Partial discharge pattern recognition for four kinds of electrode systems

The authors describe a method for recognizing partial discharge characteristics for four kinds of electrode systems. The method uses a neural network system with an input of phi -q-n distribution patterns. The theta -q-n distribution consists of the pulse count n versus the pulse height q and the phase angle phi . The learning characteristics and recognition characteristics of the neural network were investigated, and the recognition capability for a combined pattern signal input was shown. The effectiveness of the neural network system for partial discharge recognition is discussed.< >