Remaining Useful Life Predection Of Lithium Batteries Based On Long-range Dependence Model

The internal working principle of lithium-ion batteries(referred to as lithium batteries)is very complex,and because of the continuous chemical reaction of charge and discharge,its actual performance is constantly degraded.It is found that the data series of the degradation process of the lithium batteries have the non-Markov property and long-range dependence(LRD),which provide a new method for modeling the degradation remaining useful life(RUL)of lithium batteries.In this paper,a lithium batteries RUL prediction method based on Fractional Brownian Motion(FBM)model with long-range dependence is proposed.The main work is as follows:1.The Hurst exponent(H)is calculated according to the degradation process data of the actual lithium batteries,so as to judge the long-range dependence of the degradation process.According to the definition,the expression of stochastic differential equation of FBM model is derived and the drift parameter ? and diffusion parameter ? in FBM model are estimated by the maximum likelihood estimation method and Hurst exponent.Thus the FBM prediction model is established.2.The working principle and remaining useful life prediction analysis of lithium batteries are described,and the main internal and external causes of life degradation are given.Because lithium batteries usually work in series and parallel groups,the failure of independent lithium battery in the whole system will directly affect the life of other lithium batteries,so the RUL prediction of lithium batteries is usually to predict the RUL of each independent lithium battery.Then,the related concepts of RUL prediction of lithium batteries are introduced,and the capacity of lithium batteries is selected as the index to evaluate whether the life of lithium batteries reach the failure threshold.3.The prediction model of discrete FBM stochastic differential equation is verified and the parameters are optimized.The validity of FBM model in predicting RUL of lithium batteries is verified by comparing with Monte Carlo simulation.Because the two key drift parameters ? and diffusion parameter ? of FBM model depend on the calculated value of Hurst exponent.Therefore,this paper uses a Fruit-fly Optimization Algorithm(FOA)to optimize the Hurst exponent,so as to improve the estimation accuracy of drift parameter ? and diffusion parameter ?.4.Several groups of degradation data of lithium batteries are selected from NASA Ames Prognostics Data Repository as experimental research objects to carry out the RUL prediction experiment of lithium batteryies.The FBM prediction experiment was carried out with B0018 groups of lithium battery degradation data,and the FBM model is compared with the FBM model combined with FOA with B0005 and B0006 groups of lithium batteries degradation data.Finally,the validity of the FBM model combined with FOA algorithm is proved by the predicted Probability Density Function(PDF)simulation diagram and the prediction error indices(Mean Square Error and Root Mean Square Error).