Thesis

An expert system approach to on-line fault diagnosis in power system networks

Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 1992
Thesis identifier
  • T7421
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • The increased computing power which has been installed in power system controlcentres coupled with the introduction of more advanced Supervisory Control AndData Acquisition (SCADA) systems has increased the volume of data presentedto system operators. In certain instances, control centre staff have failed tokeep track of the interpretation of system conditions following faults. There is arecognised need to provide on-line aids to the interpretation of SCADA systemdata, such as fault diagnosis systems.This thesis reports on research undertaken to produce a practical fault diagnostic expert system which can cope with real-world problems encountered bypower system operators such as protection/switchgear maloperations and missing alarms. The Prolog language was used to represent the knowledge in bothlogic rules and production rules. In view of the scale of the fault diagnosis problem, hierarchical structures were developed for diagnostic conclusions and themeans of diagnosis. Two diagnostic strategies have been developed. Firstly, the“wait-and-see” strategy awaits the appearance of black-out areas and uses theintersection of related relays’ protected areas to locate faults. In contrast, the“hypothesiser” strategy compares activity obtained from qualitative simulationsof fault hypotheses with received SCADA alarms. Both strategies rank competingconclusions according to a developed “relative likelihood” measure. Significantportability has been achieved through the incorporation of qualitative modellingknowledge for power and protection system equipment which allows a distinctionbetween the diagnostic knowledge and network knowledge to be maintained. Theefficiency of the network tracing used in the expert systems was improved throughthe co-operative use of a loadflow routine. Both the desire to include maloperations and to contain, to a reasonable level, the number of possible conclusionsconsidered, were found to be contrasting. The evolved hypothesising expert system is presented as a viable solution. Case studies demonstrate the beneficialinformation provided by the fault diagnosis. They also show the advantages ofthe ultimate “hypothesiser” over the “wait-and-see” strategy, which include aninherent explanation capability.The level of knowledge in the fault diagnosis conclusions opens up potentialfor the expert system’s use as the basis of a restoration advisor.
Advisor / supervisor
  • McDonald, J. R. (James Rufus), 1957-
Resource Type
Note
  • Please note page 135 is missing from this thesis.
DOI
EThOS ID
  • uk.bl.ethos.340000
Funder

Relations

Items