Ensemble decision trees for high impedance fault detection in power distribution network

Abstract

The paper presents a new technique for high impedance fault (HIF) detection in power distribution network using ensemble decision trees (random forest). Giving the randomness in the ensemble of decision trees (DT) stacked inside the random forest (RF) model, it provides effective decision on HIF detection. The process starts with estimating the amplitude and phase of harmonic contents (fundamental, 3rd, 5th, 7th, 11th and 13th) in the HIF current signal using Extended Kalman Filter (EKF). In the next stage, random forest is trained with the amplitude and phase information of the HIF current signal to build up a highly efficient classifier for HIF detection. While testing, the proposed RF based classifier provides HIF detection with more than 99% reliability, considering extreme operating conditions of the power distribution network. The results indicate that the proposed method can reliably detect HIF in large power distribution network.