Electric vehicles are equipped with thousands of lithium-ion cells. For superior lithium-ion battery performance and guaranteeing its longer life and safety, there is a fundamental requirement for a BMS (Battery Management System). Estimation of aging rate, state of health state of charge, core temperature, and residual life in the lithium-ion cells are very important. The state of charge indicates the battery's remaining capacity, and improves battery safety by preventing high, charge or discharge. Estimating the battery charge status is one among the critical tasks of BMS. However, it is difficult to accurately estimate the state of charge because the state of charge is the internal state of the battery cell that cannot be measured directly. So, designing specific methods to estimate the state of charge is necessary. This paper presents the battery model identification and state of charge estimation algorithm based on the Kalman filter for lithium-ion batteries in electric vehicles. In this method, first, a second-order RC equivalent circuit models of the lithium-ion battery was developed. Then the Kalman filter method was used to estimate the battery parameters and battery charge status.

Introduction

With rising energy prices, increasing environmental pollution, and declining fossil fuel reserves due to over-consumption, electric vehicles have received much attention due to their many advantages, such as high power, high energy density, and high voltage, less pollution, no memory effect, longer life, and lower self-discharge rate. Lithium-ion batteries are more widely used in electric vehicles than other batteries.

Compared to other regularly used batteries, lithium-ion batteries offer high energy density, high power density, and long life, and are therefore widely used in electric vehicles (EVs). The battery system in electric vehicles is controlled by a battery management system (BMS).

With the development of electric vehicles, various methods have been proposed to estimate the battery charge status. One of the main tasks of the BMS is to estimate the battery condition, such as the state of charge.

Since measuring of the state of charge directly is impossible, it is estimated by applying battery data, such as current, voltage, and so on. In practice, the inaccurate estimation of the state of charge of the battery would lead to overcharging/discharging, which affects life and safety.

Various methods have been presented for estimating battery state of charge. Coulomb counting [1-3] is one of the simplest techniques, where state of charge is computed by integrating the measured current. Another simple approach is the open-circuit voltage technique, where the state of charge is calculated using the open circuit voltage- state of charge relation [4-6]. However, these methods are not suitable because error accumulation in the Coulomb counting method results in inaccurate current measurement, and the open circuit voltage technique cannot be used in continuous operation because the battery must be disconnected from the circuit. Impedance spectroscopy is also used to determine the state of charge of the battery by correlating the recorded impedances of the battery at different state of charge values [7]. Although this technique is time-consuming and temperature-dependent. It also requires additional laboratory testing, which drives up the cost. Hence, this technique is not suitable as well.

Neural networks (NNs) [8-10], Kalman filter (KF) , and KF extensions for nonlinear systems such as the extended Kalman filter (EKF) , the unscented Kalman filter (UKF) , and the sigma-point KF (SPKF) are other techniques for state of charge estimation which approaches outperform Coulomb counting and OCV in terms of performance [11-19].

The extended Kalman filter (EKF) technique, unlike classical estimation methods of state of charge (such as the ampere-hour integration method), does not rely on the initial value of state of charge and has no accumulated error, making it ideal for actual electric vehicles operating conditions. EKF is a model-based method, and the accuracy of the state of charge, which is estimated using this approach, highly depends on the precision of the battery model and model parameters. The characteristics of a lithium-ion batteries change due to a variety of factors and indicate considerable nonlinearity and variance over time. The battery is approximated as a linear, time-invariant system in a typical EKF technique; however, this approach presents estimation errors [11-19].

To eliminate the aforementioned problems and ameliorate the accuracy of state of charge estimation, this research presents a state of charge estimation technique that integrates time-varying battery parameters into the EKF algorithm.