Digital Signal Processing Techniques Applied to Partial Discharge Monitoring and Classification

The environment in which PD signals are recorded may exhibit varied levels of noise due to radio frequency interference, thermal activity, and electronic switching devices. To ensure reliable PD monitoring, meticulous use of signal processing techniques to isolate the signal of interest from unwanted disturbances. Section 6.1 briefly discusses the general features of PD measuring systems and their relation to the characteristics of PD signals. Then, in Sect. 6.2, we present basic techniques for signal separation and denoising based on linear time-invariant filters and wavelets. Examples illustrate both approaches applied to synthetic and real signals. The final section of the chapter focuses on classification schemes based on artificial neural networks that take PRPD diagrams as inputs. The aim is to automatically estimate different types of insulation impairments by analysing PRPD data. Classification results by using PD data from power plants are shown. A list of relevant references concludes the chapter.