Prediction of Failure Counts Caused by Weather Factors in Distribution Systems Based on Quantile Regression Forests

Abstract

Distribution systems are prone to occur failures under unfavorable weather conditions, where the probability of occurring failures in wide areas improves. The reliability of distribution is threated by weather-related weathers. In order to improve the ability of responding to weather-related failures, prediction of weather-related failure counts caused by weather factors is necessary, which can help the utility companies to make adaptive operation and maintenance plans in advance, and then help to quickly restore power supply. In this paper, Quantile Regression Forests (QRF) is used to predict the weather-related failure counts in distribution systems based on actual weather parameters. We evaluate point estimate ability of the proposed method using failure dataset from a city in Northeast China. The experimental results show that the proposed method can provide powerful point estimates. In addition, the proposed method can give a reasonable prediction interval, i.e. the prediction uncertainty can be considered in the proposed method. Considering prediction uncertainty into prediction model will make a further contribution to the operation and maintenance decision against weather-related failure risk of utility companies.