| New energy vehicle industry has been one of the emerging sectors of strategic importance in China.Although the government has published a package of green-car subsidy program to boost electric vehicle(EV)industry,the development of EV market would mostly depend on the customer acceptance.Among all consumer concerns,price,driving range,battery life and safety are and continues to be the top ones.One way to solve these problems is to design an effective battery management system(BMS).Estimation of state of charge(SOC)is the key function in BMS.It is not only related to the capacity utilization rate,battery aging and safety,but also severs as the urgent prerequisite for implementing other facilities such as battery state monitoring and balance control.However,being an internal state,the accurate and reliable SOC estimation is always difficult especially in the complex road condition.To solve this problem,we propose a novel SOC estimation solution for BMS in EVs,which is based on an accurate and simple battery model.The thesis main research results are as follows:(1)Develop a systematical solution framework for the best equivalent circuit model construction.The model structure in equivalent circuit model(ECM)is always empirical determined,which cannot guarantee to obtain a compact and simple model.In this study,a systematical solution framework for simultaneous functional form selection and parameter estimation is proposed.A bi-objective mixed-integer nonlinear programming(MINLP)model is first constructed to balance model accuracy and model complexity.To obtain global solution of such difficult model,two solution approaches,namely the explicit and implicit methods,are then developed to transfer this bi-objective MINLP model to a single objective MINLP model,respetively.The former explicitly treats the model complexity as a constraint and the latter implicitly embeds the model complexity into the objective as a penalty.Both approaches require sequential solution of the transformed MINLP model and an ideal and nadir ideal solutions based criterion is utilized to terminate the solution procedure for determining the optimal functional forms.The effectiveness of both approaches are thoroughly evaluated through pulse current discharge test(PCD)and hybrid pulse power characterization(HPPC)test of a commercial LIB.The fitting RMSEs of battery voltage are less than 7mV and the prediction RMSEs are less than 15 mV.These results illustrate that the proposed method can effectively construct an optimal ECM with minimum complexity and prescribed precision requirement.It is thus indicated that the proposed MINLP based solution framework would be greatly helpful to SOC estimation.(2)Develop a moving horizon estimation approach for SOC estimationExtended Kalman Filter(EKF)based method is most widely used one in SOC estimation.However,being a filter method,it has some intrinsic drawbacks such as slow convergence rate,high sensitivity to poor initial guess and lack of constraint handling mechanism.In this study,a novel SOC estimation method based on the moving horizon optimization is proposed.On the basis of accurate battery model developed in part one,a nonlinear state-space model for moving horizon estimation(MHE)is first constructed.Then,the effect of tuning prameters including horizon length,iteration number and covariance is studied,and an efficient interior-point method for solving MHE problem is developed.Three typical battery tests including cosnstant discharge test,HPPC test and Dynamic Stress Test(DST)are used to evaluate the effectiveness of proposed method.It is found that,no matter the initial SOC guess is extreme high or extreme low,MHE can capture the ture value in very short time under all tests.The SOC estimation errors keep low and stable during the full range and the RMSEs are less than 2% in all conditons.Compared to the EKF,the MHE can perform more accurate,reliable and robust SOC estimation of LIBs.(3)Develop a joint moving horizon estimation approach for SOC estimationThe model mismatch issues,originating from battery inconsistency,battery dynamic characteristics difference and battery aging are widely existed and would significantly deteriorate the performance of model based SOC estimation method.To solve this problem,a joint moving horizon estimation(joint-MHE)approach which is based on augmented nonlinear state-space model is proposed here.By sensitivity analysis of model parameter in ECM,the additional states in the augmented nonlinear state-space model are properly selected,and then are jointly estimated in company with SOC using the interior-point method.The effectiveness of joint-MHE is evaluated under three typical battery tests including cosnstant discharge test,HPPC test and DST.The results demonstrate that,in the face of model mismatch and poor initial SOC guess,the proposed method can both correct the model error and track the true SOC in short time.In the condition of HPPC test with 40% model parameter error,the SOC estimation RMSE of joint-MHE is still less than 2%.Compared to the joint-EKF,the joint-MHE could provide a more reliable,robust and accurate SOC estimation of LIBs under various model mismatch conditions.On basis of these works,a systematical SOC estimation framework is developed,which includs offline model constriction,update parameter selection,algorithm parameter tuning and SOC estimation realization.The results demonstrate that,even in the face of poor initial guess and serious battery inconsistency,the proposed strategy can provide the accurate and reliable SOC estimation under different working condition during the full range.This SOC estimation strategy can be easily implemented in commercial BMS as a form of software.It is reasonable that,the proposed framework would greatly benefits the effectivness and relability of BMS,and eventually help to realize low price,long driving range and high safety of EVs.As a general stratagy,the proposed method also could be used in other LIBs applcations such as portable sets,energy storage station and so on. |
PDF Full Text Request