Lithium ion battery SOC estimation based on BP-DCKF-LSTM
Accurate estimation of state of charge(SOC)is critical for battery management systems(BMS).This paper introduces an SOC estimation method integrating backpropagation neural net-work(BP),dual cubature Kalman filter(DCKF),and long short-term memory neural network(LSTM)to enhance lithium battery SOC precision.Addressing the inadequacies of conventional poly-nomial fitting method in accurately fitting open-circuit voltage(OCV)to SOC under varied tempera-tures,a BP neural network-based approach is proposed,proven to significantly improve fitting accu-racy.Moreover,considering the strengths and weaknesses of solely using model-based or data-driven methods for SOC estimation,a combined approach of DCKF and LSTM is suggested.This not only boosts estimation accuracy but also reduces parameter tuning time and training costs.Experimental validation indicates that the BP-DCKF-LSTM algorithm achieves root mean square Error(RMSE)and mean absolute error(MAE)of less than 0.5％and 0.4％,respectively,demonstrating high accu-racy and robustness in SOC estimation.
state of chargebackpropagation neural networkdual cubature Kalman filterlong short-term memory neural network
万方数据
维普
