Learning Diagnostic Rules for Power Distribution Systems

Abstract

This paper discusses the machine generation of diagnostic rules for fault diagnosis in power distribution systems. The Machine Learning technique that we have implemented processes examples of fault events with the associated diagnoses (from records of previous errors), and derives rules that correctly classify the available examples. In order to formalize relevant domain knowledge and to build adequate diagnostic rules, first order concepts had to be introduced. The paper describes the existing prototype RUDI (Learning Rules for Diagnosis) and the initial results of the test phase.