A Line-Fault Cause Analysis Method for Distribution Network Based on Decision-Making Tree and Machine Learning

摘要

Line fault is the most important factor which affects the reliability of power supply in distribution network. After failure, it is needed to find fault cause quickly and accurately, which will eliminate potential safety hazard for grid operation from the source and reduce fault probability effectively. At present, analysis for fault cause always depends on human experience, which is lack of scientific quantitative basis and poor in time effectiveness. This paper mines out the gradational structure and space-lime distribution law of fault cause from fault historical records and proposes a line-fault cause analysis method based on decision-making tree. For the decision tree analysis, according to line type, time segment, territorial area, external environment and cause analysis level related with a fault, a modeling method based on machine learning for larger-/subcategory fault cause analysis is proposed and then a multi-level model library for line fault cause analysis is established. Based on fault historical records, the proposed method builds models offline and makes decisions on-line, which improves analysis speed. And the hierarchical analysis process of decision-making tree makes fault cause conclusion more detailed and accurate. Finally, a case is given to illustrate application of the method in detail.