Partial discharge pattern recognition for three kinds of model electrodes with a neural network

Partial discharge (PD) signals have been studied for a long time as one of the significant high voltage test parameters and therefore many research works have been done on the sophistication of the PD measurement and data processing methods. PD characteristics can be derived from pulse data through statistical treatment of the data. With the aids of computer measurement systems, the φ-q-n distribution characteristics and the φ-q-n distribution characteristics have been proposed. Those characteristics add the voltage phase angle information to the conventional PD measurements. Recently, may applications of the φ-q-n distribution characteristics have been proposed. Those remarkable development in PD data processing may bring data-overflow in PD measurement. To treat a number of PD data requires the more sophisticated data processing method which fits the purpose of the high voltage tests. A neural network system was introduced to recognize the φ-q-n distribution pattern. This paper describes PD data processing as input signal for the neural network and the capability of the neural network system to recognize PD signals using four kinds of electrode systems