Research on Identification Method of Voltage Sag Type Based on MI-CNN

Accurate identification of voltage sag event types is the basis for sag location and treatment. Considering the full use of the voltage and current data when the sag event occurs, this paper proposes a voltage sag type identification method based on multi-input convolutional neural network (MI-CNN). In this method, the voltage and current data are input into the two input layers of MI-CNN respectively, and the features of the voltage and current data are extracted separately. Then the separately extracted data features are fused, and then the fused data features are further extracted. Finally, through several fully-connected layers, the classification results are output. In the simulation experiment, this paper tests three feature fusion patterns, and selects the feature vertical fusion pattern according to the optimal test results. Finally, the method proposed in this paper is compared with the other two classification methods, and the results show that the proposed method has higher identification accuracy and better anti-noise.