| With the rapid development of new energy electric vehicles,the demand for charging facilities and other equipment is also increasing.The charging process of electric vehicles and charging facilities is a process with high requirements for voltage and current,and it is easy to be affected by external factors.The charging safety of charging facilities is not only related to the safety of electric vehicles and users.It also affects the promotion and application of electric vehicle industry.Aiming at the problem of early warning of electric vehicle charging facility failure,this paper,from a practical point of view,considers the running characteristics of charging facility and the problem of data loss of charging facility,proposes a charging facility data imputation method combined with artificial intelligence and charging facility safety early warning method.Firstly,the purpose and significance of this research are expounded,and the research status of the operation process of charging facilities is summarized.The system structure and working principle of DC charging pile are analyzed.Taking DC charging pile as the research object,different charging modes are introduced.Based on the influence factors of DC pile safety,typical abnormal operation state was studied,and safety warning state indicators were selected.The theoretical model of data loss is summarized,and the data loss of charging facilities in the actual scene is studied.The random loss model is the premise of data preprocessing in this paper.A charging facility safety early warning method is proposed,and a deep learning neural network model is built as the basis of charging facility safety early warning.According to the real-time collection of charging facility data,the charging facility is input into the model,and the predicted value of model output is compared with the actual value to determine whether the charging facility is in abnormal operation.Secondly,the data loss processing operation is carried out on the collected data of charging facilities.First,on the basis of the research on traditional generative adversarial networks(GAN),the model structure is improved.A generative adversarial imputation network(GAIN)is built,and the global data generation is reduced to the imputation of local missing data positions.The effects of different parameters of the GAIN model on the optimization effects of the model are analyzed through simulation,and the optimal superparameters are selected.Realize reliable imputation of missing data of charging facilities.Finally,the algorithm and structure of typical deep neural network models are analyzed,and the Convolutional Neural Networks(CNN),Bidirectional Gate Recurrent Unit(Bi-GRU)and attention mechanism are selected,and the CNN-Bi GRU-Attention hybrid model is established as the charging facility fault warning model,and analyze the threshold indicators of fault warning;Based on the operation data of charging facilities,the model is trained,the fitting effect of the early warning model under different hyperparameters is analyzed,and the appropriate model hyperparameter optimization model is selected.Simulation analysis is carried out in the case of missing data and imputed data based on GAIN model,which verifies the effectiveness and feasibility of the proposed CNN-Bi GRU-Attention security early warning method based on GAIN data imputation. |
