Fast estimation of lithium-ion battery state of health using time series classification

Rapid and reliable estimation of the State of Health (SoH) of lithium-ion batteries is critical for large-scale diagnostics, predictive maintenance, and second-life applications. Traditional machine learning approaches often rely on full charge–discharge cycle data, making them impractical for high-throughput scenarios. In this work, we propose a novel multivariate Time Series Classification (TSC) approach based on the Random Interval Classifier (RIC) to estimate battery SoH using only partial charging cycle data, specifically temperature and voltage time series. Using Oxford and NASA benchmark datasets, we demonstrate that accurate SoH predictions can be achieved using only about 2%–3% of the measurement points from the full charge cycle. Our method preserves estimation accuracy (SoH prediction errors of about 1%) while reducing diagnostic time by a factor of 30, offering a scalable solution for real-world battery management systems (BMS) and industrial testing environments.