SOH Estimation and RUL Prediction Framework Based on ED-LightGBM and ED-BiGRU-LA

State of health (SOH) estimation and remaining useful life (RUL) prediction of lithium-ion batteries (LIBs) are of critical importance to ensure the safety of battery systems. This paper presents an innovative framework for SOH estimation and RUL prediction, i.e., light gradient boosting machine (LightGBM) is used to estimate SOH of LIBs and bi-directional gated circulation unit with logsparse attention (BiGRU-LA) is applied to predict LIBs RUL. In order to enhance the model exactness, enterprise development optimization (ED) is used to optimize the hyperparameters of above models. Firstly, it is difficult to measure the capacity of LIBs directly during operation. So, in charging phase, features are extracted from two perspectives i.e. measured variables as well as computed variables. The features exhibiting the strongest correlation with capacity degradation are identified as health indicators (HIs) by using spearman correlation coefficient. Secondly, a single battery is selected for training purpose, while the remaining batteries of the same type are employed to accurately estimate SOH through the utilization of ED-LightGBM feature processing capability. Finally, RUL is accurately predicted through the deployment of ED-BiGRU-LA, which enables the efficient processing and grasp long-term correlations within the sequential dataset. In this paper, the proposed method is evaluated by using two datasets and analyzed in comparison with other algorithms. The results show that the prediction framework demonstrates excellent performance with outstanding accuracy, robustness and generalization.