GA-based approach to optimize an equivalent electric circuit model of a Li-ion battery-pack

This article presents the optimization procedure based on genetics algorithms (GA) to obtain an equivalent electric circuit model (EECM) of a Li-ion battery pack. In the first part, a series of experimental tests in time and frequency domains were carried out. These tests were used to identify the parameters of the EECM under different State-of-Charge (SoC) for a commercial battery-pack. Each EECM consists of a voltage source connected in series with a resistor and a set of k networks composed of a resistor in parallel with a capacitor, where k = 1, 2 o 3 (1RC, 2RC, and 3RC). Subsequently, parametric identification of the EECM was performed using optimization techniques. At this stage, the topology that gives the lowest estimation error was determined, where the options analyzed were to use 1RC, 2RC, or 3RC networks. The objective function consists of minimizing the mean square error between measured and calculated impedances of the different proposed circuit models. GA was used to solve this optimization problem. The minimum error obtained was 1.07% and 1.05% for the EECM with 2RC and 3RC networks, respectively. Finally, these EECMs were implemented in Matlab®/Simulink to validate the Li-ion battery-pack model response for an electric vehicle application. A hardware-in-the-loop (HIL) simulation platform was developed to simulate the performance of an electric vehicle (EV) under different driving cycles. The results show that the GA-based approach allows obtained an EECM of low order to represent the highly dynamic behavior of a Li-ion battery with high accuracy.