Research On The Capacity Estimation And Fault Diagnosis Of Lithium-ion Batteries

In recent years,the demand for electric vehicles has promoted the rapid development of lithium-ion batteries.As the Nobel Prize winner Akira Yoshino said:"Lithium-ion batteries will play a central role in the future energy revolution."According to the report of SNE Research,an authoritative new energy research organization:In 2020,the global installed capacity of lithium-ion batteries for vehicles has reached 137GWh,a year-on-year increase of 17%.It is estimated that the annual growth rate of global Lithium-ion battery sales will exceed 15%in the next five years and will exceed 650GWh by 2025.However,there are many key scientific issues that remain unresolved,especially the theoretical methods for efficient utilization and safety management of lithium-ion batteries are far from mature.These severely restrict the healthy development of lithium-ion batteries in electric vehicles and large-scale energy storage applications.On the one hand,the battery capacity estimation method has poor applicability and large estimation errors,leading to the phenomenon of battery "virtual power" and causing anxiety about the range of electric vehicles.On the other hand,there are safety management problems such as the diagnosis of internal short circuit and the misdiagnosis between faults,which have caused many safety accidents such as battery fires and explosions.For this reason,this dissertation has mainly finished the following research works.For research objects of two types(LiFePO4,NCM lithium-ion batteries)and three levels(battery cells,parallel modules,series modules),a battery aging test platform was built,and a comprehensive battery test program was designed-Based on the experimental results,we revealed the influence of current rate,depth of charge and discharge,and inconsistency on battery capacity degradation,and explored the types and characteristics of common faults in battery systems,especially the causes and evolution of thermal runaway.Aiming at the problem of poor applicability of the battery capacity estimation method under the random operating conditions of electric vehicles,an online capacity estimation strategy suitable for any charging starting point is proposed.For three types of charging conditions with different voltage ranges,a back-propagation neural network algorithm based on the peak of the incremental capacity curve,an integrated learning algorithm based on part of the incremental capacity curve,and a linear regression algorithm based on the ampere-hour coordinate transformation are designed respectively.According to the current cycle conditions,one of the suitable methods is selected to estimate the battery capacity in the current charging cycle,and the contradiction between estimation accuracy and complexity is optimized.The experimental results show that the battery capacity can be accurately estimated within any charging voltage range,that is,the problem of missing characteristic values caused by random charging and discharging ranges are solved.Aiming at the problems that the inconsistency is difficult to quantify,and the capacity estimation error is large after the batteries are grouped,this section discusses capacity degradation path and estimation error sources under the battery capacity-electricity coordinate system.Furthermore,the evolution law of voltage change rate with battery aging in constant current charging is discovered.A data-driven method for quantifying power inconsistency is proposed,the estimation result can be used as the input of battery pack capacity estimation and optimal balancing strategy at the same time.The capacity of the series battery pack is estimated combined with the BPNN cell capacity estimation method.Experimental results prove that the proposed inconsistency quantification method avoids the error sources as battery modeling and SoC estimation,and the battery capacity can be accurately estimated within a wide cut-off voltage range.Aiming at the difficulty of diagnosing minor faults such as internal short circuits in the battery system,an online diagnosis method based on the correlation coefficient is proposed.Based on the voltage and current characteristics of the battery string,the diagnostic idea of voltage correlation is designed.The forgetting mechanism and discrete square wave correction are introduced to eliminate the influence of measurement noise and battery inconsistency.A standardized fault feature comparison method is proposed to quantitatively compare the sensitivity and robustness of the three types of diagnostic methods,including correlation coefficient,model-based,and sample entropy.The fault experiment results show that the proposed method can diagnose 0.1C internal short circuit faults in the battery system,and has strong robustness to the interference of temperature,aging,and inconsistency.It provides an ideal solution for robust diagnosis of battery minor faults under complex environmental conditions.In view of the difficulty in distinguishing multiple types of mixed fault types in power battery systems,a comprehensive diagnosis strategy for mult-faults of battery packs is proposed.The proposed cross-style measurement topology can separate the miscellaneous fault types and fault point information,and effectively extract the fault characteristics with the correlation between the cell voltages and the voltage and current.The circuit matrix analysis and experimental results show that the new strategy can effectively distinguish and diagnose miscellaneous faults such as internal/external short circuit,connection fault,voltage sensor fault,and realize the synchronous diagnosis of battery system fault type,location,and degree without increasing hardware cost.It is suitable for complex measurement noise and inconsistency conditions in real battery system and has great engineering application value.In summary,this dissertation establishes an "estimation-diagnosis-warning"intelligent management theoretical method for battery packs.It lays a solid foundation for ensuring the safe,efficient and reliable operation of lithium-ion batteries in applications such as electric vehicles and large-scale energy storage systems.