Equivalent Circuit Model Based Parameter Identification And State Estimation For Lithium-ion Batteries

In recent years,lithium-ion batteries have been widely used in the energy storage systems of electric vehicles(EVs)because of their low self-discharge rate,high energy and power densities.Nevertheless,despite the above advantages,the basic conditions of battery packs,including voltage,current and temperature,need to be continuously monitored by the battery management system(BMS)to ensure the safe and reliable operation of the battery system.Moreover,some essential parameters and state variables,such as open-circuit voltage(OCV),impedance parameters,state-of-charge(So C),and state-of-health(So H),also need to be accurately estimated by the BMS.Owing to the strong time-variable and nonlinear electrochemical reactions,physically immeasurable of state variables and disadvantages of the conventional estimation methods,the model-based battery management technique is attracting increasing attentions from the academia and the industry.To realize the optimal management of the lithium-ion battery in the EV application,the following work has been carried out:For the acquisition of battery OCV,an improved OCV characterization test is proposed based on the active polarization voltage reduction method.Compared with the conventional incremental OCV characterization test,the extra current excitation is imposed immediately after the incremental discharge/charge period,and thus the fast convergence of the battery terminal voltage is achieved,which in turn shortens the test time.Firstly,the battery's relaxation behavior is broadly divided into three regions(i.e.,fast-,middle-and long-frequency regions)according to the order of the magnitude of relevant time constants,and the third-order ECM is employed to describe this behavior.Secondly,based on the characteristic of the RC network,two couples of current pulses are proposed to actively minimize overvoltages across the long-and middle-term RC networks,which correspond to the polarization effects dominating the majority of the relaxation period.Furthermore,the parametric sensitivity of the imposed current excitation to battery model parameters is analyzed,and the parametric determination method for the imposed current excitation is provided subsequently.At last,a lithium-ion polymer battery is adopted under test to validate the proposed test procedure.Experimental results proved the feasibility of the proposed test procedure,and demonstrated advantages over the conventional one in terms of the convergence speed.For the identification of the battery impedance parameters,an improved offline battery parameter identification method is proposed based on the typical urban driving scenarios in EVs.Instead of employing the measured data from the whole rest period in the conventional offline parameter identification methods,only the prior portion of the collected data is selected,and the length of the fitted data is determined by the amplitude distribution and frequency spectrum analyses of the load current under the typical urban driving condition.In addition,the variation trend of the RC network voltage during the pulse discharge period is analyzed.According to the unsaturated phenomenon caused by the long-term RC network,a fitting equation with an improved expression of the initial RC network voltage is obtained.Finally,simulation and experimental results validated the feasibility and the advantage of the developed parameter identification method.For the estimation of the battery So C,a split battery model based adaptive extended Kalman filtering So C estimation method is proposed.The proposed split battery model is achieved by dividing the conventional augmented battery model into two parts:the RC voltage submodel and the So C submodel.The model partition helps reduce the cross interference between the RC voltages and the So C.Thus,the oscillation in the states estimation is degraded at the start-up stage,and the estimation accuracy is improved during the stable stage.Two types of lithium-ion batteries,including the Li Fe PO₄ battery and the Li NMC battery,are employed under test,and a case of a 2^nd-order ECM is analyzed.Comparative results show that the proposed method demonstrates a higher stability during the start-up stage and constrains the So C error within 1%during the stable stage,regardless of the incorrect initial So C and additional sensor noise.Moreover,the average computational time of the proposed method is far less than 1s,which verifies its feasibility in on-board applications.For the estimation of the battery aging state,especially the capacity fading,an online So H estimation method is proposed based on the dynamic characteristic of the charging current during the constant-voltage(CV)period.According to the preliminary analysis of the battery test data,the time constant of CV charging current is considered as a characteristic parameter related to the battery capacity fading.The detailed expression of the current time constant is derived based on the ECM.The quantitative correlation between the normalized battery capacity and the current time constant is established to indicate the battery So H.Specifically,for the uncompleted CV charging process,the reference correlation curve with respect to the specific data length is employed to estimate the battery So H.Besides,the logarithmic function-based prediction model is identified from the partial CV charging data to predict the reference time constant.At last,four Li Fe PO₄ batteries are employed under test to verify the feasibility of the proposed method.The results demonstrate that there exists a strong linear regression between the normalized battery capacity and the current time constant,and the relevant correlation coefficient can reach-0.9880.Moreover,the correlation function extracted from one battery is able to evaluate the So H of other three batteries with less than 2.5%absolute error except a few outliers,regardless of the adopted data size.