6
DEVELOPMENT OF AN EXPERT SYSTEM FOR PERFORMANCE EVALUATION AND DIAGNOSIS IN NUCLEAR POWER PLANTS Ki-sig ~ang* Power Engineering Research Institute, Korea Power Engineering Company,Inc 87 Samsung-dong, Kangnam-ku. Seoul Korea Se-Woo Cheon, Soon-Heung Chang Dept. of Nuclear Eng., Korea Advanced Institute of Science and Technology 373-1 Gusung-dong, Yusung-ku, Taejeon Korea ABSTRACT An expert system, called ESPED,for the performance evalu- ation and the diagnosis of nuclear power plants (NPPs) using performance parameters has been developed to support opera- tors, to improve plant performance, and to enhance operating flexibility. ESPED consists of three function modules: a performance evaluation module, a diagnosis module and a knowledge base entry module. The performance module processes major param- eters by meta facts and evaluates the performance effective fac- tors through data acquisition before a protection alarm level is reached. The diagnosis module diagnoses the cause of perfor- mance degradation to prevent further degradation, and gener- ates an operational guides using alarm logic tree and signif- cant value band. Limitation value libraries and failure identif- cation rules are stored in the knowledge base entry module. For ESPED diagnostic domain, the secondary system in NPPs is divided into five main system and subsystems. For veri- fication of ESPED, the operating results of the condenser and circulating water systems in Kori 3 & 4 NPPs have been ap- plied. I. INTRODUCTION In today's power industry, there are strong incentives to reduce power generation costs. This goal can mainly be achieved with con- dition-based maintenance and optimal operation control [l]. Although many power plants have extensive and complete in- strumentation systems, the large available database of information available may not be systematically followed, analyzed and/or stored. In addition, since the operator has not received significant information before alarm and/or trip levels are reached, he or she has to know a great deal of domain-knowledge and information beforehand in order to effectively enhance overall system perfor- mance and avert mps. In this work, a computerized operating support system for the secondary system in nuclear power plants (NPPs), called ESPED (Expert System for Performance Evaluation and Diagnosis), has been developed using expert system technology. This system was implemented on an IBM compatible PC/386. To analyze systematically the deviation between optimal and ac- tual performances of major components, ESPED has the perfor- mance evaluation function for analysis and the alarm processing function for diagnosis. To develop ESPED, the task performance by the operator is divided into the followings, according to the plant behavior [2]: - monitor and comprehend the state of the plant, - identify normal and abnormal plant conditions, - diagnose the abnormal plant condition, - predict the plant response to specific control actions, - select the best available control action, and - implement a feasible control action. Based on the operator's tasks, ESPED has been developed to support the plant operator aiming at improving net output, plant efficiency, and operating flexibility. For this purpose, the following tasks should be accomplished by ESPED: - monitor the component performance through the analysis of the performance effective factors, - detect the deviation from the optimum performance, - diagnose the cause of such deviation, and - suggest the possible operational guides to eliminate malfunc- When one of the significant performance effective factors reach- es a predefined warning level, ESPED evaluates the component performance through the performance evaluation module, in turn, diagnoses the cause and generates operational guides through the diagnosis module. tions. 11. DESCRIPTIONS OF THE DEVELOPED SYSTEM 11.1 The Structure of ESPED Electric utilities have placed considerable emphasis on enhancing certain aspects of plant performance, particularly the heat rate im- provement and the unit availability. In the application areas such as plant monitoring and control, maintenance, failure analysis, con- struction, and environmental emission control, the traditional com- putational approaches to these problems have met with marginal success. Currently, as an alternative to these approaches, the knowl- edge-based expert systems are better suitable for processing the symbolic information with acceptance and success [6]. 308 92CH3233-4 $3.00 0 1993 IEEE

[IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

Embed Size (px)

Citation preview

Page 1: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

DEVELOPMENT OF AN EXPERT SYSTEM FOR PERFORMANCE EVALUATION AND DIAGNOSIS IN NUCLEAR POWER PLANTS

Ki-sig ~ a n g * Power Engineering Research Institute, Korea Power Engineering Company, Inc

87 Samsung-dong, Kangnam-ku. Seoul Korea

Se-Woo Cheon, Soon-Heung Chang Dept. of Nuclear Eng., Korea Advanced Institute of Science and Technology

373-1 Gusung-dong, Yusung-ku, Taejeon Korea

ABSTRACT

An expert system, called ESPED, for the performance evalu- ation and the diagnosis of nuclear power plants (NPPs) using performance parameters has been developed to support opera- tors, to improve plant performance, and to enhance operating flexibility.

ESPED consists of three function modules: a performance evaluation module, a diagnosis module and a knowledge base entry module. The performance module processes major param- eters by meta facts and evaluates the performance effective fac- tors through data acquisition before a protection alarm level is reached. The diagnosis module diagnoses the cause of perfor- mance degradation to prevent further degradation, and gener- ates an operational guides using alarm logic tree and signif- cant value band. Limitation value libraries and failure identif- cation rules are stored in the knowledge base entry module.

For ESPED diagnostic domain, the secondary system in NPPs is divided into five main system and subsystems. For veri- fication of ESPED, the operating results of the condenser and circulating water systems in Kori 3 & 4 NPPs have been ap- plied.

I. INTRODUCTION

In today's power industry, there are strong incentives to reduce power generation costs. This goal can mainly be achieved with con- dition-based maintenance and optimal operation control [l].

Although many power plants have extensive and complete in- strumentation systems, the large available database of information available may not be systematically followed, analyzed and/or stored. In addition, since the operator has not received significant information before alarm and/or trip levels are reached, he or she has to know a great deal of domain-knowledge and information beforehand in order to effectively enhance overall system perfor- mance and avert mps.

In this work, a computerized operating support system for the secondary system in nuclear power plants (NPPs), called ESPED (Expert System for Performance Evaluation and Diagnosis), has been developed using expert system technology. This system was implemented on an IBM compatible PC/386.

To analyze systematically the deviation between optimal and ac- tual performances of major components, ESPED has the perfor- mance evaluation function for analysis and the alarm processing function for diagnosis. To develop ESPED, the task performance by the operator is divided into the followings, according to the plant behavior [2]:

- monitor and comprehend the state of the plant, - identify normal and abnormal plant conditions, - diagnose the abnormal plant condition, - predict the plant response to specific control actions, - select the best available control action, and - implement a feasible control action.

Based on the operator's tasks, ESPED has been developed to support the plant operator aiming at improving net output, plant efficiency, and operating flexibility. For this purpose, the following tasks should be accomplished by ESPED:

- monitor the component performance through the analysis of the performance effective factors,

- detect the deviation from the optimum performance, - diagnose the cause of such deviation, and - suggest the possible operational guides to eliminate malfunc-

When one of the significant performance effective factors reach- es a predefined warning level, ESPED evaluates the component performance through the performance evaluation module, in turn, diagnoses the cause and generates operational guides through the diagnosis module.

tions.

11. DESCRIPTIONS OF THE DEVELOPED SYSTEM

11.1 The Structure of ESPED

Electric utilities have placed considerable emphasis on enhancing certain aspects of plant performance, particularly the heat rate im- provement and the unit availability. In the application areas such as plant monitoring and control, maintenance, failure analysis, con- struction, and environmental emission control, the traditional com- putational approaches to these problems have met with marginal success. Currently, as an alternative to these approaches, the knowl- edge-based expert systems are better suitable for processing the symbolic information with acceptance and success [6].

308 92CH3233-4 $3.00 0 1993 IEEE

Page 2: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

To improve a NPP availability, the operator should know a great deal of operating informations before making an operational deci- sion. In Kori 3 & 4 NPPs, as in other NPPs around the world, about 4000 parameters are monitored using gauges, meters, strip-chart recorders, and indication lamps as well as computer print-outs. How- ever insufficient quantitative analysis of these parameters makes it difficult to determine optimal plant conditions. Normally the operat- ing conditions may depend on the operator's ability and experience.

Condenser e f f e c t i v e n e s s

I

Fig. 1 Management of selected parameter

T o t a l h e a t t r a n s f e r r e d Heat t r a n s f e r r e d i n idea l condenser

To overcome these difficulties, ESPED has been developed as the diagnostic system which moves away from exclusive reliance on the operator. This includes the following methods:

- ranking and selection of the parameters for calculation of per- formance effective factors,

1,

- management of the selected major parameters as shown in Fig.

- evaluation of the relevant performance effective factors, and - diagnosis of the cause and effect relationship by alarm logic

tree diagram. To fulfill the requirement of the secondary plant conditions, the

diagnostic function is divided into the five main systems and sub- systems as shown in Fig. 2 [7].

There are 34 major parameters selected for the performance eval- uation of the steam and power conversion system. The eight per- formance effective factors are derived from the major parameters. The definitions of plant performance effective factors are described in Table I.

Condenser pump e f f i c i e n c y

Feed w a t e r h e a t e r e f f e c t i V ~ n e s s

11.2 The Functions of ESPED

Pump w o r k P o w e r suppl ied t o pump

T o t a l h e a t t r a n s f e r r e d Heat t r a n s f e r r e d i n i d e a l h e e t e r

As shown in Fig. 3, ESPED consists of three main modules: a performance evaluation module, a diagnosis module and a knowl-

Moisture separator e f f e c t i v e n e s s

Reheater e f f e c t i v e n e s s

11.2 .I Performance Evaluation Module

T o t a l moisture i n t o separator

T o t a l h e a t t r a n s f e r r e d Heat t r a n s f e r r e d i n idee l r e h e a t e r

The performance evaluation module processes the major param- eters by meta facts with query operation. The input data are checked for validity in an options window. The options window is used to display the acceptable responses for queries during the consultation. Part of the validated data is used to calculate a performance effec- tive factor. Any deviation between the result calculated for the op- timal value and the process data is evaluated by significant bands value [5] . If the deviation is beyond predefined limitation values stored in a limitation value library, the identification of performance degradation is annunciated.

Table I. Major parameters for performance effective factor

I D e f i n i t i o n I E f f e c t i v e f a c t o r 1

Pump w o r k

Pump w o r k P o w e r suppl ied by i s e n t r o p i c FWPT Feed w a t e r pump e f f i c i e n c y

PLANTSYSTEMS

I 1 1 I 1

DATA ACQUISITION SYSTEM

ESPED/PL ANT IN'IEGRATION

& FUNCTIONAL

ARCHITECTURE

edge base entry module. Fig. 2 Plant integration and functional architecture

309

Page 3: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

cause messaEe 1 guide & diagnosis message 1

Feed w a t e r f l o w Open ing o f f l o w c o n t r o l v/v C o n d e n s e r l o a d

CRT Display Conrrol (message and picture)

S/G w a t e r l e v e l , s t e a m f l o w ( , ( 2 ) M e a s u r i n g p r e s s u r e

d i f f e r e n c e I n p u t p a r a m e t e r (31

Fig. 3 ESPED functional diagram

11.2.2 Diagnosis Module

The diagnosis module evaluates and compares the measured val- ues with the limitations and failure identification rules stored in the knowledge base. The evaluation sequence of the failure identifica- tion rules is decided automatically by the knowledge base goal. According to the diagnosis results, the diagnosis module diagnoses the causes and generates operational guides to prevent further deg- radation of performance [3]. The goals of operational guide are:

- prevention of turbine trip, and - recovery of system function.

The first operational guide is the isolation of the failed compo- nent and maintaining the system function by using an alternative path, if available. When the first operational guide can not be fol- lowed, the turbine power level is reduced to mitigate the influence of the failure, according to the power reduction procedure as the second operation guide. The operational guide network [4] corre- sponding to process state is shown in Fig. 4.

Table 11. Example of sensor validation

E s t i m a t i o n o b j e c t I I n p u t i n f o r m a t i o n I C a t a g o r i e s I

11.2.3 Sensor Validation Conventional sensor validation techniques such as comparison of

redundant sensors, limit checking, and calibration tests have been employed in NPPs. However, these techniques have serious draw- backs, e.g., the need for extra sensors, vulnerability to common

mode failures, limited applicability to continuous monitoring, etc. To alleviate these difficulties, a new sensor validation technique has been developed by using the methods called analytic redundancy and panty space [10,11]. Although the new technique has been proved feasible as far as preliminary tests are concerned, further developments should be made in order to enhance its practical ap- plicability. The techniques used for sensor validation in ESPED are classified into three categories as follows and shown in Table II:

Plant stah Goal Kind of Guide Content

System function

mitigation ailure ocwence

Turbine trip

Stan up standby mponentcbangi ng power supply

bus,etc.

Manually operate

controller,Turbi nepower level

reduction I Slop failed

Isolate failed component

prevenuon m recovery

After acluding Restan failure cause component

Shutdown Stop plant operation pocedure in d a y

Fig. 4 Operational guide corresponding to plant state 310

Page 4: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

- mass and energy conservation-law-based, - functional-relationship-based, and - options window.

C o n c l u s i o n

111. APPLICATION OF ESPED

[Then] T u b e wall gauge c o r r e c t i o n f a c t o r I 2.857 1428E-02* B W G + 0.162857 14

For application of ESPED, the condenser and associated systems in Kori 3 & 4 NPPs have been adopted.

111.1 Evaluation of Performance Effective Factor

The cleanliness factor of the condenser is defined as the ratio of actual heat transfer to the maximum possible heat transfer in the condenser with a given condition. Therefore the cleanliness factor is an accurate measure of the actual condenser performance. The per- formance evaluation module of the condenser calculates the cleanli- ness factor and vacuum pressure with a given set of data as follows:

- condenser load, - circulating water flow, - circulating water inletloutlet temperatures, - cleanliness of tube material, - thermal conductivity of tube material, and - total effective tube surface area.

The typical rule format is shown in Fig. 5.

P r e m i s e [IF1 M a t l c l a s s = Aus-SS and M a t l t y p e = 304-SS o r M a t l t y p e = 3 1 6 3 s o r M a t l t y p e = 3293s

L e f t o f " = " A r e v a r i a b l e ( c o n t r o l l i n g p a r a m e t e r s ) R i g h t o f " I " A r e v a l u e f o r t h e v a r i a b l e

Fig. 5 Typical rule format

The diagnosis module checks the correlations involving the de- sign performance effective factors and actual performance effective factors. The category correlations which are the satisfaction degrees are divided into <LLA>, <L>, dVL>, <">, uY> and &HA> de- pending on the result of comparison as shown in Fig. 6. In the case of a mismatch between the actual and the design range values, the diagnosis module is started.

111.2 Diagnosis of Performance

The diagnosis can be made when the module detects the data deviation which exceeds the permitted value bands, i.e, between <LLA> and uVL>, <"> and <"A> according to the significant value bands. It gives the diagnosis of possible cause for the devia- tion of the effective factor, and recommends the corrective opera- tional guide.

~

3 1 1

(1) I (2) I (3) N:Normalrange l(4) l(5) I(6) LLA L NH H HHA

Boundaries of value bands

LLA = low low and automatic trip action L = low NL = low limit of the normal range NH = high limit of the normal range

significant bands values

(1) Cf c 65

.............................................................................

(2) 65 C= Cf c 70 (3) 70 <=Cf c 7 5 NL = 75 NH =85 (4) 85 < Cf 4 0 H = hioh

HHA= hig; high and automatic trip action (5) 90 <= Cf > 95 (6) Cf =>95

Fig. 6 Boundaries and significant bands value

P- ur.w;uer pump motor Abnormal

water leak

Condenser vacuum

Cond pull pit level

fanAbnormal I

Fig. 7 Alarm logic tree of condenser

The diagnosis results consist of the followings: - the description of the deviation, - an explanation of the deviation, - a certainty factor (CF) to indicate the possibility of deviation, -and recommendation of operational guides to avoid subsequent

The condenser alarm logic tree with the top down approach is

- top event: turbine trip alarm => condenser vacuum abnormal - first group: process condition => circulating water flow low - second group: loss of component function => circulating wa-

damage

grouped into the three stages as follows:

ter pump motor abnormal

Page 5: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

Seeking suggestion(box-level-loW). Found suggestion(box-level-low). Found operation-guide.

Reduce Turbine Power

Condensa U Vacuum Reavery

4 Check valve llne up of

5 . Setting value 162 inch DA sysyem

1

REASONING

1. Check water box priming tank pressure. 2. Check the priming pump running conditions. 3. If the CWP discharge pressure is low,

run the stand-by CWP. 4. Check the valve line-up of D A sysyem. 5. Setting value : I 6 2 inch rule-c 4:succeed.

I

'OPTIONS I LOW

High

Fig. 8 Operational guide example IV. DEMONSTRATION

- Control Discharge Valve

- St.rtS(.ndbyCWP

The performance effective factors for condenser, condenser vac- uum pressure and cleanliness factor, have been treated by the per- formance evaluation module. Due to off-line mode, the input pa- rameters have been provided by queries for demonstration. For val- idation of sensor data, three methods as described in Sec. 11.2.3 are used to display the acceptable responses for the quries during con- sultation.

As a result of consultation, the condenser vacuum pressure was found as normal low, i.e., ab. The diagnosis module detects the cause of <NL> and gives an operational guide to support operator, according to the alarm logic tree as shown in Figs. 9 and 10.

"TZLr coldrum hx VlCYum

pmnw3P'W e)

Rule - f E If ideal-u=674.548 I 9 6 2 and cfd=XI and ti=X2 and -0.20 I 6 *0 .03 I5*X2-0.0002592*X2* X2+0.7524€-06*XZ*X2=X3 and

Then design-u=X4. The lollowing knowledge base entries are also being used. rule-f5 (a rule) rule-17 (a rule) kb-2 (a init ial data)

674.5481962'Xl*0.81*X3=X4

Fig. 9 Fired rule tracing example

Fig. 10 Diagnostic result of ESPED

V. CONCLUSIONS

The expert system, ESPED, has been developed to support the plant operator using performance parameters before making a deci- sion during degradation of system performance. ESPED consists of three function modules: the performance evaluation module, the di- agnosis module and the knowledge base entry module. The antici- pated benefits of this system include the enhanced plant perfor- mance and operating flexibility.

The developed system has been verified with the operating data of condenser and circulating water systems in Kori 3 & 4 NPPs. As described in the previous sections, ESPED can successfully calcu- late the performance effective factors for condenser, condenser vac- uum pressure and cleanliness factor using the performance evalua- tion module. The diagnosis module can diagnose the degradation cause and generate the operational guides using the failure identifi- cation rules and the procedure libraries. Also, "Why" and "How" command functions are used to confirm why these inputs are neces- sary, and to trace the fired rules. Three modules of the system have been utilized to provide the optimal combination of the evaluation and diagnosis of the system performance.

REFERENCES

[I] G. Lindbery et al., "On-Line Condition Monitoring of Power Station Component using Expert System," Roc. of Conference on Expert System Application for Electric Power Industry, Or- lando, Florida, June 1989.

312

Page 6: [IEEE Proceedings of 1992 IEEE 5th Human Factors and Power Plants - Monterey, CA, USA (1992.06.7-1992.06.11)] Conference Record for 1992 Fifth Conference on Human Factors and Power

[ 21 D.W. Miller et al. , "The Knowledge-Based Framework for a Nuclear Power Plant Operator Advisor," Trans. of the ANS,

Vol. 59, 1989. [3] A. Kaji, T. Marayama et al., "Development and Application of

an Expert System (HITREX) for Plant Operation," PrOC. of Conference on Expert System Application for Electric Power Industry, Orlando, Florida, June 1989.

[4] ~ o r i o Naito. et al., "A Real-Time Expert System for Nuclear Power Plant Failure Diagnosis and Operational Guide," Nucle-

ar Technology, Vol. 79,284-296,1987. [5] 1,s. Kim, and M. Modarres, "Application of Goal Tree-Success

Tree Model as Knowledge-Base of Operator Advisory SYS- tern," Nuclear Science and Engineering, Vol. 16467-81, 1987

[6] L.J. Valverdeef al., "Fossil Power Plant Applications of ExPen Systems," ROC. of Conference of Expert System Application for Electric Power Industry, Orlando, Florida, June 1989.

[71 Final Safety Analysis Report for Kori 3 & 4 NPPs, Chapt. 10:

Steam and Power Conversion System, and Chapt. 15: Transient Analysis, Korea Elecmc Power Corporation.

[SI Heat Exchanger Institute, Steam Surface Condenser, the 3rd edition, HEI, Inc., 1980

[gl Ki-Sig Kang, Development of Secondary Plant Performance Evaluation and Diagnosis Expert System using Performance Parameter, M.S. Thesis, KAIST, 1989.

[lo] 0.L. Deutsch el al., Validation and Integration of Critical PWR Signals for Safety Parameter Display Systems, EPRI Report,

[ 1 11 C.H. Meijer et al., On-line Power Plant Signal Validation Tech- nique Utilizing Parity-Space Representation and Analytic Re- dundancy, EPRI Report NP-2 1 lo, 198 i.

NP-4566, 1986.

3 1 3