State of Health Estimation of Lithium-Ion Battery via Bidirectional Long Short-Term Memory Neural Network with Bayesian Hyperparameter Tuning

The accomplishment of electrification targets and the successful integration of renewable energy necessitates the implementation of a Battery Energy Storage System (BESS), which not only requires proficient and optimal management but also demands strict adherence to safety protocols in their usage. Therefore, precise estimations of the State Of Health (SOH) assume paramount importance. In this context, the increasing popularity of machine learning models for SOH estimation emerges as they demonstrate a remarkable ability to discern patterns and comprehend intricate relationships within extensive battery datasets. In this research paper, a novel proposition is presented, advocating the adoption of Bayesian optimisation in conjunction with a stacked, bidirectional LSTM model in order to attain highly accurate SOH estimations. This approach is subsequently validated using one publicly available battery dataset, ensuring the robustness and credibility of the findings.