Research On State Of Health And Life Prediction Model Of Vehicle Power Batteries

In recent years,in order to reduce vehicle exhaust emissions and ease energy crisis,new energy vehicles,especially electric vehicles（EVs）,have rapid development.As EVs energy storage device,lithium-ion batteries play a significant role.With the improvement of the performance and driving range of electric vehicles,the power and capacity of lithium-ion batteries are increasing.Their safety and reliability are increasingly important.During the cycle of lithium-ion battery,its state of health（SOH）will gradually degrade,and its remaining useful life（RUL）will gradually shorten.When the battery reaches the end of life condition,if it is not replaced in time,it will seriously affect the performance and safety of the EV.Therefore,it is really important to accurately monitor the battery state of health and predict the battery remaining useful life.In order to obtain the accurate internal aging parameters of the battery,it is necessary to be removed from electric vehicles.This kind of method is expensive and difficult to be applied to the power battery on the real vehicle.Therefore,in this paper,the performance parameters that can represent battery aging and can be calculated are used for research.Firstly,the characteristics of battery aging performance parameters are studied under laboratory conditions.Then,data mining is carried out on the running data.Due to the major gap between the running data and laboratory data,the data is preprocessed first.Combined with the features of the running data,a battery model and a capacity calculation model are established,which are used to calculate the internal resistance and capacity of the battery,respectively.Finally,the capacity health indicator（HI）is selected to establish the SOH degeneration model.In this paper,the long short term memory（LSTM）is used to learn the SOH degeneration law,and the battery life prediction model is established to predict the battery RUL.The main research contents of this paper are as follows:（1）The basic performance of lithium iron phosphate battery cell is experimentally studied under laboratory conditions.Then,the features of different performance parameters are analyzed.Firstly,the rated capacity of the battery is obtained according to the static capacity calibration experiment,and the effects of discharging current and air temperature on the battery capacity are analyzed.Secondly,the internal resistance of the battery is obtained by using the hybrid pulse power characteristic test（HPPC）.The change of the internal resistance with the state of charge（SOC）of the battery during the charging and discharging process is analyzed.Finally,the aging mechanism and aging model of the battery are introduced and analyzed,and the external characteristic aging model is determined to be used for research.（2）A battery equivalent circuit model based on running data is established by data mining and improved recursive least square（RLS）.Firstly,the running data of No.361 pure electric buses in Changchun are analyzed,extracted and preprocessed.Then,the first-order RC equivalent circuit model and RLS are selected to identify the parameters of power battery.Since the power battery cannot obtain the accurate SOC-OCV（Open Circuit Voltage,OCV）curve in the actual operation process,this paper improves the RLS algorithm,obtains the RLS-OCV dynamic identification method,and uses it to calculate the battery ohmic internal resistance.Finally,the battery state of health is evaluated based on the ohmic internal resistance.（3）The mined battery charging and discharging data are studied,and the battery capacity calculation model based on running data is established.Firstly,the charging and discharging data meeting the set conditions are selected,the current integration method and variable sliding window method are used to calculate the battery capacity,and the box diagram method is used to eliminate outliers.Then,the effects of air temperature and current on capacity are considered.By analyzing the charging data,it is found that the battery adopts constant current charging mode,and the charging current is 100 A or 200 A.Therefore,this paper mainly considers the influence of air temperature on battery capacity,and makes a temperature correction to the battery capacity.Finally,the battery state of health is evaluated based on capacity.（4）In this paper,the SOH degeneration model is established first,and then the battery aging model is established combined with the LSTM neural network.Firstly,the characteristics of three health indicators（Capacity,Internal Resistance and Average Charge Voltage）are analyzed,and the capacity is selected to establish SOH degeneration model.The model considers two main influencing factors: current and temperature.Then,the battery discharging current is statistically analyzed,and five different working conditions were established to verify the accuracy of the model according to the difference in air temperature and discharging current.Finally,combined with the SOH calculated by SOH degeneration model and LSTM neural network,the battery aging model is established.The model predicts the change of battery capacity in the next four months.The results show that the model has high prediction accuracy.The proposed state of health estimation method and aging model of electric vehicles based on running data can avoid offline battery experiments.The proposed methods significantly reduce costs of power battery research.At the same time,it can effectively consider the complex working conditions in the actual environment to estimate the state of health of power batteries online.