Research on Intelligent Diagnosis and Route Selection Method of Multi-load Circuit Series Fault Arc

The circuit structure of the low-voltage power system is complex, and the load forms are various. When a series arc fault occurs due to line aging, poor contact, etc., the existing arc fault circuit breaker can usually only protect the single-phase single-load circuit, and it is difficult to detect the lower level line failure. Therefore, a household multi-load circuit series arc fault experiment platform is set up under AC 220 V to carry out arc fault experiments under various branch and trunk faults, and obtain the trunk current data when fault arcs occurred under different working conditions, which are used as samples to establish a model sample database to train the established one-dimensional convolutional neural network. After a lot of training, the established household multi-load circuit series arc fault diagnosis and line selection model has an accuracy rate of 98.6% on the validation set. Then, the continuous current data in the test set is used to verify the robustness, immunity and generalization of the model, and finally the trained model is imported into the Raspberry Pi for running time test. The results show that the detection time meets the requirements of the UL standard. However, before it is put into practice, the model still needs to be further optimized to make it more suitable to run on embedded system.