Prediction of remaining useful life and state of health of lithium batteries based on time series feature and Savitzky-Golay filter combined with gated recurrent unit neural network

Prediction of state of health (SOH) and remaining useful life (RUL) of lithium batteries (LIBs) are of great sig-nificance to the safety management of new energy systems. In this paper, time series features highly related to the RUL are mined from easily available battery parameters of voltage, current and temperature. By combining Savitzky-Golay (SG) filter with gated recurrent unit (GRU) neural networks, we developed a prediction model for the SOH and RUL of LIBs. The SG filter is used to denoise the aging features and the GRU model is used to predict RUL of LIBs with different charging strategies. Experiments and verification show that the proposed SG-GRU prediction model is an effective method for different applications, which could give out accurate prediction results under various charging strategies and different batteries with fast prediction response. The prediction model can accurately track the nonlinear degradation trend of capacity during the whole cycle life, and the root mean square error of prediction can be controlled within 1%.

Automated summary

In this paper, a prediction model for the state of health (SOH) and remaining useful life (RUL) of lithium batteries (LIBs) is developed by combining the Savitzky-Golay (SG) filter with gated recurrent unit (GRU) neural networks. Experiments and verification show that the proposed SG-GRU prediction model is effective for different applications, providing accurate prediction results under various charging strategies and different batteries. The model accurately tracks the nonlinear degradation trend of capacity during the whole cycle life, with a root mean square error of prediction controlled within 1%.