DT based intelligent predictor for out of step condition of generator by using PMU data

Abstract

Rotor angle instability of synchronous generators following large disturbances is regarded as a major threat for system dynamic security. Fast detection of out of step condition and disconnection of unstable generator is a vital requirement for avoiding any mechanical and thermal damage to the generator. In this paper, for early prediction of out of step condition of an unstable generator, a three stage decision tree based approach consisting of Fault Detector DT (FDDT), Clearance Detector DT (CDDT) and Instability Predictor DT (IPDT) is presented. The proposed approach by using a moving time window consisting of on line sampling vector data acquired by PMU, at the first and second stages detects the occurrence and clearance of the fault by FDDT and CDDT respectively. Then at the third stage IPDT will be activated for predicting out of step condition. The input data to each DT is a moving time window consisting of six consecutive sampling vector data. The nature of data used for FDDT and CDDT are electrical variables while the data used for IPDT is post fault mechanical variables of generator like angle, speed and acceleration of rotor. The proposed algorithm can be used as an out of step relay which could be trained and designed for each specific generator with respect to any fault or abnormal condition which threat generator stability. The proposed algorithm is demonstrated on generators of IEEE 39bus test system with promising result.