State-of-health rapid estimation for lithium-ion battery based on an interpretable stacking ensemble model with short-term voltage profiles

Lithium-ion batteries are playing an increasingly important role in industrial applications such as electrical vehicles and energy storage systems. Their working performance and operation safety are significantly impacted by state of health (SOH), which will decrease after cycles of charging and discharging. This paper has proposed a novel two-stage SOH estimation method that can realize SOH estimation flexibly, rapidly and robustly. In the first stage, eight typical 300-s voltage profiles are used for describing the whole charging process and multiple aging features are extracted. Then, a novel stacking ensemble model with five base models is proposed. In the second stage, a Shapley additive explanation approach is introduced to obtain the contributions of features and understand why a prediction is made, thus reducing the concern of applying black-box model. The performance of the proposed model is verified using two different battery degradation datasets and the results show that the accuracy of the proposed model is better than conventional machine learning models including lightGBM, XGBoost, RF, SVR, and GPR. In addition, with various forms of noise interference, the proposed stacking model is proved to be more robust than conventional machine learning models.

Automated summary

This paper proposes a novel two-stage SOH estimation method that can flexibly, rapidly, and robustly estimate the state of health (SOH) of lithium-ion batteries. The method uses voltage profiles and aging features in the first stage, and introduces a stacking ensemble model and Shapley additive explanation approach in the second stage. Experimental results demonstrate that the proposed model outperforms conventional machine learning models in terms of accuracy and robustness.