A multi-scale SOC estimation method for lithium-ion batteries incorporating expansion force

Abstract

Electric vehicles (EVs) have become a viable alternative to fuel vehicles, and accurate State of Charge (SOC) estimation of the lithium-ion battery is the key to guarantee EVs' safe operation. However, small data sampling interval like 1 s is usually required for exiting SOC estimation techniques, which cannot be achieved in practical EVs operation scenarios. This paper proposes a novel SOC evaluation method through the fusion of expansion force with voltage and current measurements. Specifically, long-term estimation (i.e., global trend) of SOC are established by extracting information from expansion force and voltage data with long short-term memory (LSTM) algorithm, while short-term estimation (i.e., local variation) of SOC are obtained by extracting information from current and voltage measurements with Support Vector Regression (SVR). After that, long-term and short-term (i.e. multi-scale) SOC estimations are fused to provide final SOC estimation. Various test data under different driving profiles are utilized to validate the proposed method, including NEDC, and results are compared to those from widely used SOC estimation techniques like LSTM. Results demonstrate that with sampling interval increases from 1 s to 10s, the multi-scale SOC estimation method can maintain maximum absolute error (ME) less than 2.84 %, while mean absolute error (MAE) changes from 0.32 % to 0.48 % and root mean squared error (RMSE) changes from 0.44 % to 0.64 %. It demonstrates that the proposed method can effectively alleviate the dependence on small data sampling interval, thus can provide accurate SOC estimation with various sampling intervals.