A Comparative Evaluation of Filters Applicable to Low-Complexity Processors in the Model-Based State-of-Charge Estimation of Lithium-Ion Batteries

The widespread adoption of electric vehicles (EVs) has led to a significant increase in demand for lithium-ion batteries (LIBs). Due to their inherent characteristics, the charge and discharge processes of LIBs must be closely monitored and managed by a battery management system (BMS) to ensure safe operation. Accurate estimation of the battery's state of charge (SoC) is crucial for determining remaining driving range, controlling current draw within safe limits, and providing reliable information to users. SoC estimation methods vary from simple to advanced techniques, such as deep learning, which require more memory and processing power. While processor cost is less critical for high-value EVs, it becomes vital for more affordable models. Battery manufacturers also prefer to embed SoC estimation and charge/discharge control directly into BMS hardware using low-capacity processors, necessitating resource-efficient methods. This research focuses on SoC estimation based on an equivalent circuit model (ECM), utilizing terminal current, terminal voltage, and the open-circuit voltage (OCV)-SoC curve. However, SoC estimates often show oscillations and abrupt changes, which can be reduced through filtering techniques. Some filters are computationally intensive, while others are resource-efficient. This paper compares various filters, including Moving Average Filter (MAF), Exponential Filter (EF), Butterworth Filter (BWF), Extended Kalman Filter (EKF), and Particle Filter (PF), to evaluate their impact on SoC estimation accuracy and processing time. Experiments using battery data from Supplemental Federal Test Procedure (US06), Urban Dynamometer Driving Schedule (UDDS), Los Angeles 92 (LA92), and two combined drive cycles identified the most effective filtering method suitable for low-complexity processors.