Type Identification of Short Circuit Faults Based on Alexnet

Correct identification of fault types after grid short-circuit faults is the basis for fault handling, and it is also the basis and difficulty of grid fault analysis. In view of the traditional method, it is difficult to quickly and correctly guide the action of the circuit breaker relying on the electrical quantity judgment ,so based on the deep learning, a method of using Alexnet to judge the fault type is proposed. Firstly, this paper transforms the fault time series data into Hankel twodimensional matrix, next, the three-phase Hankel matrix is superimposed into jpg type image data, after that, converts the simulated voltage data into image pixel data between 0-255 through linear transformation. Then divides the graphic data into test set and verification set and puts them into the Alexnet network where the fault recognition model is trained. Finally, the proposed algorithm is verified to have higher accuracy through batch simulation data and it provides a new solution for fault type discrimination.