Remaining Life Prediction of Lithium Battery Based on Data Preprocessing CNN-GRU Model

In order to overcome the non-stationarity and non-linearity of the capacity change of traditional lithium batteries during cyclic charging, which will be affected by the regeneration capacity change when life prediction is carried out, a neural network lithium battery life prediction technique using Whale Optimisation Algorithm (WOA) optimised Variational Modal Decomposition (VMD) and Convolutional Neural Network (CNN) combined with Gated Recycling Unit (GRU) is proposed. Firstly, the number of decomposition layers and penalty factor in the VMD were optimised by WOA, secondly, the optimised VMD was used to divide the capacity of Li-ion batteries into several components, and in the third step, the CNN-GRU neural network model was used for prediction. The WOA-VMD-CNN-GRU model is validated by NASA dataset, and the experimental results show that compared with other methods such as CNN, GRU, feed-forward neural network (FNN), the results of this study have a maximum average absolute error of no more than 0.0092, a maximum root-mean-square error of no more than 0.0096, a minimum relative accuracy of 0.9937, and a maximum prediction error of no more than 2 cycles for RUL, indicating that the study has a maximum prediction error of no more than 2 cycles. exceed 2 cycles, indicating that the study has good robustness and high accuracy.