Mars based state of charge estimation using real-life loading condition of lithium-ion battery for electric vehicle

Abstract

Battery state of charge (SoC) estimation is very crucial for the safe operation of electric vehicles (EVs). For practical application, dynamic profiles resembling the EV drive cycle profiles should be considered for SoC estimation of lithium-ion battery with high accuracy. In this paper, multivariate adaptive regression splines (MARS) method, a nonlinear and nonparametric regression approach, is used for SoC estimation of INR1865020R lithium-ion battery. The MARS model is trained by experimental dynamic stress testing (DST) and federal urban driving schedule (FUDS) profiles at 25°C, as per the US Advanced Battery Consortium (USABC) testing procedure. These dynamic profiles consider various current steps with a wide range of amplitudes for different time duration to include regenerative charging. The developed MARS model is then validated using an independent US06 highway driving schedule. The results show good agreement with the experimental data with a maximum error of 6%, suggesting the potential future application of the MARS model integrated with a battery management system for SoC estimation.