基于ASFF-DC1D-CNN的变电站典型复合故障自动化辨识OA

The multi-faults of substation are of great significance for the stability of power system.In order to automatically identify the typical faults of substation,a dual-channel one-dimensional convolutional neural network(DC1D-CNN)model based on analog-switch fusion features(ASFF)was proposed.The model was consists of two independent feature processing modules and a shared decision module,which could utilize the features of different types of data.Batch normalization layer(BN)and attention mechanism(AM)structure were added into the traditional one-dimensional convolutional neural network(1D-CNN)model.The analog features of the recording signal and the digital features of the switching signal could be efficiently fused by using dual-channel,which could enhance the ability of feature extraction and learning of the model.A typical 220 kV intelligent substation was modeled and simulated.The data obtained from the simulation were employed to train and test the model.The results indicated that the model achieved a recognition accuracy of 98.33%for typical faults.Compared with the traditional single-channel CNN,backpropagation neural network(BPNN),and recurrent neural network(RNN)algorithm,the proposed method has great efficiency and robustness.It performs better in the face of complex data.