Application Of Bayesian Optimized LSTM Model In Power Battery SoC Estimation

With the economic development and the improvement of people’s living standard,automobiles have gradually gone to thousands of families.At the same time,with the increase in car penetration,problems such as environmental pollution and energy shortage will inevitably arise.As a new type of low-carbon,environmentally friendly transportation tool,electric vehicles are the most effective way to solve the current energy consumption and air pollution.It has become an inevitable trend that electric vehicles will replace traditional fossil fuel vehicles.At present,in the research process of electric vehicles,there are still many technical problems that need to be solved urgently.As one of the most important parameters in the power battery management system.The accurate estimation of the SoC helps to improve the utilization rate of the battery and prolong the service life of the battery,which can be used as an important basis for predicting the remaining mileage of the vehicle.Therefore,accurate estimation of the SoC is one of the key issues that need to be solved urgently.Based on the analysis of the existing power battery SoC estimation methods,this paper researches on the power battery SoC estimation based on the LSTM optimized by Bayesian.The research work and results of this article include:(i)The feedforward neural network only considers the relationship between the input and output when estimating the battery SoC,which leads to the low accuracy of the estimation result.In this paper a estimation model of power battery SoC is proposed based on LSTM,which fully considers the timing characteristics of the data.The input at this time and the past time will be fully considered when estimating the SoC.The experimental results show that the LSTM-based power battery SoC estimation model error proposed in this paper is about 3%,which meets the requirements of electric vehicles for the power battery SoC estimation error of less than 8%,and each index is better than the other three models.The LSTM-based power battery SoC estimation model has higher accuracy.(ii)Aiming at the problem of hyperparameter tuning and time efficiency in the training process of LSTM model,Bayesian Optimization(BO)is introduced for hyperparameter optimization.The Bayesian optimization method is based on Bayes’ theorem.When selecting the next set of hyperparameters,the previous evaluation information can be fully utilized to reduce the number of hyperparameter attempts.Through comparison experiments with various hyperparameter optimization algorithms,the experimental results show that the proposed LSTM-based power battery estimation model proposed in this paper has shorter training time,higher prediction accuracy of the optimized model and stronger generalization ability.(iii)Based on the SoC estimation method proposed in this paper,a prototype system for power battery SoC estimation is designed and implemented.The system not only effectively strengthens the management of power battery data,but also enables real-time display of the estimated results.