Traditional low-voltage protection devices such as low-voltage circuit breakers

fuses and other protection devices cannot effectively detect series arc faults which are caused by poor contact or insulation failure

therefore

how to accurately detect series arc faults has become a hot issue in current research. In this paper

the detection method based on current waveform is used for in-depth research. By building an arc fault platform to simulate series arc faults

the data of normal and arc faults under different loads are obtained. On this basis

a neural network algorithm for multi-feature fusion is established. The model is optimized by using mini-batch gradient descent

exponential decay learning rate

and Adam’s optimization algorithm. The research results show that the accuracy and recall rate of the algorithm in this paper can reach 98% and 99%

respectively

which has a higher recognition rate than the SVM and BP neural network algorithms. The research provides a new algorithm for series arc fault identification

and expands new ideas for the research in this direction.