Research on State Evaluation Method of Energy Storage Lithium-Ion Batteries Based on Multi-Source Non-Invasive Big Data

Abstract

Lithium-ion batteries are the backbone of electric vehicles, renewable energy storage, and new emerging smart grid applications. However, the safety and the economic value of such batteries depend heavily on the proper assessment of State of Health (SOH). Conventional invasive measurements provide detailed information; however, they are difficult to apply to sealed or in-service battery packs. This makes non-invasive techniques such as voltage, current, temperature, impedance, and data-driven modeling better suited for battery management at scale. In this paper, three representative research directions, including comprehensive SOH characterization, fast impedance-based SOH estimation, and machine-learning diagnosis of degradation patterns from electrochemical impedance spectroscopy, are reviewed and synthesized. Rather than analyzing each study in isolation, the paper proposes a comparative framework to identify methodological strengths, data requirements, limitations, and potential for practical deployment. A unified SOH evaluation framework is proposed, which combines Gaussian Process Regression-Automatic Relevance Determination (GPR-ARD)-based feature selection, fast impedance calculation, lightweight Extreme Learning Machine (ELM)-based online estimation, and cloud–edge model updating. The practical feasibility is discussed in terms of computational cost, latency, bandwidth, temperature variation, data quality, cybersecurity, and real-time battery management constraints. The study concludes that, to support safer electric vehicles, second-life battery use and sustainable energy storage systems, future non-invasive battery diagnostics should combine physical interpretability, multi-source data fusion, and deployment-oriented model design.