A new method using ultrasonic for partial discharge pattern recognition

The partial discharge measurements on a power transformer are studied with a new method toward discrimination and classification of different discharge models. Based on characteristics of partial discharge ultrasonic signal are nonlinear and nonstationary, fractal theory is adopted to extract fractal parameters from ultrasonic signal of transformer partial discharge in this paper. The self-similarity character of ultrasonic signal for partial discharge is analyzed. The fractal theory and its parameters calculation are introduced. Experiments are done about several typical transformer partial discharge models. The fractal parameters (box counting dimension and lacunarity) are calculated in different model´s ultrasonic signals. The calculation result shows that fractal parameters are dissimilar to box counting dimension and lacunarity in different transformer partial discharge ultrasonic signals. The artificial neural network is used to recognize discharge patterns. The result shows the fractal method is effective for transformer partial discharge pattern recognition.