Prediction of lithium-ion battery’s remaining useful life using an improved approach combining variational mode decomposition, gaussian process regression, and gated recurrent unit

Abstract

Accurately predicting the Remaining Useful Life (RUL) of batteries is of paramount importance. This paper proposes a novel hybrid approach for predicting battery RUL by integrating the improved Variational Mode Decomposition (VMD), Gaussian Process Regression (GPR), and Gated Recurrent Unit (GRU). Firstly, the Grey Wolf Optimizer (GWO) is employed to optimize the number of decomposition layers and penalty factor of VMD. The battery capacity data is then decomposed using the optimized VMD, and the correlation between the resulting Intrinsic Mode Functions (IMFs) and battery capacity is quantitatively analyzed using correlation coefficients. For the trend component IMF₁ and the remaining components IMF_2-k, optimized GPR and GRU algorithms are utilized respectively to establish prediction models that output the respective components of battery capacity. The final prediction of the battery's RUL is obtained by superimposing these components. Finally, the NASA battery dataset is used to validate the performance of the proposed model, which is compared with other models such as standalone GRU, VMD-GRU, and VMD-GPR-GRU with a single kernel function. The results demonstrate that the proposed model can efficiently capture battery degradation trends with higher prediction accuracy and generalization capabilities compared to single models. Specifically, the model’s prediction results show that the average MAE was 0.21 %, the average RMSE was 0.36 %, and the average MAPE was 0.29 %, validating the effectiveness of the proposed research method.