Research On Remaining Useful Life Prediction Of Lithium-Ion Batteries

With the advancement of the two-carbon goal,energy storage systems have developed rapidly.As the mainstream solution of the energy storage system,lithium-ion batteries play an important role in people’s daily life,but this also puts forward requirements for the stable operation of lithium-ion batteries.In recent years,accidents related to lithium-ion batteries have occurred frequently,and research on its failure prediction and health management(PHM)has attracted attention.As an important part of PHM,the remaining life(RUL)prediction of lithiumion batteries has important research significance and value.In this paper,starting from the structure and working principle of lithium-ion batteries,the causes of battery degradation are analyzed,and the capacity that can directly reflect the degradation of battery performance is used as a health factor.The prediction accuracy of the proposed method is verified.Secondly,in view of the difficulty of parameter selection of the SVR method,a novel and highly optimized GWO algorithm is used to search the parameters of the SVR method,and the search results are assigned to the SVR model,which improves the prediction performance of the SVR model and completes the GWO-Establishment of SVR model.Then,the prediction results of this model are compared with the prediction results of the current state-of-the-art ALOSVR method.The comparison results show that the relative prediction error of GWO-SVR is 7.163%lower than that of ALO-SVR,and the prediction accuracy is higher,which verifies the effectiveness of the method.Finally,in order to solve the influence of local capacity regeneration phenomenon in Liion battery capacity degradation sequence and doped noise signal during battery data acquisition on RUL prediction,the modal decomposition method was combined with Pearson correlation coefficient method.According to the idea of modal decomposition,the decoupling of the overall capacity degradation trend and the local capacity regeneration in the capacity degradation data of lithium-ion batteries is realized.The CEEMD algorithm is selected as the decoupling method,and the capacity degradation sequence is decomposed into a series of IMF components;GWOSVR is adopted.The model predicts the IMF components processed by the Pearson correlation coefficient method separately,and integrates the prediction results to obtain the final RUL.By comparing the prediction results of the CEEMD-GW O-SVR method with the results of the RNN,PF-LSTM,and HI-GPR prediction methods,the superiority of the prediction method is verified.The RUL experimental verification is carried out,and the results prove that the CEEMD-GWOSVR method proposed in this paper has strong universality and robustness.