Under the limitation of the line load

the fault current of the series arc fault is relatively small

which makes such traditional electrical protection devices as low-voltage circuit breakers

fuses cannot operate

therefore

the detection of series arc faults has become a hot issue in the current arc fault detection research. For this purpose

the arc experiment is performed by setting up arc fault platform and the current data of typical loads during normal operation and arc faults are obtained; Then

on the basis of arc multi-feature fusion

the information gain and the gain rate of the features are combined to sort the importance of the arc features so to construct a more suitable set of series arc features; Finally

on this basis

a random forest algorithm is set up and the particle swarm algorithm is used to optimize the parameters. The research results show that compared with the traditional random forest and BP neural network

the proposed method has a higher identification rate

and both its accuracy rate and completion rate reach 99%

which can achieve effective diagnosis of series fault arc.