基于改进一维卷积神经网络的充电桩故障诊断OA

In view of the problems of difficult fault feature extraction and low fault diagnosis accuracy of elec-tric vehicle charging piles,this study proposes a charging pile fault diagnosis model that improves the whale optimiza-tion algorithm to optimize the structural parameters of MHA(Muti-Head Attention)-1DCNN(One-Dimensional Convo-lutional Neural Network)-GAP(Global Average Pooling).Based on a one-dimensional convolutional neural network and combined with a multi-head attention mechanism,it captures the important feature information of different levels of subspaces of fault data.The global average pooling layer is adopted to replace the fully connected layer in the tra-ditional one-dimensional convolutional neural network,reducing the number of parameters and improving the generali-zation ability of the model.The traditional whale optimization algorithm is improved by introducing Levy flight pertur-bation and nonlinear convergence factors,thereby optimizing the global parameters of the fault diagnosis model and a-voiding the model from falling into local optimal solutions.The simulation results show that the MHA-1DCNN-GAP charging pile fault diagnosis model based on the improved whale optimization algorithm performs well in terms of con-vergence speed and generalization.The fault diagnosis accuracy and loss value are 99.18%and 0.06,respectively.