State-of-Health Forecasting for Battery Cells using Bayesian Inference and LSTM-based Change Point Detection

With the global shift towards an ecologically conscious environment and the increasing prominence of electric vehicles, the focus on Lithium-ion (Li-ion) batteries continues to grow, and thus, amplifies the need for prognostics and health management. The estimation of State-of-Health (SOH) plays a crucial role in maintaining Li-ion batteries, and numerous predictive models have been suggested to address this objective. This paper proposes a combination between a deep Long Short-Term Memory (LSTM) network and Bayesian Inference (BI) to forecast the SOH of Lithium-ion batteries. The proposed method first detects the Change Point (CP), i.e., the point at which accelerated deterioration begins, then forecasts the long-term SOH of the battery cells after the degradation phase has started. Our model outperforms four other state-of-the-art models tested on the same public datasets.