Research On Prediction Of Battery SOC And Remaining Charging Time For Electric Vehicles

In recent years,with the rapid development of science and technology,electric vehicles have become the focus of attention in various countries.Batteries are the power source of electric vehicles and battery SOC is an important indicator of electric vehicle driving safety.The remaining charging time is an important parameter for electric vehicle drivers to arrange driving plans and the charging stations to improve management efficiency.In this paper,the battery data of electric vehicles is the research object,and the research of prediction of battery SOC and remaining charging time of electric vehicles are conducted.The main work of this paper is as follow:(1)Data analysis of electric vehicles.Firstly,the data involved in this article are described briefly,and the missing values,outliers and invalid data in the dataset are processed as well as the charging data of electric vehicles are extracted.Then,the charging and discharging characteristics of battery of electric vehicles are analyzed,including the the influencing factors of battery SOC and remaining charging time.Finally,the characteristic parameters are selected by Pearson correlation coefficient.(2)Established a battery SOC prediction model for electric vehicles based on CNN-LSTM.Since the change of battery SOC of electric vehicle is based on the time axis,all data is generated as time series data,the battery SOC of electric vehicle are analyzed based on LSTM in this paper.At first,after studying the advantages of CNN model and the model Combining CNN with LSTM,an electric vehicle battery SOC prediction model based on CNN-LSTM is established.Then based on the analysis of the influence of different parameters on CNN-LSTM model prediction result,the best parameters of the CNN-LSTM model were determined.According to the final CNNLSTM model,CNN,RNN,and LSTM models are built as comparison model to predict the battery SOC of electric vehicles under consistent experimental conditions.The experiment shows that the results based on the CNN-LSTM model is better,and its MSE is 0.3472,MAE is 0.4299,and maximum error is 2.1%.(3)A LSTM-based prediction model for the remaining charging time of electric vehicles is established.Due to the charging process of electric vehicles is a part of the application of electric vehicles,an electric vehicle remaining charging time prediction model based on time series LSTM is established.According to the established model for predicting the remaining charging time of electric vehicles,the experimental results show that the maximum error is within 5 minutes.Then in order to determine the best model parameters,parameter analysis experiments of the remaining charging time prediction model for electric vehicles are conducted.Based on the prediction model of the remaining charging time of electric vehicles,which has the best model parameters,the experimental results show that the MSE of the predicted results is less than 0.03,MAE less than0.1,and the maximum error less than 2 minutes.At the same time,CNN and RNN models are established based on consistent experimental conditions as a comparative experiment.The experimental results show that the remaining charging time prediction model of electric vehicles based on LSTM is better.