Research On Health Management Prediction Of Lithium-Ion Based On Gated Recurrent Unit

As a typical new energy battery,the lithium-ion battery is an efficient energy storage unit with high energy density and a long charging cycle.With the wide application of lithium-ion batteries in mobile phones,smart grids,electric vehicles,and other production and life,its health management has been paid more and more attention by industry and academia.The validity of the prediction method is one of the most important links in the research of management science and engineering and the application of lithium-ion battery health management.However,the data of lithium-ion batteries has the characteristics of diversity,it is difficult to complete the data analysis and the construction of the prediction model by relying solely on the basic detection method.The appearance and development of deep learning algorithms bring a new direction to management prediction.Therefore,based on the theory of lithium-ion battery health management and the data of the whole life cycle of the lithium-ion battery,the thesis adopts a deep learning model for feature extraction and prediction,carries out research on lithium-ion battery health management prediction,and provides method support for the health management of lithium-ion battery and other industrial equipment.The main research contents are as follows:Firstly,the main components of current research on lithium-ion battery health management forecasting methods were described,and the research status of related forecasting methods at home and abroad were analyzed.This was done by consulting and analyzing a large number of data and literature on lithium-ion battery health management forecasting methods.The hot places of battery prediction applications both domestically and internationally were outlined,along with the value of further study and the necessity of enhancing battery health management prediction.In this study,the concept of battery prediction research is discussed along with an analysis of the drawbacks and issues with the current research methodologies.Secondly,a gated recurrent unit model for the health prediction of lithium-ion batteries is proposed.The collection and operation conditions of public data sets and experimental data sets are briefly explained.The principles of several commonly used models are reviewed in detail,including Convolutional Neural Networks(CNN),Long Short-Term Memory Networks(LSTM),Echo State Networks(ESN),and Gated Recurrent Unit(GRU).Through the public data set of lithium-ion batteries and the experimental data set collected by our laboratory,the typical learning models CNN,LSTM,GRU,and ESN are used to predict the collected original data respectively,and the prediction effect of each model is analyzed visually.The results show that the prediction effect is better when using GRU directly.Then,considering the feature concealment in the original data of the whole life cycle of the lithium-ion battery,a gated recurrent unit enhanced prediction method based on sparse autoencoder feature extraction was proposed.Since manual feature extraction is difficult to meet the needs,in order to predict the lithium-ion battery more accurately,a sparse autoencoder adaptive feature extraction strategy is adopted.Based on the second part of the gated cycle unit prediction model,a sparse autoencoder is proposed to extract features from the original data to obtain feature vectors,and then the feature vectors are used to carry out gated recurrent unit prediction.Due to the application of an adaptive feature extraction strategy,the proposed method can obtain better prediction results to some extent.Finally,the attention mechanism is added to the previous prediction model to further improve the prediction ability of the gated recurrent unit model.After the sparse autoencoder is used for feature extraction,the attention value of each time point is generated by the attention mechanism,and then the attention value is multiplied bit by bit with the input,and the weight is multiplied by the input as the predictive input of the gated recurrent unit.Through the improved adaptive feature extraction of attention mechanism,combined with the deep learning prediction ability of gated recurrent unit,the prediction performance is effectively improved and the prediction error is reduced,which provides strong support for the lithium battery health management prediction.