Wavelet energy distribution with PCA & DBSCAN for partial discharge pulse extraction

Condition monitoring based on partial discharge diagnostics has seen a very rapid increase of interest from high voltage industry. Condition monitoring is an asset monitoring tool that can inform the user of real time health of their high voltage equipment / plant, which has now become a necessity in ageing networks that still have equipment in operation that has fulfilled its rated life. Partial discharge pulse detection and extraction is the first step of condition monitoring. This research is focused on assessing the capability of wavelet energy distribution for pulse extraction with the help of principle component analysis (PCA) and density based spatial clustering algorithm (DBSCAN). One problem that every pulse extraction method faces is that their efficiency is highly affected when a signal with high noise is processed through them. This research will also assess the robustness of the aforementioned procedure on signals with different signal to noise ratio by comparing results of each signal. Moreover, this procedure will also be vetted on its level of intrusiveness to see if it can be used for online condition monitoring.