Lithium-ion Batteries Capacity Degradation Trajectory Prediction Based on Decomposition Techniques and NARX Algorithm

Abstract

It is critical to accurately predict the remaining capacity of lithium-ion batteries to guarantee safe, reliable operation with minimal maintenance costs. However, because of the complicated and nonlinear characteristics of the battery’s degradation throughout its lifetime, predicting the amount of capacity that will still be available in lithium-ion batteries is a complex process. In addition, the phenomena of capacity regeneration have a significant impact on the accuracy of the remaining capacity projection. For this purpose, the signal decomposition method is becoming a more attractive and promising method for overcoming the difficulty of the capacity regeneration phenomena due to its simplicity and capability to accommodate the nonlinear dynamic behaviour of the lithium-ion battery. Therefore, this paper investigates the performance of three signal decomposition techniques: the discrete wavelet transforms, the empirical mode decomposition, and the variational mode decomposition techniques in analysing the capacity regeneration phenomenon. The nonlinear autoregressive neural network algorithm is developed to predict the trajectory of the future capacity of the battery. The performance of the proposed algorithms is analysed by using two datasets from NASA Ames Research centre and the centre for advanced life cycle engineering (CALCE). The comparison results demonstrate that the variational mode decomposition method combined with the nonlinear autoregressive neural network outperforms other methods with 2.385% RMSE and 1.6% MAE.