Expert System Applications to Protection, Substation ...smartgridcenter.tamu.edu/resume/pdf/j/esapp91.pdf · Electric Power Systems Research, 21 (1991) 71 - 86 71 Expert System Applications

Electric Power Systems Research, 21 (1991) 71 - 86 71

Expert System Applications to Protection, Substation Control and Related Monitoring Functions

M. KEZUNOVIC, K. WATSON and B. DON RUSSELL

Texas A&M University, College Station, TX 77843-3128 (U.S.A.)

P. HELLER and M. AUCOIN

MICON, Inc., College Station, TX 77845-9636 (U.S.A.)

(Received September 13, 1990)

ABSTRACT

Appl icat ion of expert systems to power system problems has become an area of strong research interest in the past few years. A number of papers have been pub l i shed on the sub- ject wi th an emphasis on applications that relate to the overall power system monitoring, operation and p lanning . However, less emphasis was placed on protective relaying, substation control and related moni tor ing functions.

This paper is pr imar i ly concerned with the applications that are focused on power system protection, substat ion operation, and monitoring. It gives both a survey of the present research efforts and a discussion of future possibil i t ies and trends in this area.

1. INTRODUCTION

The r ecen t pub l ica t ion of severa l ve ry in- fo rma t ive papers on exper t sys tem appl ica- t ions to power sys tem prob lems has provided an ex tens ive b ib l iog raphy [1], a state-of-the- a r t r ev iew [2], and a su rvey [3] of p resen t r e sea rch act ivi t ies . The ma in emphas i s is on the ene rgy m a n a g e m e n t sys tems (EMS) rela ted app l i ca t ions such as sys tem moni to r ing , ope ra t i on and p lanning . The p r i m a ry use of these deve lopmen t s is to ass is t sys tem dis- pa t che r s and o the r ope ra to r s to pe r fo rm the i r dai ly tasks.

The a rea of p ro tec t ive re lay ing , subs t a t ion control , and power a p p a r a t u s and sys tem mon- i to r ing has been less emphasized, a t leas t as far as the n u m b e r of publ i shed papers would sugges t [4]. I t is felt t ha t one of the ma jo r reasons for this comes f rom the rea l - t ime con-

0378-7796/91/$3.50

s t ra in t s imposed by the app l i ca t ions which requi re mos t of the func t ions to be execu ted au tomat i ca l ly . Use of the exper t sys tem ap- p roaches in this case is not yet fully under- s tood and apprec ia ted , and hence more r e sea rch ac t iv i t ies are needed to r each fu r the r conclusions .

This pape r provides d iscuss ion of severa l i m p o r t a n t app l i ca t ion issues re la ted to the use of exper t sys tems in con junc t ion wi th the sub- s t a t ion equ ipment opera t ion . Severa l applica- t ions proposed so far are su rveyed and some ongoing r e sea rch ac t iv i t i es are repor ted . Fut- ure app l i ca t ion a reas are also poin ted out. The goal of th is pape r is also to c o n c e n t r a t e on the new deve lopment s not discussed in the recen t ly publ i shed surveys [1-4] .

The first pa r t of the pape r is devoted to ana lys i s of func t iona l r equ i r emen t s imposed by the re lay ing , cont ro l and m o n i t o r i n g functions. The mos t common exper t sys tem ap- p roaches t ha t can be ut i l ized to mee t these app l i ca t ion r equ i r emen t s are also poin ted out in this section. Next , exis t ing exper t sys tem imp lemen ta t i ons are su rveyed wi th an emphasis on the un ique design cha rac t e r i s t i c s and exper t sys tem features . This su rvey is con- cluded by prov id ing a cr i t ica l overv iew of the exper t sys tem solu t ions used. Final ly , some deve lopmen t ac t iv i t ies are analyzed, po in t ing out d i rec t ions for fu r the r r e sea rch efforts. An ex tens ive b ib l iog raphy is g iven a t the end.

2. IMPLEMENTATION CONSTRAINTS

2.1. Funct ional characteristics In order to be able to classify the exis t ing

exper t sys tem app l i ca t ions as well as to

t:' Elsevier Sequoia/Printed in The Netherlands

72

propose some di rect ions for fu ture research, a detai led discussion is needed of the t ime response, dependabi l i ty /se lec t iv i ty and equipment a l loca t ion requ i rements of the protection, control and moni to r ing functions.

As is well known, there are a number of protect ion, control and moni to r ing funct ions tha t reside in a subs ta t ion [5]. The scope of this paper is to cover exper t system applica- t ions tha t pr imar i ly re la te to these funct ions. The solut ions of in teres t are the ones tha t may be implemented at e i ther the subs ta t ion or the remote contro l location.

2.1.1 Response time One of the main charac te r i s t i cs for almost

all of these funct ions is a very shor t response time. This may be re la ted to the reac t ion time, if the funct ion is a closed-loop real-t ime control such as pro tec t ive relaying. On the o ther hand, this t ime may be associated with the execut ion t ime if an open-loop operator-ini- t ia ted funct ion such as subs ta t ion switching is of interest . Finally, the t ime response may denote the database update t ime if da ta acqui- si t ion funct ions such as a larm and s ta tus mon- i tor ing are considered. A br ief overview of the t ime cons t ra in ts associated with the main groups of protect ion, cont ro l and moni to r ing funct ions is given in Table 1.

Given the s ta te of the ar t of exper t systems, it is unl ike ly tha t they could opera te in the shor t t ime required for pro tec t ive relaying. However , a pro tec t ive re lay which implements

an exper t system can be programmed in such a manne r tha t the expert system recognizes when to tu rn over the re lay ing funct ion to an ex t remely fast opera t ing algori thm. A permis- sible longer response t ime would be selected when there has been a more subtle v io la t ion of the s ta te l imitat ions.

Exper t systems also offer potent ia l for im- proving system opera t ions by ident i fying more complex s i tua t ions which happen slowly (such as the loss of the t ransmiss ion network) . It may be difficult for a human to de te rmine the act ion needed in t ime to p reven t the loss of a t ransmiss ion system. On the o ther hand, an exper t system could de te rmine the necessary act ions in t ime to p revent system breakup as the loss of a t ransmiss ion system happens over tens of seconds or minutes.

2.1.2. Dependability/selectivity The dependabi l i ty and select ivi ty of protec-

t ive relays describe the abil i ty of the re lay to ident i fy events cor rec t ly and take ac t ion based upon the in format ion avai lable to the relay. The dependabi l i ty and select ivi ty of p ro tec t ion re lays have been limited in the past by conven t iona l approaches and l imita t ions of the implementa t ion technology. This has been widely recognized as an inabi l i ty of the relaying scheme to adapt to the changes in the power system opera t ing condit ions.

With the advent of microprocessor applica- t ions in pro tec t ive re lay ing and the paral le l deve lopment in exper t systems, it is now possible

TABLE 1

Time response constraints

Category Function Order of response

(ms) (s) (min)

Type Operator

Protection Line Transformer Bus Backup Out-of-step

Control Switching V/VAR control Load shedding SCADA

Monitoring SOE Disturbance rec. Fault location Revenue metering

Closed-loop No Closed-loop No Closed-loop No Closed-loop No Closed-loop No Open-loop Yes Open-loop Yes Open-loop Yes Open-loop Yes Off-line Yes Off-line Yes Off-line Yes Off-line Yes

to implement protection systems which can analyze data in a more complex fashion and maintain or surpass the high degree of dependability and selectivity previously available with conventional approaches. This ability implies that the circuits and systems which are protected by the new relay designs can be operated at higher loading capacity and higher efficiency and reliability.

2.1.3. Equipment allocation Another major consideration is related to

the equipment allocation. The relaying equipment is located in the switchyard and/or the control house of a substation. This equipment is connected directly to the related power ap- paratus and exchange of information between different relays is quite limited. The substation control equipment is usually centralized for the entire substation and is implemented using either a dedicated substation computer or a remote terminal unit (RTU) of a supervi- sory control and data acquisition (SCADA) system. The monitoring equipment may be or- ganized in a number of different ways. One option is the use of dedicated devices such as sequence of events (SOE) recorders and digital fault recorders (DFRs). Another approach is to have substation-wide monitoring using

73

dedicated computers for alarm and status acquisition and display. Yet another approach is the use of RTUs to bring the monitoring data to the EMS control center.

One of the latest developments for substation application is an integrated control and protection system (ICPS) approach [5, 6]. In this case all of the protection, control and monitoring functions are implemented using a distributed processing system which enables close coordination among all of the functions.

A summary of some of the most important equipment implementation approaches and their characteristics is given in Table 2. It may be observed that the different equipment would require different approaches to expert system implementation. The approaches would range from adding the expert system functions in the existing designs to providing an additional computer to carry out the expert system tasks. This would affect the cost of the expert system designs as well as the possibility for hardware/ software optimization of the overall design.

Finally, expert system applications may require a brand-new approach to equipment designs which would combine the functions of several existing types of equipment with the addition of some new functions. The ICPS design is an example of such an approach.

TABLE 2

Equipment design approaches

Equipment Location Communicating Operator type with: interact ion

Protection relays Switchyard/ Switchyard/ Change of sett ings control house eng. office

Fault locators Control house Switchyard Settings, local readings

SOE recorders Control house Switchyard Settings, local readings

Digital fault Control house Switchyard/ Local/remote recorders master s tat ion readings RTUs Control house Switchyard/ Remote indicat./

master s tat ion remote control Operator and Control house Switchyard Local readings revenue meters Control functions Switchyard/ Switchyard/ Monitoring/ (synch. check, control house control house ini t ia t ion V/VAR, blocking) Integrated Switchyard/ Switchyard/ Settings/ control and pro- control house eng. office/ monitoring/ tection systems control center control

74

2.2. Expert system characteristics Exper t systems have been emerging rapidly

from labora to ry studies into a broad range of indust r ia l and business applicat ions. The mar- ket segment for exper t systems and the i r development tools has grown from about U.S. $9 mill ion in 1982 to an es t imated U.S. $1.8 bil- l ion in 1990 [7]. Many have predicted tha t exper t systems will soon be viewed as merely a new, powerful , p rogramming technique which can be in tegra ted into most computer programs.

In general , an exper t system is a program tha t mimics human expert ise in a par t icular , domain-specific decis ion-making process. These decisions may per ta in to funct ions in control / moni tor ing, debugging, design, diagnosis, in- s t ruct ion, in te rpre ta t ion , planning, or predic- tion. The mimicking of human experts should include not only the decision funct ion, but also the abil i ty to reason, or infer, even in the face of unce r t a in or missing informat ion. The abil i ty to explain to others what the reasoning is, as well as the abil i ty to act in a cor rec t and t imely ma nne r should also be included. The con t ras t between the exper t systems and more conven- t ional computer programs is i l lus t ra ted in Table 3 [8]. With this Table in mind it should be recognized tha t most cu r ren t computers base thei r a rch i t ec tu re on the use of conven t iona l programs, not exper t systems. This helps to explain why exper t systems seem to push the boundar ies of computer resources and execu- t ion times more than conven t iona l programs.

New hardware and sof tware which address the cons t ra in ts found in exper t system implemen ta t ion are con t inu ing to be developed. As the demand for these tools rise, the i r avai labi l i ty and cost is expected to become more pract ical . Some of the major character is - tics of exper t systems are reviewed next.

2.2.1. Area of application In any compar ison of compute r programs it

is important , even obvious, to keep in mind the a rea of appl icat ion. This serves as a re- minder tha t even though one tends to discuss exper t systems as if they are one lumped group of programs, the i r different areas of appl ica t ion requi re great care when the implementa t ions are compared.

2.2.2. Languages An exper t system can be developed uti l izing

any compute r language. Some languages, such as Lisp and Prolog, have fea tures which s t rongly suppor t the deve lopment of exper t systems.

These languages were developed for symbolic processing, an d commonly rely upon in- t e rpre te r s to in ter face the language to the ac tua l compute r mach ine ry where they are executed. Other languages, such as C, For- t ran, and Pascal, have few or no special con- s t ruc t ions for symbolic processing, however these special cons t ruc t ions can be developed in these languages. It may be more difficult to develop or modify exper t systems in this la t te r

TABLE 3

Contrast between expert systems and conventional programs

Characteristics Expert system Conventional program

Processing type Symbolic Numeric

Major Compare, select, sort, match Arithmetic and logic operators logic sets, context and partitioning,

pattern retrieval and recognition Program flow Nondeterministic Deterministic

Execution

Information management System refinement

Dynamic (creation of data structures makes resource allocation difficult) Representation and acquisition of knowledge is challenging Continuously open to refinement

(procedure flow and termination predictable) Static (clear concept of required resources) Algorithms usually well defined with well-structured data Little to no online refinement mechanism

set of languages. However , since they tend to be compiled languages, they will have fas ter execu t ion times.

2.2.3. Knowledge source The source of knowledge for an exper t sys-

tem can inf luence the o ther charac te r i s t i cs in its design and grea t ly affect the ease and time of development . The acquis i t ion of the domain-specific knowledge for the exper t system can be the most difficult par t of the development. In addit ion, the abi l i ty to check the knowledge base for correctness , conciseness, and thoroughness , in o ther words the verifica- t ion of the knowledge base, can be very difficult, depending upon the source of knowledge. For example, the source of knowledge could be a set of manuals , a single h u m an expert , or several human experts. The verifica- t ion of heur i s t ic knowledge in these th ree examples can be seen as to ta l ly different processes.

2.2.4. Knowledge representation Rules, frames, or logic, the common methods

for r epresen t ing knowledge in an exper t sys- tern, are discussed in many texts and articles. Acknowledging tha t a field of eng ineer ing has emerged involving the acquis i t ion and represen ta t ion of knowledge, the t radeoffs be tween knowledge rep resen ta t ions for the knowledge domain can be complex to evaluate . The choice of knowledge r ep re sen ta t i on can have major effects on the deve lopment time, cost, performance, and u l t imate success of the exper t system design. With the impor tance and non- t r iv ia l na tu re of the ac tua l choice of knowledge r ep re sen ta t i on in mind, a grossly t r ivial ana logy is given here to provide some feeling %r the choices being made.

If the knowledge necessary to plan t rave l routes can be given adequa te ly in a road map of l imited dimensions and the necessary facts are road condi t ions and goals, then a rule- based r ep re sen ta t i on can be ideal. This is not an a t t empt to t r ivial ize the power of rule- based systems, for often to t ry to view informat ion beyond the scope of the road map does not real ly help in the select ion of adequa te highways. Thus the rules descr ibing roads, connec t ions and cu r r en t condi t ions are quite sufficient wi thou t ge t t ing bogged down in o ther charac te r i s t i c s such as road construc- t ion charac ter i s t ics , grades, and speed limits.

75

However , if the rules become extensive or one must ven tu re at t imes away from the paved highways, a frame-based r ep resen ta t ion may become advantageous . More relat ionships, inher i ted charac ter i s t ics , and usual ly options, may be expressed in a frame-based system. (More rules may do the job but the execut ion times of rule-based systems tend to be more sensi t ive to the size of the knowledge base than the execut ion t imes of frame-based systems [8].) In the ana logy with maps, this may be viewed as the difference between a road map and a con tou r map. P lann ing a pa th with con tou r maps requires tha t one can view more of the overal l re la t ionships since one may not be depending on paved roads.

Finally, if the knowledge of the t e r r a in is only in the form of some major l andmark descr ipt ions and locat ions, one may have no real map to represen t it at all. In this case our logic of search ing and ident i fying landmarks , new and old, must take p recedence over de- pendence on previously paved trails. This logic r ep resen ta t ion may be much slower than the o ther techniques , depending upon the ter- rain, but if the maps are too unre l iable or nonex is ten t this may still be the best option.

2.2.5. Inference engines The discussion of inference s t ra tegies in-

volves search d i rec t ion and pat tern . W h a teve r the knowledge r ep resen ta t ion chosen, the inference process involves compar ing cu r r en t condi t ions with the knowledge base. In some problems the best d i rec t ion is to view the facts and see what kind of conclus ions are indi- ca ted in the knowledge base. This is forward chaining. In o ther appl ica t ions it may be more impor tan t to get goals and subgoals and then find or faci l i ta te the facts which suppor t these goals. This is backward chaining. In still o ther appl icat ions, the combina t ion of the two techniques is desirable. In e i ther forward or backward chaining, matches be tween the cur- ren t condi t ions and the knowledge base are searched for. One may choose to s tudy one b ranch of opt ions as extensively as possible unt i l one succeeds or is forced to abandon this t ra in of t h o u g h t and move to ano the r branch . This is a depth-first search. On the o ther hand, it may be more advan tageous to s tudy all possible b ranches in equal depth and only proceed to deeper levels in the more promising branches . This is breadth-f i rs t searching.

76

Again, it may be advantageous to combine, or switch, from one technique to another within a given problem.

2.2.6. Shells On top of many languages, knowledge rep-

resentations and inference strategies, empty shells have been developed for expert systems. These shells require only the domain-specific knowledge base to become functional expert systems. Shells usually provide easier user interfacing for the development and execution of an expert system. The tradeoffs are usually in the constraints (expert system characteristics) chosen for a part icular shell, and the costs of the shell.

2.2. 7. Hardware Just as any computer language may be used

to develop an expert system, any computer may be used for an expert system implementation as well. The performance of the expert system, as of other computer programs, will be greatly affected by the hardware. The following characterist ics make expert systems par- t icularly burdensome on the hardware: (1) the extensive use of memory for program storage and execution space; (2) the vast number of memory accesses for the symbolic compari- sons and searches. All of this usually represents execution speed bottlenecks. Some special computer architectures have been developed to support languages such as Lisp or Prolog more effectively. However, demand for these special architectures has not been great enough to manufacture them at a cost com- parable with conventional architectures. It still seems that more conventional computer architectures will remain in the forefront. However, it does not seem that the architectures will become more tuned to expert system type processing in the future (especially in their memory management capabilities).

3. SURVEY OF THE PRESENT APPLICATIONS

3.1. Funct ional description The following discussion is limited to the

functions that are in some way or another related to the relaying, control and monitoring of a power system. The main distinction between this survey and some other recent surveys [1 - 3] is that the discussion given here

focuses primarily on the functions that are directly related to the operation of the substation equipment.

3.1.1. Monitoring and diagnosis As may be expected, the most frequent ex-

pert system applications are in this area. It is well recognized that, in general, an expert system is ideally suited for diagnostic function implementation. This impression has been reflected in the power field as well.

One relatively broad area of application is monitoring of the substation equipment and diagnostics of the abnormal regimes of operation. These activities are directly related to prevention of damages that may result from undesirable equipment conditions. Some authors have proposed expert systems for electric machine diagnosis [9, 10]. One solution is called the Vibration Cause Expert and is aimed at identifying mechanical defects [9]. Another solution is based on a number of sensors providing information about different operating parameters of a turbine~generator system [10]. High voltage circuit breaker monitoring is also viewed as a potential application for expert systems. One approach has been developed to monitor HV breaker contacts, oil pres- sure and pump conditions, and insulation damage in order to diagnose the breaker operational condition [11]. Some authors are also suggesting expert system applications for diagnosis of HVDC systems by monitoring operational conditions of the converters [12, 13]. The problem of monitoring power transformers by analyzing gas characterist ics has also been recognized as an expert system application area [14-17]. Similar techniques have been suggested for monitoring overall gas insulated substation (GIS) equipment [17, 18].

Another application area that has produced a lot of different expert system solutions is the fault diagnosis area. Owing to the large number of applications already covered in the other survey papers [1-3], only the most recent (1989) applications are reported in this paper. Since this is an area where the first systems are already in online operational mode [3, 19], the references published recently reflect the maturi ty of the considerations. First, several ongoing system implementation activities are reported [20-23]. Then, a number of new techniques for the expert system approach are developed [24- 27]. Finally, some

new implementation concepts are considered, indicating that the substation computer ra ther than the SCADA master station may be used for the implementation [28, 29].

3.1.2. Substation control One of the most common control functions

in a substation is the switching function. This function is normally carried out by the operators, but it can also be automated. The approaches proposed so far have enabled data validation, execution of switching sequences and final status verification steps to be automated so that the operator is fully assisted in carrying out such a complex task.

Expert system applications have also con- centrated on supporting operators in making their decisions more reliable and secure. Sev- eral approaches for expert system applications to the switching sequence implementation have been proposed [30-32]. The recent approaches have suggested that overall substation operation can be guided by an expert system [33, 34].

One other area of interest for expert system application for substation control purposes is in distribution automation systems. Prelimi- nary results in this field have been reported [35] and initial implementation efforts under- taken in the area of feeder monitoring and protection [36].

3.1.3. Protective relaying and related functions Owing to the str ingent time-response re-

quirements, there are very few applications of expert systems in the implementation of protective relaying functions. The only known application is in the area of high impedance fault detection.

High impedance faults have eluded the application of any conventional techniques for detection. Such faults appear to mimic normal load conditions experienced on a given feeder. Substantial research in recent years [37-39] has shown that it is necessary to implement very sophisticated approaches to identifying what with conventional approaches have been seen as very subtle changes. Using expert system techniques, small differences become obvious changes which are unique to the high impedance downed-conductor fault. This approach provides for reliable, accurate and secure detection.

77

All of the other expert system applications in the protective relaying field are concerned with the functions closely related to relaying ra ther than with the relaying functions them- selves.

One area of interest is the relay setting coordination function. Some preliminary research has been reported in the past [40-42], and one of the activities has produced a proto- type system [43]. Similar activities for selection and coordination of fuses in an industrial customer environment have also been reported [44].

Yet another application of expert systems is to aid selection of appropriate algorithms for fault location. A hybrid expert system has been proposed to carry out the fault location function using a combination of known fault location algorithms and expert system support for algorithm classification and utilization [45].

Finally, expert systems are also proposed for digital fault recorder (DFR) signal analysis. Some preliminary results indicate that a combination of digital signal processing algorithms and expert system techniques will have to be used to solve this type of fault analysis problem [46, 47].

As a conclusion, it has also been recognized that integrated control and protection systems may benefit from expert system applications. However, only initial considerations have been reported so far [48].

3.2. Implementation characteristics A survey of the expert system characteris-

tics for the implementations mentioned in the previous section is given in Table 4.

3.2.1. Hardware As can be seen in Table 4, various levels of

PC class microcomputers, higher level work- stations and even microcomputers are used. Only one case indicated that specialized hardware for symbolic processing was used. These choices are made in view of many criteria, including: familiarity, availability, and afford- ability. The use of PC class machines may be introducing development and performance limitations on the expert systems but may be much more practical for the final environment and users of the systems. In addition, the industrial trend is for the PC class machines to become more and more suitable for expert system applications. This is not only due to

78

<

0

0

~'.~. ~ ~ ~ ~ i -~

' ' O?

~ ~i~ ~

~ ~ ~ ~ ° ° - ~

° • ~ > > ~ ~

Z x

o ; >

© o

;>

e~

o

<

©

rapid advances in processors , bu t also to the predic ted in t roduc t ion of AI coprocessors for PC class mach ines in the not too d i s t an t future. Therefore , a c o n t i n u a t i o n and expans ion of the t endency to see m a n y exper t sys tems developed in PC class mach ines is predic ted for the fu ture designs.

3.2.2. Software The sof tware co lumn of Table 4 shows a

spli t be tween the choice of a basic l anguage and the re l i ance upon an exper t sys tem shell. The t r adeof f on the se lec t ion of a shell is usua l ly cost versus ease and t ime of development . In m a n y ins tances the shells m a y not be ta i lo red enough to the app l i ca t ion cons t r a in t s to w a r r a n t the i r use. Mos t app l i ca t ions were developed us ing AI l anguages in con t r a s t to more conven t iona l l anguages . In app l i ca t ions wi th s t r i ngen t execu t ion t ime cons t ra in t s , compi led conven t iona l l anguages , such as F o r t r a n and Pascal , m a y be advan tageous . I f there is a lack of f ami l i a r i ty wi th a l anguage , it is s t rong ly r e c o m m e n d e d tha t a fu r the r ex- a m i n a t i o n of the AI l anguages should be performed, as they can be ex t remely advan- t ageous dur ing cons t ruc t ion and modi f ica t ion of an exper t system. I f in t e r fac ing to exis t ing p rog rams is of ma jo r concern , then m a n y envi- r o n m e n t s exis t t oday which a l low the interface of mul t ip le l anguage codes.

3.2.3. Knowledge representation The knowledge r e p r e s e n t a t i o n is genera l ly

rule based. This is a p p r o p r i a t e and n a t u r a l for m a n y of the cu r r en t appl ica t ions . Cau t ion should be exerc ised as new app l i ca t ions and expanded app l i ca t ions are developed. Rules m a y seem the s imples t to implement , but can often resu l t in pe r fo rmance l imi ta t ions in more complex systems.

3.2.4. Shells The shel l -based sys tems will t end to have

more deve lopmen t aids for user in terfaces . Wi thou t us ing the shells, some app l i ca t ions seem to be still deve lop ing the i r in ter faces , even t h o u g h they are r epo r t ing success wi th the i r decis ion processes. In some a u t o m a t i o n and rea l - t ime app l i ca t ions the user in te r face may be used only in f requen t ly and by rela- t ively h ighly t r a ined experts , thus, fewer user aids need to be provided t h a n in some user in tens ive appl ica t ions .

79

As a final commen t on useful i n fo rma t ion for the expanded deve lopmen t of exper t systems, more discuss ion on the knowledge acquis i t ion process, how the t radeoffs for the exper t sys tem cha rac t e r i s t i c s were made, and the inference s t ra teg ies in the face of uncer- ta in or miss ing da ta would be helpful.

4. RESEARCH ACTIVITIES

4.1. Power system applications This sec t ion gives a s u m m a r y of the re-

sea rch efforts u n d e r t a k e n by the Power Sys- tem A u t o m a t i o n L a b o r a t o r y a t Texas A&M Unive r s i t y and the indus t r ia l affiliates. Be- sides the ongo ing research , some fu ture direc- t ions are also out l ined. The fu ture r e sea rch d i rec t ions are based on the genera l t rends in the field and r ep resen t p roposa l s for w h a t are bel ieved to be the mos t p romis ing topics for an immedia te deve lopmen t effort.

4.1.1. Protective relaying The ongoing r e sea r ch pro jec ts in this a rea

are re la ted to deve lopmen t of new adap t ive r e l ay ing and inc ip ien t faul t de tec t ion exper t sys tem solut ions.

Resea rch has been conduc ted on the use of exper t sys tems as a pa r t of adap t ive tech- n iques for s t anda rd o v e r c u r r e n t p ro tec t ion on d i s t r ibu t ion feeders [49]. Typical ly , g round and phase o v e r c u r r e n t re lays on a d i s t r ibu t ion feeder are set accord ing to the m a x i m u m emergency load which t ha t feeder will encoun- ter. However , much of the t ime ac tua l load va lues are far below these p ickup set t ings. As a resul t , there is a subs t an t i a l loss in detect- ing faults.

With the app l i ca t ion of exper t sys tems to the problem, it is possible to develop an over- cu r r en t r e lay which adap ts to the load levels in a secure manner . The exper t sys tems can a c c u r a t e l y ident i fy o the r events occu r r ing on the feeder in an in te l l igen t manne r , to dis- c r imina te them from o v e r c u r r e n t faults . Us ing such an a p p r o a c h it is possible to m a k e sub- s t an t i a l i m p r o v e m e n t s in the app l i ca t ion of o v e r c u r r e n t r e l ay ing on a d i s t r ibu t ion feeder to de tec t more low cu r r en t faul ts w i thou t in- c reas ing nu i sance outages .

Some of the t echn iques which have been developed for h igh impedance faul t de tect ion, as men t ioned in §3.1.3, are also useful for

80

identifying incipient, low current conditions. This benefit is part icularly applicable in circuits where even a momentary outage or voltage dip caused by an overcurrent fault could result in an expensive loss of a process or other system emergency.

Incipient fault detection is one area where substation functions can be enhanced through the application of expert systems [50]. It has been determined that an expert system approach is meaningful in solving this compli- cated fault detection problem. The detection of such faults requires the processing of many different variables related to intermit tent changes in waveshape and frequency content. To complicate matters, various feeders will often have vastly different 'normal' levels of harmonics and noise conditions. To predict accurately that an incipient fault is present requires a system which can learn the normal state for each feeder and detect when a pattern of abnormalities represents a fault condition. A large number of staged downed- conductor fault tests have provided data for ongoing work in this area. Activity is under- way to develop field-hardened hardware which may be used for protection algorithm validation.

4.1.2. Fault diagnosis As was mentioned in §3, there are a number

of different expert system approaches to the problem of fault diagnosis. The most common implementation is to use a SCADA system as a way of collecting the required field equipment contact data to be arranged as the expert system database. Some of the most recent approaches suggest that a substation-based computer may be used to perform the fault diagnosis function [28, 29, 34].

A research project has been initiated to study an implementation framework for the fault diagnosis function. The main goal of this project is to define different expert system implementation approaches based on various types of equipment used to record and collect relevant field data. It is recognized that the expert system approach is heavily dependent on the type of field data collected and on the level of preprocessing performed by the devices that are used for data recording.

One approach suggested for further imple- mentat ion is a hierarchical expert system approach that utilizes both the alarm and status

E ,,~g. I R,:i;~: T

IRemot e Statiml I Sy~t~qn Otf!~ ~ ] Cvntralized ' ' t ~ : @ "~"C I Location !

Location

. I ~1 7 L

C,mn, , t ion SOgs DFRs ; RELAY> { I{I I .

( ' A S W A S W (' ( 'U, ~, \

Fig. 1. Hierarchical expert system approach. (C, contact; A, alarm; S, status; W, waveform.)

data recorded by SOE and DFR equipment as well as the analog and contact preprocessed data generated by digital relays. A block diagram of the system is shown Fig. 1. Future research activities will be aimed toward defin- ing the implementation framework based on the available equipment and knowledge related to the interpretat ion of the system operation using specific data provided by the equipment.

The implementation frameworks identified so far fall into two major categories. One is the individual equipment category where the fault diagnosis is done by implementing expert systems in the individual substation devices. The final conclusion is reached by comparing recommendations given by the individual devices. This is done at a central point where all of the data from different equipment loca- tions are collected and processed. In another category these are the integrated system solutions where various types of equipment are connected to a common database, and the conclusion is made based on these data. Typical implementation examples for the substation level are integrated control and protection systems (ICPSs) [28]. Examples of the new implementation approach for the entire power system are energy management system (EMS) control centers that use ICPS solutions in place of remote terminal units (RTUs) [51].

Preliminary work has also been done on using an expert system to perform fault diagnosis in a distribution automation system [52]. This function involves the determination of the type and location of a fault on a distribution feeder. Such information can be used in several ways, including the provision of information to trouble crews as needed, and poten- tially to an automatic reconfiguration function for the feeder where this capability is

available. By having faul t diagnosis informa- t ion avai lable to a dispatcher , it is possible to send a t rouble crew di rec t ly to the point of the fault, the reby saving considerable t ime and expense over present methods in which the line crew has to drive considerable dis tances to look for the fault. Where in te l l igent and cont ro l lab le switching devices are avai lable on the feeder, it is possible to use the faul t diagnosis funct ion to provide a means of auto- mat ica l ly reconf igur ing circui ts in a m an n e r which minimizes the ex ten t of the outage caused by the fault.

4.1.3. Field data recording and analysis The recen t u t i l i ty prac t ice in the U.S.A. is to

instal l digital faul t recorders (DFRs) to cap ture faul t waveforms. The usual approach is to have several DFRs connec ted to a mas te r s ta t ion which is located in a cen t ra l office remote from the substat ion. A common problem found in most of these appl ica t ions is the very f requent t r igger ing of the DFRs due to the high sensitiv- i ty of the t r igger ing system. As a result , normal devia t ions are recorded along with the faul ts and d is turbances of interest . Owing to such a large amoun t of data, the system opera tors have a hard t ime going th rough the recorded events t ry ing to ident i fy the faul t events. This is quite an annoy ing and t ime-consuming task and pro- duces o ther problems such as a long transmission t ime to get all of the records to the cent ra l office, and an extens ive memory r equ i r emen t for data storage.

A research ac t iv i ty was in i t ia ted under a u t i l i ty sponsorship to au tomate the process of classifying and analyzing the faul t waveforms. An exper t system is being developed to facili- ta te this task. At present , an 'appended ' approach, shown in Fig. 2, is being implemented. This approach assumes tha t all of the recorded da ta are t r ans fe r red to the mas te r s tat ion, and then a dupl ica te of the da tabase is placed

2= l :'o2- 2 f Exp. Syst.

PC

Fig. 2. An 'appended' implementation approach.

Expert

System

1 Parameter

Calculation

Data Format

Conversion

1 DFR Data

Files

Fig. 3. Expert system functional block diagram.

81

in a dedicated personal compute r which car- ries out the exper t system tasks.

The sof tware func t iona l d iagram of the exper t system implementa t ion is shown in Fig. 3. As may be observed, the exper t system database includes ca lcu la ted pa ramete r s of the recorded waveforms. The ca lcu la t ion of the paramete rs is the most impor tan t step. All of the exper t system rules are dependent on the level of in format ion tha t reflects the parameter change.

Ano the r project in this a rea is sponsored by a manufac tu re r . There is cu r ren t ly ongoing work to develop the requ i rements necessary for fu ture exper t system appl ica t ions in faul t recording [47]. Extens ive work has led to deft- n i t ions for paramete rs which will be necessary to preprocess the massive amounts of da ta ga thered in faul t recording. The abil i ty to preprocess and t rans fe r this da ta in a more t imely fashion than avai lable today will be necessary to provide proper in format ion to exper t system applicat ions. Areas s tudied include channe l informat ion, sensors, and trig- gers, as well as o the r areas which provide subs ta t ion s ta tus informat ion. Methods of r ank ing individual records c rea ted wi th in the subs ta t ion have been studied as well. This r ank ing could be used for au tomat i c processing and dec is ionmaking at a level h igher than the faul t recorder .

4.1.4. System-wide solutions This sect ion repor ts on the exper t system

implementa t ions tha t are based on in format ion obta ined from the ent i re power system. Specific proposals are re la ted to the areas of cus tomer load management , in tegra ted subs ta t ion con- t rol and system-wide pro tec t ive relaying.

82

Substantial research has been done in recent years on the various means to control customer loads for the purpose of limiting electrical demand. While many of these techniques are relatively unsophisticated, there have been some initial approaches at more intelligent techniques for load control and demand limiting [53]. In one specific area of research, techniques were demonstrated for rearranging load patterns according to regu- larly scheduled needs. In this demonstration, the air conditioning and kitchen equipment in a fast food res taurant were operated prior to the noon rush hour to precool the building and preheat the ovens. This approach enabled the controller to establish load diversity during the noon rush hour while still responding to load demands. By taking these approaches it is possible to establish some load diversity even during peak use periods so that all large loads are not on at the same time. With an expert system for this application, it is possible for the load contoller to develop a history of customer load use patterns. Over time, the expert system can develop techniques for in- tervening in precooling or preheating certain loads to anticipate certain use patterns. This approach is inherently better because the expert system adapts to the part icular use patterns of that customer and will adapt to it over time, as opposed to the technique of programming times for the load controller to intervene.

A load management scheme can also be lumped with the incipient fault detection solution to enable full load management under normal and fault conditions. In this case it is desirable to develop a system which can detect incipient faults and reconfigure circuits to critical loads so that service for these critical loads is rerouted before an overcurrent fault occurs. An expert system can perform such fault detection and load reconfiguration functions in an effective manner. In this way, overall reliability to critical loads can be improved with the potential for substantial sav- ings in cost by avoiding process downtime.

Another research activity has been proposed in the area of substation automation. As a result, several computer system architectures are under development for use in sophisticated control. EPRI sponsored work has led to a recommendation for a communication interface specification for use in control and

I

e,o

_ _ ]

elo

HOST

LOC L1

Fig. 4. Automation system architecture: HOST, main computer; LOC, local operating center; RIC, remote intelligent controller; DAC, data acquisition and command.

protection applications, for instance. Work in this area will provide the system tools necessary to implement future expert system applications in control.

Other work provides for the power system automation (PSA) archi tecture shown in Fig. 4, current ly under development. This system is comprised of distributed controllers associated with load-breaking switches. These controllers would allow monitoring of voltages, currents, and other parameters associated with each switch. They could be used in a backup protection mode in addition to being used for implementation of automated restora- tion algorithms. The PSA system would allow for the interconnected control needed for implementing expert system applications at the substation level.

Finally, a system-wide relaying scheme is also suggested as an expert system implementation. It is well known that the existing protective relaying philosophy has an inherent shortcoming due to its inability to adapt itself to changes in power system operating conditions. This has triggered a number of adaptive relaying approaches which are aimed at making adjustments in the existing schemes so that the schemes may be tuned to the changing operational conditions [54- 56]. All of the approaches are algorithmic in nature and are not implemented using expert system techniques.

However, another approach for potective relaying has been proposed recently to enable an inherent adaptabili ty [57]. This approach has been further studied in order to analyze an implementation strategy using ICPS [58].

This implementation direction has been proposed for the new EMS archi tecture where ICPS is used in place of an RTU [59].

A research activity is proposed to study the new system-wide relaying implementation using an expert system approach. The implemen- tat ion is envisioned as a hierarchical strategy where different expert system decisions are made in a distributed fashion at substations. This protective relaying implementation requires new communication facilities for system-wide data exchange as well as dedicated hardware for high-speed execution of the new relaying function.

4.2. Expert system technology Research and development in expert system

technology is widespread and diverse. Lan- guages, compilers, shells and even operating systems for expert system environments are being actively pursued. Knowledge bases con- taining 'meta-knowledge' which help steer the efficiency of problem searches and approaches by the inference engines are being actively integrated into expert systems. Tools to dis- tribute or parallelize the expert system functions onto multiple processors are beginning to appear. For that matter, machines which include multiple processors and coprocessors are becoming more common at all levels of computing machinery. The availability and costs of larger and faster memory devices are having great impact on expert system development. Even coprocessors to accelerate expert system operations are seen in many research efforts. Many of these technological advances are essential if expert systems are going to expand effectively into more online and automated real-time processes. To illustrate these concepts, four examples of technological advances for expert systems will be mentioned.

4.2.1. Real-time expert system shells The shell can be tailored for certain appli-

cations. By tailoring an inference engine in an expert system so it expects and handles con- stantly changing conditions, ra ther than a fairly static list of facts, the execution performance of an expert system can be greatly enhanced (by a factor of 6) in real-time monitoring applications [60]. The fact that the shell is a C-based shell also accelerates the execution time over similar Lisp-based shells. And finally, the ability of the shell to shuffle

83

functions out to multiple processors, when they are available, can also enhance the expert system execution times by a factor of 5.

4.2.2. Dedicated architecture Architectures tuned to the preprocessing

requirements and the flow of the data into and out of the expert system can be much more versatile and robust. Multiple processors for general use, or to maintain a specific function, such as communication, memory access, or searching, can be incorporated in an architecture. The microprocessor boom has made this not only possible, but also reasonably econom- ical. This allows architectures to interface numerical processing and expert system processing more effectively. In one such application, processors for data acquisition, processors for signal processing, processors for expert systems, and processors for communications to other systems can be integrated into a single, robust decisionmaking environment [61].

4.2.3. Customized hardware Customized coprocessors which enhance

expert system performance are being studied. Just as floating-point coprocessors, or FFT coprocessors accelerate those functions, coprocessors to accelerate search and matching operations in expert systems have received much attention. For example, much work is being pursued on accelerating the unification function in Prolog, which unifies facts to conditions in a consistent manner [62]. Several coprocessors for accelerating unification have been and continue to be investigated since a Prolog program spends an average of 60% of its time in the unification operation.

4.2.4. Automatic learner A very important aspect for expanded use of

expert systems will be their abilities to compre- hend and possibly even expand their knowledge boundaries. Currently, many expert systems have no mechanism to account for the fact that they may be working on the edge of their knowledge domain. More work needs to be done to represent their limits and to allow for automated expansion. This puts one in the domain of a learning machine, and many feel the programs called expert systems do not deserve this title until the systems can learn. The concept of unsupervised and/or supervised learning is being researched extensively, and

~4

has l e a d to n e w a reas , or o ld a r e a s b e i n g

r e v i t a l i z e d , s u c h as n e u r a l n e t w o r k s . S o m e

w o r k has b e e n done to a l l o w s t a n d a r d e x p e r t

sys t ems on a m i c r o p r o c e s s o r s y s t e m to in te r -

face to an u n s u p e r v i s e d a u t o m a t i c l e a r n e r [63].

Th i s l e a r n e r w a t c h e s t h e no i s e on a s i m u l a t e d

d i s t r i b u t i o n l i ne a n d c lass i f ies t h e no i s e pat-

t e rns . U n u s u a l l y d i s t a n t no i s e p a t t e r n s a re

f l agged as e v e n t s and s t o r e d for f u r t h e r exam-

i n a t i o n . Th i s too l can s p a w n n e w c a t e g o r i e s as

it sees n e c e s s a r y , a l t h o u g h i t t a k e s h u m a n

i n t e r a c t i o n to g i v e t h e c a t e g o r i e s n a m e s ( such

as c a p a c i t o r b a n k s w i t c h e s , a i r sw i t ch , a r c i n g

fau l t , etc.).

5. CONCLUSIONS

A n a l y s i s of t he i m p l e m e n t a t i o n c o n s t r a i n t s

i n d i c a t e s t h a t e x p e r t sy s t em a p p l i c a t i o n s in

p o w e r sys t em p r o t e c t i o n s u b s t a t i o n c o n t r o l and

m o n i t o r i n g a re q u i t e diff icult . A s u r v e y of t he

e x i s t i n g s o l u t i o n s s u p p o r t s t h i s c o n c l u s i o n

s ince v e r y few a p p l i c a t i o n s in th i s a r e a h a v e

b e e n r e p o r t e d so far. T h e o v e r v i e w of t he

r e s e a r c h a c t i v i t i e s , u n d e r t a k e n a t T e x a s A & M

U n i v e r s i t y w i t h s u p p o r t f rom i n d u s t r y , indi-

c a t e s poss ib l e d i r e c t i o n s in e x p e r t s y s t e m imple-

m e n t a t i o n s t r a t e g y a n d h a r d w a r e / s o f t w a r e

d e v e l o p m e n t s n e e d e d to m e e t t h e r e q u i r e m e n t s .

AC KNOWLEDG EMENTS

T h e r e s e a r c h a n d d e v e l o p m e n t a c t i v i t i e s

r e p o r t e d in th i s p a p e r a r e s p o n s o r e d by a var -

i e ty o f sou rces . T h e f o l l o w i n g a g e n c i e s h a v e

s u p p o r t e d the a c t i v i t i e s : t h e E l e c t r i c P o w e r

R e s e a r c h I n s t i t u t e , N A S A , and the N a t i o n a l

S c i e n c e F o u n d a t i o n . T h e i n d u s t r i a l s u p p o r t

c o m e s f rom P i re l l i , T U E l e c t r i c , H o u s t o n

L i g h t i n g a n d P o w e r , and R o c h e s t e r G a s &

E l e c t r i c . T h e E l e c t r i c P o w e r I n s t i t u t e of t h e

D e p a r t m e n t of E l e c t r i c a l E n g i n e e r i n g a t T e x a s A & M U n i v e r s i t y has a l so p a r t i c i p a t e d

in t h e suppor t .

R E F E R E N C E S

] Z. Z. Zhang, G. S. Hope and O. P. Malik, Expert systems in electric power systems a bibliographical survey, IEEE P E S Winter Meeting, Atlanta, GA, U.S.A., I990, Paper No. 89WM 212-2 PWRS.

2 C. C. Liu. T. S. Dillon and M. A. Laughton (eds.t, Expert System Applications in Pou,er Systems, Prentice Hall, U.K., 1990, pp. 383- 4(18.

3 CIGRE Subcommittee Working Group 02, An interna- tional survey of the present status and the perspective of expert systems on power system analysis and teefi- niqnes. Electra, (123) (1989) 72-93.

4 L. l,. Lai, Application of expert systems to power system protection, Proc. bzt. (?onf. Power ~8~'stem Pro- tection, Singapore. 19~9, Applied Technology Pte., Sin- gapore, pp. 806-822.

5 E. A. Udren and J. S. Deliyannides, Integrated system for substation relaying and control shows benefits, IEEE Comput. Appl. Power, 2 (1989) 21 - 27.

6 M. Kezunovic and B. D. Russell, Microprocessor applications to substation control and protection, IEEE Comput. Appl. Power, I (1988) 16-20.

7 D. Hu. C/C+ + for Expert Systems, MIS Press, Port- land. OR, 1989.

8 B. W. Wah and G. J. Li, A survey on the design of multiprocessing systems for artificial intelligence applications, IEEE Trans. Syst. Man. Cvbern., 19 (1989) 667 - 692.

9 A. Korteniemi, An expert system fault diagnostics of rotating machines, Proc. 2nd Syrup. Expert System Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst., Palo Alto, CA, pp. ] 15 - 120.

10 A. J. Gonzalez, R. L. Osborne, C. T. Kemper and S. Lowenfeld, On-line diagnosis of turbine-generators using artificial intelligence, IEEE Trans., EC-I (1986) 68 - 74.

11 Y. Z. Guo, S. Gao and Y. Yang, A diagnostic expert system for high voltage breaker faults, Proc. IFAC Syrup. Power Systems and Power Plant Control, Korea, I989, pp. 689-692.

12 K. S. Swarup and H. S. Chandrasekharaiah, FDES: fault diagnosis expert systems for HVDC systems, Proc. 2nd Syrup. Expert System Applications to Power ~vsterns (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst., Palo Alto, CA, pp. 294-302.

13 P. K. Kalra, Fault diagnosis for an HVDC system: a feasibility study of an expert system application, Electr. Power Syst. Res., 14 (1988) 83-89.

14 E. N. Dialynas, A. V. Machias and J. L. Souflis, An expert system methodology for determining the characteristics of power system component failures, Electr. Power Svst. Res., 14 (1988) 71 -82.

15 H. E. Dijk, Exformer, an expert system for transformer fault diagnosis, Proc. ~ h Power System Computation ConL (PSCC), Cascais, 1987, Butterworth, London, pp. 715 - 721.

16 K. A. Barrett, Dissolved gas-in-oil analysis: an expert system, Proc. 42nd Conf. for Protective Relay Engineers, Texas A & M Unit,., College Station, Texas, U.S,A., 1989.

17 T. Ueda, A. Maruyama, E. Mori, T. Yamagiwa and K. Nakayama, Application of expert system for diagnosis of substation equipment, Proc. 2ncl Syrnp. Expert Sys- tem Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst., Palo Alto, CA, pp. 206-210.

18 G. Tangem A knowledge based diagnostic system for gas insulated substations, Proe. 2nd Syrup. Expert ,~vs- tern Applications to Power Systems (ESAPS), Seattle,

WA, U.S.A., 198,9, Electr . Power Res. Inst. , Palo Alto, CA, pp. 20/)-205.

19 S. B. Jadid , B. J e y a s u r y a and S. A. Khapa rde , Power sys tem faul t d i agnos i s u s i ng logic p r o g r a m m i n g , Proc. Int. Conf. Power System Protection, Singapore, 1989, Applied Techno logy Pte., S ingapore , pp. 771 -788.

20 P. Heine, J. P a r t a n e n and T. Koppanen , A knowledge based sys t em for fau l t d i agnos i s for med i um vol tage d i s t r ibu t ion ne tworks , Proc. 2nd Syrup. Expert ~vstem Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst. , Palo Alto, CA, pp. 67 - 71.

21 J. J. Ke ronen , Knowledge-based even t ana lys i s in electric power ne tworks , Proc. 2nd Syrup. Expert System Applications to Power ~vstems (ESAPS), Seattle, WA, U.S.A., 1989. Electr . Power Res. Inst. , Pa lo Alto, CA, pp. 232 - 237.

22 P. F a n q u e m b e r g u e and L. Perrot , An exper t sys tem for the fau l t ana lys i s of e lectr ic power sys tems , Proc. 2rid Syrup. Expert ~,s tem Applications to Power Systems (ESAPS), Seattle. WA, U.S.A., 1.989, Electr . Power Res. Inst . , Pa lo Alto, CA, pp. 264-268.

23 Y. M. Park , An exper t sys tem for faul t d iagnos i s in a power sys tem, Proc. IFAC Syrup. Power Systems and Power Plant Control, Korea, 1989, pp. 697 - 701.

24 S. Mu to and Y. Sekine. Moda l i ty based in fe rence for power e q u i p m e n t / s y s t e m diagnos is , Proc. 2nd Syrup. Expert ~'~vstem Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr . Power Res. Inst. , Pal() Alto, CA, pp. 269- 275.

25 Y. Sekine, H. O k a m o t o and T. Sh ibamoto , Fau l t sec- t ion e s t i m a t i o n u s i n g cause effect ne twork , Proc. 2nd Syrup. Expert System Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst., Pal() Alto, CA, pp. 276-282.

26 E. Wal the r , Power ne t work d iagnos i s u s i ng causa l and tempora l r e a s o n i n g , Proc. 2nd Symp. Expert System Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst . , Palo Alto, CA, pp. 290 - 295.

27 P. Kadar , A complex method for fa i lure ana lys is , Proc. 2nd Syrup. Expert System Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 198,9, Electr. Power Res. Inst., Palo Alto, CA, pp. 283-289.

28 B. J e y a s u r y a , S. S. Venka ta , S. V. Vadar i and J. Pos t t a roosh , F au l t d iagnos i s u s i ng subs t a t i on computer , Proc. PICA Conf., Seattle, WA, U.S.A., 1989, IEEE, New York, pp. 289-295.

29 Qi-Hong Sun, Xiao-Mei Guo, Shi-Mo W a n g and Dan- Zbi Xia, A real - t ime exper t sys tem for s u b s t a t i o n faul t d iagnos is , Proc. 2nd S~mp. Expert System Applications to Power Systems (ESAPS), Seattle, WA, U.S.A., 1989, Electr . Power Res. Inst. , Palo Alto, CA, pp. 189 - 193.

30 Z. Z. Zhang , G. S. Hope and O. P. Mal ik , A knowledge- based app roach to opt imize swi t ch ing in subs t a t i ons , IEEE PES Summer Meeting, Long Beach, CA, U.S.A., I989, Paper No. 89 SM 815-2 PWRD.

31 M. Stalder , D. Niebur , L. Pa lmier i and A. J. Germond , A rule based sys tem for s u b s t a t i o n moni to r ing : the sw i t ch ing opera t ions , Proc. 1st Symp. Expert System Applications to Power Systems (ESAPS), R. Inst. Tech- nol., Stockholm, Sweden, 1988, pp. 6-1-6-8.

32 P. K. Dash, K. S. S o o r y a n a r a y a m a and H. P. Kincha , An exper t sys tem a lgo r i t hm for s u b s t a t i o n swi t ch ing

85

problem in power sys tems , Proc. Int. Conf. Power Sys- tem Protection, Singapore, 1989, Applied Techno logy Pte., S ingapore , pp. 789-805.

33 S. lto, I. Hata , A. M a r u y a m a and J. K a w a k a m i , Appli- ca t ion of exper t sys tem to 500 kv s u b s t a t i o n opera t ion guide sys tem, Proc. 1st ,~vmp. Expert ,~vstem Applica- tions to Power Systems (ESAPS), R. Inst. Technol., Stockholm, Sweden, 1988, pp. 6-14-6-21.

34. I. D a b b a g h c h i and R. J. Gursky , I)ecision suppor t sys tem for s u b s t a t i o n upera t ion , Proc. PICA Conf., Seattle, WA, U.S.A., 1989. IEEE, New York, pp. 141- 147.

35 B. D. Russe l l and K. Wat son , Power subs t a t i on au- t oma t ion u s ing a knowledge based sys t em jus t i f ies- t ion and p re l imina ry field resul ts , IEEE Trans.. PWRD-2 (1987) 1090 - 1097.

36 B. D. Russel l and K. Wat son , Knowledge base exper t sys tem for improved power sys tem pro tec t ion and diag- nost ics , Proc. 1st Svmp. Expert System Applications to Power Systems (ESAPS), R. Inst. Technol., Stockholm. Sweden, 1988, pp. 4-18-4-25.

37 C. J. Kim and B. D. Russell , Class i f ica t ion of faul t s and swi t ch ing even t s by induc t ive r e a s o n i n g and exper t sys tem methodology , IEEE Trans., PWRD-4 (1989) 1631 - 1637.

38 C. J. Kim, B. D. Russe l l and K. Wat son , A paramete r - based process tbr se lec t ing h igh impedance fau l t detec- t ion t e chn iques u s ing decis ion m a k i n g un d e r incomple te knowledge , IEEE PES Winter Meeting, At- lanta, GA. U.S.A., 1990, Paper No. 90 WM 043-0 PWRD.

39 IEEE Tutor ia l Course Text , Detection of Downed Con- ductors on Utility Distribution S~'stems, IEEE Publ . 90EH0310-3-PWR, New York, 1989.

40 S. Gora and P. Kacejko, Appl ica t ion of exper t sys tem for de t e rmina t i on of pro tec t ive devices" se t t ings , Proe. 2rid Symp. Expert System Applications to Power ~vsterns (ESAPS), Seattle, WA, U.S.A., 1989, Electr. Power Res. Inst. , Palo Alto, CA, pp. 251 - 256.

41 L. L. Lai, Deve lopmen t of an exper t sys tem for power sys t em pro tec t ion coord ina t ion , Proc. IEE Conf. Devel- opments in Power System ProtectioT~, Scotland, 1989, Inst. Electr. Eng., London, pp. 310-314.

42 S. J. Lee, S. H. Yoon, M. C. Yoon and J. K. Yang, An exper t sys tem for pro tec t ive re lay se t t i ng of t r ansmis - s ion sys tems , Proc. PICA Conf., Seattle, WA, U.S.A., I989, IEEE, New York, pp. 296-302.

43. S. J. Lee, S. H. Yoon, M. C. Yoon and J. K. Yang, PROSET: an exper t sys tem for pro tec t ive re lay se t t ing , Proc. IFAC Syrup. Power Systems and Power Plant Control, Korea, 1989, pp. 1003- 1007.

44 M. Hat ta , H. Sa toh and Y. Serizawa, Appl ica t ion of the exper t sys tem to the de t e rmina t i on of the opera- t ional coord ina t ion of swi tch gears and power fuses for shor t c i rcui t protect ion, Proc. 1st Syrup. Expert System Applications to Power Systems (ESAPS), R. Inst. Technol., Stockholm, Sweden, 1988, 6-27-6-34.

45 A. A. Girg is and M. B. Johns , A hybr id exper t sys tem for faul ted sec t ion ident i f ica t ion, fau l t type classifica- t ion and se lec t ion of fau l t a lgor i thms , IEEE Trans., PWRD-4 (1989) 978- 985.

46 H. M. Sendula , A. M a k k i and A. R. Sanders , Auto- mated in t e rp re t a t i on of digi tal faul t recorder data , Proc. Fault and Disturbance Conf., New Orleans, LA,

86

U.S.A., 1989, sponsored by Transient Recorder User Council.

47 R. P. Heller and B. M. Aucoin, Expert system data requirements for analysis of faults and disturbances, Proc. Fault and Disturbance Conf., New Orleans, LA, U.S.A., 1989, sponsored by Transient Recorder User Council.

48 J. Trecat and W Jianping, Expert system applications in substation monitoring, back-up protection and control, Proc. IEE Conf. Developments in Power System Protection, Scotland, 1989, Inst. Electr. Eng., London, pp. 75 - 78.

49 C. L. Fluornoy, An algorithm for adaptive phase overcurrent protection of power distribution feeders, M,S. Thesis, Texas A&M Univ., 1989.

50 M. Aucoin, B. D. Russell and C. L. Benner, High impedance fault detection for industrial power systems, Conf. Record IEEE Industry Applications Soc. Annu. Meeting, San Diego, CA, U.S.A., 1989, IEEE Publ. 89CH2792-0, pp. 1788-1792.

51 M. Kezunovic and B. Perunicic, Prospects for integrated control and protection system applications in energy management system implementation, NSF Conf. Computer Relaying, Virginia State Univ., Blacks- burg, VA, U.S.A., 1987, Paper No. III-3.

52 R. Balakrishnan and A. Pahwa, A computer assisted intelligent storm outage evaluator for power distribution systems, IEEE PES Winter Meeting, Atlanta, GA, U.S.A., 1990, Paper No. 90 WM 162-8 PWRD.

53 R. P. Heller B. D. Russell, L. W. Perry and G. W. Gerloff, Analysis of energy management system control philosophies utilizing collected field data, Proc. Energy Conservation Technology Conf., Houston, TX, U.S.A., 1983, Energy Utilization Dep., Texas Industrial Commission, Austin, TX, p. 106.

54 G. D. Rockefeller et al., Adaptive protective relaying on transmission power systems for improved reliabil-

ity, DOE Rcp. ORNL/Sub/85-22012/I, U.S. Dept. of' Energy, Washington, DC, 1988.

55 A. G. Phadke, S. H. Horowitz, J. S. Thorp and D. T. Rizy, Integrated hierarchical computer system for adaptive protective relaying and control of electric transmission power systems, DOE Rep. ORNL/Sub/85- 22005/1, U.S. Dept. of Energy, Washington, DC,1988.

56 A. K. Jampala, S. S. Venkata and M. J. Damborg, Adaptive transmission protection: concepts and com- putational issues, IEEE Trans., PWRD-4 (1) (1989).

57 J. Zaborszky, M. Ilic and G. Huang, Inherently adaptive selective protection based on the nested protection unit principle and hypothesis testing, NSF Conf. Computer Relaying, Virginia State Unit,., Blacksburg, VA, U.S.A., 1987, Paper No. III-2.

58 M. Kezunovic, Integration of a system-wide protective relaying scheme in a distributed processing energy management system design, CIGRE Symp. Digital Techniques in Power, Bournemouth, U.K., 198.9. CIGRE, Paris, Paper No. 2B-09.

59 M. Kezunovic, G. Huang and A. Abur, The ICPS application in a new distributed approach to EMS implementation, Preprints llth World IFAC Congr.. Tallinn, USSR, 1990, pp. 25-30.

60 G. Cook and K. Watson, A parallelized rule-based system shell for monitoring applications, Proc. ASME Conf. Computers in 'Engineering, San Francisco, U.S.A., 1988.

61 B. D. Russell, EPRI Phase I Res. Rep. Project 1472-4, Electric Power Res. Inst., Palo Alto, CA, 1989.

62 F. Sibai, K. Watson and M. Lu, A parallel UNIFICA- TION machine, IEEE Micro, 10 (4) (1990) 21 - 33.

63 K. Watson, B. D. Russell and K. McCall, A digital protection system incorporating knowledge-based learning, Proc. IEEE Intersociety Energy Conversion Engineering Conf., Washington DC, 1989, Vol. 1, pp. 239 - 244.

Documents

Expert System Applications to Protection, Substation ...smartgridcenter.tamu.edu/resume/pdf/j/esapp91.pdf · Electric Power Systems Research, 21 (1991) 71 - 86 71 Expert System Applications