Research On Life Of Lithium-Ion Battery Based On Deep Learning

Lithium-ion batteries have been widely used in business,science and technology,military,aerospace and other fields due to their outstanding advantages such as high energy density,energy saving and environmental protection.Therefore,research on the remaining useful life（RUL）of lithium-ion batteries has important practical significance for more efficient and safer use of lithium-ion batteries.There are many research methods for the remaining useful life（RUL）of lithium-ion batteries,among which the research method of lithium-ion battery life based on deep learning stands out due to its strong applicability and strong generalization.This article focuses on the research of the remaining useful life（RUL）for lithium ion batteries based on the Bi-directional Long Short-Term Memory network model（Bi LSTM）.（1）The physical structure of lithium-ion batteries is clarified,and its working principle is summarized,and then the meaning of its common physical parameters is introduced.Aiming at the problem that it is difficult to calibrate and update the battery model parameters in real time for the measurable parameters such as charge and discharge current,terminal voltage,discharge depth,and discharge rate in the actual working process of lithium-ion batteries,this paper uses AMESim model simulation software to analyze the aging of lithium-ion batteries.（2）In view of the defect that the current lithium-ion battery data set contains a lot of redundant information and is relatively complex,and cannot accurately track the latest information of lithium-ion batteries,this paper uses a dimensionality reduction method to process the original data set of lithium-ion batteries,and then this data set is normalized,and a lithium-ion battery target data set suitable for neural network algorithm input is established.（3）In view of the long-term dependence of the cyclic neural network and the"gradient disappearance"problem caused by backpropagation during training,this paper adopts the Bi LSTM algorithm to effectively alleviate the above problems,and trains the lithium-ion battery data set through the Bi LSTM network.Results were evaluated using modeling root mean square error（RMSE）,cycle error(E_RUL),and prediction accuracy（RA）.Because the Bi LSTM algorithm has the ability to update in real time,the prediction accuracy of the remaining useful life（RUL）of lithium-ion batteries has been effectively improved.（4）Aiming at the problem that the traditional network is not sensitive to the input data,it is difficult to extract key information and filter secondary information and irrelevant information,this paper adds an attention mechanism structure to the traditional Bi LSTM network model.Finally,the modeling root mean square error（RMSE）,cycle error(E_RUL),and prediction accuracy（RA）are used to evaluate the training effect of the model,and the model effect is optimized.The experimental results show that the Bi LSTM network model after adding the attention mechanism can achieve a more accurate prediction of the remaining useful life（RUL）of lithium-ion batteries.