Research On PSO-LSTM SOC Estimation Method For Lithium Ion Batteries

With the proposal of the national dual carbon strategy,new energy vehicles have become a research hotspot.As the core component of new energy vehicles,the accurate estimation of state of charge(SOC)of power batteries directly affects the performance of the whole vehicle.Aiming at the problem of SOC estimation accuracy of lithium-ion battery,by selecting 18650 ternary lithium-ion battery as the research object,a SOC estimation method of lithium-ion battery using particle swarm optimization(PSO)to optimize long short term memory(LSTM)neural network model is proposed.Finally,the accuracy of the method is verified by experiments under different working conditions.The main research contents are as follows:(1)According to the characteristics of lithium-ion battery,the battery performance parameters such as lithium-ion battery capacity,battery charge discharge ratio and battery voltage were analyzed.The characteristics of battery temperature,discharge ratio and internal resistance are mainly studied,and their effects on battery SOC estimation are analyzed.The lithium-ion battery model was analyzed to lay a foundation for the research on SOC estimation of lithium-ion battery.(2)According to the estimation accuracy of lithium-ion battery SOC,a lithium-ion battery SOC estimation method based on LSTM algorithm was adopted to establish the lithium-ion battery SOC estimation model based on BP,PSO-BP and LSTM neural network.The training results and estimation accuracy of the three models were compared according to the experimental data.The results show that the LSTM neural network model has higher estimation accuracy.(3)The PSO algorithm is used to obtain the optimal parameters of the model in the repeated iteration process,and the PSO-LSTM prediction model was established.The comparative analysis under constant current conditions in the laboratory shows that PSO-LSTM neural network has superior accuracy in SOC estimation of lithium-ion batteries.Aiming at the SOC estimation performance of lithium-ion battery based on PSO-LSTM model under actual working conditions,the simulation research on SOC estimation accuracy of lithium-ion battery based on LSTM and PSO-LSTM models under different working conditions and temperatures is carried out.The experimental results show that the maximum estimation errors of PSOLSTM model under DST and US06 working conditions are 1.871%and 1.76%respectively,which is 1%～2%less than that of LSTM;At different temperatures,both LSTM and PSOLSTM models show good estimation accuracy at high temperatures.The maximum error of PSO-LSTM estimation is 1.53%,which is 1%less than that of LSTM model.Simulation results show that PSO-LSTM model has good generalization ability and estimation accuracy.(4)By collecting the voltage and current data in the lithium-ion battery discharge experiment,as the input value of PSO-LSTM neural network model,the SOC of lithium-ion battery is verified online.The SOC estimation curve and prediction error diagram of EKF,LSTM and PSO-LSTM algorithm models are compared and analyzed by designing experiments under constant current condition and DST condition.The experimental results show that the estimation accuracy of the optimized PSO-LSTM algorithm model is improved by 2.1%and 1.5%compared with EKF and LSTM algorithm models respectively.