Expert system for the integrated management of a power utility's communication system

1208 IEEE Transactions on Powcr Delivcry, Vol. 14, No. 4, Octobcr 1999 EXPERT SYSTEM FOR THE INTEGRATED MANAGEMENT OF A POWER UTILITY'S COMMUNICATION SYSTEM C. Leon', M. Mejias', J. Luquel, F. Gonzalo2 Facnltad de 1nform:ltica. Universidad dc Sevilla. 4 1012. SEV1LI.A. Spain 'Scvillana de Electricidad. Avd de la Borbolla sin. 4 1013. SEVILLA. Spain I Abstract: hi this papcr we present a nile-based cxperl systeiii applied io the hilt diagnosis in tlic conununicatioii system of a power ulility. diCficull to develop ci"Cectivc TMN systems. I The cxprrl system is part oSthc system dcvoted lo the inaiiagemciil of il powcr utility's conuiiunicalion system, Iluit we called NOMOS. systelns 121, slipportiug in uliiquc all Actually, NOMOS is managing the coiiuiiunication systems VI' Sevilleiie tlc filcctricidad a1111 IJni6n Elbclrica I'cnusa (two major Spanish powcr utilities) T~ facc ~l~~~~ problellls the integrated TMN management solulions i nrplcincnted by differenls vendors. The use of expert system teciiiiology is an alternative in order to improve the quality oC service and to mitigate soine of the lellls, ~o,lunull~cations syste,n faanlt d,agllosis, existing coinniunicalioii system Iiianageincnt difficulties [1] [4] Man;lgcrnc,~~, Commuiiicaiio~i cqoipmcnt, SCADA syslem, Is in this coiitexl that milst be located the Project NOMOS. Conunuiiiulion system operations and m;inageinciil, I'rolocols, accomplished by two Spanish power utilities: Sevillana de Communicalion stendads, Electricidad (CSE) ;itid IJnioii Eldclrica Fciiosa (UEF). in collaboration with tlic Deparlniciil of Electronic Teclinoloby or the University of Seville. I. INTRODUCTION Traditionally, power titilitics over the world llavc Owned In lliis papcr, we present the main cliaracteristics of the private comminiication syslems, utilized liiailily [Or aclivities NOMOS systein. Also, we will describe Ihe NOMOS expcrl directly related with their produclivc process. During lastyears, systcm applied to Ihc network faull diagnosis of one of tlic CSE diKercnl Caclors 111 have made ticccssaly to Clllpb cotnintinicalion systcms, llic microwave network. Finally we Tclecoiiiiiimdcalion Mallagcliiclll Network Systcllls (TMN). 111 will expose some of the results obtained during tlic cxpert a short relation wc considered tbe following Factors: system operation. -Change from an analog cnvironoient to a digital one. -1ncrcasing integration of the cominiinicatioii systems. -Rising complcxity of the coniniunicalion systems, from botli point of view, the size and the managenicnt. -1ncrcasing strategic value or the commuliications. -Priority of decreasing llic operation and maintenance COSIS. It is also iiccessaiy to consider the prcsencc of a great wiricly of solutions offercd by vendors supparling n~anagemcnt for their own equipmcnts only. Other problem are the high cost of equipmcnts, the fraginentary inaiiageineiit of tlic conrmunication system and the need of qualified personiiel working full-lime in management. These problems inakc ~E~+~o-~wRD-o-I 1.1998 A paper recommended and approved by the IEEE power System Communications Committee Of the lEEE power Engineering Society for publication in the iEEE Transactions on power D~I~V~V. Manuscript submitted July 22, 1998; made available for printing December 2, 1998. I[. THE NOMOS SYSTEM NOMOS is an integrated cxperl systcin applied to tlic nlailagenient of a power utility coniinuiiication syslein [5]. Tlic NOMOS system Ins to intcgratc the inanagcineiit inforination generated Cram these cornininiication syslcnis (Fig. 1). NOMOS uses an cxpert system whicli liclps the operator locating and resolving iielwork failures. The tnain NOMOS systcin features are' Pig. I. iixainple ofNOMOS system screen 0885-8977/99/$10.00 0 1998 IEEE 1209 The standard GDMO [9] is used to define the managed ob,jccts iii lhc NOMOS system. In these standards, a double liicrarchical stnicture is defined: the type stnicture and the object stmetiire. The object classes are hierarchically stmctured in such a way that the lower ob.jcct classes inherit the reatures from upper object classes (object oriented technology). On the other side the objects are stmcturcd in a fiinctiooal reliance tree. Both structures do not overlap usually. Unfortuirately. the GDMO do not have syntax orieiitcd to its aiitoinatic reading. In llic NOMOS system an automatic processing oilhe object types is required For this reason, based on GDMO. a modified synlax lo dclinc tlic ob.jecr types has bccu proposed. in such a \\ay tllat it allows its aulomatic processing. Fig. 2 shows iiii esaiiiplc or a dialog box where the System Object Classes of a network are clclincd in NOMOS. Tlierc is an Object Class called "MDM 3418: Third Order MUX". Ir the "Modify" buttoii is pressed, Ilicn llic dcCiiiitiorr or an Object Class dialog box (Fig, RJ will be shown. That dialog box allows describing tlifEcrciit characterislics or llie olject class, such as allributes. actions, xicl so 011. V. EXPERT FAULI' MANAGEMENT As an cxaiiiplc or tlie expert reatures iiicluded iiito the NOMOS syslcm we describe tlie case of the microwave coiiiiiitiiiicatioii syslciii fault diagnosis. A. lhe Micmisavc (hiiiiiiunicnliuil .\.v.siem Tlic CSE iiiicrowavc coinuiunicalioii syslcm is coinposcd or cligilal Iransuiission/receplioir equipment via radio in lhc UHF and SI-IF bands. digitxl iiiultiplexors and auxiliary cquipmeiil. which eiiiploy llic Pulsc Code Modulatioii technique. and a iiiiiitber of arisiliary equipment, installed into 39 stations covering ail area or 100.000 ko2. The management of the iiiicrowavc syslcin was made, before -Aiitotnated conversion between communicalion protocols. -Description of network cleineiils cinployiiig iiitcriiatioiial standards (IS0 and ITU-TJ, rnaiiily GDMO (Guidelines for Defiriitioii of Maiiaged 0b.jectsJ. -An easy definition of tlic rules lo achieve the expert faull diagnosis. NOMOS is today iinpleinciilcd Cor the integration or six difrerciits private conimunication systems, bcloiigiiig to two different power utilities: Tlic microwave iiclwork, the Cross- Conliecl Digital iielwork of TITAN and the PABXs network ALCATEL 4000. all tlicrri beloiiging to CSE. The network or lransmission cqoipmenl of both NOKIA and ALCATEL, ;ind. tlic modem network of TELDAT. all bcloiigiiig to UEF. Ill. PROTOCOLS A first key issue or tlie iiitegralcd management or a coiiunuiiicatioii syteiii is the iiitegralioii ol the different prcsciit and rimire protocols used by the dirkrent devices prcscrit in tlie iietwork. To get this target we have taken advaiitaigc or the use of ESTELLE [61, which is based on tlie Formal Description Techniques (FDT), and is staiidardized by [SO. The integration orthe protocols is obtained by the dcfioilion ofevery protocol using ESTELLE. wliich provides aiitoinatically the corresponding cxcculable code. Every parlicular protocol iiscs CMLS 171 sciviccs lo set the comiiiunicalion with the NOMOS system CMIS services havc becu standardiml by IS0 lor its iietwork open managing architecture. This way of iiiiplcniciitation does not ineaii that the several irlailaged stibiietworks intist bc operated using tlic CMlP standardized by IS0 as llic inanaging protocol. Iiistead. il iniplies that cvciy pirlicolar prolocol is adapted 10 llic CMlS services. 111 this way, on every single protocol, a service adapter layer is added orleering 10 llic TMN syslciii a coilinion API (Application Program Interface). In this coiiiiiioii API it can bc plugged so many new protocols as required, sincc all or them are seen in the same way by NOMOS, achieving a Cull iiitegralioii. 1V. OBJECTS Otlier key issue of tlic integrated managemenl of a coiriiniiiiiCatioiis sysleiii is tlie integration of Ihe dirrcrcnt equipments (objects) 181 in a single TMN system. One of tlic main problems to get this integration is tlie very different classcs olobjccts used in tlic coiiiinuiiicatioii systcius. Every oiic has its owii reatures and behaviors. and no coiiiinoii dcfiiiilioii can bc used €or most oflhcm. IS0 has scI scver;il standards to define, in a compatible way. any object in tlie comiiiimicatioii systems. Pig. 2: System Oli,iccl Classes Dialog Box 1210 Pig. 3: Ol,jact Definition Dialog Hon NOMOS operation, using a sipenision system called SSC (Communication Siqmvisory Syslcmn). This systcln can monitor, in rcal time, tlic network's main paramctcrs, making use ortlic information supplied by a SCADA system, forrncd by a Control Center (placed on the miin CSE building), and RTUs installed into diKcreiit stations. The use of a SCADA system is due to llic inanagcment limitations or the microwave communication eqnipment. The SSC allows the operator to acquire ii~ormnation (alarms or digital and atlalog parameters ormeasure) registcrcd 011 each RTU. Starting from the supplicd infornution, the operator is able to inidertake actions through tlic SSC in order to solve tlie failures that could appear or to send a technician to rcpair the stations equipment. Usually the operator does 1101 supervise the SSC constantly. Under iiortnal system conditions, the operator controls periodically tlc list of active alarms looking for tlw prcscme of possiblc Cailures, unless the opcrator recives a warning abont relcvaiit failurcs in the communications, which need lo bc resolved at once. In tlic SSC operation, two tnaiii problems arisc: I) Tlic SSC shows to the operators a great aiiiount of rednndallt information in the case of a relevant coiiiinniiicatioii equipment failnrc. So operators usually are not able to resolvc specific faulty sitnations. 2) Tlic iicetl of high-qualified operators, becausc tlic SSC does not provide cxplicil infornlatioii about tlic railtires that can bc the origin of each alarin or group of alarms. B. A~OhfOS expeuf,fuult diagnu.sis In order to solvc these problcms, NOMOS cxpeit system acts collcc~ing the SSC infortnatioii and applying niles that conlains the opcrators' knowledge. Tlicsc iules carry ant the following tasks: * Rcdinidaiit alarms are filtered in order to avoid the salmation of tlic capacity of tlie operator. This Altering is not accoinplished iiidiscriiiinatel~~. Those alarms that show a service degradation or Ihat can mask certain Mores are no1 climinaled. * Sytetn faihircs arc dctcctcd using llie iiifonnafion collected by tlic alarms. * A scvcrity index is created to givc priority lo the a ' 1 arms. * Soiiic recoininendations are shown to tlie network operators in order to resolve Lilurcs or to improvc tlic system operation To leacli lhc cslablislied objectives we llave nscd a Rapid Prototypiiig nictliod 1101. Wcliaveused the Stnicturcd Interview as knowledge acquisition method [ 11 I, representing the declaratory kiiowledge in the [orin of&icts (alarins collcctiiig by llic SSC) and tlic procedural kiiowledgc in the form of If-Then niles (iiirormation supplied by tlie expert about alarms processing). To implcmciii tlie expert system included into NOMOS, tlic shell ART-IM of Iilference Corporation was selected. It lias been employed a workstation SUN SpircStation 10 to prograin tlie cxpert system. In thc casc 01 the iiiicrowave network, the resoltant expert system has about 200 nilcs. The expert rulcs arc specific for each iiiaiiagciirciit cnvironmeiit but tlicy liavc soiiic coiiiiiioii characieristics: * Thc rules arc valid Cor each alarm or gronp of alarms regardlcss tlic place where alarms occur. * The relatioiisliip among niles is based 011 tcinporal siiniiltmeity criteria Thcrcfore, several alarms arc not rclatcd if they takc place witliin a temporal interval greater than a ceriain fixcd value. * The rnlcs arc structured in a rnodular way to make easy to add new cxpcrl knowledge. In the casc ol the cxpert system for thc microwave network, we can show soinc types ofmles: Zi .Supres,sioi? ru/e.s. Tlicsc niles were designed to filtcr those alarins that appcar and disappear in a reiterativc way in short tiriie intcnds. The nientioiied filtering is cssciitial if we wishcd the system opcrating in real tiinc. It 1x1s the mission of incrcasing the process spccd, rcducing the set of fact on which arc applied the rulcs oftlie expert system. Tlic filtcr also permits to reduce the quantity oCrcdundaiit inforination that is displayed to the operator. An example of supressioii rnlcs is showed in tlie Tab. I Tlic liltcriiig only affects to tlic cxpert system, becaiisc the NOMOS operator has available a Log oTAlarnis where all tlic 1211 indications from llic SSC are collccted. Bcar in nulid that expcrt system acts as a liclp systcin iiitcgratcd in an integrated inairagcment ciivironineiit. The opcrator is the one who decides about tlic convcnieiicc of followiiig tlic espcrt syslcrn adviccs. 2) Rules dealing with alarms excluded from the ,filtering process. There is a set of rnlcs that act on the alarms Lllat spcciiically havc been exclndcd from the lillering, due to llic kict that the appearance and disappearancc of such alarins gives inCoriiuilion about possible railnrcs in tlrc microwave communication systern.Tlicsc rules usually cvalicitc llic repeated appearance and disappcarancc of an alarm, throughout a givcii lime iiilcrval. 3) Rule.? dealing with ,failure.r in !he auxilliary equipnrent. Tlic alarm associated lo sucli Wilurcs iisually appcar relatcd to alarms on tbc traiisiiiissioiilreceptioii eqiiipmcnt because a hilure in soinc oC the physical devices (sourccs, wavc-bvides, ctc) will CBUSC problcrns in traiisinission or reception 4 i?ule.r to check the pre.rewce of an alarm during extewded time intervds. In certain cascs, it is necessary to evaluate the presence of a specific ;ilarin during a given period of tiine in order to identiry a railure. An example or these niles is SSR59 (Tab. 2). In this nile the voltage in a geiicral distribution panel (TENSION CUADRO DIS GEN) mist be lower lhaii an previonsly spccified limit during at Icast 30 minulcs to determine tlic existence of a failure in thc power sopply nnit. C. .!VOMO,S exper! ,sy,slenr re.sult,v In this subscctioii wc suniinaiisc soine results ortlic opcration during the last year of a NOMOS cxpcrl systcrii applied to microwave coimnunicatioii syslern fadl diagnosis. We dividc Ihis analysis in difCercnt scclions: f) Fikring EJectiveness. For the last year. system filters on average. a loral or 93.7% of the alarms coining from the SSC cveIy day. In thc worst case, this figure is greater than 89%. Each failurc produccs, on average, 13 significant alarms and 200 fillercd alarms. . . .... . . . . . RULE 'IO CI1ECK AN KI.I\RM DIJIIINO CI~I~IAIN INII1l<\;\L 2J Localion Precisiun.The oblaiucd results show that, in the 50% of tbc CBSCS, thc expert systcin ofleers only one location. (understood as conimunication equipmclil, or coininuiiicatioii cquipmcnt eleinciit. and llic slatioii wlmrc it is installed), (or a dctcrniincd failure. In a 39. I% orllic cases it indicates a doriblc location. and only iii 10.9% of the cases, il oCfers a triple location. Never it is proposed inore than tlircc possible locatioiis for a Pailurc. So wc have, on average a total of I .6 indications of locatioii for each Pailnre. In order to evaluate thc certainty rcgarding Llic localion, wc delilie a /.ocu!im precision (LW paraiiieter (1): In such way, irtlic Number of Failiircs is equal to the Number of Indications, thc precision will be 100%, wlrilc if the Nnrnbcr of Alariiis is equal to the Nnnber or Indications ofrered by the cxperl system, the prccision will be 0%. Dnriiig last year system operation. wc obtained, on average, I3 alarm Cor each hihire. As the avcraagc iininber of indications rcsultcd of being I ,6, then the location precision is 95.2%. 3) Locullon ,Siicces~. hi addition Io [lie niimbcr of locations with failure that is reprcsciited by a deteriniiicd group oralarnis, the cxact placc wberc the fault has been produced is always defiiicd by onc of the locations oked by the expert system. In coiisccnencc tlic localion .siiccecs is 100%. 4) Processing lime. Oiic of tlic targets to start the devclopinent stage of thc expert system, was the possibility of real tiine operation. A year of system operation allows LIS to ,1212 confirm that the time for processing groups of tip lo 600 alarms is belov, to 5 seconds (Fig. 4). If wc lake into accomi( llral tllc biggest number observed during NOMOS cm'rciit opcratioii is 260 alaflns per minute, then llle rcquirements for real limc opcration are fulfilled. [SI IS0 10165-4. "Sluclurd ofmaniigemcnl inhmmtian. Pail I' Guidelmen for tho deliiritioii of ~maiiagcd oI>jccts''. i ltil ti. I,. ;vlc(;lawaiid ti. Marbisoll-Uripg~,~~novisd~eA~~ur,s~lion:P~incrpl~,s und (iiiidelrne.7, I'rcntico-Hall International, 1989. ll I] I?. M. ,\wad, niiiiding Experi sJUlem.?, Wcrt Publishing Company, 1996. VI. CONCLUSION VIII. BIOGRAPHIES NOMOS offers a solulioti to the growing demand or an integrated inaiiagernetit systems for power utilitics communication systems. To sum up, within Ll~c dcvclopcd sysfcvn, it is relevant the expert sysleiii tcchnoloby for fault diagnosis which has been employed. System operation during lasl ycar shows NOMOS performances in ordcr lo improvc the managenicnt of CSE and UEF coitimunic;itio~i systems. 151 [71 181