Remaining Useful Life Prediction Of Lithium-ion Battery Considering The Capacity Regeneration Phenomenon

In recent years,lithium-ion battery has been widely used in various fields due to its superior performance,but its own safety and reliability problems still exist,so it is necessary to conduct online monitoring and comprehensive management of the working status of lithium-ion battery in the application.As a key technology in battery management,safety guarantee and predictive maintenance,Remaining Useful Life(RUL)of lithium-ion battery has been attracting much attention,and has become one of the hot issues in battery system failure prediction and health management technology research.As a complex electrochemical system,the performance of lithium-ion battery will gradually degrade with the use of time.Among the many parameters that characterize the degradation state of the battery,the battery capacity can most directly reveal the battery’s life.Therefore,the study on capacity as a health indicator has been widely used in the prediction field of RUL of lithium-ion battery.In addition,the battery degradation process is not monotonous,the battery work process includes charging,discharging and rest stages.On the whole,with the continuous use of lithium-ion batteries,side reactions occur between the electrode and the electrolyte,making the battery capacity gradually show a downward trend.However,when the battery is in a static state after charging and discharging,the remaining reaction products will have a chance to dissipate,thus increasing the available capacity for the next cycle.This transient capacity recovery phenomenon is called capacity regeneration phenomenon,which is inevitable in the normal use of every battery.The existence of this capacity regeneration will change the trend of capacity degradation curve.Therefore,considering the capacity regeneration phenomenon in the prediction of lithium-ion battery RUL can improve the prediction accuracy of the prediction model,and it is very necessary in the practical application.In view of this,in order to improve the prediction accuracy of forecasting model and make the model more in line with the actual application,this article isbased on the internal reaction mechanism of lithium ion battery,through the study of the deep mining and analysis of capacity,for the lithium-ion battery RUL prediction method research of considering capacity regeneration phenomenon,in order to improve the prediction accuracy of lithium-ion battery RUL,and resolve the problem of the prediction accuracy greatly influenced by predicting the starting point,at the same time for the application of lithium-ion battery RUL prediction method study provides a new technical ideas.The main research work of this paper includes the following contents:(1)In view of the existence of different degrees of regeneration phenomenon in the shelving stage of lithium-ion battery capacity,this paper proposes EMD decomposition and wavelet decomposition respectively for multi-scale decomposition of lithium-ion battery capacity data,thus depth profiling capacity degradation characteristics of the process.It is found that the trend part which characterizes the normal decline of capacity and the oscillation part which characterizes the regeneration are distributed at different scales.(2)The degradation process of lithium-ion batteries is complicated,so it is usually difficult to build an accurate mechanism model to predict RUL.Therefore,the prediction of lithium-ion battery RUL based on EMD decomposition combined with Elman neural network was proposed to capture the capacity regeneration phenomenon through decomposition and partial modeling,and to improve the prediction accuracy of lithium-ion battery RUL.(3)Adjacent cycle between the lithium-ion battery degradation state highly correlated,general neural Network model is hard to learn with time dependent degradation characteristics,therefore put forward based on a Nonlinear autoregressive time series forecasting ability of neural Network(Nonlinear Auto Regressive Network,NAR)lithium-ion battery RUL prediction method,combined with the data after wavelet decomposition model,improved the prediction to the late,the deflection of a more stable and accurate prediction of RUL.(4)Aiming at the two kinds of neural network based lithium-ion battery RUL prediction method proposed in this paper,a lithium-ion battery RUL prediction system is designed and implemented by MATLAB programming software.The system mainly consists of two modules: prediction based on Elman neural network and prediction based on NAR neuralnetwork,and the function of prediction at different prediction starting points is set respectively.The system applies the theory to practice and has certain practical application value.