Research On Remaining Life Prediction Method Of Lithium-ion Battery Based On Transfer Learning

Lithium-ion batteries have been widely used in various fields as energy storage devices due to their unique advantages such as environmental protection and reliability.Taking electric vehicles as an example,the power battery is one of its most important parts,and its development has a decisive effect on the cruising range and driving safety of electric vehicles.However,in the process of continuous charge and discharge and the irreversible electrochemical reaction of lithium-ion batteries,the capacity will gradually decrease,thereby affecting the safety of system operation.Therefore,the remaining useful life(RUL)prediction of lithium-ion batteries is a hot issue in the field.There are three main methods for predicting the remaining service life of lithium-ion batteries: model-based,data-driven,and fusion-based methods.This paper takes transfer learning as the core,and studies the prediction of the remaining service life of lithiumion batteries through the fusion method.The main contents are as follows:First of all,this article uses Long Short Term Memory Network(LSTM)to solve the problem of long-term degradation trend modeling,so that the prediction model is suitable for the nonlinear non-Gaussian system of the lithium-ion battery,and capacity of lithium-ion battery is long-term forecast.LSTM is used to perform feature learning on the existing capacity data and predict the future degradation trend to determine the failure threshold period to obtain the RUL.The experimental results prove the effectiveness of the model and provide a basic model for subsequent research.Secondly,in view of the unpredictable situation of battery cells under different working conditions,this paper proposes an LSTM model based on transfer learning,which transfers the model trained with source domain data to the target domain data to solve the problem of mapping under different working conditions.This makes the model applicable to a wider range,can predict the RUL under other working conditions,and reduces the design cost of the model.And try to predict the battery capacity degradation trend with fewer data,and realize the small sample size RUL prediction.Finally,in order to solve the problem of insufficient prediction accuracy and the inability of LSTM prediction to provide uncertainty expression,this paper proposes a model based on transfer learning fusion of LSTM and particle filter(PF).The PF is used to model the operating state of the battery itself.However,if the PF algorithm has no observations in the prediction stage,the state cannot be updated and the prediction accuracy is low.Therefore,this article uses the LSTM model based on transfer learning to predict the lithium-ion battery capacity for a long-term to obtain the predicted value.The predicted value is used as the observed value of the particle filter prediction model,and the predicted value is continuously adjusted and updated during the iterative process to obtain more accurate prediction results.,And provide an expression of uncertainty.