A wavelet-based method to discriminate between inrush current and internal fault

With the development of power systems, the content of the second harmonic current can be comparable to that produced in the inrush current. The conventionally used second harmonic current restrained method becomes unreliable for transformer protection. To obtain some new approaches on discrimination between inrush current and internal fault, transformer models with enough precision of computing inrush current and short-circuit current are firstly described. After that, Daubechies family wavelets are selected as a mother wavelet to analyze the inrush current and short-circuit current. The results show that the characteristics of inrush current are significantly different from those of short-circuit current. Based on the analyzing result, the back-propagation neural network is trained to discriminate the inrush current and short-circuit current. The training results presented in this paper show that wavelet based discrimination method is efficient with good performance and reliability.