Research On State Estimation And Life Model Of Lithium Ion Battery

In recent years,widespread attention has been paid to the electric vehicle industry and the industry has been developing rapidly.Battery is the key component of new energy vehicles,the research on state estimation and life model of the battery can ensure reliable and stable operation of electric vehicles and has important practical significance.The state estimation of the battery mainly includes the SOC estimation,the SOH estimation and RUL prediction.In this dissertation,a depth research on state estimation and life model of batteries is conducted,taking the lithium battery as the object.The main results achieved are as follows:1 、 Based on the second-order RC network equivalent model,research on SOC estimation was conducted.To solve the particle degeneracy phenomenon of the PF and improve the accuracy of SOC estimation,a novel method based on SCDPF was proposed.The result of the SCDPF algorithm was in comparison with that of EKF,UKF and UPF algorithms under the NEDC condition which showed that the SCDPF algorithm had higher accuracy and faster convergence speed,the maximum error never exceeded 1.21% at 25℃.2、Research on model-based SOH estimation and RUL prediction algorithms was conducted.A double exponential model of battery capacity was established.Combined with the model,a novel method was proposed based on SCDPF for SOH estimation and RUL prediction.The method can solve the particle degeneracy phenomenon and improve the accuracy and reliability of the prediction.It was validated by datasets from the University of Maryland and NASA.Compared with the PF and UPF algorithms,the SOH estimation obtained by the proposed SCDPF algorithm was closer to the actual value,the maximum SOH estimation error of A5 was only 0.61%.RUL prediction value was also closer to the real RUL,which means the algorithm has higher accuracy and stability.3、Research on data-driven SOH estimation and RUL prediction algorithms was conducted.A hybrid method based on VMD,SSA and LSSVR model was proposed to improve the accuracy of SOH estimation and RUL prediction.By using datasets from the University of Maryland and NASA,the proposed algorithm was verified and compared with SVR models optimized by PSO and ABC.It can be seen from the comparison that the proposed algorithm had a more accurate prediction,the minimum root mean square error of the CS37 battery was 0.0041.With the advance of the prediction point,the prediction accuracy of the proposed VMD-SSA-LSSVR method does not change much.The prediction range was enlarged remarkably and the prediction was more stable.4、Research on battery life degradation model was conducted.A series of calendar and cycle life degradation experiments were carried out.Considering the influence of various factors(temperature,charge-discharge rate,depth of discharge,storage SOC)on battery life,and taking the effect of both calendar aging and cycle aging into account,a combined multiple factor degradation model was proposed.The model was verified on the experiment platform based on the 18650,and the comparison between experimental and predicted data proved the accuracy of the proposed model,the margin of error never exceeded 1%.