Estimation Of Battery Pack State For Electric Vehicles Using Model-data Fusion Approach

Battery is the bottleneck technology of electric vehicles (EVs), it is meaningful both intheory and practical application to implement the research on the state estimation ofbatteries, which is very crucial to optimize the energy management, extend the cycling life,reduce the cost and safeguard the application of batteries in EVs. However, the batteries,with a strong time-variable, nonlinear characteristics in it, are further influenced by suchrandom factors as driving loads, operation environment, et al, in the application in EVs.The real-time, accurate estimation of their state is challenging. For realizing theoptimization use of the lithium-ion battery in EVs, the detailed work has been carried out.The battery test platform has been set up and a systematic test program has beendesigned accordingly. A comprehensive battery test database with two kinds of lithium-ionbatteries has been established. Through the elaborate analysis with the electrochemicalimpedance spectroscopy (EIS) measurements showed that the impedance characteristic ofthe battery is sensitive to its operating condition and maximum available capacity, but it isnot sensitive to the charge/discharge history and the State-of-Charge (SOC).To achieve an accurate model of lithium-ion battery, the general mathematical modelwith n order RC networks has been put forward and then the five steps model parameteridentification approach has been proposed accordingly, which includes model parameteridentification and optimization, complexity and accuracy balance algorithm through theAkaike information criterion (AIC). Then, the model structure and model parameter havebeen determined and estimated. Finally, the problem of model accuracy has been solved.To solve the problem of battery model identification, a data-driven model parameteridentification method has been proposed for predicting the real-time characteristic ofbattery with the online measurements of battery current and voltage. Because of the onlineparameter identification process, it has the potential to realize the accurate prediction forbattery against uncertain battery aging states, operating environment and cell types. What’smore, it has solved the problem of poor extensibility and reliability caused by the constant and inaccurate model parameters for battery model.To achieve accurate battery SOC estimation, a data-model fusion based adaptive SOCestimation approach has been proposed. It uses the open circuit voltage to correct the SOCestimation error, the adaptive extended Kalman filtering (AEKF) to design the stateestimator and the three-dimensional response surface to build the open circuit voltagemodel. It has achieved the accurate SOC estimation against uncertain battery operatingenvironment, health status and driving cycles.To achieve a reliable battery peak power capacity estimate, a multi-parameterconstraint dynamic continuous estimation method for battery peak power capacity has beenproposed, it has considered the constraints from the design current, voltage and power, thedemanded operating range of SOC and the continuous period of required power. It has thepotential to solve the problem of the conservative estimation determined by the traditionalmethod and ensure the safety operation of battery. To solve the peak power capacityestimation error caused by the inaccurate SOC estimate, the joint estimation method forbattery SOC and peak power capacity has been proposed finally.To achieve accurate estimations for battery multi-state, the multi-scale AEKFalgorithm has been proposed for achieving accurate joint estimation for battery capacity,SOC and peak power capability. The proposed approach uses the macro time scale toidentify battery capacity and model parameter, the micro time scale to estimate batterystates of SOC and peak power, and uses the three dimensional space response surfacemodel to correct estimation errors of the capacity and SOC. It has achieved accurateestimations of multi-states of battery.To achieve accurate state estimation for battery pack, the multi-level multi-scaleAEKF has been proposed and applied to the state estimation of multi-cell series connectedbattery pack. With the parameter estimation objective of the inconsistency of the cellparameters of capacity and resistance and the state estimation target of cell SOC and peakpower capacity, the estimations of battery pack parameters and the states have beenobtained. It has solved the problem of accurate state estimation of battery pack. Therefore, to deal with the problem of joint estimation of battery capacity, SOC andpeak power capacity and provide a theoretical support for the energy management of EVs,this study has proposed a multi-scale AEKF algorithm, a multi-parameter constraintdynamic continuous peak power capacity estimation method and the joint estimationapproach for battery capacity, SOC and peak power capacity with multi-scale AEKFalgorithm. To overcome the problem of the balancing between the complexity and accuracyof battery model, model parameter update, the recalibration and robustness of the stateestimation for battery pack, the five steps recalibration method for battery modelparameters and multi-scale multi-level AEKF has been proposed, which possessesimportant applicable value in engineering.