A neural network approach for on-line estimation of partial discharge location in power transformer using advanced correlation technique

This paper presents a neural network approach for online estimation of partial discharge (PD) location using advanced correlation techniques in power transformers. Ultrasonic sensors detect ultrasonic signals generated by a PD and the proposed method calculates time difference between the ultrasonic signals at each sensor pair using the cross-correlation technique applied by moving average and the Hamming window. The neural network takes distance difference as inputs converted from time difference, and estimates the PD location. Case studies show that the proposed method using advanced correlation technique and a neural network estimates the PD location better than conventional methods.