A residual service life prediction of lithium-ion batteries based on decomposition algorithm and fully connected neural network

Abstract

The challenges faced in battery health management are caused by the occurrence of the capacity regeneration process (CRP) during battery degradation. This article suggests a combination method to predict the remaining useful life (RUL) of lithium-ion batteries, considering CRP. The proposed method starts by breaking down the original data into multiple intrinsic mode function (IMF) components using the improved complete ensemble empirical mode decomposition with the adaptive noise (ICEEMDAN) method. Then, the IMF components are categorized into high-correlation components (HC), which indicate the primary deterioration pattern of the battery, and low-correlation components (LC), which indicate CRP, based on the Pearson correlation coefficient (PCC). Next, the dataset is split into training, validation, and testing sets through data segmentation. The HC and LC data are separately utilized to train and predict using feedforward neural networks (FNN)-I and FNN-II, respectively. Throughout the experiment, the HC and LC data are treated as multiple sets of new capacity data, effectively enhancing the diversity of the dataset. Finally, the predictions from both models are combined to obtain the final capacity degradation curve, and the battery’s RUL is determined. Experiments are conducted on two distinct datasets, achieving a mean absolute error (MAE) of less than 1.31% and a root mean square error (RMSE) of less than 1.74%.