Separation of Multiple PD Sources in MV Switchgears during AC Withstand Voltage Test Stage Based on Time-Frequency Representation

Multiple-source partial discharge (PD) phenomenon commonly occurs in the process of alternating current (AC) withstand voltage test on medium voltage (MV) switchgear. This phenomenon can significantly reduce the accuracy of in-field PD diagnosis as their signals are mixed together. Accordingly, this paper proposes an automatic tool to separate multiple PD signals in a row of MV switchgears. In the tool, the short-time Fourier transform is used to transform the time-domain waveform into the time-frequency spectrogram, which can be verified to have better performance to distinguish different types of PD. Three feature parameters named time center, frequency center, and center moment are defined in the spectrogram in order to reduce the complexity of subsequent signal separations. Signals of multiple PD sources are automatically separated via the Gustafson-Kessel clustering algorithm. The feasibility of the proposed method is validated on four artificial PD models in three 10-kV switchgears.