| Recently,protecting the environment and reducing carbon emissions have become the focus of society,and the electric vehicle(EV)technology can help solve such problems,which has attracted attention from various countries.With the introduction of relevant national support policies and traffic improvement,a large number of EVs have been put into operation,and their charging safety drawbacks have gradually emerged.According to the analysis of spontaneous combustion accident of EVs during charging,it is found that overheating of on-board batteries is an important cause of such accidents.To solve the problem,this thesis proposes a method for monitoring the charging status and safety warning of EVs.Firstly,this method uses a deep learning method to explore the deep level features of EVs charging data,and fully trains the model;Secondly,the sliding window method is used to correct the impact caused by incorrect data to ensure the accuracy of prediction;Finally,the reasonable pre-warning and alarm thresholds are set by the residual analysis method to quickly complete the pre-warning task and ensure the charging safety of EVs.The specific research contents are as follows:(1)Model selection stage: Starting from the research background,analyze the causes of EVs spontaneous combustion accidents.Starting from existing warning methods both domestically and internationally,and comparing warning methods in various fields,choosing three hybrid models: ConvLSTM,CNN-BiLSTM and AT-CNN-BiLSTM to complete the following experiments.(2)Method determination stage: By analyzing the composition and charging method of EVs batteries,as well as the composition and working principle of the charging system,a constant current and constant voltage charging method and the processing dataset method are selected.Starting from the research on the monitoring method of EVs charging status,the main research content of this monitoring method is elaborated;Furthermore,a safety warning method for EVs charging is proposed,and the design concept and evaluation criteria of this warning method are described;On this basis,a framework and process of the experimental are proposed for both monitoring and early warning methods.(3)Method validation stage: In the monitoring of EVs charging status,this article selects test set 4 with the best prediction effect to ensure the accuracy of EVs charging data prediction;In the EV charging safety warning work,select test set 1 with the worst prediction effect to verify the feasibility of this method.Based on the analysis of experimental results,the three hybrid models proposed in this article have all completed the monitoring charging status and safety warning work,ensuring the charging safety of EVs. |
