The Fusion Capacity Degradation Estimation And Prediction For Lithium-ion Power Battery

Because the automobile vehicles consume a lot of petroleum resources and produce harmful exhaust gas which results in environmental pollution,the popularization of new energy vehicles is imperative.New energy vehicles mainly include battery electric vehicle(BEV),plug-in hybrid electric vehicle(PHEV),fuel cell electric vehicle(FCEV),etc.At present,the new energy vehicles sold in the market are mainly BEV and PHEV,hereinafter referred to as "Electric Vehicles".Lithium-ion power battery is the key chemical energy storage device of electric vehicles.But the battery cell life will gradually degrade in the long-term use,which seriously affects the driving range and power output of electric vehicles.How to estimate the current battery capacity and predict lifetime closed-loop based on online data becomes the difficulty and focus of the vehicle battery management system(BMS).In this paper,the fusion capacity degradation estimation and prediction for lithium-ion power battery is studied,which is mainly from four aspects: the experimental design of battery durability life,the online battery capacity estimation based on data-driven method,the capacity prediction based on the discrete Arrhenius aging model,and the fusion capacity estimation based on the double extended Kalman filters.The fusion technique,which can receive high accuracy and good reliability,could provide technical support for the state of health(SOH)estimation and lifespan prediction.First of all,the battery durability lifetime tests are designed.The realistic vehicle working conditions are complex and unpredictable,so the designed durability experiments should be able to simulate the realistic battery lifespan degradation to the greatest extent,and at the same time,considering the limited postgraduate learning time,it must be able to accelerate the battery aging process.Finally,the three manufacturers#1,#2,and #3 batteries are selected to carry out the different-constant-temperature experiment,the alternating-temperature experiment 1 and the alternating-temperature experiment 2,which provide different aging data and path for the subsequent algorithm to verify the applicability of the fusion algorithm.Secondly,the first capacity estimation technique based on data-driven method is proposed,which is the online capacity identification method based on charging curve sections.By zooming and translating the constant-current charging voltage curves in the different aging stages,and taking the charging curve of the fresh battery as the benchmark,the difference between the time when the zooming and translating charging curves reach each voltage point and the time when the initial charging curve reaches the same voltage point can be obtained.Two voltage characteristic points are identified by the genetic algorithm,which together constitute the fixed voltage window(FVW).After that,the time interval of each aging charging curve passing through FVW is intercepted,and based on the initial capacity and initial time interval,the current battery capacity will be identified.The experimental results show that the method can follow and approximate the real capacity values with the maximum estimation error less than 4%.Then,the second capacity estimation technique based on data-driven method--the joint capacity estimation method based on the third-order extended Kalman filter(EKF)algorithm is proposed.Based on the first-order resistance-capacitance(RC)equivalent circuit model,the method of model parameters identification between zones is adopted.Combined with the optimization algorithm,the model parameters needed in each zone are achieved,which will be for the whole battery lifespan.Afterwards,the third-order EKF algorithm is constructed,which takes capacity as a state space variable.Because it utilizes the characteristics of EKF algorithm’s closed-loop feedback correction,the inaccurate initial capacity value can be gradually corrected to the real capacity value,and the average value of the last section of capacity estimation values is taken as the current capacity estimation value.The experimental results show that the method can also approximate the real value,and the maximum estimation error is less than 4%,with high accuracy and good convergence.Furthermore,a discrete Arrhenius aging model is proposed based on the Arrhenius model.As a kind of empirical model,Arrhenius model can be used to predict the capacity aging trajectory in the next cycle with the updated model parameters.In order to adapt the model to alternating-temperature conditions,the Arrhenius model is transformed into a discrete Arrhenius aging model.Then,the applicability and accuracy of the proposed discrete Arrhenius aging model is verified by the real aging data of the manufacturers #2 and #3 batteries.The experimental results show that this method is suitable for the capacity degradation characteristic for these batteries.Finally,the fusion capacity estimation algorithm of the double extended Kalman filters is designed.The current battery capacity is estimated based on data-driven method,which will be triggered only when the specific conditions or working conditions are met.At the same time,the proposed discrete Arrhenius aging model estimates the battery capacity,which needs a lot of capacity aging data to fit the model parameters,and usually the model parameters fitted in the laboratory are quite different from the actual model parameters due to the complex and unpredictable working conditions in the real vehicle applications and the capacity inconsistency aging between the batteries.Therefore,the two technical routes have their own advantages and disadvantages.In this paper,the double extended Kalman filters algorithm is designed to give full play to their advantages and avoid their disadvantages.One Kalman filter is used to update the model parameters and predict the capacity degradation of the next cycle to achieve continuous prediction;the other Kalman filter is used to fuse the estimated value based on data-driven method and the value of the discrete Arrhenius aging model to achieve a fusion capacity estimation value,and as the start capacity for the next cycle simultaneously.The experimental results show that the fusion algorithm can effectively correct the inaccurate model parameters,and the fusion capacity estimation and prediction errors are within 4%,which indicates its high estimation accuracy and good reliability.