Co-estimation Of Lithium-ion Battery States Based On A Fractional-order Model

With increasingly severe global warming and energy crisis,pure electric vehicles with higher energy efficiency have received extensive attention and on-going investments across the globe to make them key application solutions for clean energy future.As electric vehicles always work in a complex environment,to ensure the safety and reliability of batteries,they are always equipped with well-designed battery management systems(BMSs).One of the critical functions of a BMS is to keep monitoring internal battery states precisely,like state-of-charge(SoC)and state-of-health(SoH).Accurate SoC and SoH indication can not only promote a safer and more efficient working condition for batteries,but also share useful information for battery second-life use.Therefore,accurate co-estimation of SoC and SoH has become one of the most important issues to address in the battery control community.Reported SoC and SoH co-estimation methods in the existing literature usually fall short in convergence,accuracy,and robustness.With this research gap in mind,this thesis presents a novel multi-scale co-estimation observer of lithium-ion battery parameters,SoC,and SoH,based on a nonlinear fractional-order model.The specific investigations in the thesis are outlined as follows.The first chapter demonstrates the state of the art of automotive batteries and their BMSs.A comprehensive review of battery modeling,SoC and SoH co-estimation methods are presented.Based on these,research gaps are sufficiently exposed and summarized.Second,the battery testing system and programs are introduced.Some basic concepts of fractional-order calculus and modeling principle of fractional-order electric model are presented.Moreover,an optimization-based method is employed for battery parameters identification offline.Model validation is performed by a comparison against integer-order counterparts.In the third chapter,a linear regression form of the fractional-order model is first derived.Based on this regression model,the recursive least-squares(RLS)method can be employed for online parameter identification.The fractional-order extend Kalman filter and adaptive filter are presented and then coupled together to synthesize an adaptive fraction-order extended Kalman filter(AFOEKF).Then,a SoC and parameter online co-estimation framework utilizing RLS and AFOEKF are well summarized.The accuracy and convergence of the proposed co-estimation framework are verified by a comparison with other four different methods.Furthermore,by comparisons with a dual adaptive fractional-order extend Kalman filter(DAFOEKF),the accuracy and robustness of the proposed method are sufficiently studied under different aging states for different cells.In the fourth chapter,the basic principles of the full information estimator,moving horizon estimation,and a modified moving horizon estimation are presented.Battery capacity and fractional order are employed as indicators of SoH to capture battery aging dynamics.Considering its slow varying characteristics,a sliding window off-line estimation method is devised for the SoH estimation.Moreover,the effectiveness of the proposed method is verified considering different sliding window horizons and physical constraints.The impact of the off-line SoH estimation on the online SoC estimation is examined.Verifications for different cells and different aging states are also carried out.