Cable fault recognition using multiple wavelet neural networks

The aim of this paper is to realize the online fault recognition instead of the current state-of-the-art of offline fault recognition of the power cable. A novel online fault recognition method of a multiple wavelet neural networks is proposed based on the wavelet energy function of the traveling wave. In the new method, the difference of the zero-order currents between the fault cable and the normal cable are selected as the original fault recognition features. Then the sub-band energy functions of wavelet package of the original features are designed and calculated to serve as the input vectors of the neural networks. Furthermore, an intelligent cable fault diagnosis system is designed. The system consists of 4 wavelet neural networks to recognize the fault and 1 wavelet fault locator to determine the fault position. Finally, the simulation results show that the cable fault recognitions can be implemented correctly. The fault classes include the 1-phase ground faults, the 2-phase short circuit faults, the 3-phase short circuit faults, and the open circuit faults. This presents the theoretical support for the online fault diagnosis of power cable.