Classification Of Partial Discharge Patterns Using Fractal Geometry

Partial discharge (PD) pattern recognition is an important diagnostic tool to get the dielectric deterioration degree of electrical insulation. A PD pattern recognition approach of artificial partial discharge sources based on neural networks is proposed in this paper. A commercial PD detector is used to measure the three-dimension PD patterns. Two fractal features (fractal dimension and lacunarity) were extracted from the raw PD patterns. The box counting technique is used to generate the fractal features. Fractal dimension is a constant value for each pattern. On the other hand, each pattern has several values for lacunarity depends upon the size of the box. Therefore each PD pattern is characterized by several quantities. The performance of neural networks based recognition system depends upon the number of input features as well as upon their contents of information. This paper is an attempt to improve the performance of PD recognition systems by minimizing number of input features. This can be done through selecting the value of lacunarity which has the maximum ability to classify different PD sources. A wide range of lacunarity values were tested. The obtained results show that a very narrow range of lacunarity values has the maximum ability for PD