Research On Degradation Modelling And Residual Life Prediction Of Lithium-ion Battery

As an important kind of energy storage device,lithium-ion batteries have many desirable characteristics,such as high energy density,low self-discharge rate,no memory effect,long cycle life and low pollution,etc.Due to these excellent characteristics,lithium-ion batteries have been widely used both in civil and military fields.Similar to other performance-degraded products,the performance of lithium-ion batteries decreases gradually with the number of charging/discharging cycles increasing.For example,after being put into use,the usable capacity of a lithium-ion battery will degrade over time.When its performance deteriorates to a certain level,lithium-ion batteries will not be able to satisfy the power consumption requirements of loadings.From the view of utilization efficiency and safety,it is necessary to prevent the occurrence of premature replacement and overuse of lithium-ion batteries.To reach this end,the prognostics and health management of lithium-ion batteries should be properly arranged,and one importance thing is to predict the residual life at current time.In engineeringpractice,the usable capacity is usually adopted to characterize the healthy performance of a lithium-ion battery.In general,a lithium-ion battery fails when its usable capacitydeclines to 70%-80% of the rated capacity.Intuitively,referring to theories and methodsof residual life prediction based on performance degradation data,we can predict the residual life of a lithium-ion battery by properly modeling the degradation process of the usable capacity.However,unlike other health indicators,the usable capacitycannot be measured directly.In the laboratory,the ampere-hour integration method can be used to estimate the usable capacity since requirements of full charging and discharging can be satisfied by controlling laboratoryconditions.However,in operating fields,due to the restriction of powerconsumption requirements and uncontrollable environment conditions,there exists partial charging,incompletedischarging or excessive discharging and so on.In such a situation,it is imposable to estimate the usablecapacity by using laboratory-oriented methods,such as the ampere-hour integration method.Therefore,it is necessary to explore available estimation methods of usable capacity in operating fields.In addition,for the purpose of capturing the performance degradation process of lithium-ion batteries effectively,the environmental stress accelerating effects should be taken into account if the operation stress profile is unsteady.Moreover,in some special situations,the issue of degradation modeling and residual life prediction for lithium-ion batteries based on multi-performance characteristics should be investigated,since competitivedegradation exists and each degradation process can lead to a failure.In this paper,taking the above-mentioned issues as entry points,the performance degradation modeling and residual life prediction of lithium-ion batteries are studied.The main contributions are summarized as follows:(1)Methods of residual life prediction of lithium-ion batteries based on time interval characteristics.Since the usable capacity is difficult to be estimated by laboratory-oriented methods in operation fields,an available usable capacity estimation model is presented,which utilizes the time intervals in the charging phase as characteristic variables.Through correlation analysis,the time interval of equal charging voltage difference is used to establish the usable capacity prediction model.The prediction accuracy is affected by the values of voltage drop and cut-off point in the primary function vector.For the purpose of avoiding subjectivity in determining these two kinds of ranges,an optimal feature extraction method is proposed with the objective of minimizing the model error.Then,the exponential model is used to describe the degradation process of the usable capacity over time.The issue of residual life prediction of lithium-ion batteries is realizedunder the framework of the particle filter method,in which values gained by the usable capacity estimation model are regarded as online observations.(2)Methods of residual life prediction of lithium-ion batteries based on the tail end characteristics of the charging voltagecurve and the charging current curve.With the deterioration of lithium-ion batteries,both the charging voltagecurve and the charging current curvechange accordingly,especially the tail end of each curve presents obvious regularity.As a matter of fact,the regular change is related to the deterioration of usable capacity.For this reason,the tail end characteristics of voltage and current curves in charging are modeled,and parameters of each model are regarded as characteristic variables,based on which an usable capacity estimation model is presented through regression analysis.Then,the Wiener process with a time-scale transformationis adopted to capture the degradation of the usable capacity loss rate over time.After that,the online estimated degradation data is used to update the proposed degradation model,and the residual life is predicted accordingly.(3)Methods of residual life prediction of lithium-ion batteries considering theaccelerating effect of discharge rate.As an environmental stress,higher discharge rate can accelerate the performance degradation process of lithium-ion batteries.Therefore,it is necessary to consider theaccelerating effect of discharge rate in degradation modeling.For lithium-ion batteries with linear degradation trend of usable capacity,the Wiener process with a linear drift is chosen to model the changing process of the usable capacity loss rate over time.To model the accelerating effectof discharge rate,an accelerated function,which is determined from a statistical point of view,is used to describe the effect of discharge rate on the drift coefficientof the proposed degradation model.Then,the drift coefficient is regarded as a hidden state,and the kalman filter method is used to update the degradation model based on the online usable capacity degradation data.After that,the residual life of lithium-ion batteries is predicted according to the updated degradation model.(4)Methods of degradation modeling and residual life prediction of lithium-ion batteries with two competitive performance characteristics based on Gumbel Copula function.Both the usable capacity and the usable energy are used to characterize the performance of lithium-ion batteries.An bivariate Wiener process is used to track the evolution of two performance characteristics,in which the degradation process of each performance characteristic is captured by the Wiener process with a linear drift and the dependency between them is modeled by the Gumbel Copula function.Then,a state space model referring to the drift coefficient and the correlation parameter of the Gumbel Copula function is given,based on which the degradation model is updated through the particle filter method by using the online degradation data.After that,the residual life of lithium-ion batteries is predicted according to the updated degradation model.Since the parameter of the Gumbel Copula function is updatediteratively,the proposed degenerated modelhas the ability of describing the dynamic dependency.By doing this,the generalization ability of the proposed degenerated model is improved.