Hilbert–Huang transform and decision tree based islanding and fault recognition in renewable energy penetrated distribution system

Abstract

A fast protection algorithm based on Hilbert–Huang transform (HHT) is proposed in this paper for islanding and fault detection, classification, and location in a distribution system penetrated by a solar renewable energy source. The three-phase current signals measured at the substation are utilized to extract the instantaneous features from the first level residue using Hilbert Transform (HT) after empirical mode decomposition (EMD). The feature indices computed per phase for each event using minimum post-disturbance data are then fed to a decision tree machine learning (DTML-1) model to classify disturbances into four classes: normal, fault, islanding, and switching transient. Once a fault is determined, another DTML-2 model classifies the fault into one of the seven types of faults, namely 3P-ABC (three-phase faults), 2P-AB, 2P-BC, 2P-AC (two-phase faults), 1P-AG, 1P-BG, and 1P-CG (single-phase faults). Subsequently, seven DT models based on the fault type locate the IEEE 13 bus system’s zone of fault. Variations in noise levels, DG capacity, and fault incidence angles test the proposed algorithm to achieve a good accuracy level in less time. Switching transients like capacitor switching, feeder on/off, and transformer excitation/de-excitation are also successfully classified within a quarter cycle contaminated with noise levels of 20 dB SNR.