An expert system for the analysis of faults in an electricity supply network: problems and achievements

Lengua Española

12 pages
4 views

Please download to get full document.

View again

of 12
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
An expert system for the analysis of faults in an electricity supply network: problems and achievements
Transcript
  See discussions, stats, and author profiles for this publication at:https://www.researchgate.net/publication/222719213 An expert system for the analysis of faultsin an electricity supply network: Problemsand achievements  Article   in  Computers in Industry · September 1998 DOI: 10.1016/S0166-3615(98)00087-6 CITATIONS 17 READS 37 2 authors , including:Phillip BurrellLondon South Bank University 17   PUBLICATIONS   178   CITATIONS   SEE PROFILE All content following this page was uploaded by Phillip Burrell on 27 May 2014. The user has requested enhancement of the downloaded file.  Ž . Computers in Industry 37 1998 113–123 An expert system for the analysis of faults in an electricity supplynetwork: problems and achievements Phillip Burrell  ) , Dave Inman  1 Knowledge Based System Centre, South Bank Uni Õ ersity, Borough Road, London, SE1 0AA, UK  Received 19 March 1997; accepted 10 February 1998 Abstract FAUST 3 is an Expert System for diagnosing faults occurring in an electricity supply network. Its task is to diagnose, inreal-time, faults which happen on a large and highly interconnected network comprising items of equipment of several types.A complex representation of the network as a collection of ‘objects’ is the means by which diagnosis is made possible. Thesystem performs its diagnosis by means of a ‘constrained generate and test’ approach. Hypotheses are formed, and thenexamined to find the one most likely to account for the available evidence. The system uses causal reasoning, heuristicknowledge and techniques such as the use of blackboards for communication amongst system modules. q 1998 Published byElsevier Science B.V. All rights reserved. Keywords:  Blackboard; Causal-reasoning; Electricity-network; Fault-analysis; Objects; Real-time; Simulation 1. Introduction The UK supply industry has a number of indepen-dent Regional Electricity Companies whose task it isto maintain the electricity supply for their region.Although the detailed operation of each companymay vary, the overall functioning of each is the sameand that is, to ensure that all customers have a secureelectricity supply. This involves both pre-emptivemaintenance and fault correction.The supply is controlled by the company at bothArea and District levels. Each of these has controlrooms where the control engineers monitor the elec- ) Corresponding author. Tel.:  q 44-171-815-7408; e-mail:phillb@sbu.ac.uk  1 Tel.:  q 44-171-815-7408; e-mail: dave@sbu.ac.uk  tricity supply network by means of telemetry signalsfrom the various items of equipment in network.These telemetry signals indicate changes of state andalarms from this equipment, and these form threemajor types: Ø  Reconfiguration of the network. Ø  Pre-emptive maintenance. This is referred to as an‘outage’. Ø  Faults on the network.The first two of these require no action on behalf of the control engineer past that of noting how thenetwork is reconfigured. For the third, the controlengineer must decide if the network has been leftunstable and whether any part has been shut down.Appropriate action then needs to be taken to eitherre-configure the network or to despatch appropriateengineers to repair the fault, or a combination of both. 0166-3615 r 98 r $19.00 q  1998 Published by Elsevier Science B.V. All rights reserved. Ž . PII: S0166-3615 98 00087-6  ( )P. Burrell, D. Inman r Computers in Industry 37 1998 113–123 114 The nature of diagnosing the faults can often bedifficult, in that the sequence of telemetry signalsreceived by the control engineer are not always inrelation to the time the fault occurred, or that someitem of equipment fails to respond when it should orwhen there are multiple, simultaneous faults. In or-der to assist control engineers with difficult faults,the Fault Analysis Using Simulation and Test project Ž .  w x FAUST was developed 4,6 . This was part of theLarge Demonstrator Project on Mobile InformationSystems established by the UK government’s Alvey w x Program 19 .FAUST’s tasks were to identify faults in an elec-tricity supply grid by the interpretation of the grid’sautomatic protection system and to confirm zones,isolated for the purpose of maintenance, by theobservation of other switching activities. Messages,which form the basis for diagnosis, arrive by meansof telemetry signals from items of equipment beingcontinually monitored from within a region of thegrid under consideration. FAUST can cope withseveral different faults simultaneously and eventswhere some of the incoming data is incomplete, duethe loss of messages in transit.FAUST presents its information to the userthrough a multi-window environment, displaying thediagnoses as they are formulated, the incoming mes-sages on which they were based and graphics relatedto the section of the network under consideration.The paper presented here is an enhanced version w x of a previous paper presented by the authors 12 .This paper describes more fully, the overall develop-ment of FAUST. The paper describes the problemsassociated with the analysis of faults occurring onthe supply network and how these were representedas system knowledge. The overall operation of thesystem is then described together with the specifica-tion for on-line tests and system validation. Thepaper finally concludes with the lessons learned fromits development. 2. The network protection system Ž . The Circuit Breaker CB is the most importantsingle piece of equipment that is used to automati-cally protect the network against faults. Other de-vices, which we describe here as actors, are used toisolate parts of the network but these devices, suchas line isolators, are only used once the circuitbreakers have operated.Because the protection system of each ElectricityCompany varies in terms of specific equipment used,we describe here the protection system so as to showeach has a similar way of monitoring and controllingthe process of maintaining the network. The controlprocess is split broadly into three parts: the sensors,the control logic and the actors. 2.1. Sensors There are a wide variety of sensors employed onhigh voltage supply networks. They consist primarilyof voltage and frequency sensors used to detectabnormal conditions on the network. For example avoltage drop, or frequency drop might indicate amismatch between supply and demand. In additionthere are temperature sensors, pressure sensors andswitch sensors. The former two are primarily con-cerned with transformer malfunction, and early de-tection is necessary to prevent expensive damage.The latter is used to detect unusual activity at aremote site, typically for security, such as doorsopening. For fault diagnosis we are concerned pri-marily with voltage and frequency sensors. 2.2. Actors If abnormal conditions are detected, then actionneeds to be taken. This consists of isolation of thefaulty part of the network. CB’s are the primarymeans of doing this, and due to the high current thatthey can take, and associated cooling, are expensive.Cheaper auto isolators are thus used to isolate partsof the network that are already dead. They do notneed to handle any current, and so are much cheaper.Finally fault throwers are used to trip the circuitbreakers. They do this by earthing a part to thenetwork, usually transformers. These actors can iso-late until action is taken, or can have a time delaybefore they attempt to re-connect the appropriate partof the network. Some actors are linked so that isola-tion at one point in the network automatically iso-lates other parts of the network.  ( )P. Burrell, D. Inman r Computers in Industry 37 1998 113–123  115 2.3. Control logic The logic used to activate the actors in the desiredsequence falls into several categories. These aredesigned to cause only those actor nearest to thefault to trip, so isolating the most constrained areapossible. 2.4. Time magnitude Each circuit breaker has a time delay that can beset. CBs near the supply have a longer delay toensure that CBs nearest to the fault on the supplyside trip first. 2.5. Current magnitude As for time delay, so CBs can have a thresholdvalue set for the fault current. As the fault current ishigher near to a fault, by setting this thresholdappropriately, only those CBs near to the fault willtrip. 2.6. Distance measurement  Each CB can detect the distance between itself and the fault. If the fault is further away than thenext CB, it does not trip. 2.7. Power direction comparison If the direction of current entering and leaving azone changes, then a fault must exist, and the zone isisolated. 2.8. Phase comparison As for power direction, so the phase of the zoneends should be the same. If not the zone is isolated. 2.9. Combination logic Most single logics are rarely used alone, as theyare less reliable than combination logic. Thus timeand current magnitude, and time and distance mea-surement are often used together. 3. The reasoning strategy The most commonly used method for diagnostic w x expert systems is by the use of rules. Cooper 8 , w x w x Sakaguchi and Matsumoto 16 , Liu et al. 14 and w x Gann et al. 10 use these methods applied to variousproblems in electricity distribution. These types of ‘surface’ systems have their limitations in that theycannot diagnose faults that are not included in the w x rule base. Komai and Sakaguchi 13 , Cardozo and w x w x Talukdar 7 and Talukdar et al. 18 include rule-based simulators in their systems, producing ‘deep’causal models. These have an advantage over‘surface’ systems in that they have the ability tohandle incomplete and inconsistent information.FAUST uses a hybrid of ‘surface’ and ‘deep’ strate-gies whereby a ‘surface’ model is used to producepreliminary hypotheses about the distribution system.These hypotheses are then simulated using a ‘deep’causal model and the results compared against realobservations.This is achieved during a one-minute cycle inwhich a stream of telemetry messages arrive inreal-time and are then filtered in order to select thosethat are potentially useful. Messages are selected onthe basis of originating from a possible fault orpossible maintenance and are distributed accord-ingly. Hypotheses are generated using a hybrid strat-egy, mentioned above, in conjunction with an inter-nal model of the network, and time related con-straints of the messages. The system executes bymeans of a number of program modules whoseinputs and outputs communicate through a numberof ‘blackboards’. A blackboard, in this context, is amulti-clause Prolog predicate. A summary of thereasoning process cycle is shown in Fig. 1. 4. The network database The system possesses an internal representation of the grid in the form of different types of objects. Thehierarchy is arranged so that objects which are lowerdown can inherit parameters and methods from thosehigher up. The database stores, in a hierarchy of textfiles, the information necessary to build the network model. Each line of the textfile describes one object.  ( )P. Burrell, D. Inman r Computers in Industry 37 1998 113–123 116Fig. 1. Summary of the reasoning process. When the network model is constructed, the textfiles are loaded, starting at the highest object in thehierarchy, and objects are created, as the data is readin. Four of the object types, in descending order of hierarchy, are: ‘Boards’, ‘Districts’, ‘Substations’and ‘Circuits’, ‘Substations’ and ‘Circuits’ are com-posed of ‘Electricalitems’, and these contain infor-mation about individual item of plant. ‘Electricali-tems’ are divided up into the hierarchy shown in Fig.2.
Advertisement
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks