Comparison study between hybrid Nelder-Mead particle swarm optimization and open circuit voltage—Recursive least square for the battery parameters estimation

Abstract

Recently, the whole world has turned to the rechargeable battery as the primary source of electric vehicles (EVs). Meanwhile, the majority of the automobile manufacturers quantify lithium as a key material for the development of battery chemistry. Currently, Lithium-ion (Li-ion) batteries are the dominant technology for battery electric vehicles (BEVs). The real value of Lithium is that they deliver high and fast energy densities in a small volume or configuration. Thus, the description, identification and modeling of the battery, during its operation, are essential actors for the success of EVs development. However, modeling Li-ion-based energy storage devices has not yet been easy enough due to the complexity of chemical reactions in real time. In this work, the modeling of Li-ion cells is performed with dynamic equivalent electrical circuit model (ECM). This study aims to exploit the advantages of two hybrid estimation methods for the identification of Li-ion battery model parameters. The first method is hybrid optimization algorithm PSO-NM which is composed of a particle swarm optimization algorithm (PSO) and a Nelder–Mead algorithm (NM). On the other hand, the second hybrid model, namely the OCV-RLS algorithm, benefits from the advantages of direct measurement of open circuit voltage (OCV) and of a Least Square Recursive (RLS) to obtain an optimum overall performance estimate. The validation results support the reliability of these proposed hybrid methods to identify the Li-ion battery model. In addition, they constitute a reference for the comparison of the two algorithms PSO-NM and OCV-RLS to solve the problem of Li-ion battery SOC estimating.