Fault arcing detection for building low voltage lines based on CMAC neural network

The value of arcing fault current is too small to make the traditional circuit breaker to cut off the power supply,and there are loads which have the similar characters to arcing fault in the circuit,so arcing fault is one of the major causes of electrical fire.Simplex criterion used to detect fault arcing has the shortcoming of high miscarriage rate.In this paper,the fault arcing of building low voltage distribution lines is simulated through building an experiment platform and the characteristics of fault arcing are extracted.CMAC neural network is used to build a model to fuse two criteria which are the differences of the mean values of sample points per cycle and the wavelet high frequency coefficient to overcome the uncertainty and limitation of simplex criterion,and the presented method of the multi-information fusion can improve the accuracy of identifying fault arcing effectively.