Advanced Signal Processing Techniques for Partial Discharge Measurement

Various diagnostic methods have been deployed in utilities for assessing transformer insulation conditions over the past decades. Among these methods, Partial discharge (PD) measurement is widely adopted due to its capability for providing continuous online condition monitoring of a transformer without disturbing its normal operation. The noise distorts PD signals, which can significantly influence the performance of the later steps of PD signal analysis. De‐noising can be achieved by using hardware, but software‐ based methods provide an adaptive means of removing various types of noise. Some typical software‐based methods are signal decomposition, blind processing, and graphics‐based techniques. These techniques are detailed in this chapter. PD source separation is executed to separate the PD sources and reveal individual PD patterns related to each source. This can be performed by utilizing the time and frequency characteristics of PD signals to form distinctive clusters in a time/frequency (TF) map, with each cluster corresponding to a single PD source.