What is it about?

This article proposes, designs and implements an online framework for estimating the state of charge (SoC) of battery cells using system identification methods. The methods combine two modified nonlinear optimization algorithms (modified Genetic Algorithm and modified Levenberg Marquardt) that are adapted to estimate the battery cell parameters. Then a linear recursive Kalman filter is used to estimate the state parameters of the battery cell. Furthermore, a new statistical approach is developed to deal with the hysteresis effects in the cell. The SoC estimation in the electric vehicle (EV) is challenging because the battery can have hundreds of cells with varying load currents and short time requirements for SoC estimation to prolong the battery pack lifetime. Therefore, accurately estimating the SoC of the cells in a battery pack is crucial for their effective use. The framework is robust, optimal and feasible in time-constrained environment with reasonable accuracy.

Featured Image

Why is it important?

The battery SOC is a vital parameter that reflects the battery's performance. Accurate SOC estimation can protect the battery from overcharging or discharging, enhance the battery life, and enable the application to make rational control strategies for energy-saving purposes.


 To further improve the SOC estimation accuracy using the model-based method, the battery model parameters should be estimated simultaneously as they change with the battery aging.  To balance the SOC estimation accuracy and the computational load, the sampling time should be adaptively adjusted according to the situation.  To achieve both high SOC estimation accuracy and low computational time, a hybrid algorithm that combines different algorithms should be developed and tested.  To achieve high online SOC estimation accuracy in real-time applications, a combination of model-based and data-driven methods should be explored in future studies.  To take advantage of the machine learning algorithms, a digital twin concept could be applied for online battery SOC estimation.  To enhance the efficiency of the BMS, a combined state estimation method that uses both data-driven and model-based approaches could be developed.  To reduce the overall cost of xEVs, a cloud-based BMS that leverages machine learning algorithms could be designed.

Dr. Mohammad Rezwan Khan
Aalborg Universitet

Read the Original

This page is a summary of: An online framework for state of charge determination of battery systems using combined system identification approach, Journal of Power Sources, January 2014, Elsevier, DOI: 10.1016/j.jpowsour.2013.07.092.
You can read the full text:




The following have contributed to this page