| In order to accurately predict the state of charge(SOC) of lithium-ion batteries in vehicle which is running, this paper takes the lithium iron phosphate battery as the research object. Considering voltage data, current and internal resistance as the influence factor, the paper analyzes the characteristic of and voltage, internal resistance and capacity in detail. Equivalent circuit simulation model of battery is built, besides, parameter identification and precision verification for the model are presented. Lastly, the paper focuses on the research of SOC prediction of lithium iron phosphate battery.According to nonlinear characteristic and random fluctuations of the change of power battery SOC in working condition, the prediction model of extended kalman filter(EKF) is used for the prediction of power battery SOC. Based on the working condition of UDDS in vehicle simulation software which is called as ADVISOR, the paper takes the current data in working condition as input current of EKF prediction model, the initial value of SOC is predicted and corrected gradually, eventually power battery SOC is predicted. By comparing the prediction curve and real curve of SOC, which indicate that the average prediction error of EKF is 0.033. However, EKF prediction error is linear increasing trend apparently.It’s worth noting that prediction objects of Markov prediction model is dynamic system with random changes, state transition probability reflects the mutual influence among random factors, which will benefit prediction error correction. Based on the above analysis, the paper makes improve on disadvantage which EKF prediction method will make cumulative prediction error increasing linearly with time, by combining prediction model of Markov and EKF, using EKF-Markov method to prediction of vehicle lithium-ion power battery. Firstly, the initial value of SOC is got by EKF prediction model, then make use of Markov model to get state division of prediction error and state-transition matrix. Lastly, using EKF-Markov method to predict again. The paper design test validation scheme which based on the working condition of UDDS to get data samples of power battery SOC. Comparative analysis before and after improvement revealed that prediction average error of EKF-Markov is 0.0072, which reduce 78.18% compared with EKF,which indicates that EKF-Markov has a accurate prediction for power battery SOC at certain moment in future. In practical application, by EKF-Markov method, the state of power battery SOC in future could be predicted relatively accurately, which provide reference information accurately for judgment of the next start time point to charge battery in advance.The paper focuses on SOC prediction of lithium iron phosphate battery and makes improvement on the prediction algorithm of EKF, which reduce prediction error effectively. The research work in the paper will help people make reasonable scientific decisions in practical application, on the other hand, which has significance to BMS(BMS, Battery Management System) management systems which are being built or to be built. |
