Research On SOC Estimation Algorithm For Lithium Ion Batteries Of Electric Vehicles

In order to maximize the performance of lithium batteries,the state of charge(SOC)is used as an momentous norm to judge the property of the battery.Accurate SOC not only represents the cruising range,but also an important basic parameter for thermal management technology and balanced control.The use of limited measurable parameters to achieve accurate estimation of battery SOC has always been the core problem and technical difficulty of battery management systems.The estimation accuracy of SOC mainly depends on the accuracy of the lithium battery model and the SOC estimation algorithm.This article cuts in from these two points,establishes the optimal variable-order model,designs the SOC estimation algorithm based on multi-state division,and uses the lithium battery test and test system The accuracy of the lithium battery model and the accuracy of the SOC estimation algorithm are verified.An optimal variable order model is established.The model uses the Bayes information of the terminal voltage error as the basis for the order change of the Thevenin model and the DP model.The modified humpback whale optimization algorithm is used to hunt the optimum solution in whole situation of Bayes information sequence.Realize the solution of the optimal variable order sequence of the model.Dynamical capacity experiment,open-circuit voltage experiment,hybrid power pulse experiment,pulse discharge experiment and EPA city dynamic experiment were carried out.Complete dynamic capacity modeling and joint based on the above electrical characteristic test Multi-model SOC-OCV curve calibration and model impedance arguments recognition.The model of battery was discretized and the arguments of the model were recognited online by the recursive least square method with forgetting factor.According to the model terminal voltage innovation The working state is divided into steady state,failure state and transition state.An extended Kalman filter based on the improvement of the beetle search algorithm is designed,and the filter is used to estimate the SOC in the steady state;in the failure state,the extreme learning machine is used to estimate Lithium battery SOC;linear fusion algorithm is used to estimate lithium battery SOC in the transition state;the hardware-in-the-loop simulation system and the lithium battery test system are combined to achieve rapid control prototype verification for the accuracy of the SOC estimation algorithm.The test results show that the established lithium battery optimal variable-order model can improve the terminal voltage prediction accuracy,eliminate redundant items in the model and prevent the occurrence of model over-fitting.The average error of the optimal variable-order model is maintained at about 0.016V under pulse discharge conditions,which is 18.3%and 4.1%lower than the errors of the traditional Thevenin model and DP model,respectively;the error remains at 0.01 under EPA urban dynamic conditions About V,it decreased by 40.3%and 15.7%respectively.The designed method for segregating the working state of lithium batteries and the SOC calculation algorithm based on multi-state segregation can prevent the effect of model lose efficacy on the precision of SOC estimation.The precision of SOC calculation is immensely boosted.The average error of the SOC estimation algorithm based on multi-state division under pulse discharge conditions is maintained at about 1.0%,which is reduced by 25.5%and 6.3%respectively compared with a single steady state or a failure state;the error remains at about EPA urban power conditions About 1.0%,compared with a single stable state or a failure state,it is reduced by 1.9%and 12.2%,respectively.The experimental consequence reveal that the model established and the SOC calculation algorithm designed in this thesis have lofty precision and can achieve the practical project demandment.