Lithium Iron Phosphate Battery Modeling And Health State Estimation Study

New energy vehicle technology has become one of the main development direction in future.As the power battery manufacturing technology and battery management technology of mature gradually,the importance of the battery health status evaluation is more and more prominent.In order to protect the driver's driving safety,ensure the electric vehicle to run efficiently and safely,we need to accurately estimate the battery SOH,replace the old batteries in time,providing real-time accurate battery health information to the driver's.Based on this,this paper to estimate the battery state of health problem do the following several aspects work:(1)Design the standard of the HPPC(Hybrid Pulse Power Characteristic,HPPC)experiment,and studied the attenuation mechanism of lithium-ion batteries.Compared with other lithium batteries,lithium iron phosphate battery there is a voltage flattens"plateau",which creates a battery SOH estimation is difficult;Based on this,then on the basis of experiment,study the different magnification and different temperature of the nature of the lithium iron phosphate batteries;Finally,based on the working principle of lithium-ion batteries and the experimental data,analysis the main factors influencing the lithium battery attenuation is:the lithium battery in use process,the change of the battery voltage,temperature change,current change and the change of SOC,etc.(2)Several kinds of lithium battery model and several common battery parameter identification method were studied,meanwhile,the electrochemical model and the merits and demerits of the equivalent circuit model were analyzed.The second-order RC equivalent circuit model can not only reflect the battery activation in the course of chemical reaction,polarization and other dynamic characteristics,also can reflect the cell concentration in the process of static diffusion effect,compared to the actual battery external characteristic,approximation degree is higher,but a moderate amount of calculation,so put the second-order RC model as the research of lithium ion model.In HPPC,on the basis of experimental data,respectively with genetic algorithm and least square method and least square method with forgetting factor for the established lithium battery second-order equivalent circuit model for parameter identification,and comparative analysis between the three error and error change trend,the conclusion shows that with forgetting factor recursive least squares identification result accuracy is much better than the front,both in the whole discharge cycle experiment the RMSE is far less than that in the process of GA and the LS.(3)The effect of BP neural network algorithm and support vector machine(SVM)algorithm to estimate the battery SOH were studied.Based on the battery cycle test data,we selection of discharge time T,E and voltage integral cycles as input characteristic variables,with battery SOH as output.First of all,the battery SOH was predicted by BP neural network,the maximum error of the prediction can remain below 0.7%,can effectively track the change of battery SOH;Then,based on support vector machine(SVM)algorithm battery SOH estimation method is studied,in order to achieve more accurate estimates of the effect,we must looking for the penalty factor C and kernel function algorithm optimization coefficient of g,Then,using the K-CV method,genetic algorithm and particle swarm algorithm to find the optimal parameters of C and g,comparing the BP neural network algorithm and parameter optimization of the SVM algorithm is concluded that the prediction error,the results show that the SVM algorithm to predict the maximum error of under 0.25%,obviously superior to the BP neural network algorithm;Finally,study the only selection of discharge voltage integral time T,E as input variables,battery SOH as output,using the optimized SVM algorithm to predict the maximum error of drop to below 0.065%,which proves the importance of selecting input feature.Taking lithium iron phosphate battery SOH estimation as research subject,found the battery SOH estimation method based on support vector machine(SVM)can accurately predict the cycles of the rest of the battery.Battery SOH estimation method based on support vector machine(sSVM)has the advantages of good reliability,high precision,good real-time,solve the problem of estimate electric vehicle power battery SOH is difficult in engineering application.