Battery State of Health (SOH) estimation—the process of quantifying remaining capacity and performance relative to a battery's rated specifications—has become a foundational capability in electric vehicle fleet management, energy storage systems, and next-generation battery management systems (BMS). As tracked by organisations such as the IEA, the global electrification of transportation is accelerating demand for accurate SOH estimation as a critical determinant of vehicle range, resale value, safety, and second-life repurposing.

Among the 70+ patent and literature records retrieved spanning 2011–2026, the field spans three primary technical paradigms. Model-based methods use equivalent circuit models (ECMs), electrochemical models, and physics-informed approaches combined with adaptive observers such as Extended Kalman Filter and recursive least squares to identify parameters like ohmic resistance and capacity that track SOH. Data-driven and machine learning methods—including LSTM, BP, ELM, DNN neural networks, SVR, and GPR—learn SOH mappings from historical charge-discharge data without explicit physical modelling. Hybrid/fusion methods combine both paradigms, often incorporating Kalman filtering, adaptive observers, or knowledge graph frameworks to achieve robustness under real-world operating variability.

Health indicators (HIs) used across the dataset include incremental capacity (IC) curves, voltage integrals over fixed intervals, constant-voltage (CV) charging duration, electrochemical impedance spectroscopy (EIS) features, charging time segments, and open-circuit voltage (OCV) characteristics. The patent landscape analysis reveals that data-driven methods are now the dominant innovation axis, though real-world generalisation remains the key differentiator.

Data-driven methods are now the dominant innovation axis, but real-world generalisation remains the key differentiator. R&D teams should prioritise training datasets from diverse operating conditions (temperature, C-rate, usage pattern) rather than optimising further on controlled laboratory benchmarks. Patent landscape analytics can identify white-space opportunities in underserved operating regimes.

Partial charging data compatibility is a near-term competitive frontier. Methods that extract reliable SOH from incomplete, non-standardised charging segments—such as CV-phase duration features or short random charging segment methods—are particularly valuable for real-world fleets where full cycles are rarely observed. This aligns with guidance from IEEE standards bodies on real-world battery testing protocols.

Cloud-edge hybrid architectures are becoming the deployment standard. The convergence of IoT telematics, connected car platforms, and cloud-based model serving—as evidenced by GEOTAB, APPAREO IOT, and iCar filings—signals that IP strategists should evaluate defensibility not just in estimation algorithms but in data pipeline and model update architectures. The data security and IP protection implications of cloud-based model serving are particularly relevant for enterprise deployments.

Second-life battery SOH screening represents an underserved but growing IP space. With EV batteries reaching end-of-first-life in increasing volumes, rapid non-destructive SOH assessment methods—particularly EIS-based and partial-discharge approaches—are commercially strategic and currently under-patented relative to in-vehicle SOH estimation. Jurisdiction strategy should weight EP and KR heavily: the density of active patents in the European Patent Office and Korean jurisdiction, combined with active filers from Audi, GS Yuasa, Samsung SDI, Hyundai, and LG, makes freedom-to-operate analysis essential. Organisations such as the EPO provide public search tools, but comprehensive FTO analysis requires the full citation and family data available through PatSnap's customer platform.