Lithium ferro phosphate (LiFePO4) battery state of charge estimation using unscented Kalman Filter

Abstract

Throughout the past year, the development of Electric Vehicles (EVs) has been rapidly increasing due to shortage of unsustainable energy source and global climate warming. Battery is one of the key technologies applied in EVs that also contributes to the restriction of EVs expansion. Lithium Ferro Phosphate (LiFePO4) is one of the lithium-ion batteries that is widely used due to its high energy density, long lifespan, high efficiency, fast charging characteristic and low self-discharge. For battery management system (BMS), the state of charge (SOC) estimation of the battery is an indispensable parameter that need to be essentially considered. The accuracy of SOC estimation is very crucial to monitor the charging and discharging operation of the battery pack for optimizing the performance and prolong the lifespan of the battery. Since the battery stores the energy in the chemical state, and this chemical energy cannot be directly accessed, then the SOC estimation becomes very complex. This also includes many uncertainties and noises contribute a challenge in determining the accuracy of the SOC estimation. The objectives of this project focus on the development of the LiFePO4 battery model using Equivalent Circuit Model (ECM) to predict the SOC by using Unscented Kalman Filter (UKF) algorithm. Several battery ECMs with up to three level of RC pairs have been studied to compare the accuracy of the model. The battery ECM parameters were estimated using MATLAB Parameter Estimation Tool by utilising the dynamic behaviours of the LiFePO4 battery from the experimental data. The dynamic characteristics of the LiFePO4 battery have been experimentally studied by using Constant Discharge Test (CDT), Pulse Discharge Test (PDT) and Random Charge and Discharge Test (RCDT). The SOC estimation by using UKF algorithm was implemented by using battery ECM from one RC pair until three RC pairs. Then, the accuracy of the battery ECMs were analysed by using error analysis such as Mean Absolute Error (MAE), Mean Square Error (MSE) and Root Mean Square Error (RMSE). From the result of error analysis, the most accurate battery ECM was selected to be implemented in the UKF algorithm to estimate the SOC of the LiFePO4 battery The results from the simulation are then validated by comparing to the real SOC by using Coulomb Counting method Then, the performance of the UKF algorithm was compared to the Extended Kalman Filter (EKF) and Particle Filter (PF) by using error analysis of MAE, MSE and RMSE. From the result of the error analysis, the most accurate algorithm for estimating the SOC is determined.