| As the auxiliary power source of hybrid electric vehicles(HEVs)and the only power source of electric vehicles(EVs),the lithium-ion battery storage system is made up of several batteries connected in series,parallel,and hybrid to meet the power and mileage requirements of users.The batteries are electrically connected to form a group,which is not a simple multiple of numbers.The performance and safety of the battery pack depend on the worst battery.The accident of EVs are always related to battery fault.Therefore,the safety management and healthy operation of the battery system are issues that urgently need to be solved in the application field of lithium-ion batteries.Strengthen the research on active safety technology of battery system,detect and isolate faults,and avoid accidents caused by continuing to work after the battery fails.As one of the key functions of the battery management system,the fault diagnosis technology of the lithium-ion battery pack plays a crucial role in ensuring the safe and effectual operation of the battery system.This dissertation mainly carried out the following researches:(1)The least square method is employed to accurately distinguish battery parameters,and the double extended Kalman filter is modified to estimate the state of the battery online.For the ternary lithium-ion battery,predicated on Thevenin equivalent circuit model,the mathematical model of the battery is settled.The characteristic parameters of the lithium-ion battery are identified by the least square method.To solve the problem that the parameters are not easy to update and correct online,the dual extended Kalman filter is corrected to accurately estimate the open-circuit voltage of the battery online.Experimental tests have verified the accuracy of the method.(2)The virtual connection fault of the lithium-ion battery pack is investigated.A diagnostic method for a series-connected battery pack based on MSE and modified Z-score is proposed.Firstly,a novel cross-voltage test method is adopted to isolate the virtual connection from the internal resistance increases of the battery.The virtual connection causes the terminal voltage of two adjacent batteries to fluctuate synchronously.Predicated on the Thevenin model and the MATLAB/Simulink platform,the mean square errors of the experimental voltage real-time and the simulated voltage are calculated to preliminarily detect abnormal voltage.The voltage detection abnormal coefficients of each battery are calculated by using the modified Z-score,which determines the abnormal voltage and the location of the fault.The temperature rise rate of the battery is employed as an auxiliary fault diagnosis indicator to realize the detection,location,and classification of the virtual connection fault.The characteristics of the connection and internal resistance increase fault are compared and analyzed,and an isolation method for a parallel-connected battery pack is proposed in combination with the least square method.The results of the fault experiment show that the distribution of branch current under the conditions of connection and internal resistance increase is similar,but there are still differences.The distribution of branch current in the parallel-connected battery pack is the result of the competition among the impedance effect,the charge accumulation effect,and the self-balancing effect.The voltage difference of the battery itself under fault conditions is clarified by collecting the tab voltage and the total voltage.The reasons for the capacity loss of the battery pack under the two types of faults are revealed by the analysis of the differential capacity curve.The temperature of the battery is affected by the branch current and electrochemical effects.To isolate and locate the virtual connection from the internal resistance increase fault in the battery pack,firstly,the parallel unit is considered as a"large battery".Whether the unit has an increased resistance fault is determined by the internal resistance of the "large battery" estimated by the least square method.Secondly,the type of fault is determined according to the STD of the estimated internal resistance.Thirdly,combining the tab voltage and surface temperature,the specific fault battery is located in the parallel-connected unit.(3)The fault characteristics of the external soft short circuit in lithium-ion battery pack are revealed,and a quantitative diagnostic method based on the DEKF and linear fitting is proposed.Owing to the non-linear characteristics of the lithium-ion battery,even if the short circuit resistance is constant,the voltage difference between the battery with a soft short circuit and other normal batteries will not always increase,which follows the rule of first increasing,then decreasing and then increasing.The concepts of the "median cell" and the "minimum cell" are proposed.The SOC difference between the "median cell" and the "minimum cell" is accurately estimated based on the modified dual extended Kalman filter and OCV-SOC curve.The SOC difference curve before the inflection point is linearly fitted,and the short circuit current and the short circuit resistance are estimated.The results show that the relative error between the estimated short circuit resistance and measured resistance is about 6%,which is employed to accurately and quantitatively evaluated the risk of the fault.This method achieves the quantitative diagnosis of the external soft short circuit within one discharge cycle.Experimental tests under dynamic conditions verify the accuracy of the method.(4)Combining the battery model and the data-driven method,a parallel PCA-KPCA method for multi-fault diagnosis in lithium-ion battery pack is developed.Firstly,the battery PCA model is established based on the OCV and ohmic internal resistance estimated by the least square method and the terminal voltage monitored in real-time.The "median cell" is regarded as the normal battery for the model training.The fault detection online is realized based on the PCA contribution value.Owning to the typical nonlinear characteristics of the lithium-ion battery,the parallel PCA-KPCA reconstruction technology is used to estimate the fault waveform of the faulty battery.The cause of the fault is clarified based on the fault waveform.Experimental tests verify the accuracy of battery inconsistency evaluation and the fault of the virtual connection fault and external soft short-circuit fault.While,the experimental results also show that some types of fault will not cause abnormalities in all battery characteristics,for example,the virtual connection fault will not cause abnormal changes in open-circuit voltage.Therefore,combining the multiple characteristic indicators,terminal voltage,open-circuit voltage,and ohmic internal resistance,the reliability,universality,and robustness of fault diagnosis are increased through fault signal reconstruction based on the lithium-ion battery model and multivariate statistical analysis methods. |
PDF Full Text Request