The main drawback of fault arc detection method based on arc current characteristics is that there is only one information source

and it can lead to false action and refused action of arc fault circuit breaker. In order to solve this problem

this paper utilizes the electric field signal and magnetic field signal of arc radiation and arc cur-rent as the input signals

and proposes a fault arc detection method based on neural network and D-S evidence theo-ry. Firstly

modulus maximum method is used to obtain the characteristics of arc radiation signal under various load current

and waveform similarity method is employed to get the characteristics of arc current signal. Secondly

The BP neural network is used to establish the mapping relationship between arc fault and each characteristic

and the fault arc identification rate of each characteristic variable is obtained. Finally

all of the identification rates are tak-en as the input of D-S evidence theory method to realize multi characteristic information fusion. The experimental re-sults under typical load prove that the proposed method can effectively improve the detection accuracy of fault arc.