Predicting the state of charge of lithium ion battery in e-vehicles using Box-Jenkins combined artificial neural network model

Abstract

Abstract This study presented a novel hybrid model for predicting the state of charge (SoC) of lithium-ion batteries in electric vehicles that combines Box-Jenkins approach with artificial neural networks (ANN). Unlike existing approaches that use either linear or nonlinear models, the suggested method mixes the linear Auto-Regressive Moving Average (ARMA) model with a nonlinear Multi-Layer Perceptron (MLP) network. This integration explores the inherent non-stationarity in SoC data by using the Battery Performance Index (BPI), which normalises SoC for improved time-series analysis. The hybrid model outperformed conventional models, with a R2 of 0.947. Furthermore, it exploited four critical battery parameters—charge rate, voltage, depth of discharge, and energy density—to provide a more precise SoC prediction than previous techniques. The results demonstrate the hybrid model’s stability and capacity to capture complicated battery dynamics, establishing it as a significant step forward in SoC estimate for electric vehicles.

Keywords

• State of charge
• Lithium ion batteries
• e-vehicles
• Box-Jenkins
• Artificial neural network