Co-estimation of Li-ion battery states using an improved dynamic model for electric vehicles

In electric vehicle applications, accurately estimating the states of lithium-ion batteries, particularly the state of charge (SoC) and state of health (SoH), is essential for optimizing performance, ensuring reliability, and maintaining precise control over all operating conditions. This paper presents an asymptotic observer for the real-time estimation of all battery states, accounting for nonlinearities caused by hysteresis, the polynomial relationship between internal voltage and SoC, and the effects of Warburg impedance.The proposed estimation method is based on an advanced Extended Kalman Filter (EKF) that incorporates a third-order approximation of the Warburg impedance within a fully integrated lithium-ion battery model. In addition, the EKF performance is systematically compared with Particle Filtering (PF) and Long Short-Term Memory (LSTM) approaches, providing a benchmark against both advanced stochastic filtering and data-driven machine learning techniques. Simulation results provide a comparative analysis of the proposed filter’s accuracy, robustness, and computational efficiency in state estimation.To validate the practical relevance of this approach, the paper compares simulation results with experimental data obtained from actual battery tests. Discrepancies between the simulated and experimental outcomes are analyzed, with particular attention given to model simplifications, sensor inaccuracies, and environmental influences. Furthermore, voltage and capacity estimation are investigated under three ambient temperature conditions (0 °C, 25 °C, and 45 °C), highlighting the influence of temperature on the accuracy and robustness of the estimation framework. This comparison underscores the strengths and limitations of the filtering methods and offers valuable insights into their applicability in real-world EV battery management systems (BMS).The findings emphasize the critical importance of selecting suitable estimation techniques to enhance the performance, reliability, and lifespan of electric vehicle batteries. The integration of model-based and data-driven estimators, together with multi-temperature validation, demonstrates the robustness and adaptability of the proposed framework for practical BMS deployment.