Fault Prediction Method for Distribution Network Outage Based on Feature Selection and Ensemble Learning

Abstract

In this paper, a fault level prediction method for distribution network based on feature selection and ensemble learning is proposed to solve the problem of high, redundant and low accuracy rate of level prediction in power distribution network. First, by preprocessing the fault data of the distribution network, the 23 kinds of initial fault features for distribution networks involving weather, load and equipment are summed up, and a new method of fault grade division for distribution network is proposed. Combined with grey relational analysis, a fault feature selection method which improves the characteristic that Relief-F algorithm can't eliminate redundancy is proposed. Finally, a strong classifier is used to predict the fault level for distribution network based on support vector machine improved by ensemble learning. Then, the proposed method is compared with the other feature selection and prediction methods. Through the actual example analysis, it is verified that the proposed method can effectively improve the prediction accuracy of distribution network fault, and has practical application value.