Research On Lithium-ion Battery Rul Model Driven Prognosis Method

Lithium-ion batteries are widely used in various fields as energy-providing equipment because of their high energy density,long service life,and no pollution to the environment.The most mature fields include aerospace systems,electric vehicles,and military communication equipment.In recent years,various countries have strongly supported the research work of lithium-ion batteries,the most important part of which is the battery management system.The battery management system collects relevant data inside the battery,calculates battery health indicators,remaining life,etc.,to guide relevant personnel to better use the equipment to prevent the occurrence of dangerous events.Since the battery belongs to a nonlinear system,the internal chemical reaction principle is complicated,which makes the prediction of the remaining life of the battery low.The relevant personnel have carried out a lot of research on this,and proposed various methods to improve the prediction accuracy of the remaining life.Most of the current predictions use a single model,such as polynomial model,exponential model,Verhulst model,which often cannot fully describe the capacity degradation process.This paper proposes to integrate multiple models through integrated learning to predict the remaining life of the battery and improve the accuracy of remaining battery life prediction.The main research contents of this paper are as follows:Firstly,for the current common capacity degradation model,because the capacity degradation process cannot be accurately described,this paper looks for features with high capacity correlation characteristics for capacity prediction.In the feature selection,the Pearson coefficient,the maximum information coefficient(MIC),the distance correlation coefficient and other correlation coefficients are used for feature selection,and the feature importance ordering is performed by using the recursive elimination feature method and the Las Vegas Wrapper method.Secondly,since the health factor cannot be obtained after the predicted starting point,the correlation vector machine(RVM)and the neural network algorithm are used in the prediction of the health factor.Then,based on the predicted health factors,linear regression is used to estimate the parameters of the capacity degradation model,and finally,battery capacity prediction and battery remaining life prediction are performed.Thirdly,the current capacity degradation model is studied,and multiple capacity degradation models are used to improve prediction accuracy using the idea of integrated learning.In this paper,we use random forest algorithm and eXtreme Gradient Boosting(XGBoost)integrated learning method to carry out model fusion,through parameter tuning,and finally through the experiment to compare the two methods.