| Lithium-ion batteries have been extensively used in new energy industries,including electric cars and electrochemical energy storage,as a result of the thorough promotion of the national "dual carbon" plan.As the latest energy storage carrier with the most complete industrial chain and the largest installed capacity,it has advantages such as long cycle life,fast response speed,and flexible topology structure.Specifically,power battery systems which are the core energy component of electric vehicles and energy storage systems,are usually made up of hundreds or even thousands of lithiumion cells connected in series and parallel,which are used to meet the voltage and capacity level requirements of the systems they serve.However,due to the limitations of the current manufacturing process,there is inevitably a certain inconsistency between the produced battery cells,which makes the overall performance of the battery pack cannot simply equivalent to the cloning and superposition of individual battery performance.It will also be influenced by various factors such as grouping,topology,and matching methods.Accordingly,the most widely used lithium iron phosphate/graphite battery is taken as the research object,improving battery grouping performance is taken as the core goal,and the series-parallel structure optimization of the battery pack is taken as the main method.The emphasis of the paper is on selecting the optimal topology structure of the battery pack for “available capacity” in the short-time scale,and for “cycle life” in the long-time scale.1.In the short-time scale,the available capacity of the battery pack will be affected by the unbalanced current distribution of the short-board battery and the branch circuit at the same time.The difficulty of the calculation lies in the effective tracking of the current change for the branch circuit where the short-board battery is located.For this reason,a calculation method for the discharge performance of series-parallel battery packs based on "branch current recursive calculation" is proposed.Based on the classic Rint model,the "hybrid structure" is equivalent to a "dual-battery structure in which the target battery and other batteries are connected in parallel as a whole",and combined with real-time calculation of the state of each battery under inconsistent distribution of battery parameters,and then achieves recursive updating of each branch current.Furthermore,through multiple simulations based on Monte Carlo,the available capacity and energy distribution patterns of each topology under given parameter conditions are obtained,and the selection of the optimal topology is ultimately completed.The accuracy of the algorithm is verified by charging and discharging experiments of battery packs under different topologies.2.In the long-time scale,the capacity of battery packs after multiple cycles is not only constrained by the capacity loss characteristics of the cells,but also by the increasing inconsistency between the individual states after aging.To obtain the overall aging trajectory of the battery pack under given topology conditions,it is necessary to construct a battery pack capacity loss behavior model based on real-time tracking of the state of charge(SOC)and the state of health(SOH).Therefore,the influence function of battery capacity loss on the overall capacity and various characteristic parameters based on the classic dynamic behavior model of batteries is firstly added,and the characterization of cell’s aging behavior is achieved.And then,combined with parameter inconsistency distribution experiments,the overall aging behavior of the battery pack is modeled through the combined modeling method in the Simscape environment.Finally,the real-time updates of SOC,SOH,and various model parameters are achieved through the established model,and the capacity trajectory of the battery pack under individual cut-off conditions is calculated.The results of battery pack aging experiments under different topologies indicate that the root mean square error of the simulation prediction results of the built model is less than 1%.3.Based on the obtained battery pack capacity loss model,aiming at the contradiction between the richness of topological combinations contained in large-scale battery packs and the limited number of topologies that can be preset by the algorithm,a battery pack arbitrary topology cycle life prediction method based on the learning of typical structure calculation results is proposed.This method completes the calculation of the cycle life under arbitrary topological conditions and the selection of the best serial-parallel combination structure through the cyclic aging simulation of the typical structure and the machine learning derivation of the atypical structure.Experiments show that the root mean square error of the proposed derivation prediction method is controlled within 1%.At the same time,the experimental results after 1200 cycles prove that the capacity attenuation rate of the optimal series-parallel hybrid structure calibrated by the derivation method is 3.1% higher than that of the traditional parallel cell module,accounting for the attenuable capacity range of the electrical vehicle(15.5%of 20%SOH),which fully reflects the unique advantages of the new method in selecting the optimal structure. |
