Coal Fired Power Plant Operations

Transcription

1 ADVANCED OPTIMISATION OBJECTIVES The objective of this project was to develop an online software tool to control the operation of local optimisation packages and provide whole plant optimisation. The specific activities were; To further develop individual software optimisers for a single coal-fired electricity generating unit. To design and build an integrated overall optimiser to supervise and control the operation of these local optimisers to the overall benefit of the operation of the unit. To understand how to handle potential conflicts among the optimisers and operate plant with several different optimisers running simultaneously. Coal Fired Power Plant Operations PROJECT SUMMARY 363 CLEANER FOSSIL FUELS PROGRAMME CLEANER COAL R&D PROGRAMME SUMMARY The purpose of this software optimisation project was to develop an integrated approach to unit optimisation and develop an overall optimiser that is able to resolve any conflicts between the individual optimisers. Financial support for the project was provided by the UK s Department of Trade and Industry (DTI), the US Department of Energy (DOE) and the Electric Power Research Institute (EPRI) together with the participant companies, principally E.ON UK and Southern Company in the USA. A substantial demonstration project has been conducted at Hammond Power Station, owned by Southern Company, over recent years.

2 Figure 1. Schematic showing convergence of the unit optimisation algorithm. (courtesy of E.ON UK plc) The success of the boiler optimisation models has suggested that on-line optimisation can be used in other parts of the power station. However, although each local optimiser is able to perform its task well, individually, there will be occasions when the individual packages will provide conflicting advice. A number of optimisers were installed and further developed at Hammond Power Station and two different unit optimisers to handle potential conflicts between the individual optimisers were developed and tested as part of the project. Potential conflict between optimisers can seen to be reduced by the use of a high level cost function or prioritising the objectives of different optimisers or adding rules to optimisers. BACKGROUND In recent years the efforts to reduce nitrogen oxide (NO x ) emissions from power stations have resulted in operational modifications including the fitting of low NO x burners. However, these modifications are expensive and generally have an adverse effect upon plant performance, resulting in an increase in unburnt carbon. To reduce these adverse effects, on-line optimisers have been developed as an enhancement to the power station's digital control system (DCS). GNOCIS (Generic NO x Optimisation Control System) is the main combustion optimiser used on coal-fired plant within the UK. This is a neural network based optimiser with on-line retraining that takes various control parameters such as mill feeder speeds, excess oxygen, burner tilt and load as inputs and predicts the resultant NO x emissions and carbon-in-ash levels. In fact the models are usually used in reverse with boiler control settings being provided by the model to optimise the emissions. The GNOCIS optimiser was previously developed in a joint project with Southern Company Services and EPRI in the USA. This unit optimisation project again brought together the same partners. The success of the boiler optimisation models has suggested that on-line optimisation can be used in other parts of the power station, eg thermal efficiency, electrostatic precipitator (ESP) and sootblowing. Although each local optimiser is able to perform its task well, individually, there will be occasions when the individual packages will provide conflicting advice. The purpose of this project is to develop an integrated approach to unit optimisation and develop an overall optimiser that is able to resolve any conflicts between the individual optimisers.

3 This approach is preferred to designing from first principles a single optimiser for the whole unit, since it provides significant flexibility and allows other optimisers to be added later. Although it is important to resolve the conflicts in the advice from different optimisers, it was expected that this will only affect the advice for a few control variables. Thus at the outset of the project, the unit optimiser was perceived as a relatively small piece of software that interfaces easily with the existing optimisers. The individual optimisers will retain the main interface to the control system with the unit optimiser providing slight adjustments to some of the recommended settings. SOFTWARE FOR OPTIMISING POWER PLANT OPERATION The following individual optimisers were considered during this project: Thermal efficiency Package. Boiler gas side optimisation (GNOCIS). Boiler steam side optimisation. Electrostatic Precipitator (ESP) Optimisation Intelligent Sootblowing System (ISBS) Thermal Efficiency Package The main difficulty with providing a real time thermal efficiency calculation is the accurate measurement of both the amount of coal that enters the boiler and its heat content. The calculation procedure starts with a detailed (ultimate) coal analysis that is generally not available in real time. It was recently suggested that the coal composition could be determined from an on-line analysis of the flue gas composition. This work was done on a test combustor at the Southern Research Institute, Birmingham Alabama. Since all values are available in real time, the heat content of the coal can be calculated in real time. The revised calculation process for the on-line method was compared to the off-line ultimate analysis and there was good agreement between the two methods. Thus the on-line coal analysis made an on-line thermal efficiency calculation possible. Boiler Gas Side Optimisation The boiler gas side optimisation used during the project was a standard GNOCIS installation. GNOCIS provides recommendations for the excess oxygen, secondary air and mill loading. Increasing the excess oxygen will increase the air flow through the ESP and most likely this will reduce the performance of the ESP. Thus, although the NO x emissions may be reduced, there may also be some slight increase in the dust emissions. Boiler Steam Side Optimisation The boiler steam side optimisation was a new optimisation package developed during the project. The steam side optimiser considers the boiler superheat temperature, superheat pressure and reheat pressure. These steam side conditions have a significant effect upon the thermal efficiency and set points for these parameters are generally defined within the control system. This optimiser was developed since some generating units cannot achieve their set points, especially during lower load operation. In these cases it is useful for an optimiser to show the best operating settings that can be achieved. In general though, when designing plant optimisers care must be taken not to fight the control system and avoid providing recommendations that will conflict with the normal boiler controls.

4 Electrostatic Precipitator Optimisation The electrostatic precipitator optimiser installed at plant Hammond during this project was the EPRI ESPert package. ESPert continuously receives and interprets data from the ESP control system, the continuous emission monitoring (CEM) system, and the boiler controls. ESPert provides an on-line estimate of ESP performance, including opacity and the probable causes of any divergence between measured and predicted opacity. Provided considerable plant and ESP data is available then ESPert can effectively model the performance of the ESP and predict the outlet conditions. However, it should be noted that ESPert requires an analysis of the coal and as indicated previously with regard to the on-line heat rate model this may not be available. suggest that a black box modelling technique like neural networks or expert systems will allow easier and better modelling of the ESP. Intelligent Sootblowing System When coal fired plant operates then the heat transfer surfaces gradually become fouled and plants are fitted with soot blowers that are used periodically to clean the surfaces. If the sootblowers are used too much then operating costs will increase due to the steam requirement of the sootblowers. Maintenance costs will also increase as a result of increased tube erosion. The tube erosion may also affect plant reliability. It is thus useful to have an intelligent sootblowing adviser that can determine when and where to sootblow within the boiler. As part of this project an expert system was developed to provide advice on the use of sootblowers. This system used readily available plant data like the superheater spray flows and the backpass damper position. The main cost of the sootblowing is related to the frequency with which it is used. This model was well received by the operators at Plant Hammond and showed that the plant could be operated as effectively with only 50% of the sootblowing that was normally used. Use of the sootblowing model will thus improve the performance of the unit. Unit Optimiser Figure 2. Screen for the intelligent sootblowing advisor showing the cleanliness of different tube banks.. (*courtesy of E.ON UK plc) Since installing ESPert at plant Hammond as part of this project there has been substantial software development of the model to improve its predictions. The results The unit optimiser is the new software that was developed within the project to resolve potentially conflicting advice from different individual optimisers. If several local optimisers give conflicting control settings for the unit then the total unit costs need to be considered to optimally resolve the situation.

5 Figure 3. Schematic showing the information flow between the unit and the individual optimisers. (*courtesy of E.ON UK plc) Whereas each local optimiser has its own objective function to minimise and only has knowledge of its own local restricted environment, the unit optimiser has to integrate the advice from all the local optimisers to produce an overall control strategy. In order to do this some means of compromising individual advice must be found. Since the objective functions for the local optimisers involve different high-level plant variables a common basis needs to be found to enable appropriate recommendations to be made. This common basis is usually costs and a unit cost function has to be defined in terms of high level plant variables such as NO x, carbon in ash, boiler efficiency, etc. It is important that costs can be associated with the high-level plant variables otherwise it is not possible to fully define a unit cost function. CONCLUSIONS Analysis of the flue gas composition for a coal-fired power station has enabled an on-line thermal efficiency optimiser to be produced. An intelligent sootblowing adviser has been developed for a coal-fired plant. The fuzzy logic exert system used readily available plant data to provide advice on the use of the sootblowers. There has been substantial development of individual power plant optimisers during this project. In the cases where the optimiser includes fundamental modelling then development has been difficult. A stable unit optimiser algorithm has been developed to resolve potential conflicts between different optimisers. This requires the definition of a unit cost function and it is important that costs can be associated with the high-level plant variables in the individual optimisers. Potential conflict between optimisers can be reduced by either prioritising the objectives of different optimisers or by adding rules to further constrain individual optimisers. POTENTIAL FOR FUTURE DEVELOPMENT The next step in the development of the unit optimisation software should be a demonstration on a coal fired power station in the UK. A typical demonstration may include three separate optimisers, namely:

6 Emissions Minimisation - GNOCIS minimisation of NO x and carbon in ash Thermal efficiency optimisation - On line PROATES model including advice on burner tilts. ESP Optimisation - The volume air flow has a significant effect upon ESP performance. It is clear that there are major conflicts among these optimisers. In particular GNOCIS and PROATES both include burner tilt level (where available on corner fired units) in their optimisations with possible conflict. GNOCIS also advises on the excess air level in the boiler. As the excess air level is increased then the volume flow through the ESP increases and again there will be a clear conflict between the optimisers. All three of the UK optimisers mentioned here are at an advanced state of development and a unit optimisation study could now been done on a UK coal fired plant using the methodology described here. The unit optimisation software should also be demonstrated on a gas fired power station in the UK. The use of an on-line neural network model for resolving combustion instability problems with a gas fired plant has been demonstrated on an instrumented gas turbine test rig. An on-line efficiency model using PROATES has been installed on some UK gas turbine plant. Gas turbine components are both very expensive and capable of causing significant subsequent damage within the gas turbine if they fail. Another optimiser looks at maintaining the integrity of these components. Good operation of a gas turbine thus requires balancing several objectives, namely maintaining combustion stability, thermal efficiency and plant integrity. There is thus a complex multi-objective function problem to be solved by the unit optimiser when applied to gas turbines. COST The total cost of this project is 483,000, with the Department of Trade and Industry (DTI) contributing 31,000. The balance of funding was provided by the particpants. DURATION 57 months - August 1999 to March 2004 CONTRACTOR E.ON UK plc Power Technology Ratcliff-on-Soar Nottingham NG11 0EE Tel: +44 (0) Web: COLLABORATORS Southern Company Services inc. US Department of Energy Electric Power Research Institute Further information on the Cleaner Fossil Fuels Programme, and copies of publications, can be obtained from: Cleaner Fossil Fuels Programme Helpline, Building 329, Harwell International Business Centre, Didcot, Oxfordshire OX11 0QJ Tel: +44 (0) Fax: +44 (0) helpline@cleanercoal.org.uk Web: Crown copyright. First printed December Printed on paper containing a minimum of 75% post-consumer waste. DTI/Pub 05/941