavoidable imperfections in the mixing of fuel and air. Excess air is always necessary in a real plant and can be as high as 10 per cent above the 'stoichiometric' ratio to achieve complete combustion. Without sufficient air flow to the furnace, incomplete combustion results in the formation of black smoke, poisonous carbon monoxide and the danger of unburnt fuel accumulating within the boiler. In contrast, excess air may generate unwanted NOx and SOx emissions and reduce the efficiency of the boiler by carrying useful heat out the chimney, as well as increasing ID/FD fan requirements. The continuous flow of air to the boiler furnace is achieved using forced draft (FD) fans to force air into the furnace and induced draft (ID) fans to extract the combustion gases. The internal draft pressure (furnace pressure) is maintained just below atmospheric pressure to prevent hazardous gases from escaping through observation portholes, soot-blower openings and other orifices in the furnace. The natural ingress of air through these openings is referred to as 'tramp air'.

Overseeing the combustion process is the burner management system (BMS) which regulates the extremely hazardous process of firing the fuel. To ensure safety, numerous sensors supply data on current operating conditions. These include a UV flame detector, and furnace pressure, air flow, oxygen, NOx and CO sensors, to name but a few.

1.3.2 Boiler control subsystems

Boiler control systems exist in a hierarchial arrangement. As previously stated, resid- ing at the top is the master control signal, which determines the steam load for the unit. Control systems or loop controllers at lower levels derive their demand or setpoint values from the master controller. A short list of other (boiler) control loops is as follows:

Coal pnlveriser control regulates the supply ofpulverised fuel to the boiler from coal mills. The time delay between coal entering the mill and reaching the boiler, along with the startup time of additional mills, is often a limiting factor when the unit is required to respond quickly.

Drum level control is closely linked with feedwater control. The 'swell' and 'shrinkage' effects, resulting from changes in steam demand, are confusing to simple single-element controllers. Typically, a 'three-element' controller is used, which combines drum level, steam flow and feedwater flow signals.

Steam temperature control regulates the temperature of the steam exiting the boiler after the superheater and reheater stages. The long time delays associated with these loops make for challenging control.


In addition to the control systems previously described there are many others that are essential for operation: generator excitation, burner angle, cooling water flow rates, LP/HP feed heating, flue gas recycling, etc.

1.3.3 Plant instrumentation

More than any other aspect of power generation, control and instrumentation equipment has changed unrecognisably in the last hundred years. When power stations were first constructed in the 1880s control was typified by the steam governor, where a simple mechanical flywheel with rotating weights was connected to a hydraulic system through a series of sliding linkages and springs. Since then, pneumatic and then analogue electrical equipment have been introduced for general plant control. How- ever, the radical transformation in control came with the advent of the microprocessor, leading to stand-alone devices being adopted for individual loop control in the 1970s and early 1980s. The microprocessor permitted new and innovative control solutions to be considered, so as processing power advanced, system functionality grew. Today these isolated control systems have evolved into distributed control systems (DCSs), with the capability to control entire power stations. Although distributed control systems are used primarily for loop control their processing power and flexibility has allowed them to handle many other data management applications.

1.3.4 Distributed control systems

Over the course of the last two decades, distributed control systems have become the domain of large industrial processes and power plants. Indeed, it is their ability to handle control on large-scale systems that distinguishes them from their smaller programmable logic controller (PLC) and PC-based counterparts. The dichotomy between high-end PLC systems and DCS installations is, however, uncertain as both have similar functionality and network topologies. The distinguishing features of a DCS can be summarised as:

Size - capacity to handle many tens of thousands of signals. Centralised administration - complete control of distributed units from one single node on the network. Data management - ability to handle and store tens of thousands of data points in real time.

A simplified generic DCS network is illustrated in Figure 1.5. At its core, the DCS has a dual redundant, bidirectional, high-speed communications network, which facilitates the transfer of vast amounts of data between nodes. Connected to a typical network, four distinct types of device can be identified, each with unique functionality. Arranged in hierarchial order they are:

The engineering workstation provides complete control over the DCS. Typical tasks may involve programming the distributed control units and adding/removing spare I/O capacity to the network.


Plant

*~put /output

0 f f ice~etwork ~ ~ ~

'~ Data ~[ l~ l I I~ l

:::::q Distributed control units ~ " atwaomr=t Operator workstation

Figure 1.5 Simplified distributed control system

The data management workstation is usually assigned the task of managing the process database containing all the process data points or 'tags' on the DCS. Time stamping new data as it arrives on the network may also be performed. The operator workstation provides high-resolution mimics of the plant, allowing the plant operator to control the unit using a human-machine interface (HMI). Distributed control units are responsible for implementing plant control. They are directly connected to plant signals and can usually operate independently of the rest of the DCS. Like most other parts of the DCS, dual or triple redundancy is employed to ensure availability of control equipment at all times.

1.4 External influences

The environment in which power stations operate has undergone a radical shakeup over the last two decades and still remains in a state of flux. No longer can a station be operated in isolation, where unit efficiency and good engineering practice are the main considerations. Managing a station today involves juggling a myriad of conflicting external factors. On one side there may be shareholders antic- ipating a profitable return on their investment, and on the other, environmental legislation forcing the procurement of emissions reducing plant and equipment. As already suggested the two areas which have had the most significant influence on station management are liberalisation of the energy markets and environmental legislation.


1.4.1 Power system deregulation

During the years between the end of World War II and the 1970s, utility management in the United States and Europe focused on the major task of building new power plant and improving the transmission and distribution grids to meet the demands of rapidly growing economies. The only competition that existed was between individual concerns trying to install the largest generating unit of the day or the most thermally efficient unit. Beyond this, utility managers remained cooperative with their colleagues. The business and technological strategies they employed were all very similar and governed mainly at national level. The level of cooperation extended to national research and development organisations, for example, the Central Elec- tricity Generating Board (CEGB) in the United Kingdom and the Electric Power Research Institute (EPRI) in the United States. These organisations engaged in col- laborative research and openly shared their findings, as few secrets existed among their members.

The United States was the first to truly witness the 'winds of change' for the regulated utilities. Growing discontent from the general public was fuelled by the continued price increases in electricity, the spiralling costs of large generation plant and widespread fears about nuclear generation. In an attempt to placate the public the United States introduced the Public Utilities Regulatory Policies Act in 1978 to allow unregulated generators to supply the grid. By doing so it was hoped that non- conventional and independent sources of power would appear. These independent power producers (IPPs) were not allowed to sell to end users but it was mandatory for local regulated utilities to purchase their generated output. This measure proved sufficiently successful that by 1993 some 50 per cent of new generating capacity in the United States was being constructed by IPPs.

These changes challenged the long-held belief that electrical generation and distribution was a natural monopoly. With this realisation the next step was to open up the market whereby customers could, in theory, benefit from increased competition.

In the United Kingdom, the Electricity Act of 1989 legislated for the breaking up of the nationalised CEGB industry into smaller privately owned companies. A 'pool' system was introduced, where generators competed against each other for contracts to generate electricity (Hunt and Shuttleworth, 1997). The new legislation separated the product (generated electricity) from the transportation medium (the transmission grid). In doing so, costs could be unbundled into an 'energy' and a 'delivery' component. These arrangements were replaced in March 2001 by the New Electricity Trading Arrangements (NETA) in an attempt to facilitate greater market freedom for generators and suppliers, particularly in the wholesale market. Plans exist to extend these arrangement to Scotland by creating the British-wide Electricity Trading and Transmission Arrangements (BETTA).

The United States has gradually been moving towards increased competition. Here, privatisation is not considered an issue, as the majority of electricity companies are investor-owned utilities that are territory based. In 1992 the Energy Policy Act permitted wholesale customers the choice of supplier, and obligated the relevant utilities to transmit power across their networks. Their restructuring model resembles that which was implemented in the United Kingdom.


In Europe, the European Commission (EC) is similarly endeavouring to liberate the electricity markets of its 15 member states. Amendments to Directive 96/92/EC on March 2001 committed member states to be fully open to competition by January 2005. Unfortunately, loopholes in the legislation allow individual countries to opt out entirely or comply in a piecemeal fashion. As of February 2000, approximately 60 per cent of EU customers have a choice of electricity supplier (Lamoureux, 2001). The two member states who led the way, and have already taken action to deregulate their electricity industries, are Belgium and the United Kingdom.

1.4.2 Environmental factors

From an environmental perspective, burning fossil fuel releases undesirable and harmful emissions into the atmosphere - carbon dioxide (CO2), carbon monoxide (CO), sulphur dioxide (SO2), nitrogen dioxide (NO2), nitric oxide (NO), etc. In addition to the above, particulate emissions may be produced, especially when burning coal and heavy oil.

Oxides of sulphur (SOx) are formed when the fossil fuel itself contains sulphur. Coal burning is the single largest man-made source of sulphur dioxide, accounting for almost 50 per cent of annual global emissions, with oil burning accounting for a further 25-30 per cent. The simplest approach to reduce SOx emissions is to burn fuel with a low sulphur content. Hence, the current popularity in burning natural gas and low sulphur oil and coal in conventional power stations. Alternatively, flue gas desulphurisation plant 'scrubbing' is available, but expensive.

Oxides of nitrogen (NOx) are formed during high-temperature burning of fossil fuel. Nitric oxide (NO) and nitrogen dioxide (NO2) are formed when nitrogen from either the fuel or air supply, and oxygen from the air, combine. Minimising NOx formation requires correct design of the furnace bumer, and optimised boiler control. However, the higher the combustion efficiency the higher the formation of nitric oxide. Nitrogen dioxide also contributes to the formation of ground-level ozone (O3) when mixed with volatile organic compounds (VOCs) in a sunlight-initiated oxida- tion process. In addition, both NOx and SOx, when absorbed into the atmosphere, slightly increase the acidity of the precipitation that falls to earth. Hence, the term acid rain.

Carbon dioxide and ozone (formed from NO2) are classed among the greenhouse gases (GHGs) that are contributing to global warming. Recent attempts at environmental legislation, in particular the Kyoto Protocol, have focused on reducing emissions by setting stringent targets. The treaty specifies that industrial countries have until 2010 to reduce their GHG emissions by particular percentages below 1990 levels. The European Union committed to cutting its emissions to 92 per cent of its 1990 level, and the United States to 93 per cent. During this time it was assumed that emissions for industrial countries would substantially increase, without any restrictions in place. James Markowsky, on behalf of the American Society of Mechanical Engi- neers (ASME), at a luncheon briefing on Capitol Hill, Washington, 1999, pointed out that to meet that goal the United States would have to retire most of its coal-burning plant, which then currently generated approximately 56 per cent of the country's


500+

400

e~ o

~ 300

200

g~ lOO

North South and Europe Former Africa Asia America Cent. America Soviet Union Pacific

Figure 1.6 Fossil fuel reserves as of 2001

electric power (ASME, 1998). Given today's technology, the only way to generate the power the United States requires and still meet the emissions standard is to burn gas.

Unfortunately, burning gas is not a sustainable long-term solution. From Figure 1.6, based on data from the BP Statistical Review Of World Energy (BP, 2001), it can be seen that proven reserves of gas are limited. If the current rate of consumption of gas continues, world reserves will be severely depleted within 40-50 years. The total reserves for coal, however, are substantially larger than those for oil and natural gas combined. This would suggest that future technology may focus on 'clean coal' plant, as viable alternatives to fossil fuel are somewhat limited. IGCC plant typify emerging technology aimed at combating the emissions problem associated with coal. Here, the fuel is gasified and cleaned, before being burnt in a conventional combined cycle plant. Gasification technology has also been combined with fluidised bed designs. In a typical demonstration plant, situated in the city of Lakeland, Florida, a carboniser receives a mixture of limestone, to absorb sulphur compounds, and dried coal. The coal is partially gasified to produce syngas and char/limestone residue. The latter is sent to a pressurised circulating fluidised bed, where it joins a stream of crushed fresh coal which is burned in the boiler furnace (DOE, 2001a,b). Gasification and gas reforming, i.e. the production and separation of gas into carbon monoxide and hydrogen, appear to be the most promising technologies at present (DOE, 1999). The former produces a gas stream that can be burned for electric power, while the latter offers a source of hydrogen for a fuel cell or chemical process.

Renewable sources (biomass, wind, hydro, tidal, solar, etc.) have been presented as part of any future solution to energy needs. In November 1997 the European Commission set itself a target of doubling renewable energy supply from 6 to 12 per cent by 2010. Similarly, the United Kingdom, for example, established a target of 10 per cent renewable generation by 2010. Indeed, the UK Cabinet Office Performance and Innovation Unit proposed a target of 20 per cent renewables


by 2020 in March 2002. More recently at the UN World Summit on Sustainable Development, in Johannesburg, September 2002, a pledge was made to increase 'substantially' the use of renewable energy in global energy consumption. World- wide, the United Nation's Intergovernmental Panel on Climate Change (IPCC) postulated a 'coal intensive' scenario with renewables contributing 65 per cent of the primary energy by 2100 (IPCC, 1996). However, on shorter time-scales, it may remain for nuclear fission (or perhaps someday, fusion) to meet growing energy needs (VGB, 2001). It is also worth noting that electricity consumption is projected to grow by 75 per cent relative to 1999 figures (DOE, 2002) by 2020.

1.5 Plant technology developments

Many power stations view the DCS as a direct replacement for older stand-alone analogue or digital controllers. Hence, the control systems used in the DCS are often simply a copy of what had been used in the past. In most cases loop control is implemented using single-input single-output (SISO) linear structures in the form of PI or PID controllers. For sequence control, the DCS provides an abundance of logical function blocks (AND, OR, XOR, etc), with programming software allowing these to be tied together to create multilevel control programs. However, from the vast amount of real-time data available only a small proportion is typically used, usually to alarm the operator of plant faults and occasionally to drive simple data trending for fault finding or management summary reports. Minimal advantage is taken of the high-speed communication network for plant-wide control schemes or supervisory layers.

An enlightened view of distributed control systems, however, reveals that the constraints of former mechanical and analogue solutions are gone. A new vista is opening in power plant control and management, with novel and innovative approaches being given consideration. It is now possible to implement non-linear multiple-input multiple-output (MIMO) model-based control, coordinated plant control (trajectory following and optimisation) or pseudo-intelligence in the form of expert systems, artificial neural networks (ANN), data mining and genetic algorithms for supervisory control. These new technologies are embracing all aspects of power plant operation, from intelligent maintenance, environmental protection, data management systems, intelligent alarm management, fault diagnostics, productivity management, purchas- ing and accounting. The potential list of applications is virtually endless (DOE, 2001 a; Oluwande, 2001; Lausterer, 2000).

1.6 References

ASME: 'Technology implications for the US of the Kyoto protocol carbon emission goals'. ASME general position paper, ASME, December 1998

BP: 'BP statistical review of world energy' 50th edition, London, June 2001 DOE: 'Vision 21 program plan'. US Department of Energy, Federal Energy

Technology Center, April 1999


DOE: 'Environmental benefits of clean coal technologies'. US Department of Energy, Topical Report Number 18, April 2001a

DOE: 'Software systems in clean coal demonstration projects'. US Department of Energy, Topical Report Number 17, December 2001b

DOE: 'International energy outlook 2002'. Energy Information Administration, US Department of Energy, March 2002

DTI: 'Innovative supercritical boilers for near-term global markets'. Department of Trade and Industry, London, United Kingdom, Pub. URN 00/1138, September 2000

GOIDICH, S.: 'Efficient power operational flexibility: The once-through supercritical boiler'. Foster Wheeler Review, Autumn 2001, Foster Wheeler Energy Corporation, pp. 11-14

HUNT, S. and SHUTTLEWORTH, G.: 'Competition and choice in electricity' (John Wiley, Chichester, 1997)

IPCC: 'Working group II to the second assessment report, intergovemmental panel on climate change, climate change 1995: impacts, adaptations and mitigation of climate change' (Cambridge University Press, 1996)

LAMOUREUX, M.A.: 'Evolution of electric utility restructuring in the UK', 1EEE Power Engineering Review, June 2001, pp. 3-5

LAUSTERER, G.K.: 'Knowledge-based power plant management - the impact of deregulation on it solutions', lEE Control, Proceedings of lEE Control 2000 Conference, Cambridge UK, September 2000, pp. 1-8

OLUWANDE, G.A.: 'Exploitation of advanced control techniques in power generation', Computing and Control Engineering Journal, April 2001, pp. 63-67

SMITH, J.W.: 'Supercritical (once through) boiler technology' (Babcock & Wilcox, Barberton, Ohio, US 1998, BR- 1658)

VGB: 'Research for a sustainable energy supply - recommendations of the Scientific Advisory Board of VGB PowerTech e.V.'. July 2001

Part 1

Modelling and simulation

Chapter 2

Modelling of power plants

A. Leva and C. Maffezzoni

2.1 Introduction

Modelling power plant processes may be approached from different points of view, depending on the purpose for which the model is intended. Here, we shall restrict the presentation to the case (most interesting for engineering) where the model is built to allow system simulation over a rather wide range of operation (non-linear model) and is based on first principles and design data. This specification naturally leads to a model structuring approach based on the representation of plant components and of their interconnections, with evidence given to variables and parameters corresponding to well-defined measurements or physical entities. Possible experimental data are, generally, not used for system identification but for model validation, which may also include some model tuning. The models here are referred to as dynamic, that is, they are able to predict transient responses, even for large process variations. Since power plant dynamics operate on a range of time scales, it is advisable to focus on the use of a dynamic model over a defined horizon. For simulation models representing an entire power plant or a large subsystem, it is quite common to seek model accuracy over an intermediate time-scale, i.e. in the range of a few tenths up to a few thousands of a second. This will be the implicit assumption in the description of the basic models. Finally, we shall limit the scope of this chapter to power plants based on the firing of a fossil fuel, i.e. conventional thermal and gas turbine plant, possibly equipped with heat recovery boilers.

There is a long track record of research and engineering effort in this area, dat- ing back to the pioneering work of Chien et al. (1958), passes through the earlier engineering-oriented works of Caseau et al. (1970), Weber et al. (1976), Modular Modelling System (1983), Lausterer et al. (1984), Maffezzoni et al. (1984), and leads to presently available simulation codes (APROS; ProTRAX; Cori et al., 1989; SIMCON-X, 1994a,b). With reference to the survey papers of Carpanzano et al.


(1999) and Maffezzoni (1992), the objective here is to review the important knowledge (concerning both methods and applications) accumulated along that track and to transfer it to the unifying framework of object-oriented modelling, which appears to be most effective in dealing with real-size engineering problems and in sharing modelling knowledge among diverse users.

So, the chapter is organised as follows: first the basic concepts of object-oriented modelling are introduced with reference to the typical structures met in thermal power plant, then a review is presented of basic models for typical power plant components. Subsequently, the task of defining a realistic model of the plant distributed control system (DCS) is investigated; some remarks about the application of dynamic decoupling and methods of model validation are then reported.

2.2 Model structuring by the object-oriented approach

2.2.1 Foreword

Object-oriented modelling (OOM) is a widely accepted technique which has already produced both modelling languages (Mattsson and Andersson, 1992; Maffezzoni and Girelli, 1998; Elmqvist et al., 1999) and software packages (Piela et al., 1991; Elmqvist et al., 1993; gPROMS, 1998). The approach is based on a number of paradigms, among which a fundamental role is certainly played by the following:

The definition of physical ports (also referred to as terminals) as the standard interface to connect a certain component model, in order to reproduce the structure of the physical system.

The definition of models in a non-causal form permitting reuse, abstraction and unconditional connection.

The mutual independence of the model interface (the physical ports) and its internal description.

State of the art OOM is well represented by the development of the Modelica project (1999), a recent international effort to define a standard modelling language. At present there are well-assessed methods to treat lumped parameter components (LPC) while, on the contrary, there are no unified solutions to describe distributed parameter components (DPC), which are quite important in power plant modelling. As such, in the remainder of the chapter we shall adopt the Modelica language as the formalism for writing the described models (whose specification manual is available at www.modelica.org) with the minimum extension required to cope with DPC's, for example heat exchangers, according to the approach proposed by Aime and Maffezzoni (2000).

2.2.2 Classification of plant components and of physical ports

From the point of view of model structuring, it is convenient to look at the typical layout of a fossil-fired power plant (Maffezzoni and Kwatny, 1999). In the case of a classical Rankine-cycle unit, the principal subsystems are the steam generator

Modelling of power plants 19

(or boiler), the steam turbine, the condensed water cycle and the electrical subsystem. The structure of the power station's electrical subsystem is not relevant to the principal characteristics of a power unit, so in the following the electrical subsystem will be drastically simplified by considering only the electromechanical balance of the alternator.

Modelling power units by aggregating component models is very convenient because it reflects the physical plant layout and enhances reuse of modelling software. Plant components may be classified first by looking at the subsystem they belong to, then considering the nature of the process transformations that they implement. Structuring by modules is, to a certain extent, a matter of choice: defining large modules implies a simpler aggregation structure, while defining small modules implies a larger reuse when the plant structure changes. Here, the 'size' of basic modules is chosen according to the best practice employed in engineering dynamic simulators: this choice is compatible with packages like APROS (www.vtt.fl) and LEGO (Coil et al., 1989), widely used for power plant simulators, and DYMOLA (www.dynasim.se), appropriate for implementing an object-oriented approach to physical system modelling.

2.2.2.1 Boiler components The boiler is the most complex subsystem, and it can be split into a pair of interacting circuits: the water-steam circuit and the air-gas circuit. There are components which take part in one circuit only (one-side components, OSC) like pumps, valves, headers, etc., while there are components, namely heat exchangers, which take part in both circuits (two-side components, TSC), being devoted to heat transfer from the combustion gas to water and steam. To define the structure of an OSC it is convenient to introduce a physical port through which a component may interact with another, i.e. the thermo- hydraulic terminal (THT), which is specified through the following Modelica script:

type Pressure = Rea l (quant i ty = "Pressure" , d i sp layUn i t = "Pa", un i t = "Pa") ;

type MassF lowRate = Rea l (q~ant i ty = "MassF lowRate" , d i sp layUn i t = "kg/s" , un i t = "kg /s" ) ;

type Entha lpy = Rea l (quant i ty = "Entha lpy" , d i sp layUn i t = " J /kg" , un i t = " J /kg" ) ;

connector THT

Pressure p;

f l ow MassF lowRate w;

Entha lpy h;

end THT ;

The Modelica language (like many others) allows type definitions to enhance the clarity of the simulation code. This is illustrated in the previous script by defining Pressure, MassFlowRate and Enthalpy. Throughout this work we shall assume that all the required types are defined in this way, hence we shall refer to types like Quality, AngularVelocity, and so forth. We do not report their definitions because of space limitations and because they can be immediately deduced from those given. Note also that Modelica has an extensive library of predefined types. For instance, the process scheme of Figure 2.1 a suggests the model structure of Figure 2. lb.

The connection between the pump THT Out (outlet) and the valve THT In (inlet) means that the pressure and enthalpy at the pump outlet coincide with the pressure


a< ut

Figure 2.1 Simple process scheme and its model structure

Gas

Figure 2.2 Typical heat exchanger configuration

and enthalpy at the valve inlet, respectively, while the mass flow rates at the pump outlet and valve inlet sum to zero. Pressure and enthalpy are effort variables while mass flow rate is a flow variable.

The interaction between a couple of neighbouring OSCs can always be modelled by the direct connection of two THTs. The Modelica script stating the aggregation of Figure 2.1 is as follows:

connect (PUMP. Out, VALVE. In) ;

where, of course, PUMP and VALVE are instances of suitable elementary models stored in some library. The THT can be used both for OSCs belonging to the steam- water circuit and to the air-gas circuit. Considerably more complex however, is TSC structuring, because in this case we need to model, in a modular way, the variation in heat transfer configurations of boiler tubular heat exchangers. A typical situation is depicted in Figure 2.2, where gas flowing through the boiler back-pass exchanges heat with the principal bank (A) disposed in cross-flow and with an enclosure panel (B) disposed in long-flow.

The model of a complex heat exchanging system like that of Figure 2.2 could be structured according to the very general approach proposed by Aime and Maffezzoni (2000), where spatially distributed heat transfer configurations are introduced. More pragmatically, one can exploit two common properties of boiler heat exchangers:

Gas flows have negligible storage with respect to metal wall and steam-water flows.


Figure 2.3 Modular structure for heat exchanging system

The gas ducts may be split into a cascade of gas zones, where the gas temperature can be assumed to be almost uniform or linearly varying.

Zone splitting is guided by the structure of the tube bank: typically a gas zone extends to include one row or a few rows of tubes. The model structure corresponding to the situation of Figure 2.2 is sketched in Figure 2.3.

The connection between a bank and a gas zone takes place through a distributed heat transfer terminal (dHT), which consists of the bank walls' temperature profile Tw(x, t) and the heat flux profile ~0(x, t) released to the wall. Tw(x, t) and qg(x, t) are functions of time t and of the banks' tube abscissa x. A typical finite-element discretisation of Tw(x, t) and ~0(x, t) replaces such functions with their interpolating approximations obtained from two vectors of nodal temperatures and fluxes. Thus, in simulation code, Tw(x, t) and ~0(x, t) are represented by two vectors of suitable dimensions, denoted in the following by Tw (t) and (t), respectively.

In Figure 2.3, it has been assumed that the transfer of energy from one gas zone to the adjacent one is solely due to mass transfer, as implicitly established by the connection of two THTs. There are, however, high-temperature gas volumes (either in the furnace or in other parts of the back-pass) where radiation heat transfer from one gas zone to those adjacent is not negligible. This requires the introduction of a specific port that extends the THT to permit transfer of heat independent of mass flow. In principle, at a boundary surface between two fluid volumes, we may have transfer of energy by convection (expressed by the group wh), radiation and/or diffusion; mechanical work is already included in the product wh. We may call this physical


Furnace ~ Zones ~ Zones ~ Furnace zone inter- inter- zone i- 1 action action i + 1

Furnace zone

i

I I .T r Gas wall interaction

I [aHT] Furnace wall

tubes

Figure 2.4 Interactions between furnace zones and walls

port the thermo-hydraulic and heat transfer terminal (THHT), which will consist of the following variables:

the mass flow-rate w, enthalpy h, pressure p and heat rate Q at the interface between the gas volumes;

the radiation temperature Tr (e.g. the flame temperature) of the lumped gas volume.

For instance, a model of the furnace zone, where there are neither burners nor sec- ondary air inputs, can be structured as shown in Figure 2.4, where the heat transfer terminal (HT) consists of the following two variables:

Qw, the heat rate to the furnace wall Tr, the radiation temperature of the gas zone.

Terminals dHT, THHT and HT are defined in the Modelica language as follows:

connector parameter

flow

end dHT; connector

extends

f low end THHT; connector

flow end HT;

dHT Integer vectorSize=l; Temperature tempProfi le[vectorSize] ; HeatFlux heatFluxProfi le[vectorSize] ;

THHT THT; Temperature Tr; HeatFlow Q;

HT Temperature Tr; HeatFlow Qw;


Note the distinction between He a t F lux (W/m 2) and He a t F 1 ow (W). These scripts employ the Modelica extends clause, which allows us to define a model or a connector by adding elements to a previously defined one. Such clauses are available in any object-oriented modelling language.

It should be noted that the situation of Figure 2.3 is a particular case of that depicted in Figure 2.4: in Figure 2.3 the interaction between two adjacent zones is very simple (there is no heat transfer besides convection) so that zone interaction becomes trivial and is omitted; the gas-wall interaction is summarised by a heat transfer coefficient, directly incorporated in the gas zone.

To fully understand the role of the model ZONES INTERACTION, we provide here a possible implementation:

WL + WR = 0 (2.1)

PL -- PR = 0 (2.2)

hL -- hR = 0 (2.3)

QL + QR = 0 (2.4)

QL = K(T 4 - T 4) (2.5)

where the subscripts 'L' and 'R' denote variables belonging to the left and fight terminals, respectively, and K is the radiative heat transfer coefficient.

2.2.2.2 Steam turbine components

For the purpose of power plant simulation, turbines are generally modelled as lumped parameters. When accurate modelling is required, it is usual to split a turbine into a number of cascaded sections, a section being in turn composed of a number of cascaded stages. The extension of a section is typically dictated by some physical discontinuity along the steam expansion, such as the presence of steam extraction or a change in the stage design (e.g. when passing from impulse to reaction stages).

Interaction between a turbine section and other components at its boundary may simply be modelled by THTs as shown in Figure 2.5, while a mechanical terminal (MT) is needed to represent the power transfer to the shaft.

The MT consists of the two variables w (angular speed) and r (torque). The same approach can be used for gas turbines, where the situation is even simpler because there are no present physical discontinuities. The MT is represented in Modelica as follows:

connector MT Angu larSpeed omega ;

f low Torque tau; end MT ;

2.2.2.3 Condensate cycle components

The condensate cycle is composed simple compact components like valves, pumps, headers, etc. and of more complex heat exchangers and/or storage tanks. Model


Steam I extract ion

I

Turbine Turbine Turbine section IH[~HI section [H[~I Header section

I ]

Shaft

I

Alternator

Turbine section

~-- - - -~- - ] _

I IMTI-

Figure 2.5 Interactions between turbine sections and boundary components

Figure 2.6

I Low pressure I

turbine

I _

Cling ~ I THT ' ~ Cling water Condenser water

discharge pump piping ~ ~ _

I

I T.TI I Extraction

pump i

Model structuring example for the condensate cycle

structuring is quite obvious for the former components, whereas non-trivial questions arise for complex heat exchangers such as the condenser, deaerator, and low-pressure and high-pressure heaters. There are two possible approaches: to build the heat exchanger model as the aggregate of simpler objects or to directly define the heat exchanger as an elementary (indivisible) component. Since the internal structure of the condensate heat exchangers is quite complex while the design of such large components is highly repetitive, it is advisable to follow the latter approach. In this case, physical ports between cycle components are still THTs. An example of structuring is given in Figure 2.6.


I Fuel I storage

I ] THT ] ~ Valve position Fuel valve

I

~ ~ 1 THT 1 ~ T~Exhaust Atmosphere Compressor Combustion Gas turbine header chamber

I I I MT I Shat~ I MT I

Figure 2.7 Model with an input control terminal

2.2.2.4 Gas turbine components

Gas turbines have become increasingly important in power generation, because of the outstanding efficiency achievable by combined cycle plants. The essential components are:

the gas turbine the compressor the combustion chamber.

Model structuring is generally simple, as sketched in Figure 2.7, where ICT denotes an input control terminal, that is a control port where a command signal is issued.

Internal modelling of basic components is, however, a non-trivial task, especially because it is often very important not only to predict power release of the turbo- alternator but also concentration of pollutants to the atmosphere. This is discussed in the next section.

2.2.3 Aggregation of submodels

Reuse of models for different case studies is enhanced by modularity, structuring of elementary models as non-causal systems and standardisation of physical ports (or terminals). However, it is common practice to repeat plant or subsystem designs from one power unit to another. For instance, we may store in a library the model of an economiser, which may result from aggregation of a block scheme like that of Figure 2.3. Of course, by aggregating model objects internal physical terminals disappear (they saturate with one another) so that the global model of the economiser (including its enclosure) may look as in Figure 2.8.

Reusing an aggregate model implies that the model structure and model equations are not accessible to the user (they cannot be changed when using the aggregate); what are specific to a given instance of the aggregate model are the model


Connection to enclosure tubes

Connection to gas circuit

Connection to high-pressure

condensate circuit

Economiser

Connection to enclosure tubes

~T ~ Connection to gas circuit

~T HT Connection to steam drum

Figure 2.8 Aggregation of model objects (economiser)

parameters which must be transferred from the internal submodels to the resulting aggregate.

2.2.4 Internal model description

The typical internal structure of a simple (non-aggregate) model using the Modelica language may be as follows:

model SIMPLEMODEL parameter ConnectorType VariableType

equation

end SIMPLEMODEL;

ParamType paramName; connectorName; variableName;

When an aggregate model AGGREGATEMODEL is obtained by composing two or more simple models (S IMPLEMODEL1, S I MPLEMODEL2 in the example), it can be defined by a Modelica script of the form:

model AGGREGATEMODEL SIMPLEMODEL SIMPLEMODELI (paramName=paramValuel); SIMPLEMODEL SIMPLEMODEL2 (paramName=paramValue2); connect(SIMPLEMODELl.connectorName,SIMPLEMODEL2.connectorName);

end AGGREGATEMODEL;

It is worth noting that the section equation in the format of the simple model is generally constituted by switching differential-algebraic equations (DAEs), i.e. alternative sets of DAEs that describe the system dynamics under different conditions. (Maffezzoni and Girelli, 1998; Maffezzoni etal., 1999). Moreover, both for the DAEs and the logical conditions, some quantities may be evaluated by referring to functions


of one or more variables, including look-up tables. Partial differential equations may be treated by finite element or finite difference approximations written as implicit matrix equations (Quarteroni and Valli, 1997). All these constructs are compatible with the Modelica language (and with several other modelling formats.)

2.3 Basic component models

2.3.1 The evaluation of steam and other fluid physical properties

2.3.1.1 Steam properties The large majority of models considered in this chapter require that various thermodynamic properties are evaluated starting from a couple of state variables. It is quite common that the computation of a global power plant model requires many thousands of steam property evaluations at each time step. So, an efficient and accurate treatment of water-steam properties is crucial for power plant simulation. Moreover, it is generally required that one or more properties be evaluated from at least three different couples of entry variables, typically (p, h), (p, S) and (p, T), where p is the pressure, h enthalpy, S entropy and T temperature. Finally, for dynamic modelling, not only standard thermodynamic quantities, like enthalpy, entropy, pressure and density are needed, but also viscosity, con- ductivity and thermodynamic partial derivatives (in particular specific heats at constant pressure Cp and constant volume Cv) or line derivatives along the saturation curve .

Water-steam properties can be computed using steam tables, based on (very complex) empirical formulas (Properties of Water and Steam, 1989), from which the required derivatives can be obtained by symbolic manipulation. These formulas, however, cannot be used for modelling in their current form, because entry variables are not those needed, formulas are very complex and, being non-linear, they should be used in a Newton-like algorithm to determine even a single property.

Therefore, the sole practical approach is to build a suitable grid in the thermodynamic plane, with a convenient number and disposition of nodes with respect to the saturation curve, so as to allow smooth and accurate approximation over the whole plane. Then, the 'model of the steam-water fluid' is constituted by large look-up tables with the required entry variables, yielding either a single property or a vector of properties, depending on the application. Within the object-oriented modelling environment, the s team- tab les may be treated either as a set of functions or as a simple model to be incorporated by any component model, when required.

In this work the first option has been chosen. The steam tables are implemented as a set of functions receiving two parameters. In the case of the full tables these are the (p, h), (p, S) or (p, T) couple; in the saturation tables these are the pressure and a two-valued parameter stating whether the liquid or vapour properties are required. This can be a boolean parameter, although in the examples presented herein its value


is represented by the strings l iqu id and vapour for better clarity. There is one function for each property and for each input couple. For example, the functions

SteamPHtablesEnt ropy (p, h) SteamPStablesDdensityDpressure (p, S ) SteamSATtablesDensity (p, "vapour" ) SteamSATtablesDenthalpyDpressure (p, "liquid" )

return the steam entropy at pressure p and enthalpy h, the partial derivative of the steam density with respect to pressure at pressure p and entropy S, the saturated vapour density at pressure p and the derivative of the saturated liquid enthalpy with respect to pressure computed along the saturation curve at pressure p, respectively. Several functions like these are used in the equations of the models that will be introduced from now on, and the syntax should be self-explanatory. For simplicity it will be assumed that all these functions are overloaded for vector treatment, so that for example if in the first of the previously mentioned functions the 13 argument is a scalar while h is a vector, the corresponding vector of entropies is returned. This is not difficult to implement in modern programming languages.

2.3.1.2 Air and flue gas properties

Similar to the steam case, it is also necessary to evaluate various properties of air or flue gas including enthalpy, density, specific heat, etc. from the state variables. As is known, the 'state' required to determine any property of a gaseous mixture comprises the gas composition, as well as some thermodynamic variables like the mixture pressure and temperature. Accurate modelling of gas dynamics would require balance equations for the different species involved, accounting also for the chemical kinetics, at least in the equilibrium state.

However, when the combustion gas composition undergoes such limited variation so as not to affect the relationships among the relevant thermodynamic properties significantly, and/or the accuracy required for the model is not very high, then one can evaluate the gas properties as if the composition was equal to a constant (nominal) value. In this case, the gas functions depend on two state variables only. The same simplification can be used for the air, provided a suitable 'nominal' composition is employed.

As a result, for air and flue gases this work adopts the same solution as for steam: two sets of functions similar to those presented are used, where the function names' prefix is Gas or A i r in lieu of S team and the rest of the syntax is the same. The Gas and A i r functions employ two different nominal compositions, which can be coded in the functions themselves or as a global variable, leaving only a couple of thermodynamic variables as arguments.

2.3.2 The boiler's water-steam circuit

2.3.2.1 Heat exchanger segment

This describes the dynamics of the fluid flowing into a tube bundle and of the metal wall, interacting with the external fluid (gas). For one-phase flow, the classical way to build such a model is by the equations of mass, energy and momentum for the fluid


stream and the energy equation for the metal wall:

Ow a i~ + ~x = 0 (2.6)

A Oh Oh P i -~ - Ai~---Pt + W~x = witPi (2.7)

dz Cf wlwl ai ~x -'l- p Aig-~x + -~ooi--~i = O (2.8)

0Zw AwpwCw ~ ---- OgetPe -- O)i~0i (2.9)

where p, p, h, w are the density, pressure, enthalpy and mass-flow rate of the fluid (depending on the tube abscissa x, 0 < x < L, and on the time t), Tw is the wall (mean) metal temperature, ~oi the heat flux from the metal wall to the fluid, ~Pe the heat flux from the external gas to the wall, Ai the tube (bundle) internal cross-section, o) i the corresponding total perimeter, We the total external perimeter, z the tube height, g the acceleration due to gravity, Cf a frictional coefficient, and Aw, Pw and Cw the metal cross-sectional area, the density and specific heat capacity. Note that, in the momentum equation, the effects of inertia and of kinetic energy variation along x have been neglected, as normal.

To complete the model, a suitable correlation for evaluating ~0i is needed. Assuming turbulent flow in the tubes, ~oi is usually expressed as:

qgi = yi(T - Tw) (2.10)

where T is the fluid temperature at (x, t) and Vi is the heat transfer coefficient for turbulent internal flow (Incropera and Witt, 1985).

Since the pressure drop and mass storage in a heat exchanger tube have definitely faster dynamics with respect to thermal energy storage, it is convenient to build the model with the following approximations:

The pressure p(x, t) can be considered 'nearly constant' along x, 0 < x _< L, i.e. p(x, t) ~- p(L, t) :--- po(t), for the sake of the evaluation of p and Op/Ot in equations (2.6)-(2.8).

The mass flow-rate w(x, t) can be considered 'nearly constant' along x, 0 _< x


Equation (2.1 1) can be written as:

O(po(t), h(., t)) dp(t) - wi(t) - Wo(t) + F(po(t), h(., t), h(., t)) (2.13) dt

where the functions 69 and F correspond to the second and first integral of (2.11), respectively, while h(., t) and/;t (., t) are the enthalpy profile and its derivative on the whole domain 0 < x < L.

Equation (2.12), in turn, can be written as:

Po - Pi + gL~(po, h(., t)) + kwilwilO(po, h(., t)) = 0 (2.14)

where ~p and ~ are suitable functions and k = (LCfogi)/(2A~), Cf obtained from suitable correlations. If any spatial approximation of the enthalpy profile is assumed, for example the finite element approximation

N

h(x, t) ~ ha(x, t) := Z hj(t)otj(x) j= l

then 69, F, ~ and fi become functions of the nodal enthalpy vector H(t) = [hl (t), hz(t) . . . . . hu(t)] ' and of its derivative i-/(t). Applying the finite element approximation with a Petrov-Galerkin type method (Morton and Parrot, 1980; Quarteroni and Valli, 1997), to equations (2.7) and (2.9), one obtains a couple of N-vector equations to be used for the computation of the fluid and metal temperature profiles. In the most general case, if weak boundary conditions are imposed (Quarteroni and Valli, 1997), such equations take the form

AI-I + w iBH + EH : A dp V - DT + DTw + MhIN dt

KTw : G(T - Tw) + C~e

where T and Tw are the fluid and metal nodal temperature vectors, ~e is the nodal external heat flux vector, and the matrices A, B, C, D, E, G, K and vectors V, M depend on dimensional data, on the fluid properties and on the specific finite element method chosen (applied to the interpolating functions otj and on the weighting functions used for computing the residual over the whole domain 0 < x < L). Details are omitted for brevity and can be found in Lunardi (1999), but it is important to note that weak boundary conditions do not constrain the first and last element of H to equal the terminal enthalpies. Moreover, the matrix E and vector M enforce the boundary condition on the side where the fluid enters the heat-exchanger segment (with enthalpy hiN). As such, they have only one non-zero element. The position of this element depends on the sign of the flow rate, but even despite this, the vector equations are affected only by the input and output structure, while the state variables (vector H) remain the same and cannot undergo change. Assuming a positive direction flow that from the 'i ' terminal to the 'o' terminal, and indicating by m and e the non-zero


elements of M and E, this means that

i fw >0

h lN=hi , M=[m 0 .. . 0]', E=d iag(e , 0 . . . . . 0), ho=H(N)

while if w < 0

h lN=ho, M=[0 . . . m]', E=diag(0 . . . . . 0, e), ho =H(1) .

Hence, this model can be implemented seamlessly in any language assuming a conditional equation construct like the i f clause in Modelica.

It is also worth noting that, with the adopted approach, a heat-exchanger segment uses a coarse approximation for the pressure and flow-rate profiles along the tube (one node for each segment) and a more accurate approximation for the enthalpy/temperature profiles. This really corresponds to the nature of the process dynamics, where hydraulic phenomena in heat exchangers are characterised by a much simpler spatial distribution with respect to thermal phenomena.

The external flux vector ~e is one of the variables included in the dHT (Figure 2.3); a convenient correlation as a function of the external gas properties and, possibly, of the wall temperature is naturally incorporated in the model of the gas zone corresponding to the heat-exchanger segment. A possible Modelica formulation of the model is given by the following script:

model HeatExchangerSegment parameter Length L ,om

Documents

Thermal Power Plant & Simulation