Abstract:
Electric Vehicles (EVs) have attained substantial interest in the recent years, since they are efficient, sustainable and zero-carbon emitting means of transportation as compared to the conventional fossil-fuel powered vehicles. As EVs are becoming popular, the use of Lithium-ion (Li-Ion) battery technology is exponentially increasing due to its good charge/discharge performance, large capacity, increased energy and current density and optimum power support. Li-Ion batteries are the bottleneck of EVs but they have complex electrochemical reactions thus they are required to be accurately monitored and controlled. To ensure the safe operation, improved driving range, optimized power management strategy, prolonged service life and decreased cost of the batteries, a Battery Management System (BMS) is essential. BMS are the brains behind the battery technology and its key functions are to estimate State of Charge (SOC), State of Health (SOH), to estimate capacity and state of function (SOF). However, the major task of BMS is State of Charge (SOC) estimation. SOC indicates the available capacity of the battery that can be withdrawn from it and can be used to prevent it from going into deep charging or discharging and to operate the battery in such a way that aging effects are reduced. Various online SOC estimation methods for batteries have been developed over the past years that are classified into two categories; direct measurement method and model-based method. Recent studies on SOC estimation focus on model-based methods with improved accuracy. These include linear Kalman filter (LKF), Extended Kalman Filter (EKF), Unscented kalman filter (UKF), fuzzy logic and neural networks. Kalman filter provides an efficient computational recursive method to evaluate the SOC. This research studies uses the Thevenin equivalent circuit theory to model the transient behaviour of the Li-Ion battery and the SOC is evaluated using four methodologies including Coulomb counting, LKF, EKF, UKF methods. First, the battery is mathematically modelled and then the estimation is done via all four methods in MATLAB/Simulink. A comparison is made between all four methods which shows that SOC determination via kalman filters gives error less than 1%.
