An Investigation into Multi-Source Partial Discharge Discrimination within a Power Transformer Model

Download

PDF Restricted to Registered users only Download (821Kb) | Request a copy

Description/Abstract

This paper investigates a new multi-PD-source discrimination method using machine learning technique, namely support vector machine (SVM). A bushing-tap-RFCT (Radio Frequency Current Transducer) system is used to simulate a power transformer and coupling sensor. Some artificial PD sources are used to generate multi-PD signals. Obtained experimental results from different PD sources are processed using accepted approaches such as φ-q-n patterns, pulse sequence analysis and wavelet transform. The processed data are also used as the input vector of the SVM. Initial results indicate that, by using appropriate feature parameters, the automatic identification results obtained by combining the SVM technique and time domain information are very encouraging.