Research On RUL Prediction Method For Lithium-Ion Battery Based On CEEMDAN’s Optimized Neural Network

As China’s promised carbon peak and carbon neutrality date approaching,China’s support for the new energy industry has gradually strengthened,which has further promoted the development of China’s new energy industry,and the related battery industry has also been greatly improved,among which lithium-ion batteries have gained the greatest attention due to their better performance.Safety accidents caused by lithium-ion batteries occur from time to time,and once they occur,the losses caused are incalculable.Therefore,the study of the health status and remaining useful life(RUL)prediction of lithium-ion batteries can greatly improve the safety factor of lithium-ion batteries during operation and effectively protect people’s life safety and property safety,so the study of lithium-ion battery life prediction is of great practical significance.In this paper,taking lithium-ion batteries as the research object,a series of studies on the RUL prediction method of lithium-ion batteries are carried out,and the main research work is as follows:Aiming at the problem that the existing fusion methods do not fully apply the hidden features of lithium-ion battery life degradation data,a method of using feature extraction methods to decompose the original lithium-ion battery life degradation parameter signal and then predict it is proposed.The most suitable method for extracting the life degradation data characteristics of lithium-ion batteries is introduced and compared in detail,and the most suitable method for predicting lithium-ion battery RUL,the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN),is selected for the next prediction.Aiming at the problem that the existing fusion methods do not fit enough neural networks in predicting the life degradation of lithium-ion batteries,a multi-model comparative experiment is proposed.A variety of neural networks and machine learning methods that are more suitable for lithium-ion battery RUL prediction are introduced and compared in detail,and the most suitable model for lithium-ion battery RUL prediction,Bidirectional Gated Recurrent Unit(Bi GRU),is selected as the prediction model of the whole method.Aiming at the inefficiency of the neural network model and the unsatisfactory final model when manually adjusting parameters,a method of using the automatic optimization model to optimize the neural network model is proposed to optimize the learning rate and the number of hidden layer neurons that affect the prediction accuracy.After many experimental comparisons,it is found that each Hyperband(HB)automatic optimization model is more suitable for lithiumion battery RUL prediction than Bayesian optimization model,so HB optimization model is selected to automatically optimize the Bi GRU model.Finally,in order to verify the suitability of the proposed CEEMDAN-HB-Bi GRU lithiumion battery RUL prediction method in predicting lithium-ion battery RUL,a validity experiment is carried out,and the final prediction results are obtained,the error of the results is 0,1,0charge-discharge cycles,respectively.By comparing the prediction results of other algorithms,it is found that the algorithm proposed in this paper has high prediction accuracy.