STATE OF CHARGE (SOC) ESTIMATION TECHNIQUES: REVIEW VARIOUS SOC ESTIMATION TECHNIQUES, INCLUDING MODELBASED, DATA-DRIVEN, AND HYBRID APPROACHES

State of Charge (SoC) estimation plays a crucial role in battery management systems (BMS), directly impacting the performance, safety, and longevity of lithium-ion batteries. This study presents a comparative review of three major categories of SoC estimation techniques: model-based, data-driven, and hybrid methods. The review is driven by the need to evaluate the accuracy, robustness, and practical applicability of these methods across various real-world conditions, including different temperature profiles, battery chemistries, and aging states. The research methodology involved a structured literature search, selection of 45 peerreviewed studies published between 2018 and 2025, and systematic data extraction. Model-based approaches, particularly those using Kalman filters and equivalent circuit models, demonstrated computational efficiency but showed sensitivity to parameter drift and aging. Data-driven techniques, including LSTM networks, Gaussian Process Regression (GPR), and Random Forests, offered high accuracy—often achieving <2% RMSE—but required large, diverse datasets. Hybrid methods, such as AEKF-LSTM and UKF-PSO-LSTM models, consistently achieved the highest accuracy (RMSE <1%) while balancing robustness and adaptability. The findings suggest that while model-based methods are suitable for resourceconstrained systems, hybrid approaches offer the most promising results in terms of overall performance and reliability. These insights can guide future BMS development
and inform system-level design choices in electric vehicle and energy storage applications.