Study On Estimation Of State-of-Charge Inconsistency In Lithium-ion Battery Pack

With the rapid development of the global economy,there is an increasing demand for automobiles,which brings concern about energy and environmental issues.Thus,many countries have proposed green travel strategies for sustainable development.Compared with traditional vehicles with internal combustion engines,new energy vehicles have developed rapidly due to their economy and environmental friendliness,and gradually achieve high sales in automotive market.Sine power battery technology is one of the key technologies in new energy vehicles,an efficient battery management system is essential for safe and efficient operations of vehicles under complex road conditions and environments.In order to prevent over-charging and over-discharging,and alleviate mileage anxiety,battery SOC need to be accurately estimated.A battery pack typically consists of a large number of cells connected in series or parallel.Due to manufacturing errors and different operation conditions,there exist inconsistencies between the battery cells.Because the performance of the battery pack is dependent on individual cells,estimating all cells’SOC quickly and accurately will be beneficial to the pack SOC estimation and efficient operation of battery management system.In order to achieve a trade-off between accuracy and computational complexity during SOC inconsistency estimation,this thesis develops a new method for estimation of SOC inconsistency in lithium-ion battery pack by combing both model-based and data driven approaches.The factors that influence battery pack inconsistency and parameter propagation are analyzed,which provide basis for establishing a battery pack model.The bisecting k-means algorithm is adopted to divide cells into different levels and reduce the objects that need to be considered.Then,a battery pack model that considers parameter discrepancies is established and verified against simulation and real test data.Comparison between the established model and other models have also been made and analyzed to highlight the strength of the proposed model.The main contents of this thesis are as follows:First,the causes of the inconsistencies of the battery pack are divided into internal causes and external causes.The propagation mechanism,main factors,and specific manifestations for parameter inconsistency are analyzed from the cell and pack level.Furthermore,the coupling between internal parameters and external parameters as well as the evolution mechanism of mutual-coupling between internal and external parameters are summarized in detail.The inconsistency parameters that have a large influence on battery pack performance are determined,which can be important to battery management.Then,commercial software,including Autolion-st and Matlab,are used to build a simulation model of a battery pack composed of 12 cells connected in series.The initial SOC of battery pack is set to be inconsistent,with typical normal and Will distribution.Dynamic discharge conditions(urban conditions and high-speed conditions)are used as the input simulate the real vehicle operation,and simulation charge and discharge database can then be obtained.The charging and discharging data of a new energy vehicle battery pack(2p86s)collected in real-time undergoes several pre-processing operations such as data deduplication,data trapping,etc.The real-time capacity and reference SOC of battery cells can be calculated based on the data characteristics and OCV-SOC curve.Then,the real-time running database can be obtained.Finally,4 features are extracted from the simulated charging data to form a 12×4 sample set.The bisecting k-means algorithm is adopted and the optimal clustering number is selected as 3 through the change of SSE and Silhouette Coefficient.A battery pack model considering SOC inconsistency is established and used to estimate cells SOC.Meanwhile,6 features are extracted from the real-time running charging data to form a 86×6 sample set.The bisecting k-means algorithm is adopted and the optimal clustering number is selected as 7 through the change of SSE and Silhouette Coefficient.A battery pack model considering SOC and internal resistance discrepancies is established and used to estimate the SOC of each individual cell.Comparison between the established model and other three battery pack models have been made in terms of estimation accuracy and computational complexity,which demonstrate that this method is more efficient for actual BMS operation.