Analyzing Computational Complexity and Footprint: A Comparative Study of State of Charge Estimation across Various Equivalent Circuit Models for Batteries

This study delves into the computational complexities and footprints associated with SoC estimation using Extended Kalman Filter (EKF) across a few Equivalent Circuit Models (ECMs): Rint, first-, second-, and third-order models. Through a comparative analysis, we consider the computational demands imposed by each model. The research focuses on understanding the interplay between computational complexity and footprint, shedding light on the trade-offs inherent in choosing an appropriate ECM for SoC estimation. The ECMs under scrutiny are evaluated for their suitability in practical applications. Notably, the results yield a surprising revelation: the footprint in terms of RAM and Flash required by all models remains relatively the same despite the highest optimization level, but the first, second, and third-order models demand 15%, 76%, and 88% more computational time, respectively, compared to the Rint model. Though accuracy is not an issue, the mean error of SoC estimation is consistently below 4 % for all models. In addition, the second and third-order models cause more estimation errors and computational complexity. The results from this research can provide guidance for Battery Management System (BMS) design to trade-off between computational complexity and estimation accuracy.