Simulation of partial discharges and implementation of noise elimination techniques

Abstract

High Voltage Electrical Equipment such as Power Cables, Transformers, Circuit Breakers, Instrument Transformers, Insulators, Bushings, Generators, Motors, Surge Arrestors and Capacitors are the vital part of any power system distribution & transmission network. These High voltage products are highly prone to failures due to partial discharges, during operation. An impurity or void present in the insulation of an electrical equipment results in localized dielectric breakdown called partial discharges within the insulation. Partial discharge measurement is the most important diagnostic test to locate incipient fault in electrical equipment during service. However, on-line partial discharge measurement at site, are hampered with external and internal disturbances, which decrease the sensitivity of the measurements. The three common types of disturbances which are hampering the partial discharge signals are narrow band interference, broadband noise and pulse-shaped interference. The measurement of exact partial discharge magnitude and correlation of the same with the condition of insulation with respect to location, decides the life expectation of the Electrical equipment. The discrete Wavelet transform (DWT) and Wavelet Packet analysis are found to be powerful signal processing tools to denoise the PD signals. In this paper, the noise elimination through wavelet techniques is carried out on the simulated PD signal using the mathematical model of void in the XLPE power Cable. The imperfections in the insulation, such as voids, are modeled by using the improved ABC Capacitive model. The Partial discharge current pulses developed due to voids are captured in the simulated void model of the cable. Then, these current signals are mixed with white Gaussian noise and DSI signals to simulate the noisy signal. The effectiveness of noise elimination using Wavelet and Wavelet packet denoising methods are analyzed by extracting the PD signals from the noisy signal. The mother wavelet selection based on the minimum entropy criteria and the level dependent hard thresholding method was adopted. The original Partial discharge signals were retrieved with entropy based mother wavelet selection and level dependent hard thresholding method. In both wavelet and wavelet packet methods, wavelet packet is able to reproduce the original signal in a better manner. The same analysis is carried out on one of the laboratory measured PD signal and the effectiveness of the entropy based mother wavelet selection & level dependent hard thresholding is verified.