Wavelet Based Fault Classification for Partially Observable Power Systems

Conventional fault classifications methods mostly assume that system is full observable. Then they extract the features from the measured or estimated fault line signals. In this paper, a new WAMS/PMU based fault classification algorithm is proposed which is successful to deal with large scale power systems especially when allocated phasor measurement units cannot ensure full observability over the entire network. In the proposed algorithm, first the faulty subnetwork is detected by discriminant analysis method. Afterwards, detection of fault subsystem is followed by a classification algorithm based on pattern recognition techniques that utilizes wavelet transform as a feature extraction tool. Proposed method is applied to IRAN 128-bus transmission network that is simulated by DIGSILENT and MATLAB. Results are given to validate algorithm performance when various fault types and fault locations are considered. Simulation results show that the proposed algorithm is able to classify fault type perfectly even in the cases that fault is located in unobservable lines and extracting features from fault line is impossible.