Research On Estimation Of State-of-Charge And State-of-Health For Li-ion Batteries

In the context of the global energy crisis and environmental pollution,lithium-ion batteries have become the most promising energy storage devices in the future due to their various excellent characteristics.In order to ensure the safe use of lithium-ion batteries and prevent overcharge or over-discharge,it is necessary to design a battery management system(BMS)to monitor and control them.The estimation of state of charge(SOC)and state of health(SOH)is the core function of BMS.Based on the experimental data,this paper analyzes the working characteristics of lithium-ion batteries and establishes a second-order RC network equivalent circuit model.Aiming at the problem that the error of the estimation result of square-root unscented Kalman filter(SRUKF)algorithm is increased due to system time-varying noise,this paper introduces Sage-Husa algorithm to estimate the noise variance matrix online in the recursive calculation process of the filtering algorithm,and the adaptive square-root uscented Kalman filter(ASRUKF)algorithm is obtained.The validation results shows that the ASRUKF algorithm can effectively reduce the influence of timevarying noise on the estimation result and improve the estimation accuracy.In order to realize SOH estimation and reduce the estimation error of ASRUKF algorithm caused by capacity and internal resistance drift,this paper establishes capacity and internal resistance model,and realizes capacity and internal resistance estimation by extended Kalman filter(EKF)algorithm.The ASRUKF algorithm is combined with the EKF algorithm to form a double Kalman filter(DKF)algorithm.The DKF algorithm can update the capacity and internal resistance in real time during the SOC estimation algorithm,so it can correct the model error and reduce the influence of the capacity change on the estimation result.So it can get more accurate estimation results under various working conditions.In order to effectively compensate the influence of the actual physical quantity of temperature and aging on the estimation result,this paper uses the data-driven method to estimate the state of the lithium-ion battery.A dynamic recurrent neural network(DRNN)is designed based on the nonlinear autoregressive with exogenous inputs(NARX)structure.The DRNN has global feedback,which can realize dynamic mapping from input to output.The particle swarm optimization(PSO)algorithm is improved by self-adaptive weight method,and the self-adaptive weight particle swarm optimization(SWPSO)algorithm is obtained.The DRNN is trained by the SWPSO algorithm,which improves the error convergence speed and avoids the network falling into local optimum.The verification results show that SWPSO-DRNN can obtain effective estimation under different temperature and aging conditions,and the estimation accuracy is higher and the generalization ability is stronger.