Application of DA-Preconditioned FINN for Electric Power System Fault Detection

Abstract

This paper proposes a hybrid method of Deterministic Annealing (DA) and Fuzzy Inference Neural Network (FINN) for electric power system fault detection. It extracts features of input data with two-staged precondition of Fast Fourier Transform (FFT) and DA. FFT is useful for extracting the features of fault currents while DA plays a key role to classify input data into clusters in a sense of global classification. FINN is a more accurate estimation model than the conventional artificial neural networks (ANNs). The proposed method is successfully applied to data obtained by the Tokyo Electric Power Company (TEPCO) power simulator.