Discovering decision algorithm of distance protective relay based on rough set theory and rule quality measure

• Main Author: Othman, Mohamad Lutfi
• Format: Thesis
• Language: English
• Published: 2011
• Online Access: http://psasir.upm.edu.my/id/eprint/42311/1/FK%202011%2091R.pdf; http://psasir.upm.edu.my/id/eprint/42311/
• Institution: Universiti Putra Malaysia

A distance protective relay performance analysis is only feasible when the hypothesis of expected relay operation characteristics as decision rules is established as the knowledge base. This has been meticulously done by soliciting the relay knowledge domain from protection experts who are usually constrained by their experience and expertise, strenuously manually perusing tremendous amount of data in event report and traditionally relying on such intelligent electronic devices as digital fault recorders, sequence of event recorders and SCADA’s remote terminal units that are lacked of detailed protection information. Thus, this thesis addresses these issues with the objective of intelligently divulging the knowledge hidden in the recorded event report at a relay device level using a data mining strategy based on Rough Set Theory, Genetic Algorithm and Rule Quality Measure under supervised learning within the framework of Knowledge Discovery in Database (KDD) in order to discover the relay’s decision algorithm (prediction rules) and, subsequently, the association rule. The KDD approach was applied on a simulated event report recording ‘extracted’ from a numerical distance relay that had been modeled to emulate an actual distance protective relay device used by TNB, a Malaysian utility company. The high prediction accuracy rate and the close-to-unity area under curve (AUC) value of ROC curve of the discovered relay decision algorithm (prediction rules) from the Rough-Set-Theory-and-Genetic-Algorithm data mining verified the algorithm’s generalized ability to predict as well as discriminate future unknown-trip-status relay events. Subsequently, by post-pruning the relay prediction rules using a Rule Quality Measure known as G2 Likelihood Ratio Statistic as well as the rule-interestingness-and-importance-judgment, a rationalized relay association rule had been discovered. The relay association rule had also been verified as being a reliable hypothesis of relay operation characteristics that was much sought after and easily understood by the protection engineers. The discovered decision algorithm and association rule from the Rough-Set based data mining had been compared with and successfully validated by those discovered using the benchmarking Decision-Tree based data mining strategy. With the association rule in hand, a distance relay performance analysis Expert System called Protective Relay Analysis System (PRAY) had been designed. PRAY had successfully demonstrated how useful it was in implementing the discovered hypothesis as the Expert System’s rule base in the validation and diagnosis analyses of distance protective relay operations and misoperations.