Research On Remaining Useful Life Prediction Of Lithium-ion Battery And Screening Method Of Recycling Battery

In recent years,energy conservation,emission reduction and environment protection have attracted great concern of international community.Electric vehicles(EV)can reduce fuel consumption and exhaust emissions,it is an important development direction in the automobile industry.As we know that lithium-ion(Li-ion)power battery is one of the key components of EVs,and the remaining useful life(RUL)prediction of power battery is crucial for the reliability of EVs.Up to now,most of the RUL prediction approaches are based on battery capacity.The online prediction of battery RUL is difficult due to the difficulty of online capacity measurement.Meanwhile,with the increase of decommissioned batteries,the reuse and recycling of decommissioned batteries become urgent.Screening is one of the key technologies for the echelon use of battery.The existing screening methods are not so suitable for the screening of decommissioned batteries in large quantities due to their complex detection processes.In this thesis,we propose a novel RUL prediction method for Li-ion power battery,which is based on the battery operating data.Furthermore,a screening method for decommissioned batteries is also put forward.The major works of this thesis are as follows:(1)An online capacity estimation method for Li-ion batteries is proposed based on battery operation data.Considering that the state variables(including current,voltage and temperature)are easily monitored during battery discharge process,four health indicators(HI)of Li-ion battery capacity are extracted accordingly,i.e.battery calculated resistance(CR),temperature change rate(TR),duration time of equal discharging voltage difference(DEDVD)and sample entropy of discharge voltage(Samp En).After that,the generalized regression neural network(GRNN)is used to establish the mapping relationship between the HIs and battery capacity so as to achieve the estimation of battery capacity.The feasibility and validity of the proposed method are verified with a practical case study.(2)Based on the capacity estimation research,a RUL prediction method is proposed.The capacity degradation and the RUL of batteries are predicted with non-linear auto-regressive(NAR)dynamic neural network and auto-regressive integrated moving average(ARIMA)method respectively.Case study shows that the prediction results of NAR model are not stable;corresponding to this,the prediction results of ARIMA model are more stable.(3)On the basis of battery historical operation data,a decommissioned battery screening method is proposed.In this method,the performance of batteries is analyzed based on their operation data.After that,the decommissioned batteries are sorted with the approaches of analytic hierarchy process(AHP),correlation degree theory and affinity propagation clustering theory.The proposed method can simplify the detection processes of decommissioned batteries,improve the screening efficiency and reduce the cost of battery echelon use effectively.The main contributions of this thesis include:(1)An effective online battery capacity estimation method and an online RUL prediction method are proposed;(2)By combining battery screening method with the battery capacity estimation and RUL prediction methods,the screening efficiency for decommissioned batteries is improved greatly;(3)On the basis of battery life cycle data recording,the research of this thesis covers the entire life of Li-ion power battery;(4)Both the proposed RUL prediction method and battery screening method are based on the battery historical data,the physical and chemical reactions in batteries are negeleted.Therefore,the proposed methods are also suitable for some other types of batteries.