Partial Discharge Location With Gradient-Descent Total Least-Squares Euclidean Direction Search Algorithm for Cable Systems

Based on the total least-squares (TLS) model, the gradient-descent TLS Euclidean direction search (GD-TLS-EDS) algorithm is proposed when both input and output signals are corrupted by noises. Taking advantage of the effectiveness of the EDS algorithm, the GD-TLS-EDS algorithm has improved performance and comparable computational complexity. Moreover, based on the time difference of arrival technique, the traditional location methods may suffer from performance degradation in the location problem with noisy inputs, the GD-TLS-EDS algorithm is exploited to mitigate the noise from the original and reflected partial discharge signals and achieve the one-step location result by estimating the time difference in cable systems. Simulation and experimental studies demonstrate the GD-TLS-EDS algorithm has improved location accuracy as compared with the existing algorithms.