State of charge estimation for lithium‐ion batteries based on square root sigma point Kalman filter considering temperature variations

Abstract

The battery management system (BMS) in electric vehicles monitors the state of charge (SOC) and state of health (SOH) of lithium‐ion battery by controlling transient parameters such as voltage, current, and temperature prevents the battery from operating outside the optimal operating range. The main feature of the battery management system is the correct estimation of the SOC in the broad range of vehicle navigation. In this paper, to estimate real‐time of SOC in lithium‐ion batteries and overcome faults of Extended Kalman Filter (EKF), the Square‐Root Sigma Point Kalman Filter is applied on the basis of numerical approximations rather than analytical methods of EKF. For this purpose, the Hybrid Pulse Power Characterisation tests are combined with the non‐linear least square method that acquired the second‐order equivalent circuit model parameters. Then, the newly developed method is tested with an 18,650 cylindrical lithium‐ion battery with a nominal capacity of 2600 mAh in four different ambient temperatures. Finally, the accuracy and effectiveness of the two proposed methods are verified by comparing with results of pulse discharge and dynamic driving cycle tests. The comparison results indicate the error of the proposed algorithm is about 0.02 under the most test conditions.

An efficient field programmable gate arrays based real‐time implementation of smooth variable structure filter to estimate the state of charge of Li‐ion battery in electric vehicle application

“…The low convergence speed is due to the strong non‐linear characteristic of the battery in the low SoC region [43]. Moreover, in this region, the variation in the model parameters is in important with a non‐linear relationship with the SoC [17]. The convergence of the implemented ‘ SVSF Core ’ is therefore confirmed in real‐time despite the wrong values of ‘SoC Init’.…”

“…On the other hand, many indirect estimation methods have been developed in the recent research literature. They are based on state observers such as the Kalman filter (KF), the Extended KF (EKF), [14, 15], the Unscembled KF (UKF) [16], the Sigma‐Point KF (SPKF) [17–19], the Splice KF (SKF) [20], the Lunenberger observer [21], the Sliding Mode Observer (SMO) [22], the nonlinear model based $${\mathrm{H}}_{\infty }$$ [23], the Smooth Variable Structure Filter (SVSF) [24–26] etc. These indirect methods have proven to be stable and accurate for tracking and estimating the online or offline battery's SoC.…”

Enhanced EKF and SVSF for state of charge estimation of Li‐ion battery in electric vehicle using a fuzzy parameters model

“…A comparative study is carried out with recent SoC estimation algorithms with online parameters identification [18,[22][23][24]. The methods used in these references have already been presented in the introduction.…”

“…Two indicators are used throughout this paper to evaluate the precision of the SoC estimation algorithms: the mean absolute error (MAE) and the RMSE. They are computed using the following formulas [22]:…”

“…They are simple and easy to implement, which makes them very suited for online applications [21]. However, the ECM parameters are not constant and vary with many internal and external factors such as temperature, SoC level, ageing and C-rate [22]. The battery's behaviour is also nonlinear notably in the region of low SoC [23].…”