Abstract: Lithium-ion batteries have the advantages of high energy density, long cycle life and being free from memory effect, and are widely used in electric vehicles and energy storage. In recent years, how to deal with retired batteries has become an urgent development problem for the new energy vehicle industry. It is an effective way to solve this problem by applying retired batteries to the energy storage system on a large scale. When the retired lithium-ion batteries are grouped, the performance difference between individual batteries will cause a rapid attenuation of module performance, shortening the module cycle. When a lithium-ion battery is assembled from these multiple retired cells, the difference in the performance of each cell will cause a rapid attenuation of battery performance. The attenuation of battery performance will reduce cycle life and even cause safety problems. In order to reduce performance difference of decommissioned batteries after grouping and improve the clustering effect of K-means algorithm, this paper takes 100 of the decommissioned 18 650 lithium batteries purchased in the same batch as the research objects for charge-discharge tests and internal resistance tests. The charge and discharge experiments and internal resistance tests of 100 retired lithium-ion cells are conducted. A consistency sorting method is proposed based on dynamic curve. Firstly, in terms of parameter selection, considering that the internal resistance of a battery and the internal resistance of a connector will consume part of the electric during the charging and discharging process, and the electrochemical polarization and concentration polarization of Li+ insertion and detachment will also cause a part of energy loss, the difference between the charging energy and the discharging energy can be used to describe the chemical polarization and concentration polarization in the general testing process. A consistent sorting and reorganization method for retired lithium-ion batteries considering dynamic curve characteristics is proposed, and the multi-parameter sort is carried out by using six indicators, including capacity, energy difference, charging voltage, discharging voltage, charging resistance and discharging resistance. In consideration of the correlation between the sorting variables, in order to reduce the sorting variables and simplify the calculation, factor analysis is applied to sort variables of retired batteries, and the batteries are classified by using the inter-group connection clustering method and the square Euclidean distance as the measurement standard. In terms of sorting methods, the dynamic changes of battery parameters during charging and discharging are taken into account through the dynamic characteristic sorting method. Combining with multi-parameter sorting, higher consistency can be achieved. The existing dynamic sorting methods generally use voltage curve for sorting, but this method cannot reflect current, capacity and other performance parameters. The voltage curve during constant voltage charging cannot show the change trend of battery energy and capacity; the energy curve at the time of shelving cannot represent the change of battery voltage. In addition, the consistency of voltage and capacity should be taken into account during battery combination to avoid energy waste. Therefore, on the basis of multi-parameter sorting, energy difference is used to represent the difference of cell polarization. Then, the K-means algorithm is used to carry out dynamic curve sorting based on voltage deviation and capacity deviation. Finally, in terms of algorithm improvement, to solve the problem of an uncertain K value, the voltage and energy curves are normalized respectively. According to an analysis of frequency distribution histogram and frequency distribution curve, the K value of clusters is determined. The voltage standard deviation of 100 retired lithium-ion batteries before sorting is 0.043 1, of Class I batteries after sorting is 0.203 7, and of Class Ⅱ batteries after sorting is 0.011 1. It can be seen that the voltage consistency of type I battery is poor, indicating that performance degradation of the battery pack is caused by performance degradation of very few batteries. The results of experiment show that the method can effectively improve inconsistency of the battery. The consistency of charging voltage increases by about 60%～94%. The consistency of discharge voltage increases by about 10%～41%. The consistency of capacity increases by about 54%～67%.