State of health prediction of lithium-ion batteries based on autoregression with exogenous variables model

The gradually decreasing capacity of lithium-ion batteries can serve as a health indicator for tracking their degradation. Therefore, it is important to predict the capacity of future cycles to assess the health condition of lithium-ion batteries. According to electrochemical theory and the characteristics of the data curves, this paper proposes several ideas for feature extraction. A novel fusion prognostic framework is proposed, in which a data-driven time series prediction model is adopted and combined with extracted features for lithium-ion battery capacity prediction. The proposed method is based on an autoregression with an exogenous-variable model that can self-adaptively update at each cycle and then predict the state of health in the next cycle and cycles in the near future. Under the assumption that the historical capacity data is available, the experimental results showed that by using the proposed autoregression with exogenous variables model, the root mean square error, mean absolute error, and mean absolute percentage error of the prediction results were 0.000963, 0.000562, and 0.000584, respectively, which indicated that the prediction results were precise.

Automated summary

This paper proposes a prediction method based on feature extraction and data-driven models for predicting the capacity of future cycles of lithium-ion batteries to assess their health condition. Experimental results demonstrate precise prediction results.