Signal processing tools for evaluation of Partial Discharge Measurement Data of Power Cables

Abstract

Due to growing and aging power cable grids, there is a rising need for partial discharge (PD) measurement techniques to identify incorrect installed or damaged cable joints or cavities in paper-insulated lead-covered (PILC) cable. While cable manufacturer or high voltage laboratories are able to measure in shielded environments, where noise levels are insignificant, PD detection and location at power cable lines on-the-field are hampered variously by different interferences and noise. The time domain reflectometry (TDR) is a standard technique for the location of PD in power cables. There, pulses traveling along a cable and their reflections are observed. For offline power cable PD measurement methods, where cables under test are disconnected from service, it is essential to determine the delay time between direct and reflected PD pulse to estimate the ignition point (PD site). With the help of algorithms for arrival time estimation, like peak or onset detection, it is possible to gather such delay times in an automated way, assuming a good signal-to-noise-ratio and a sufficiently high sampling frequency. To get rid of noise, modern de-noising techniques, like the wavelet transform, can be used. The Paper is presenting and comparing enhanced signal processing tools for arrival time estimation and de-noising concerning location accuracy, noise sensitivity and robustness for automated cable PD location software.