Denoising and Extraction of Partial Discharge Pulse Characteristics Using Wavelet Denoising and Local Maxima Method

Partial discharge (PD) pulse characteristics describe the nature of an insulation defect, its severity, and aging behavior of electrical insulation. However, extracting these characteristics requires sophisticated tools in order to denoise measured PD data, detect actual PD pulses, and analyze the corresponding characteristics. This paper proposes an effective tool that utilizes a simple hybrid approach based on wavelet denoising and the local maxima technique that denoises PD data acquired using a high-frequency current transducer. After de noising, it detects PD pulses from a time series dataset and identifies corresponding primary features like peak pulse amplitude and location. In addition, it can extract pulse shape (rise time, fall time and pulse width), and pulse sequence (voltage interval and time interval) features. Developed tool is computationally efficient as it precisely detects maj or PD events and corresponding features from a large data set while rejecting noise. Efficiency of tool is checked by implementing it on various data sets obtained from experimental investigation performed on three distinct types of defects, i.e., corona, surface, and internal discharge. Results show that this tool can be used for the identification of defect types, their progression, and aging studies to estimate the remaining lifetime based on changing PD behavior.