A partial discharge monitoring method, system and product based on time-frequency unmixing and time series decomposition

Abstract

This application discloses a partial discharge (PD) monitoring method, system, and product based on time-frequency demixing and time-series decomposition for the field of PD monitoring. The method includes: S100, acquiring a PD monitoring pulse signal, preprocessing it to obtain a PD time-series data sequence, and converting it to a time-frequency diagram; S200, demixing the time-frequency diagram in both the time and frequency domains to obtain frequency domain data in the time length direction and time-series data distinguished by frequency domain; S300, decomposing the time-series data into trend terms, seasonal terms, and residual terms using a time-series decomposition algorithm; S400, performing long-term monitoring of PD health based on the trend terms of the target frequency domain time-series data; real-time monitoring of sudden and random anomalies of PD based on the residual terms of the target frequency domain time-series data; and verifying partial discharge modes based on the seasonal terms of different frequency domain time-series data. This application enables efficient, highly reliable, highly generalizable, and highly robust intelligent identification and monitoring of PD signals.

Images