State-of-Charge Estimation Of An LiFePO4 Battery System

The core requirement of battery management system of electric vehicle is the need to have a clear knowledge of current state of charge (SOC) of the battery pack. SOC is used as a criterion of health state in both balance management system and security management of the battery pack. More consideration is needed in the energy management of Lithium iron phosphate battery because of its own characteristics:over-discharge would cause irreversible damage to the battery while over-charge would give rise to overheat or even explosion. It is still a worldwide problem to measure the SOC of the Lithium iron phosphate accurately under working condition.This thesis starts from the analysis of the internal electrochemical reaction of Lithium iron phosphate. SOC-OCV curve of the battery model is measured and ohm resistances is calculated from the charge-discharge experiments under different SOC states and current working conditions. A second-order RC equivalent battery model is established. A good basis for the next simulation of SOC is provided.In this thesis, an extended Kalman filter based SOC estimating algorithm is implemented based on the second-order RC model. And the efficient of this algorithm in the SOC estimating system is proved. The gap in performance under different working conditions is analyzed. Moreover, limitations and disadvantages of the normal extended Kalman filter algorithm are analyzed.Since the estimation of the lithium iron phosphate battery packs’SOC is inaccurate under complex working conditions, this thesis finds that one of the main error sources is the modeling error. A model information based noise compensating extended Kalman filtering algorithm is implemented in this thesis. It is a method that adapts to the working conditions of the battery packs. Based on the discharging characteristic of the battery pack under working conditions, this algorithm extracts the features for the model classification, and compensates the extended Kalman filter model according to the parameters of the charge and discharge of the batteries for optimization estimation. This algorithm can be used in the estimating model of the SOC when the measurement noise model of the system is known. It can effectively avoid divergence of the SOC estimation and poor correctional performance of the SOC, ignoring the error of system model that caused by the multi factors.In practical battery system, the measure noise model is usually unknown, and it is easily effected by external disturbance. That is why the noise compensating extended Kalman filtering algorithm has difficulty to achieve good result. Based on the study of the mechanism of Kalman filter, this thesis makes the filter system to adapt the change of the external noise model with robust Kalman filter. According to the bad performance of the single robust adaptive filter algorithm when an initial error exists, a multi- models-based robust adaptive Kalman filtering algorithm is proposed on the basis of the idea of multi adaptive model filtering. This algorithm has a better precision and convergence, and it is suitable for complex working conditions.