State of Health Resistance Estimation Based on a Feed Forward Neural Network

Meeting electrification goals and effectively integrating renewable energy requires deploying Battery Energy Storage Systems (BESS). Accurate estimations of the State Of Health (SOH) are crucial. Until now, the predominant focus has been on batteries' capacity SOH (SOHc). However, an equally crucial aspect of dependable battery storage is the battery's internal resistance, which determines the resistance SOH(SOHr). Given the complexity of battery data, machine learning models have become highly adopted for SOH estimation owning ot their capacity to discern patterns and relationships. This research proposes a novel method combining Bayesian optimization with a stacked, bidirectional LSTM model to achieve highly precise SOHr estimations. The efficiency of this approach is verified using a private battery dataset.