State-of-Charge Estimation of Lithium-Ion Batteries Based on Data-Model Fusion Method

Li-ion batteries (LIBs) State of Charge (SOC) estimation is essential to guarantee the secure and dependable use of the battery as well as to achieve rational battery energy management. In this paper, data-driven with battery equivalent circuit models are integrated to achieve battery SOC estimation. More convincing criteria are chosen to demonstrate the reliability of the estimation. First, preliminary SOC prediction is achieved using current with voltage from the cell dynamic stress test (DST) dataset, combined with a nonlinear autoregressive structure with exogenous inputs (NARX) and an extreme learning machine (ELM) stochastic learning algorithm. Then, on the basis of the constructed 2RC equivalent circuit model (ECM), SOC prediction and the polarization voltage of the equivalent circuit are used as state variables to achieve battery SOC estimation using extended Kalman filtering (EKF). Finally, using the dynamic US06 highway driving program as well as federal urban driving schedule (FUDS) test datasets, in various temperatures (range of -10 degrees C to 50 degrees C with 10 degrees C bins), the precision and robustness of the presented SOC estimation approach are verified. The outcomes demonstrate that the presented approach is more realistic and reliable at each temperature compared to the SOC estimates obtained using data-driven only.