锂离子电池建模研究综述与展望：当前研究的深入分析与未来发展方向

Abstract With the rapid development of global energy transition and low-carbon technologies, lithium-ion battery, as the core energy storage unit, is highly dependent on accurate battery modelling for its performance enhancement and safety management. Battery modelling has gone through a development process from mechanism-driven to data-driven, and from single-scale to multi-scale fusion, forming three main technology systems: Firstly, the equivalent circuit model (ECM), based on the Thevenin framework, uses RC networks to fit battery external characteristics. With hysteresis module embedding and genetic algorithm optimization, it enables millisecond-level responses in BMS real-time control, showing engineering application advantages. However, its modelling logic is limited to port characteristics, lacking deep physical mechanism explanation. Secondly, the physical field model, based on porous electrode theory and partial differential equations, accurately describes lithium-ion transport and electrochemical kinetics, supporting new battery material research and development. Yet, its high computational complexity hinders fast calculation despite mechanistic precision. Lastly, data-driven models leverage data-driven approaches for strong generalization in nonlinear tasks like SOC/RUL prediction. Hybrid architectures improve cross-scenario accuracy via multimodal fusion but suffer from weak interpretability and poor small-sample adaptability. This paper systematically compares the modelling principles, computational costs, prediction accuracies, and typical applications of these three types of models, and analyses the engineering adaptation advantages of the equivalent circuit model, the mechanistic depth of the physical field model, and the data-driven potential of the black box model. Meanwhile, this paper also points out the common challenges faced by traditional models in terms of novel battery system adaptability, multi-field coupling modelling complexity, and deployment of edge computing devices. The research outlook will focus on multi-scale hybrid modelling and data-driven fusion, combined with current large model applications, to provide theoretical support and technical paths for battery R&D, system design and full life cycle management.