Lithium-ion Battery Model Parameters Estimation Using Modified Grey Wolf Optimization for E-mobility Applications

Abstract

Accurate estimation of battery internal model parameters and consequently state of charge (SOC) prediction is crucial in any battery powered systems. Especially, it is the fundamental need in Electric Vehicles (EVs), smart grids, and energy storage systems. The accuracy in Identification of model parameters will affect the Battery Management System (BMS), battery safety, characteristics, and performance [1]. To estimate the parameters accurately and easily, we require effective, simple, and robust parameters estimation algorithms. In this abstract, we have proposed a new method for parameters estimation using the modified Grey Wolf Optimization (GWO) for Lithium-ion Batteries (LIBs) in EV applications. Second order RC equivalent circuit Model is considered for NMC battery. The parameters estimation and non-linear relation of OCV-SOC are obtained from the experimental data as shown in Fig. 1. This proposed method produces fast, robust, and efficient identification of parameters.

GWO is a revolutionary meta-heuristic optimization method. Meta heuristic optimization approaches are in high demand for tackling optimization difficulties due to its simplicity, flexibility, derivation-free mechanism, and avoidance of local optima [2-4]. GWO primary assumption is to imitate grey wolves cooperative hunting behaviour in the wild. GWO stands apart from the competition in terms of model structure. For the optimization task, it performs well [4].

At the moment, identification of battery parameters is a challenging task because of the factor that the battery is a complex non-linear device, and parameters are affected by several factors. The development of a battery model is essential for effective utilization of battery energy, identification of operating limitations, developing Fast charging algorithms, and safe charging/discharging. An accurate battery model is essential in the design of efficient BMS. In any battery powered system, the battery modelling plays a key role since it accurately reflects the chemical reactions that have occurred inside the battery. We can accurately predict the battery characterization/ estimation of battery parameters [5]. Estimation of battery parameters and states, are pivotal for superior management and control of battery usable capacity, for safe operation and to prolong the useful life of batteries. Batteries have many known parameters like current, voltage, temperature which can be directly accessed from the experiment or from the sensors. Battery states (SOC, SOH) are unknown parameters that can’t be extracted directly from the experiments. These states are estimated from the known parameters. Accurate estimation of battery parameters and SOC is a demanding task in real time applications because batteries are nonlinear, time-dependent electrochemical devices, and depend on several influential internal and external conditions [6]. So robust, efficient, and low complexity battery models are required to connect these unknown parameters with the known parameters to find the SOC, SOH, and to effectively use and manage the LIBs [7].

The equivalent circuit model (ECM) is the most widely used battery modelling technique in EVs to model the LIBs. It is an easy and simple model that uses the electrical circuit components to describe the chemical reactions that occur inside the battery components such as resistors, capacitors, inductors, voltage source, etc. [8]. In this model, we can accurately reflect the chemical reactions like charge transfer reactions, diffusion process and each chemical reaction is represented by a particular electrical component [7].This model avoids the complexity in determination of parameters and is suitable for real-time applications as it provides moderate accuracy and less complexity. ECM gives a flexible trade-off between accuracy and complexity and easier to Identify the battery parameters like voltage, current, and the temperature. In recent times, second-order (2RC) ECM is a widely used model as it gives the best results when compared with the other ECM models in terms of complexity and accuracy. Research is going on in this field to update the parameters based on frequency [9].