A novel method for SoC estimation of lithium-ion batteries based on previous covariance matrices and variable ECM parameters

Lithium-Ion battery powered electric vehicles (EVs) offer many benefits, such as having high energy efficiency, requiring lower maintenance, and being cheaper to run. Besides, they play a crucial role in decarbonization. These advantages will probably make it an indispensable option for both drivers and governments in the future decades. However, the bottleneck of them is the batteries. In particular, accurate estimation of state of charge (SoC) of batteries, which refers to the remaining driving range, is one of the most notable challenges for EVs. With this in mind, in this paper, a novel Kalman filter-based estimation method is proposed to increase the accuracy of the SoC. The proposed method considers not only the current but also the previous covariance matrices since abrupt changes in the nonlinear dynamics of the battery may lead to incorrect estimation. Consequently, smoother state transitions are provided, and more accurate SoC estimation is possible. The improved method is supported by the 2-RC Thevenin equivalent circuit model, whose parameters are described as a function of the SoC and temperature. The battery model and proposed method are tested with three different driving cycles to prove the efficiency. According to the results, the proposed method can minimize the RMSE of SoC estimation up to at least 10% and provides better SoC estimations for compact EVs.