LSTM-based State of Charge Estimation for Electric Vehicle Battery Management Systems under Varying Temperature Conditions

Abstract

In this article, a Long Short-Term Memory (LSTM) network module was proposed for estimating the State of Charge (SOC) of a Lithium-Ion battery used in electric vehicles as part of the intelligent Battery Management System (EV-BMS). The method employs a time-series algorithm utilizing voltage, current and temperature as inputs and was trained with data from real driving cycles under a range of temperature conditions. This method avoids explicit battery modeling and is able to adjust to the complex nonlinear behavior of the system, which differs it from all traditional methods that do involve direct modeling of the battery system. Studying UDDS, US06, and Mixed drive cycles demonstrated high predictive accuracy, with MAE ranging from $\mathbf{1. 2 \%}$ to 3.4% and RMSE from $\mathbf{1. 6 \%}$ to 4.9% across all temperature conditions. The visual inspection and statistical analysis confirm the robustness and generalization ability of the model. These findings illustrate the applicability of the developed LSTM-based SOC estimator for real-time incorporation into Electric Vehicle Battery Management Systems and as an enabling component for smart grid-oriented energy management strategies.