Research Progress and Prospects of Model-Based and Data-Driven State-of-Charge Estimation for Lithium-Ion Batteries

With the rapid development of new energy vehicles and electrochemical energy storage systems, lithium-ion batteries are widely used in transportation electrification, grid-scale energy storage, distributed energy systems, and portable power supplies owing to their high energy density, long cycle life, fast power response, and mature engineering applicability. State of charge (SOC) is one of the most important state parameters in a battery management system (BMS), as it reflects the remaining available capacity of a battery and provides essential information for energy management, charge–discharge control, safety protection, fault warning, and lifetime optimization. Accurate SOC estimation is therefore of great significance for improving battery utilization, ensuring operational safety, and extending service life. However, SOC cannot be directly measured by sensors and must be inferred from measurable variables such as terminal voltage, current, temperature, impedance, and historical operating data. In practical applications, SOC estimation is strongly affected by battery nonlinear characteristics, open-circuit voltage hysteresis, polarization effects, temperature variation, C-rate conditions, aging degradation, capacity fading, parameter drift, and cell-to-cell inconsistency, which makes high-accuracy, robust, and real-time estimation challenging under complex dynamic profiles, wide temperature ranges, and full-life-cycle service conditions. This paper systematically reviews the research progress of model-based and data-driven SOC estimation methods for lithium-ion batteries and constructs an integrated analytical framework covering battery modeling, parameter identification, estimation method classification, hybrid fusion strategies, and future development trends. First, the battery modeling foundations for SOC estimation are introduced, with emphasis on battery operating characteristics, equivalent circuit models, and electrochemical models. Equivalent circuit models, such as the Rint model, Thevenin model, PNGV model, and second-order RC model, are widely used in online estimation because of their simple structure, clear physical meaning, and low computational cost, while electrochemical models, such as the pseudo-two-dimensional model and single-particle model, can describe internal ion diffusion, charge transfer, and concentration distribution more accurately but are limited by complex equations and high parameter requirements. Second, the key parameters affecting SOC estimation are summarized, including the open-circuit voltage–SOC relationship, ohmic resistance, polarization resistance, polarization capacitance, diffusion-related parameters, effective capacity, Coulombic efficiency, and temperature-dependent coefficients. Their offline identification, online identification, and adaptive updating methods are analyzed, showing that accurate parameter identification and dynamic correction are important foundations for improving SOC estimation accuracy under complex operating conditions. Third, SOC estimation methods are classified into direct methods, model-based methods, data-driven methods, and hybrid fusion methods. Direct methods are simple but sensitive to initial errors and error accumulation; model-based methods provide good interpretability and dynamic tracking ability but depend strongly on model accuracy and parameter reliability; data-driven methods can learn complex nonlinear mappings from operating data but still face challenges in interpretability, generalization, and online deployment; hybrid fusion methods combine physical models, neural networks, adaptive filtering, transfer learning, uncertainty quantification, and physics-informed mechanisms, becoming an important direction for high-accuracy and robust SOC estimation. Finally, this paper discusses the major challenges and prospects of lithium-ion battery SOC estimation from the perspectives of full-life-cycle adaptation, cross-temperature and cross-profile generalization, battery-pack inconsistency handling, physics–data fusion, uncertainty quantification, digital twin applications, and embedded real-time deployment. Overall, this review provides a systematic reference for SOC estimation research and offers guidance for the development of advanced, reliable, and intelligent BMSs.