Volume 20: Sustainable Energy Solutions for a Post-COVID Recovery towards a Better Future: Part III

Surrogate Modelling for Battery State-of-Charge Estimation in Electric Vehicles Based on Pseudo-2-Dimensional Model and Gradient Boosting Machines Min Hua , Quan Zhou , Chongming Wang , Cetengfei Zhang, Bin Shuai1 , Hongming Xu

Abstract

Lithium-ion batteries are the main power source of electric vehicles (EVs). Prediction of battery State-of-Charge (SoC) for EV is important but challenging because battery SoC cannot be directly measured through onboard sensors. This paper proposes a surrogate model
for battery SoC evaluation based on a Pseudo 2-Dimensional (P2D) model, offering increased physical insight and predictability than the conventional Resistance-Capacitor (RC) model in a computationally efficient way. By simulating battery performance under different cycles using COMSOL, the proposed P2D model demonstrates its strong representation capability quantified by Root Mean Square Error (RMSE), which can be controlled below 0.03 under all studied conditions while providing physical and analytical characteristics in
battery operation. Furthermore, based on the simulated data from the P2D model, the proposed surrogate modeling using Gradient Boosting Machines (GBMs) is proposed to build the recurrent model for the voltage and SoC prediction using previous voltages. The results from GBR with Root Mean Square Error (RMSE) 0.0387 are close to training data with RMSE 0.0258.

Keywords Lithium-ion Battery, Surrogate Battery Modelling, State-of-Charge, Gradient Boosting Regression