The state-of-the-art in state of charge estimation methods for lithium-ion batteries: Ongoing evolution and future perspective

The state of charge (SOC) of a lithium-ion battery (LiB) is one of the most critical state variables in battery management systems (BMS), providing essential information for safe battery operation, charge–discharge management, and energy management in electric vehicles (EVs). SOC characterizes the remaining usable capacity of the battery during its current operational cycle. Accurate SOC estimation not only helps prevent adverse operating conditions such as deep discharge, thereby ensuring system safety and reliability, but also contributes to improved energy utilization and prolonged battery lifespan. This paper presents a comprehensive and engineering-oriented review of lithium-ion battery SOC estimation methodologies. Conventional methods, model-based approaches, data-driven algorithms, hybrid estimation frameworks, and emerging intelligent estimation strategies are systematically analyzed from the perspectives of estimation accuracy, robustness against temperature and aging variations, computational complexity, real-time feasibility, and onboard BMS deployment capability. Compared with existing review studies that primarily focus on algorithm classification and estimation precision, this review establishes a unified multi-dimensional comparative evaluation framework to assess the practical applicability and engineering trade-offs of representative SOC estimation methods. In addition, recent advances in intelligent battery state estimation, including transfer learning, physics-informed machine learning, multi-state joint estimation, and cross-condition generalization, are comprehensively discussed to highlight future development trends toward adaptive and intelligent battery management systems. The review further summarizes the key technical challenges associated with practical deployment, including parameter drift, battery inconsistency, computational constraints, and real-world operating uncertainties. Finally, future research directions are outlined from the perspectives of physics-data fusion, cloud-edge collaborative estimation, digital twin technology, and next-generation intelligent BMS architectures. This review aims to provide researchers and engineers with systematic methodological guidance for selecting and developing SOC estimation techniques under different practical application scenarios.