Gradient boosted regression model for the degradation analysis of prismatic cells

Developing an accurate predictive model in a battery management system is a challenging task. Tree-based models are widely used to deal with non-linear problems because of their relative ease and prediction capabilities. We propose a gradient boosted regression (GBR) model with the artificial bee colony (ABC) algorithm to analyze the capacity degradation of prismatic cells. The ABC algorithm is used to obtain the optimal parameters of the GBR model. The proposed model is validated by six prismatic cells. The results show that the proposed model provides better prediction accuracy than long short-term memory (LSTM), empirical mode decomposition-based LSTM (EMD-LSTM), Elman-based LSTM and random forest regression (RFR) models. Besides, the effect of optimal hyperparameters for LSTM and proposed models is provided. The average calculation time including the time to find optimal model parameters for all datasets is 2.05 min. For four unseen datasets, the mean absolute percentage errors (MAPE) of the proposed model are obtained as 0.70%, 0.62%, 0.87%, and 0.46%. The results show that our proposed model can reliably predict the capacity degradation of prismatic cells.