Retired Power Battery Matching and Recombination Method, Service System and Storage Medium

Abstract

The present application relates to a method, a service system and a storage medium for matching and reorganizing retired power batteries. The method includes: after preprocessing various target index parameters corresponding to the target information of each target retired power battery cell to generate corresponding various standard characteristic parameters, using a consistency sorting model to process the various standard characteristic parameters corresponding to each target retired power battery cell to obtain classification label data corresponding to multiple sorted sets of retired power battery cells; based on the pre-obtained battery demand parameters, multiple sets of retired power battery cells and the corresponding classification label data, using an improved artificial bee colony algorithm to perform iterative battery reorganization configuration planning to generate corresponding reorganization configuration information; after determining the production order demand information corresponding to the battery demand parameters, using a preset heuristic algorithm to perform production scheduling planning on the production order demand information and the reorganization configuration information, and performing power battery reorganization according to the generated target production scheduling planning information.

Description

Technical Field

[0001] This application relates to the technical field of secondary utilization of retired power batteries, in particular to a method for matching and restructuring retired power batteries, a service system and a storage medium. Background Technique

[0002] Power batteries are widely used in fields such as electric vehicles. As the usage time goes by, their performance will gradually decline, and finally reach a retired state where they can no longer meet the usage requirements of the vehicle. After retirement, the power battery still has a certain remaining capacity and can be used in scenarios such as energy storage. In related technologies, the methods for recycling and reusing power batteries include disassembly and recycling and secondary utilization. Because secondary utilization can extend the life cycle value of the battery and achieve full utilization of resources, secondary utilization is becoming more and more popular in the recycling of power batteries.

[0003] In related technologies, when performing secondary utilization processing on retired power batteries, due to significant differences in parameters such as capacity, internal resistance, and health status of retired power batteries, the performance consistency of different batteries is poor. Based on the existing sorting methods that mainly rely on static indicators, the overall performance of the battery cannot be comprehensively measured, making it difficult to balance sorting accuracy and timeliness, resulting in low matching efficiency for restructuring and producing retired power battery cells and difficult inventory management.

[0004] Currently, for the problems of insufficient sorting accuracy, low restructuring and matching efficiency, and difficult inventory management when restructuring and matching retired power battery cells in related technologies, no effective solutions have been proposed. Summary of the Invention

[0005] Embodiments of this application provide a method for matching and restructuring retired power batteries, a service system and a storage medium, so as to at least solve the problems of insufficient sorting accuracy, low restructuring and matching efficiency, and difficult inventory management when restructuring and matching retired power battery cells in related technologies.

[0006] In a first aspect, an embodiment of the present application provides a method for matching and reorganizing retired power batteries, including: obtaining target information of a plurality of target retired power battery cells to be reorganized, where each of the target information includes a plurality of target index parameters, and the target index parameters are used to characterize a characteristic parameter of the corresponding target retired power battery cell, and the characteristic parameter is at least used to characterize the current storage state of the target retired power battery cell; after preprocessing the plurality of target index parameters corresponding to each target retired power battery cell to generate corresponding standard characteristic parameters, using the trained consistency sorting model to process the plurality of standard characteristic parameters corresponding to each target retired power battery cell to obtain classification label data corresponding to a plurality of sorted retired power battery cell sets, where the classification label data includes a reorganization adaptation level for characterizing the priority of the retired power battery cell set applied to the cascade scenario during the reorganization process, and the consistency sorting model is a machine learning model trained based on the fuzzy c-means clustering algorithm and is trained to generate classification label data corresponding to the corresponding retired power battery cell set according to the input battery characteristic parameters; based on the pre-obtained battery demand parameters, a plurality of the retired power battery cell sets, and the corresponding classification label data, using an improved artificial bee colony algorithm to perform iterative battery reorganization configuration planning to generate corresponding reorganization configuration information, where the reorganization configuration information includes the target information corresponding to the retired power battery cell corresponding to the reorganized power battery pack; after determining the production order demand information corresponding to the battery demand parameters, using a preset heuristic algorithm to perform production scheduling planning on the production order demand information and the reorganization configuration information, and performing power battery reorganization according to the generated target production scheduling planning information to obtain a plurality of reorganized power batteries.

[0007] In a second aspect, an embodiment of the present application provides a service system, including a memory and a processor, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for matching and reorganizing retired power batteries described in the first aspect.

[0008] In a third aspect, an embodiment of the present application provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for matching and reorganizing retired power batteries described in the first aspect above.

[0009] Compared with the related art, the method, service system and storage medium for matching and restructuring retired power batteries provided by the embodiments of the present application obtain the target information of a plurality of target retired power battery cells to be restructured, and each of the target information includes a variety of target index parameters, and the target index parameters are used to characterize a characteristic parameter of the corresponding target retired power battery cell, and the characteristic parameter is at least used to characterize the current storage state of the target retired power battery cell; after preprocessing the various target index parameters corresponding to each target retired power battery cell to generate corresponding various standard feature parameters, using the trained consistency sorting model to process the various standard feature parameters corresponding to each target retired power battery cell to obtain the classification label data corresponding to the sorted multiple retired power battery cell sets; based on the pre-obtained battery demand parameters, the multiple retired power battery cell sets and the corresponding classification label data, using the improved artificial bee colony algorithm to perform iterative battery restructuring configuration planning to generate corresponding restructuring configuration information, and the restructuring configuration information includes the target information corresponding to the retired power battery cells corresponding to the restructured power battery pack; after determining the production order demand information corresponding to the battery demand parameters, using the preset heuristic algorithm to perform production scheduling on the production order demand information and the restructuring configuration information, and performing power battery restructuring according to the generated target production scheduling information to obtain multiple restructured power batteries, solving the problems of insufficient sorting accuracy, low restructuring and matching efficiency, and difficult inventory management in the related art when restructuring and matching retired power battery cells, adopting fuzzy clustering sorting, multi-objective optimization matching and heuristic algorithm scheduling and restructuring arrangements to achieve high-precision consistency sorting of retired power batteries, efficient order demand matching, optimized production scheduling, and lean inventory management, improving the utilization efficiency of retired battery resources, and reducing the environmental pollution caused by waste batteries.

[0010] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more concise and understandable. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0012] Figure 1 is a hardware structure block diagram of a terminal of the method for matching and restructuring retired power batteries according to an embodiment of the present application;

[0013] Figure 2 is a flowchart of the method for matching and restructuring retired power batteries according to an embodiment of the present application;

[0014] Figure 3It is a structural block diagram of a retired power battery matching and recombination device according to an embodiment of the present application. Detailed implementation manners

[0015] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be described and explained below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments provided in the present application without making creative efforts belong to the scope of protection of the present application. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacturing or production changes made on the basis of the technical content disclosed in the present application are only conventional technical means and should not be understood as the content disclosed in the present application being insufficient.

[0016] Referring to "embodiment" in the present application means that the specific features, structures or characteristics described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those of ordinary skill in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments without conflict.

[0017] Unless otherwise defined, the technical terms or scientific terms involved in the present application should be of the ordinary meaning understood by those of ordinary skill in the technical field to which the present application belongs. The terms "a", "one", "kind", "the" and the like involved in the present application do not indicate a quantity limitation and can represent a single or plural number. The terms "including", "comprising", "having" and any variations thereof involved in the present application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device including a series of steps or modules (units) is not limited to the listed steps or units, but may further include unlisted steps or units, or may further include other steps or units inherent to these processes, methods, products or devices. The "multiple links" involved in the present application refer to two or more links. "And / or" describes the association relationship of associated objects and indicates that three relationships can exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. The terms "first", "second", "third", etc. involved in the present application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0018] The method embodiments provided in this embodiment may be executed on a terminal, a computer, or a similar computing device. Taking running on a terminal as an example, Figure 1 is a hardware structure block diagram of the terminal for the retired power battery matching and recombination method according to the embodiments of the present application. As Figure 1 shown, the terminal may include one or more ( Figure 1 only one is shown in the figure) processors 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the above terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only illustrative and does not limit the structure of the above terminal. For example, the terminal may further include more or fewer components than Figure 1 shown in the figure, or have a different configuration from Figure 1 shown in the figure.

[0019] The memory 104 can be used to store computer programs. For example, software programs and modules of application software, such as the computer program corresponding to the retired power battery matching and recombination method in the embodiments of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely provided relative to the processor 102, and these remote memories may be connected to the terminal through a network. Examples of the above network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0020] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network may include a wireless network provided by a communication provider of the terminal. In one instance, the transmission device 106 includes a network adapter (abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 may be a radio frequency (abbreviated as RF) module, which is used to communicate with the Internet wirelessly.

[0021] This embodiment provides a retired power battery matching and recombination method running on the above terminal, Figure 2 is a flowchart of the retired power battery matching and recombination method according to the embodiments of the present application. As Figure 2 shown, the process includes the following steps:

[0022] Step S201: Obtain the target information of multiple target retired power cells to be recombined. Each target information includes multiple target index parameters, and the target index parameters are used to characterize a characteristic parameter of the corresponding target retired power cell. The characteristic parameter is at least used to characterize the current storage state of the target retired power cell.

[0023] In this embodiment, before recombining the retired power cells, it is necessary to perform consistency sorting on the target retired power cells to be recombined, so that the recombined power battery meets the corresponding order requirements. In order to perform consistency sorting on the target retired power cells, the relevant parameters corresponding to each target retired power cell need to be used. The relevant parameters include static parameters and dynamic parameters. In this embodiment, the static parameters include cell internal resistance, cell remaining capacity, and cell open-circuit voltage. The static parameters are used to reflect the current state of the battery and are essential parameters. The dynamic parameters include charge-discharge rate, depth of discharge, and state of health (SOH). The dynamic parameters are used to evaluate the actual usage performance and remaining life of the cell.

[0024] Step S202: After preprocessing the multiple target index parameters corresponding to each target retired power cell to generate corresponding multiple standard feature parameters, use the trained consistency sorting model to process the multiple standard feature parameters corresponding to each target retired power cell to obtain the classification label data corresponding to the sorted multiple retired power cell sets. The classification label data includes a recombination adaptation level used to characterize the priority of the retired power cell set applied to the cascade scenario during the recombination process. The consistency sorting model is a machine learning model trained based on the fuzzy c-means clustering algorithm and is trained to generate the classification label data corresponding to the corresponding retired power cell set according to the input battery feature parameters.

[0025] In this embodiment, before inputting the multiple target index parameters into the trained consistency sorting model, it is necessary to preprocess the multiple target index parameters first, that is, process the multiple target index parameters into a data format that the consistency sorting model can recognize and use to provide accurate data support for the consistency sorting model to sort the target retired power cells. The consistency sorting model is trained using a sample data set in the corresponding data format. In this embodiment, after preprocessing and converting to generate the corresponding standard feature parameters, input the multiple standard feature parameters into the consistency sorting model to cluster the multiple standard feature parameters corresponding to each target retired power cell group, group the retired power cells with similar performance into one category, and allow a certain degree of ambiguity, that is, each retired power cell can partially belong to multiple categories. This ambiguity helps to achieve flexible grouping in the case of large differences in the performance of retired power cells and ensure the robustness of the clustering.

[0026] In this embodiment, a consistency sorting model of machine learning is used to cluster and sort multiple target retired power cells. Through the clustering process of the consistency sorting model, the retired power cells with similar performance are grouped into the same category to facilitate subsequent recombination and matching. In this embodiment, the classification label data of each target retired power cell marks the target retired power battery as a certain category, and the cells within this category have high performance consistency in terms of capacity, internal resistance, discharge rate, and health status. Using the classification label data as the sorting result for the subsequent cell recombination and matching process ensures that the recombined battery pack maintains a high degree of similarity in performance, providing guarantees for the stability and lifespan of the battery pack.

[0027] Step S203: Based on the pre-acquired battery demand parameters, multiple sets of retired power cells, and the corresponding classification label data, use the improved artificial bee colony algorithm to perform iterative battery recombination configuration planning to generate corresponding recombination configuration information, where the recombination configuration information includes the target information corresponding to the retired power cells corresponding to the recombined power battery pack.

[0028] In this embodiment, after determining the classification label data corresponding to the target retired power battery cell, according to the previously obtained battery demand parameters (corresponding to order demand data), and then based on the order demand parameters and the sorting results characterized by the classification label data, iterative battery recombination configuration planning is performed to recombine and match a suitable battery pack; in some alternative embodiments, the battery demand parameters are obtained through demand prediction based on the analysis of historical order data, current market trends, and application scenario requirements; in this embodiment, during the iterative process of battery recombination configuration planning, the following several target matching operations are also performed: capacity matching, which is used to ensure that the capacity of the battery pack allocated and recombined is as close as possible to the capacity demand of the order, avoiding waste or shortage of capacity; discharge rate matching, by selecting batteries that meet the discharge rate requirements of the order to ensure the performance stability of the battery during application; life matching, by preferentially selecting battery packs with a higher state of health (SOH) to meet the order's demand for battery life and improve the long-term reliability of the configuration plan; internal resistance matching, according to the order's requirement for the internal resistance of the battery pack, preferentially selecting batteries with appropriate internal resistance to reduce the internal resistance difference, so as to achieve the internal resistance balance of the battery pack and improve the overall efficiency and safety; in this embodiment, the iterative battery recombination configuration planning is based on the sorting results of the retired power battery cells in the inventory, comprehensively analyzing the standard characteristic parameters of the retired power battery cells in the inventory (including capacity, internal resistance, discharge rate, and state of health (SOH)) to perform precise screening and intelligent matching configuration; in this embodiment, through the screening of the standard characteristic parameters, it is ensured that the recombined battery pack can meet the order requirements and improve the utilization efficiency of the retired power battery cells, that is, the iterative battery recombination configuration planning is to classify the capacity, internal resistance, and other indicators of each retired power battery cell in detail and select a battery pack that meets the performance standards based on the order's requirements for capacity, life, and discharge rate, so as to meet the application scenario of cascade utilization; and the embodiment of the present application adopts the introduction of the artificial bee colony algorithm to iteratively optimize and generate recombination configuration information by simulating the behavior of bees collecting honey.

[0029] Step S204, after determining the production order demand information corresponding to the battery demand parameters, perform production scheduling planning on the production order demand information and the recombination configuration information by using a preset heuristic algorithm, and perform power battery recombination according to the generated target production scheduling planning information to obtain multiple recombined power batteries.

[0030] In this embodiment, after generating the recombination configuration information, a heuristic algorithm is used for production scheduling arrangement, and this production scheduling arrangement is based on the recombination production requirements of the battery pack, the allocation of equipment resources, and the time requirements of the order. By reasonably arranging production resources and optimizing the task sequence, the efficiency and stability of the production process are guaranteed.

[0031] Through the above steps S201 to S204, the target information of multiple target retired power battery cells to be recombined is obtained. Each target information includes multiple target index parameters, and the target index parameters are used to characterize a characteristic parameter of the corresponding target retired power battery cell. The characteristic parameter is at least used to characterize the current storage state of the target retired power battery cell; after preprocessing the multiple target index parameters corresponding to each target retired power battery cell to generate corresponding multiple standard feature parameters, the trained consistency sorting model is used to process the multiple standard feature parameters corresponding to each target retired power battery cell to obtain the classification label data corresponding to the sorted multiple retired power battery cell sets; based on the previously obtained battery demand parameters, multiple retired power battery cell sets and the corresponding classification label data, the improved artificial bee colony algorithm is used to perform iterative battery recombination configuration planning to generate corresponding recombination configuration information. The recombination configuration information includes the target information corresponding to the retired power battery cells corresponding to the recombined power battery pack; after determining the production order demand information corresponding to the battery demand parameters, the production order demand information and the recombination configuration information are used to perform production scheduling using a preset heuristic algorithm, and the power battery recombination is performed according to the generated target production scheduling information to obtain multiple recombined power batteries, solving the problems of low sorting accuracy, low recombination matching efficiency, and difficult inventory management in the related technology when recombining and matching retired power battery cells. Fuzzy clustering sorting, multi-objective optimization matching, and heuristic algorithm scheduling and recombination arrangements are adopted to achieve high-precision consistency sorting of retired power batteries, efficient order demand matching, optimized production scheduling, and lean inventory management, improve the utilization efficiency of retired battery resources, and reduce the environmental pollution caused by waste batteries.

[0032] It should be noted that the retired power battery matching and recombination method in the embodiment of the present application realizes high-precision consistency sorting, efficient demand matching, optimized production scheduling, and lean inventory management of retired power battery cells by performing fuzzy clustering sorting, multi-objective optimization matching, and heuristic algorithm production scheduling.

[0033] It should be further noted that in this embodiment, when making production scheduling plans, reasonable allocation of current production resources is adopted, including resources such as assembly equipment, testing equipment, storage space, and personnel. The recombinant production requirements of each battery pack are matched with the available resources to ensure the maximization of equipment utilization rate. Specifically, different production steps (such as charge and discharge testing, internal resistance testing, battery assembly, etc.) are arranged according to production requirements and equipment conditions to ensure that the equipment requirements for each step are reasonably met. At the same time, in task arrangement, a heuristic algorithm is used to optimize production scheduling to ensure a reasonable order of tasks. This heuristic algorithm quickly calculates the optimal task scheduling order through simulating human experience judgment. In the initialization stage, the initial scheduling order generated by the production scheduling plan gives priority to orders with urgent delivery times and ensures the continuity of production tasks. Then, the order tasks are sorted according to urgency, task complexity, and equipment availability to determine the scheduling order of each order in the production process. At the same time, the heuristic algorithm also makes fine-tuning of the task order in the local optimization stage to ensure smooth connection of each production step on which the tasks depend and avoid production interruptions caused by resource conflicts or task waiting. In the process of multiple iterative improvements, the production arrangement is also dynamically adjusted according to task priorities and resource utilization to improve the overall production efficiency and resource utilization rate. Moreover, in the scheduling process, the timeliness of order delivery is emphasized, and urgent orders are given priority to ensure on-time delivery. For orders with a longer cycle, the distribution of production tasks is balanced to ensure the balanced utilization of resources in each production link. If there is a shortage of production resources or equipment failures, tasks will be automatically adjusted or the production order will be rearranged to ensure production continuity and on-time order delivery.

[0034] In some of these embodiments, based on the pre-acquired battery demand parameters, multiple sets of the retired power battery cells, and the corresponding classification label data, an improved artificial bee colony algorithm is used to perform iterative battery recombination configuration planning to generate corresponding recombination configuration information, including the following steps:

[0035] Step 21, according to the classification label data corresponding to multiple sets of the retired power battery cells, use the improved artificial bee colony algorithm to perform battery recombination configuration to generate first recombination configuration information. Among them, the first recombination configuration information includes the target retired power battery cells sorted for the first power battery pack, and the first power battery pack is used to represent the power battery pack currently recombinantly configured.

[0036] Step 22, obtain the first fitness corresponding to each first power battery pack from the first recombination configuration information. Among them, the first fitness is calculated by an adaptation function constructed based on the classification label data corresponding to the target retired power battery cells assigned to each first power battery pack, and is used to characterize the excellent degree of performance of the first power battery pack.

[0037] Step 23: Determine the target fitness of the required battery pack corresponding to the battery demand parameters, and based on the difference between the first fitness and the target fitness, perform iterations of local search, neighborhood optimization, and global search of the improved artificial bee colony algorithm on the first recombination configuration information until the difference is not greater than a preset threshold to generate the recombination configuration information. In the global search stage, the improved artificial bee colony algorithm introduces a dynamic neighborhood adjustment and a local optimal escape mechanism.

[0038] In this embodiment, generating the first recombination configuration information corresponds to the initialization stage of battery recombination configuration using the improved artificial bee colony algorithm. In the initialization stage, several retired power battery cells are randomly selected from the retired power battery cell set to be recombined into the first power battery pack, and the configuration scheme corresponding to the first power battery pack is used as the initial solution. Before configuring the initial solution, a fitness value will be assigned to the initial solution to be configured according to the matching degree of its cell capacity, discharge rate, life, and internal resistance with the order requirements. The higher the fitness value, the better the matching degree, that is, when the initial solution is randomly generated, it needs to be matched with the assigned fitness value. After assigning the corresponding fitness value, a configuration scheme close to or equal to the assigned fitness value is randomly selected, that is, multiple retired power battery cells with approximate performance consistency are selected. It should be understood that the first fitness value of the configured initial solution is not necessarily the assigned fitness value. The first fitness value is calculated based on the adaptation function constructed from the classification label data corresponding to the multiple retired power battery cells configured. It may be equal to or close to the assigned fitness value. By solving the fitness value corresponding to the allocation scheme and based on the difference between the assigned fitness value and the solved fitness value, the solution corresponding to the power battery pack is configured and adjusted, that is, iterations of local search, neighborhood optimization, and global search of the improved artificial bee colony algorithm are performed until a power battery pack that adapts to the order requirements and has excellent performance is obtained. At this time, the difference between the fitness value corresponding to the power battery pack and the preset target fitness is not greater than the preset threshold. It should be noted that in this embodiment, when using the improved artificial bee colony algorithm to generate the initial solution, a dynamic fitness value update mechanism is introduced, and the order prediction result is considered during initialization. For parameters with a relatively large possible future demand, a higher fitness value will be assigned and set.

[0039] Through steps 21 to 23, the improved artificial bee colony algorithm is adopted to find the optimal solution corresponding to the optimal combination of the power battery pack by simulating the process of bees collecting honey, and then generate the recombination configuration information corresponding to the power battery pack including multiple optimal combinations.

[0040] In some embodiments, according to the classification label data corresponding to multiple retired power battery cell sets, the battery recombination configuration is performed using the improved artificial bee colony algorithm to generate the first recombination configuration information, which is realized through the following steps:

[0041] Step 31: Obtain the coding information and classification label data corresponding to all retired power cells in multiple sets of retired power cells. The coding information is generated based on the target information of the retired power cells.

[0042] Step 32: Randomly assign codes to multiple pieces of each coding information to obtain multiple coding bodies corresponding to multiple first power battery packs. One coding body corresponds to the recombination configuration result of one first power battery pack.

[0043] Step 33: Determine the battery capacity, discharge rate, service life, and internal resistance of the battery corresponding to the classification label data of each target retired power cell, and calculate the first fitness corresponding to each coding body based on the adaptation function constructed from the battery capacity, discharge rate, service life, and internal resistance. When constructing the adaptation function, the weights corresponding to the battery capacity, discharge rate, service life, and internal resistance are dynamically updated based on the battery demand parameters.

[0044] In this embodiment, the adaptation function is constructed based on the capacity, discharge rate, life, internal resistance of the battery pack, order demand constraints, and recombination battery constraints. Among them, the function for capacity matching is: CapacityMatch = |C order -C group |, where C order is the required capacity of the battery pack in the order, and C group is the capacity of the recombined power battery pack. Through capacity matching, it is ensured that the capacity of the allocated battery pack is close to the required capacity of the order, avoiding affecting the order performance due to insufficient capacity or causing resource waste due to excessive capacity; the function for discharge rate matching is: DischargeRateMatch = |D order -D group |, where D order is the required discharge rate of the battery pack in the order, and D group is the discharge rate of the recombined power battery pack. Through discharge rate matching, the discharge rate of the recombined battery pack is made to match the required discharge rate of the order, ensuring that the output power in the actual application of the order meets the requirements; the life matching function is: LifeMatch = |L order -SOH group |, where L order is the life requirement of the battery pack in the order, and SOH group is the health state of the recombined power battery pack. Through life matching, it is ensured that the battery pack maintains stable performance during the usage cycle to meet the service life requirements of the order. Battery packs with a high life matching degree are preferentially allocated to orders with higher life requirements; the internal resistance matching function is: ResistanceMatch = |R order -Rgroup |,R order R is the requirement for the internal resistance of the battery pack in the order group is the built-in average value of the reorganized power battery pack. By controlling the built-in range of the battery pack, the balance of the internal groups of the battery pack is ensured, the discharge efficiency and the use safety are improved; in this embodiment, the corresponding order requirement constraint is set to ensure that the capacity, discharge rate, life and internal resistance of each allocated power battery pack meet the minimum requirements of the order; the constraint for the reorganized battery is set to ensure that all the batteries entering the reorganization process must come from the set of retired power battery cells in the same group in the consistency sorting module, so as to ensure that the batteries in the same group are highly similar in performance parameters.

[0045] It should be noted that according to different application scenarios and order requirements, different target weights are set. For orders that require high-power output, the weight of discharge rate matching is increased, so as to give priority to ensuring the discharge capacity of the reorganized battery pack; for orders with higher requirements for life, the weight of life matching is increased, and the battery cells in better health states are preferentially allocated, so as to reorganize a power battery pack with a long service life.

[0046] Step 34: Generate the first recombination configuration information from multiple coding bodies and the first fitness value corresponding to each coding body.

[0047] Through steps 31 to 34, the generation of the initial solution using the improved artificial bee colony algorithm is realized, providing a data basis for solving the optimal power battery pack.

[0048] In some embodiments, according to the difference between the first fitness value and the target fitness value, iterative local search, neighborhood optimization and global search based on the improved artificial bee colony algorithm are performed on the first recombination configuration information, which is realized through the following steps:

[0049] Step 41: Based on the artificial bee colony algorithm, generate the first neighborhood solutions for all the coding bodies in the first recombination configuration information, and randomly permute the target retired power battery cells corresponding to the randomly selected first neighborhood solutions to generate multiple second coding bodies, where the randomly permuted target retired power battery cells are the target retired power battery cells in the same set of retired power battery cells, and the second coding bodies are used to represent the configuration results of the reorganized power battery pack in the local search stage.

[0050] In this embodiment, the artificial bee colony algorithm is used to simulate the honey collection of bees, which is defined as the employed bee stage. Thus, a new configuration scheme is generated within the neighborhood of the coding body corresponding to each initial solution, that is, the first neighborhood solution is generated. By fine-tuning the battery cell combination in the configuration scheme (the first neighborhood solution) and dynamically adjusting the neighborhood range, the fitness value corresponding to the first neighborhood solution is improved. In this embodiment, by adaptively adjusting the neighborhood size, the local optimum problem caused by too small a neighborhood range is avoided, and at the same time, the search range for high-quality solutions is expanded. In this embodiment, when fine-tuning the first neighborhood solution, it can be to replace some of the target retired power battery cells of a first neighborhood solution. For example, for the battery cells of a power battery pack corresponding to a first neighborhood solution being battery cell 1, battery cell 2, battery cell 3, battery cell 4, battery cell 5, and battery cell 6, after fine-tuning, battery cell 1 and battery cell 6 are replaced with battery cell 7 and battery cell 9 respectively, thus forming a new power battery pack. It can also be to replace some of the first neighborhood solutions. For example, the first neighborhood solution a and the first neighborhood solution b are respectively replaced with the first neighborhood solution f and the first neighborhood solution g. It can be understood that the replacement of the neighborhood solution is the replacement of the relevant power battery pack. In this embodiment, after fine-tuning the first neighborhood solution, the power battery pack corresponding to the second coding body is generated accordingly.

[0051] Step 42, after selecting the second coding bodies with fitness values greater than the fitness threshold from multiple second coding bodies according to the fitness corresponding to the second coding body, determine the candidate coding bodies among the selected multiple candidate coding bodies that do not belong to the preset optimal coding body pool, generate the second neighborhood solutions for the determined candidate coding bodies, and use all the second neighborhood solutions and the candidate coding bodies among the multiple candidate coding bodies that belong to the optimal coding body pool as the third coding body, where all the coding bodies stored in the optimal coding body pool are the optimal coding bodies obtained based on the memory mechanism, and the third coding body is used to represent the configuration result of the power battery pack recombined in the neighborhood optimization stage.

[0052] In this embodiment, the observing bee stage is simulated, and according to the fitness value of each second coding body, the observing bee selects the solution with a higher fitness (corresponding to the second coding body with a fitness value greater than the fitness threshold) and conducts a deeper search in its neighborhood. In this embodiment, a fitness threshold determination mechanism is added in the observing bee stage. The second coding bodies with fitness values lower than the fitness threshold are directly excluded, and the neighborhoods of the high-fitness solutions are concentrated for optimization. At the same time, a memory pool mechanism is introduced in the neighborhood search, and the historically excellent coding bodies are stored in the optimal coding body pool to prevent good coding bodies from being accidentally eliminated in the neighborhood search, so as to improve the accuracy of the matching scheme.

[0053] Step 43: For the target retired power battery cells corresponding to the randomly selected third coding body, perform a random global search within the set global search range to generate candidate coding bodies that meet the local optimal escape mechanism, and use the generated multiple candidate coding bodies as the candidate recombination configuration information corresponding to the current iteration. The global search range is determined by introducing a random factor and a large-range search mechanism. The local optimal escape mechanism is used to indicate that when the generated candidate coding body is the optimal solution, the global search range stops expanding.

[0054] In this embodiment, in the simulated scout bee stage, after the coding bodies with smaller fitness values are eliminated, the scout bees perform a global search by dynamically generating new coding bodies to maintain the diversity of the solution space and prevent the algorithm from falling into local optimality. In this embodiment, by introducing an escape mechanism based on adaptive control, the global search range of the scout bees can be adaptively adjusted according to the distribution of the current solution, perform a large-range search when deviating from the global optimum, and gradually approach the coding body corresponding to the optimal power battery pack. It can be understood that the local optimal escape mechanism adopted in this embodiment is that when the generated candidate coding body falls into a local optimal solution, it will change a part of the current solution to help the algorithm jump out of the local optimal solution and explore a broader solution space (that is, expand the global search range).

[0055] Step 44: Repeat the step of generating the corresponding candidate recombination configuration information until the difference between the fitness of the candidate recombination configuration information and the target fitness is not greater than the preset threshold, and use the candidate recombination configuration information with the difference not greater than the preset threshold as the recombination configuration information.

[0056] In this embodiment, multiple iterations of optimization are performed to continuously update the fitness value of the corresponding coding body, and gradually converge to the optimal coding body through alternating optimization in each stage. It can be understood that the improved artificial bee colony algorithm adopted in this application embodiment improves the convergence speed and global search ability through the local optimal escape mechanism, adaptive neighborhood, and memory pool mechanism, and finally outputs the optimal matching scheme that meets the order requirements.

[0057] In some of these embodiments, the heuristic algorithm includes a single-parent genetic algorithm based on simulated annealing. For the production order demand information and recombination configuration information, a production scheduling plan is performed using the preset heuristic algorithm, which is implemented through the following steps:

[0058] Step 51: From the recombination configuration information, obtain the battery pack numbers corresponding to the target power battery packs for recombination, perform random linear coding on the multiple battery pack numbers to obtain an initial set of linear coding bodies. Each first linear coding individual in the initial set of linear coding bodies is used to represent a production scheduling plan information for recombining all target power battery packs. The production scheduling plan information includes the production line and assembly time assigned to each target power battery pack.

[0059] Step 52: Obtain the first production scheduling information corresponding to each first linear coding individual obtained by decoding the initial linear coding set, and calculate the recombination fitness corresponding to each first linear coding individual according to the assembly time of each target power battery pack extracted from the first production information. The recombination fitness is used to characterize the length of the processing time for the recombination of the target power battery pack.

[0060] Step 53: Determine whether the recombination fitness values corresponding to all the current linear coding individuals generated in the current iteration are less than a preset recombination fitness threshold. Here, the current linear coding individuals include either the first linear coding individuals or the linear coding individuals generated in the previous iteration.

[0061] Step 54: In the case where it is determined that the recombination fitness values of all the current linear coding individuals are not less than the recombination fitness threshold, perform simulated annealing operation iterations based on all the current linear coding individuals and the single-parent genetic algorithm based on simulated annealing until the recombination fitness value of the generated target linear coding individual is less than the recombination fitness threshold, and use the production scheduling information obtained by decoding all the target linear coding individuals as the target production scheduling information. Here, the simulated annealing operation includes selection, mutation, accepting a new solution based on the Metropolis criterion, and cooling.

[0062] In this embodiment, the goal of production scheduling is that the recombination schemes of the same order are completed as much as possible at the same time, the recombination scheme has the maximum benefit, and the remanufacturing cost is the minimum. The constraints of production scheduling include that each recombination scheme needs to be remanufactured according to a fixed process flow, and there is a precedence relationship between the process flows. There are professional machines for each process, and there is a corresponding buffer in front of the machines. One machine can only process one recombination scheme at the same time. The processing times of different recombination schemes on different machines are different. Different recombination schemes generally have different delivery times, and different recombination schemes have different benefits.

[0063] In this embodiment, a single-parent genetic algorithm based on simulated annealing is used for production scheduling, and the specific process is as follows:

[0064] Step 1: Initialization

[0065] In this embodiment, generate an initial linear coding body set, that is, an initial population, and set an initial temperature T_0 and a cooling rate α. Each first linear coding individual represents a recombination scheme, and determine its allocation on different production lines and the production start time.

[0066] Step 2: Fitness evaluation

[0067] In this embodiment, the fitness value of each first linear-coded individual is calculated, where the fitness value reflects the length of the machine processing time, and the goal is to minimize the machine processing time.

[0068] Step 3: Selection.

[0069] In this embodiment, the first linear-coded individuals with higher fitness values are selected as parental individuals to ensure that individuals with higher fitness have more opportunities to enter the next generation.

[0070] Step 4: Mutation operation

[0071] In this embodiment, a mutation operation is performed on the selected first linear-coded individuals. The simulated annealing algorithm controls the mutation range and probability through temperature. At a higher temperature, a larger mutation range is allowed, randomly changing the production line allocation and production start time of the individuals. As the temperature gradually decreases, the mutation range narrows, causing the algorithm to gradually converge and avoiding excessive search.

[0072] Step 5: Acceptance criterion

[0073] In this embodiment, whether the solution corresponding to the newly generated coded body by the mutation operation is accepted depends on its fitness value and the current temperature; in this embodiment, the Metropolis criterion is adopted to receive the new solution, which can accept a worse solution at a higher temperature to avoid falling into a local optimum. If the new solution is better than the current solution, the new solution is directly accepted. If the new solution is worse than the current solution, the new solution is accepted with a probability of P = e^(-∆f / T), where ∆f is the fitness difference and T is the current temperature.

[0074] Step 6: Cooling

[0075] After each iteration, the temperature is gradually decreased according to the cooling rate T = αT, causing the search range to gradually converge.

[0076] Step 7: Repeat Steps 1 to 6 until a preset stop condition (such as the number of iterations, temperature threshold, fitness threshold, etc.) is reached.

[0077] In some embodiments, multiple target index parameters corresponding to each target retired power cell are preprocessed to generate corresponding multiple standard characteristic parameters, which are achieved through the following steps:

[0078] Step 61: In the target information of each obtained target retired power cell, obtain the characteristic parameters corresponding to each target retired power cell, where the characteristic parameters include at least one of the following: cell capacity, cell internal resistance, open circuit voltage, depth of discharge, and state of charge.

[0079] In this embodiment, professional testing equipment is used to conduct detailed performance tests on each retired power battery cell, and key parameters are collected, including raw data such as capacity, internal resistance, open-circuit voltage, depth of discharge (DOD), and state of charge (SOC). Capacity reflects the energy storage capacity of the battery, internal resistance reflects the electrical conductivity of the battery, open-circuit voltage is used to preliminarily judge the electrochemical performance of the battery, and DOD and SOC reflect the dynamic performance of the battery under different discharge conditions. The data of each retired power battery cell is stored in the form of a record file, forming a complete battery characteristic data set for subsequent processing and analysis.

[0080] Step 62: Filter, denoise, add and delete missing values, and delete outliers from the characteristic parameters to generate multi-dimensional index parameters. Among them, adding and deleting missing values includes filling in missing values and deleting missing values. Filling in missing values includes filling in with one of the average value method, median method, and interpolation method. One of the following methods is used to detect the outliers to be deleted for outlier deletion: mean method, standard deviation method, box plot method, and outlier detection method.

[0081] In this embodiment, after data collection, the data set needs to be preprocessed. Since the usage history and status of different retired power battery cells are different, the collected data may have missing values and outliers, and the data noise collected will affect the accuracy of sorting. Data preprocessing is used to improve the data quality and ensure the accuracy of sorting. In this embodiment, the method for dealing with missing values will be selected to fill in or delete according to the situation. Among them, the method for filling in missing values includes using the average value, median, or interpolation method to estimate the missing values to ensure the integrity of the data set. If there are many missing data and they cannot be reasonably filled, the data samples with many missing values can be selected to be deleted. In this way, the integrity of the data is ensured, and the parameter data of each retired power battery cell is complete. In this embodiment, the methods for dealing with outliers include the mean method, standard deviation method, box plot method based on statistics, or other outlier detection methods. The significantly abnormal values are deleted or appropriately adjusted to avoid misleading the grouping results and ensure the accuracy of the data.

[0082] Step 63: Normalize the multi-dimensional index parameters and extract key features to obtain standard feature parameters corresponding to the characteristic parameters.

[0083] In this embodiment, the index parameters in different dimensions are normalized to standardize all the data within a unified scale range, which is convenient for the processing of the fuzzy clustering algorithm. The purpose of normalization is to unify the numerical ranges of different parameters into a standardized interval to eliminate the influence between different dimensions. The normalized data is more suitable for the processing of the fuzzy clustering algorithm and can improve the stability and accuracy of the clustering effect. In this embodiment, after normalizing the index parameters, feature extraction is performed, that is, after normalizing the index parameters, the key features of each battery cell are extracted from the corresponding index parameters, including indicators such as capacity, internal resistance, DOD, and SOC. By directly inputting the key feature indicators into the fuzzy clustering algorithm, it is used to characterize the health state, discharge capacity, and consistency of the battery.

[0084] Through the above steps 61 to 63, the preprocessing of various target index parameters corresponding to the target retired power battery cells is realized, which can provide the accuracy for sorting the retired power battery cells.

[0085] In some of the embodiments, the training of the consistency sorting model includes the following steps:

[0086] Step 71, obtain a preset training sample set, where the training sample set includes the characteristic parameters corresponding to multiple sample retired power battery cells.

[0087] Step 72, perform corresponding preprocessing on the characteristic parameters in the training sample set to generate standard characteristic parameters corresponding to each sample retired power battery cell.

[0088] Step 73, after selecting the initial clustering center from the standard characteristic parameters corresponding to all the sample retired power battery cells, based on the initial clustering center and the standard characteristic parameters corresponding to all the sample retired power battery cells, perform iterations of membership matrix calculation and clustering center update corresponding to the fuzzy c-means clustering algorithm until the membership matrix converges to generate a consistency sorting model.

[0089] In this embodiment, after preprocessing, normalizing, and feature extraction of the characteristic parameters corresponding to the preset training sample set, the extracted standard feature parameters corresponding to the sample retired power battery cells are input for sorting based on the fuzzy c-means clustering algorithm (FCM). Based on the FCM algorithm and the standard feature parameters of each sample retired power battery cell, the battery cells with similar performance are clustered into one category, allowing a certain degree of ambiguity; in this embodiment, at the beginning of training, the initial clustering centers are first set. The setting method of the initial clustering centers can be randomly selected or based on the preset positions of the feature means. The initial clustering centers provide reference points for the subsequent clustering process. As the algorithm iterates, the clustering centers will gradually be adjusted to more reasonable positions, thereby improving the clustering effect; then, the membership matrix is calculated. In this embodiment, the membership degree of each standard feature parameter in each category is calculated according to the distance from each standard feature parameter to the clustering center. The membership matrix records the membership degrees of all battery cells in each category. The membership degree of each standard feature parameter in each category is calculated according to the distance between each standard feature parameter and the clustering center; it can be understood that the membership matrix records the membership degrees of all battery cells in each category, and the corresponding values range from 0 to 1, indicating the membership strength of the battery cell to a certain cluster. The membership matrix enables each corresponding battery cell to belong to multiple categories with different membership degree values, thus being able to more accurately reflect the performance differences between the battery cells; during the training and sorting process, after calculating the membership matrix, the clustering centers are updated according to the membership degrees. The update method is to calculate the weighted average of all standard feature parameters in each category to generate new clustering centers. The continuous adjustment of the clustering centers makes the distribution of the battery cells gradually approach the positions where the clustering centers are located, realizing the gradual refinement of the clustering results; thereafter, iterative updates are performed, that is, the steps of calculating the membership matrix and updating the clustering centers are repeated until the change in the membership matrix is less than the preset convergence threshold or the maximum number of iterations is reached. At this time, the clustering result is regarded as converged, and the final grouping result is obtained. Through the above repeated iterations, the stability and accuracy of the clustering are ensured, and the sorting result reaches the optimal.

[0090] In some of these embodiments, obtaining the battery demand parameters is achieved through the following steps:

[0091] Step 81, obtaining historical battery demand order data, where the historical battery demand order data includes order parameters and time parameters. The order parameters include the quantity of the battery demand order, the capacity of the required battery pack, the discharge rate of the battery pack, and the service life of the battery pack. The time parameters at least include the order timestamp.

[0092] Step 82: Clean the order parameters and time parameters corresponding to the historical battery demand order data to generate the first order data. The data cleaning includes at least one of the following: using the methods of mean filling, interpolation or elimination to process the missing values of the order parameters and time parameters; using one of the mean method, standard deviation method and box plot method to detect the outliers of the order parameters and time parameters, and deleting the detected outliers.

[0093] Step 83: Process the first order data using feature engineering to generate order feature data, and use the pre-trained order demand prediction model to process the order feature data to predict and output the battery demand parameters. The order demand prediction model is a classification prediction model trained based on the CART algorithm, and is trained to predict and output the order demand corresponding to the input first feature data.

[0094] In some alternative embodiments, the order demand prediction model is trained using the following steps:

[0095] Step 1: Collect historical battery demand order data

[0096] In this embodiment, the battery demand orders in the past period of time are collected. The data of the battery demand orders includes order parameters such as the quantity of battery demand orders, the capacity of the battery pack required, the discharge rate of the battery pack, and the service life of the battery pack, and time parameters including the timestamp of the order and other associated time features (such as season, month, week number, etc.).

[0097] Step 2: Data cleaning

[0098] In this embodiment, the battery demand order data is preprocessed to process missing values, outliers and duplicate values to ensure the integrity and consistency of the data. Among them, for missing values, mean filling, interpolation or elimination can be used; for outliers, statistical methods or box plot methods are used for detection and processing.

[0099] Step 3: Feature engineering

[0100] In this embodiment, for the first order data, feature engineering is used to extract features and create new features to improve the prediction effect of order demand prediction. For example, new time features (such as the beginning of the month, the middle of the month, the end of the month, etc.) are generated according to time, and information such as the type and application scenario of the order can also be extracted as corresponding input features.

[0101] Step 4: Data splitting

[0102] In this embodiment, the order feature data generated by feature engineering is divided into a training set, a validation set and a test set.

[0103] Step 5: Model training

[0104] In this embodiment, the model training process is to train a decision tree model through input features and historical battery demand order data, so as to learn the relationship between features and order demand quantities.

[0105] Step 6, Feature selection

[0106] In this embodiment, according to the division criterion of the CART algorithm, select the feature that can most reduce the mean square error as the splitting point, and construct a tree structure layer by layer. Among them, the feature selection process will find the most important feature combination in the training data so that the model can generate accurate prediction values.

[0107] Step 7, Initial node splitting

[0108] In this embodiment, starting from the root node, select the best feature and splitting point through the criterion of minimizing the mean square error, and divide the data set into two subsets. Each subset forms a new node, and these nodes will become the branches of the tree.

[0109] Step 8, Node recursive splitting

[0110] In this embodiment, repeat the above process for each new node, continue to find the best splitting feature and splitting point to minimize the mean square error of each subset. The CART algorithm continuously splits each node until the stopping condition is reached.

[0111] Step 9, Node stopping splitting

[0112] In this embodiment, the recursive process of the CART regression tree ends when the stopping condition is met: the reduction in the mean square error after node splitting is less than a preset threshold.

[0113] Step 10, Calculate the mean square error of the node

[0114] In this embodiment, for each subset after splitting, calculate its mean square error, and perform a weighted average on the mean square errors of all nodes to obtain the overall mean square error of this split.

[0115] Step 11, Generate leaf nodes

[0116] In this embodiment, when the stopping condition is reached, the decision tree stops further splitting. It can be understood that in the regression task, the output value of each leaf node is usually the average order demand quantity of the data of this node, which also represents the predicted value of this node.

[0117] Step 12, Post-pruning processing

[0118] It is understandable that decision trees are prone to overfitting, especially when the data has a large amount of noise. After the tree is constructed, pruning is used to simplify the tree structure. Post-pruning is usually based on cost complexity pruning. By calculating the cost complexity at the leaf nodes and selecting the pruning scheme with the lowest cost, overfitting is reduced. Among them, the node complexity cost is the mean squared error of the node plus the complexity penalty term. After training is completed, the performance of the decision tree model is evaluated to ensure its accurate prediction of order demand and generalization ability.

[0119] Step 13, Model evaluation. Use the validation set to evaluate the prediction effect of the model and calculate the R² score of the validation set.

[0120] Step 14, Hyperparameter tuning. After model evaluation, further improve the model performance through hyperparameter tuning. Decision tree hyperparameters include: maximum depth, minimum number of samples for splitting, minimum number of samples in leaf nodes, etc. Random search is used to select the best parameter combination.

[0121] Step 15, Cross-validation. Perform cross-validation on the training set. Divide the data into multiple subsets, conduct multiple trainings and validations, and take the average performance metric as the final result.

[0122] Step 16, Model testing. Use the test set data to conduct the final test on the decision tree model and calculate the R² evaluation metric.

[0123] After training the order demand prediction model, the trained decision tree model will be deployed and combined with real-time order data for order demand prediction; it is understandable that the trained order demand prediction model will also be monitored, that is, regularly monitor the prediction error of the decision tree model and analyze the change in model performance. If the prediction error increases significantly, it indicates that the market environment has changed and the existing order demand prediction model is no longer suitable and needs to be updated. The update is carried out by incremental learning or retraining. Among them, for slightly changing scenarios, the incremental learning method is adopted to update the model parameters with a small amount of new data; for significantly changing situations, the model is retrained with new historical order data to ensure the accuracy of the model; during the process of using the order demand prediction model, a feedback learning mechanism is also adopted, that is, by collecting the difference between the actual order demand and the model prediction value, the model is continuously optimized through the feedback mechanism; data feature optimization processing is also adopted, and the input features are continuously optimized through feature selection and feature importance evaluation to improve the prediction effect of the model.

[0124] In this embodiment, for the target index parameters corresponding to the external characteristic parameters (for example: types of retired power battery packs, brands of retired power battery packs, types of retired power battery packs), it refers to the relatively important index parameters selected from the preset index parameters and assigned to the corresponding target retired power cell groups when the power batteries are retired and determined to be used for recombination. When receiving the target retired power cell groups to be warehoused, the target information of each target retired power cell group already has the target index parameters corresponding to the external characteristic parameters; while the target index parameters representing the internal characteristic parameters can be measured when the power batteries are retired into the target retired power cell groups, or can also be measured when warehousing into the warehouse.

[0125] This embodiment also provides a matching and recombination device for retired power batteries. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, terms such as "module", "unit", "sub-unit", etc. can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0126] Figure 3 It is a structural block diagram of a matching and recombination device for retired power batteries according to an embodiment of the present application. As Figure 3 shown, the device includes an acquisition module 31, a sorting module 32, a planning module 33, and a processing module 34, where

[0127] The acquisition module 31 is used to obtain the target information of a plurality of target retired power cells to be recombined. Among them, each target information includes a variety of target index parameters, and the target index parameters are used to represent a characteristic parameter of the corresponding target retired power cell, and the characteristic parameter is at least used to represent the current storage state of the target retired power cell.

[0128] The sorting module 32 is coupled to the acquisition module 31. After preprocessing the various target index parameters corresponding to each target retired power cell to generate corresponding various standard characteristic parameters, it uses the trained consistency sorting model to process the various standard characteristic parameters corresponding to each target retired power cell to obtain the classification label data corresponding to the sorted multiple retired power cell sets. Among them, the classification label data includes a recombination adaptation level used to represent the priority of the retired power cell set applied to the cascade scenario during the recombination process. The consistency sorting model is a machine learning model trained based on the fuzzy c-means clustering algorithm and is trained to generate the classification label data corresponding to the corresponding retired power cell set according to the input battery characteristic parameters.

[0129] The planning module 33, which is coupled to the sorting module 32, is configured to perform iterative battery recombination configuration planning by using an improved artificial bee colony algorithm based on the pre-acquired battery demand parameters, multiple sets of retired power battery cells, and the corresponding classification label data, and generate corresponding recombination configuration information, where the recombination configuration information includes the target information corresponding to the retired power battery cells corresponding to the recombined power battery pack.

[0130] The processing module 34, which is coupled to the planning module 33, is configured to perform production scheduling planning on the production order demand information and the recombination configuration information by using a preset heuristic algorithm after determining the production order demand information corresponding to the battery demand parameters, and perform power battery recombination according to the generated target production scheduling planning information to obtain multiple recombined power batteries.

[0131] In some embodiments, the planning module 33 further includes:

[0132] A configuration unit, configured to perform battery recombination configuration by using an improved artificial bee colony algorithm according to the classification label data corresponding to multiple sets of retired power battery cells, and generate first recombination configuration information, where the first recombination configuration information includes the target retired power battery cells sorted for the first power battery pack, and the first power battery pack is used to represent the currently recombined power battery pack;

[0133] An acquisition unit, which is coupled to the configuration unit, is configured to acquire the first fitness corresponding to each first power battery pack from the first recombination configuration information, where the first fitness is calculated by an adaptation function constructed based on the classification label data corresponding to the target retired power battery cells assigned to each first power battery pack, and is used to characterize the performance excellence degree of the first power battery pack;

[0134] A determination unit, which is coupled to the acquisition unit, is configured to determine the target fitness corresponding to the required battery pack corresponding to the battery demand parameters, and perform iterative local search, neighborhood optimization, and global search based on the improved artificial bee colony algorithm on the first recombination configuration information according to the difference between the first fitness and the target fitness until the difference is not greater than a preset threshold, and generate recombination configuration information, where in the global search stage, the improved artificial bee colony algorithm introduces a dynamic neighborhood adjustment and a local optimal escape mechanism.

[0135] In some of these embodiments, the configuration unit is further configured to obtain the coding information and classification label data corresponding to all the retired power cells in a plurality of sets of retired power cells, where the coding information is generated according to the target information of the retired power cells; randomly assign codes to each of the plurality of coding information to obtain a plurality of coding bodies corresponding to a plurality of first power battery packs, where one coding body corresponds to the recombination configuration result of one first power battery pack; determine the battery capacity, discharge rate, service life, and internal resistance of the battery corresponding to the classification label data of each target retired power cell, and calculate the first fitness corresponding to each coding body based on an adaptation function constructed based on the battery capacity, discharge rate, service life, and internal resistance. When constructing the adaptation function, the weights corresponding to the battery capacity, discharge rate, service life, and internal resistance are dynamically updated based on the battery demand parameters; generate the plurality of coding bodies and the first fitness corresponding to each coding body into first recombination configuration information.

[0136] In some of these embodiments, the determination unit is further configured to generate first neighborhood solutions for all the coding bodies in the first recombination configuration information based on the artificial bee colony algorithm, and randomly permute the target retired power cells corresponding to the randomly selected first neighborhood solutions to generate a plurality of second coding bodies, where the randomly permuted target retired power cells are the target retired power cells in the same set of retired power cells, and the second coding bodies are used to represent the configuration results of the power battery packs recombined in the local search stage; after selecting the second coding bodies with fitness values greater than the fitness threshold from the plurality of second coding bodies according to the fitness corresponding to the second coding bodies, determine the candidate coding bodies that do not belong to the preset optimal coding body pool among the selected plurality of candidate coding bodies, generate second neighborhood solutions for the determined candidate coding bodies, and use all the second neighborhood solutions and the candidate coding bodies that belong to the optimal coding body pool among the plurality of candidate coding bodies as third coding bodies, where all the coding bodies stored in the optimal coding body pool are the optimal coding bodies obtained based on the memory mechanism, and the third coding bodies are used to represent the configuration results of the power battery packs recombined in the neighborhood optimization stage; perform a random global search on the target retired power cells corresponding to the randomly selected third coding bodies within the set global search range to generate candidate coding bodies that satisfy the local optimal escape mechanism, and use the generated plurality of candidate coding bodies as the candidate recombination configuration information corresponding to the current iteration, where the global search range is determined by introducing random factors and a large-range search mechanism, and the local optimal escape mechanism is used to represent that when the generated candidate coding body is the optimal solution, the global search range stops expanding; repeat the step of generating the corresponding candidate recombination configuration information until the difference between the fitness of the candidate recombination configuration information and the target fitness is not greater than the preset threshold, and use the candidate recombination configuration information with the difference not greater than the preset threshold as the recombination configuration information.

[0137] In some of these embodiments, the heuristic algorithm includes a single-parent genetic algorithm based on simulated annealing, and the processing module 34 further includes:

[0138] An encoding unit, configured to obtain, from the recombination configuration information, the battery pack numbers corresponding to the target power battery packs of the recombination configuration, perform random linear encoding on the multiple battery pack numbers to obtain an initial set of linear encoded bodies, where each first linear encoded individual in the initial set of linear encoded bodies is used to represent a production scheduling plan information for recombining all the target power battery packs, and the production scheduling plan information includes the production line and assembly time allocated to each target power battery pack;

[0139] A calculation unit, coupled to the encoding unit, configured to obtain the first production scheduling plan information corresponding to each first linear encoded individual obtained by decoding the initial linear encoding set, and calculate the recombination fitness corresponding to each first linear encoded individual according to the assembly time of each target power battery pack extracted from the first production information, where the recombination fitness is used to represent the length of the processing time for recombining the target power battery packs;

[0140] A judgment unit, coupled to the calculation unit, configured to judge whether the recombination fitness values corresponding to all the current linear encoded individuals generated in the current iteration are less than a preset recombination fitness threshold, where the current linear encoded individuals include one of the first linear encoded individuals and the linear encoded individuals generated in the previous iteration;

[0141] An operation unit, coupled to the judgment unit, configured to, when it is judged that the recombination fitness values of all the current linear encoded individuals are not less than the recombination fitness threshold, perform simulated annealing operation iterations based on all the current linear encoded individuals and the single-parent genetic algorithm based on simulated annealing until the recombination fitness value of the generated target linear encoded individual is less than the recombination fitness threshold, and use the production scheduling plan information obtained by decoding all the target linear encoded individuals as the target production scheduling plan information, where the simulated annealing operation includes selection, mutation, receiving a new solution based on the Metropolis criterion, and cooling.

[0142] In some of these embodiments, the sorting module 32 further includes:

[0143] An extraction unit, configured to obtain, from the target information of each target retired power cell, the characteristic parameters corresponding to each target retired power cell, where the characteristic parameters include at least one of the following: cell capacity, cell internal resistance, open circuit voltage, depth of discharge, and state of charge;

[0144] A generation unit, coupled to the extraction unit, is configured to filter, denoise, add and delete missing values, and delete outliers from the characteristic parameters to generate multi-dimensional index parameters. Among them, adding and deleting missing values includes filling in missing values and deleting missing values. Filling in missing values includes filling in using one of the mean method, median method, and interpolation method. Detecting outliers to be deleted for outlier deletion adopts one of the following methods: mean method, standard deviation method, box plot method, and outlier detection method;

[0145] A processing unit, coupled to the generation unit, is configured to perform normalization processing and key feature extraction on the multi-dimensional index parameters to obtain standard feature parameters corresponding to the characteristic parameters.

[0146] In some embodiments, the retired power battery matching and recombination device further obtains a preset training sample set. Among them, the training sample set includes characteristic parameters corresponding to multiple sample retired power battery cells; preprocesses the characteristic parameters in the training sample set to generate standard feature parameters corresponding to each sample retired power battery cell; after selecting initial cluster centers from the standard feature parameters corresponding to all sample retired power battery cells, based on the initial cluster centers and the standard feature parameters corresponding to all sample retired power battery cells, perform iterations of membership matrix calculation and cluster center update corresponding to the fuzzy c-means clustering algorithm until the membership matrix converges to generate a consistency sorting model.

[0147] In some embodiments, the planning module 33 is further configured to obtain historical battery demand order data. Among them, the historical battery demand order data includes order parameters and time parameters. The order parameters include the quantity of battery demand orders, the capacity of the required battery pack, the discharge rate of the battery pack, and the service life of the battery pack. The time parameters at least include the order timestamp; perform data cleaning on the order parameters and time parameters corresponding to the historical battery demand order data to generate first-order data. Among them, data cleaning at least includes one of the following: using the mean filling, interpolation, or deletion method to process the missing values of the order parameters and time parameters; using one of the mean method, standard deviation method, and box plot method to detect outliers of the order parameters and time parameters and delete the detected outliers; use feature engineering to process the first-order data to generate order feature data, and use the trained order demand prediction model to process the order feature data to predict and output battery demand parameters. Among them, the order demand prediction model is a classification prediction model trained based on the CART algorithm and is trained to predict and output the order demand corresponding to the input first feature data according to the input first feature data.

[0148] This embodiment also provides a service system, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0149] Optionally, the above-mentioned reconfiguration system may further include a transmission device and an input / output device, wherein the transmission device is connected to the above-mentioned processor, and the input / output device is connected to the above-mentioned processor.

[0150] Optionally, in this embodiment, the above-mentioned processor may be configured to execute the following steps through a computer program:

[0151] S1. Obtain the target information of multiple target retired power battery cells to be reconfigured. Each target information includes multiple target index parameters, and the target index parameters are used to characterize a characteristic parameter of the corresponding target retired power battery cell. The characteristic parameter is at least used to characterize the current storage state of the target retired power battery cell.

[0152] S2. After preprocessing the multiple target index parameters corresponding to each target retired power battery cell to generate corresponding multiple standard feature parameters, use the trained consistency sorting model to process the multiple standard feature parameters corresponding to each target retired power battery cell to obtain the classification label data corresponding to the sorted multiple retired power battery cell sets. The classification label data includes a reconfiguration adaptation level used to characterize the priority of the retired power battery cell set applied to the cascade scenario during the reconfiguration process. The consistency sorting model is a machine learning model trained based on the fuzzy c-means clustering algorithm and is trained to generate the classification label data corresponding to the corresponding retired power battery cell set according to the input battery feature parameters.

[0153] S3. Based on the previously obtained battery demand parameters, multiple retired power battery cell sets, and the corresponding classification label data, use the improved artificial bee colony algorithm to perform iterative battery reconfiguration configuration planning to generate corresponding reconfiguration configuration information. The reconfiguration configuration information includes the target information corresponding to the retired power battery cells corresponding to the reconfigured power battery pack.

[0154] S4. After determining the production order demand information corresponding to the battery demand parameters, use a preset heuristic algorithm to perform production scheduling planning on the production order demand information and the reconfiguration configuration information, and perform power battery reconfiguration according to the generated target production scheduling planning information to obtain multiple reconfigured power batteries.

[0155] It should be noted that the specific examples in this embodiment can refer to the examples described in the above-mentioned embodiments and optional implementation manners, and will not be repeated here.

[0156] In addition, in combination with the retired power battery matching and reconfiguration method in the above-mentioned embodiments, an embodiment of the present application can provide a storage medium to implement. A computer program is stored on the storage medium; when the computer program is executed by a processor, any one of the retired power battery matching and reconfiguration methods in the above-mentioned embodiments is implemented.

[0157] Those skilled in the art should understand that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0158] The above embodiments only represent several implementation manners of the present application, and the description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for matching and reorganizing retired power batteries, characterized in that: include: Obtaining target information of a plurality of target retired power cells to be reassembled, wherein each of the target information includes a plurality of target indicator parameters, the target indicator parameters are used to characterize a characteristic parameter of the corresponding target retired power cell, and the characteristic parameter is at least used to characterize a current storage state of the target retired power cell; After preprocessing the multiple target indicator parameters corresponding to each of the target retired power cells to generate the corresponding multiple standard feature parameters, the multiple standard feature parameters corresponding to each of the target retired power cells are processed using the trained consistency sorting model to obtain classification label data corresponding to the selected multiple retired power cell sets, wherein the classification label data includes a reorganization adaptation level for characterizing the priority of the retired power cell set in the echelon scenario during the reorganization process, and the consistency sorting model is a machine learning model trained based on a fuzzy c-means clustering algorithm, and is trained to generate classification label data corresponding to the corresponding retired power cell set according to the input battery feature parameters; Based on the pre-acquired battery demand parameters, the multiple retired power cell sets and the corresponding classification label data, the improved artificial bee colony algorithm is used to iterate the battery reorganization configuration planning to generate corresponding reorganization configuration information, wherein the reorganization configuration information includes the target information corresponding to the retired power cell corresponding to the reorganized power battery group; After determining the production order demand information corresponding to the battery demand parameter, the production order demand information and the reorganization configuration information are scheduled using a preset heuristic algorithm, and the power battery is reorganized according to the generated target production scheduling information to obtain a plurality of reorganized power batteries, wherein based on the pre-acquired battery demand parameter, the plurality of retired power cell sets and the corresponding classification label data, the battery reorganization configuration planning is iterated using an improved artificial bee colony algorithm to generate the corresponding reorganization configuration information, including: According to the classification label data corresponding to the plurality of retired power cell sets, the improved artificial bee colony algorithm is used to perform battery reorganization configuration to generate first reorganization configuration information, wherein the first reorganization configuration information includes the target retired power cells selected for the first power battery group, and the first power battery group is used to represent the power battery group currently reorganized and configured; Acquire a first fitness corresponding to each of the first power battery packs from the first reorganization configuration information, wherein the first fitness is calculated based on an adaptation function constructed based on the classification label data corresponding to the target retired power cells allocated to each of the first power battery packs, and is used to characterize the performance excellence of the first power battery packs; Determine the target fitness corresponding to the demand battery group corresponding to the battery demand parameter, and according to the difference between the first fitness and the target fitness, perform local search, neighborhood optimization and global search iterations based on the improved artificial bee colony algorithm on the first reorganization configuration information until the difference is no more than a preset threshold, and generate the reorganization configuration information, wherein, in the global search stage, the improved artificial bee colony algorithm introduces dynamic neighborhood adjustment and local optimal escape mechanism.

2. The method according to claim 1, characterized in that According to the classification label data corresponding to the plurality of retired power cell sets, the improved artificial bee colony algorithm is used to perform battery reconfiguration to generate first reconfiguration configuration information, including: Acquire the coding information and the classification label data corresponding to all the retired power cells in the plurality of retired power cell sets, wherein the coding information is generated according to the target information of the retired power cells; Randomly assigning codes to each of the plurality of coding information to obtain a plurality of coding bodies corresponding to the plurality of the first power battery packs, wherein one coding body corresponds to a reorganization configuration result of one of the first power battery packs; Determine the battery capacity, discharge rate, life span and battery internal resistance corresponding to the classification label data corresponding to each of the target retired power cells, and calculate the first fitness corresponding to each of the encoded bodies based on the adaptation function constructed based on the battery capacity, the discharge rate, the life span and the battery internal resistance, wherein when constructing the adaptation function, the weights corresponding to the battery capacity, the discharge rate, the life span and the battery internal resistance are dynamically updated based on the battery demand parameters; The plurality of encoding bodies and the first fitness corresponding to each encoding body are generated as the first recombinant configuration information.

3. The method according to claim 2, characterized in that According to the difference between the first fitness and the target fitness, the first reorganization configuration information is iterated by local search, neighborhood optimization and global search based on the improved artificial bee colony algorithm, including: Based on the artificial bee colony algorithm, a first neighborhood solution is generated for all the code bodies in the first reorganization configuration information, and the target retired power cells corresponding to the randomly selected first neighborhood solutions are randomly replaced to generate multiple second code bodies, wherein the randomly replaced target retired power cells are the target retired power cells in the same set of retired power cells, and the second code bodies are used to characterize the configuration results of the power battery group reorganized in the local search phase; After selecting the second coding body whose fitness value is greater than the fitness threshold from the plurality of second coding bodies according to the fitness corresponding to the second coding body, determining the candidate coding body that does not belong to the preset optimal coding body pool among the selected plurality of candidate coding bodies, and performing second neighborhood solution generation on the determined candidate coding body, and using all the second neighborhood solutions and the candidate coding body that belongs to the optimal coding body pool among the plurality of candidate coding bodies as the third coding body, wherein all the coding bodies stored in the optimal coding body pool are optimal coding bodies obtained based on the memory mechanism, and the third coding body is used to characterize the configuration result of the power battery pack reorganized in the neighborhood optimization stage; For the target retired power cell corresponding to the randomly selected third code body, a random global search is performed within a set global search range to generate a candidate code body that satisfies the local optimal escape mechanism, and the generated multiple candidate code bodies are used as candidate reorganization configuration information corresponding to the current iteration, wherein the global search range is determined by introducing random factors and a large-scale search mechanism, and the local optimal escape mechanism is used to indicate that when the generated candidate code body is the optimal solution, the global search range is stopped from being expanded; Repeat the step of generating the corresponding candidate recombination configuration information until the difference between the fitness of the candidate recombination configuration information and the target fitness is not greater than a preset threshold, and use the candidate recombination configuration information whose difference is not greater than the preset threshold as the recombination configuration information.

4. The method according to claim 1, characterized in that The heuristic algorithm includes a parthenogenetic algorithm based on simulated annealing, and the production order demand information and the reorganization configuration information are subjected to production scheduling using a preset heuristic algorithm, including: From the reorganization configuration information, obtain the battery pack number corresponding to the target power battery pack of the reorganization configuration, perform random linear coding on the multiple battery pack numbers, and obtain an initial linear coding body set, wherein each first linear coding individual in the initial linear coding body set is used to represent a production scheduling information for reorganizing all the target power battery packs, and the production scheduling information includes the production line and assembly time allocated to each target power battery pack; Obtaining first production scheduling information corresponding to each first linear code individual obtained by decoding the initial linear code set, and calculating the reorganization fitness corresponding to each first linear code individual according to the assembly time of each target power battery pack extracted from the first production scheduling information, wherein the reorganization fitness is used to characterize the processing time length for reorganizing the target power battery pack; Determine whether the recombination fitness values ​​corresponding to all current linear coding individuals generated by completing the current iteration are less than a preset recombination fitness threshold, wherein the current linear coding individual includes one of the first linear coding individual and the linear coding individual generated by completing the previous iteration; In the case where it is determined that the recombination fitness values ​​of all the current linear coding individuals are not less than the recombination fitness threshold, a simulated annealing operation is iterated based on all the current linear coding individuals and the parthenogenetic algorithm based on simulated annealing until the recombination fitness value of the generated target linear coding individual is less than the recombination fitness threshold, and the production scheduling information obtained by decoding all the target linear coding individuals is used as the target production scheduling information, wherein the simulated annealing operation includes selection, mutation, receiving a new solution based on the Metropolis criterion, and cooling.

5. The method according to claim 1, characterized in that: Preprocessing the multiple target indicator parameters corresponding to each of the target retired power cells to generate corresponding multiple standard feature parameters, including: In the target information of each of the target retired power cells obtained, characteristic parameters corresponding to each of the target retired power cells are obtained, wherein the characteristic parameters include at least one of the following: cell capacity, cell internal resistance, open circuit voltage, depth of discharge, and state of charge; The characteristic parameters are filtered, denoised, missing values ​​are added and deleted, and outliers are deleted to generate multi-dimensional indicator parameters, wherein the missing value addition and deletion include missing value filling and missing value removal, the missing value filling includes filling by using one of the average method, the median method and the interpolation method, and the outlier deletion uses one of the following methods to detect the deleted outliers: mean method, standard deviation method, box plot method, outlier detection method; The multi-dimensional index parameters are normalized and key features are extracted to obtain the standard feature parameters corresponding to the characteristic parameters.

6. The method according to claim 1, characterized in that Training of the consistency sorting model includes: Obtaining a preset training sample set, wherein the training sample set includes the characteristic parameters corresponding to a plurality of sample retired power cells; Preprocess the characteristic parameters in the training sample set accordingly to generate the standard characteristic parameters corresponding to each of the sample retired power cells; After selecting the initial cluster center from the standard characteristic parameters corresponding to all the sample retired power cells, based on the initial cluster center and the standard characteristic parameters corresponding to all the sample retired power cells, the membership matrix calculation and cluster center update corresponding to the fuzzy c-means clustering algorithm are iterated until the membership matrix converges to generate the consistency sorting model.

7. The method according to claim 1, characterized in that Get battery requirement parameters, including: Acquire historical battery demand order data, wherein the historical battery demand order data includes order parameters and time parameters, the order parameters include battery demand order quantity, required battery pack capacity, battery pack discharge rate, and battery pack life span, and the time parameters include at least an order timestamp; Performing data cleaning on the order parameters and the time parameters corresponding to the historical battery demand order data to generate first order data, wherein the data cleaning includes at least one of the following: processing missing values ​​of the order parameters and the time parameters by using a mean filling, interpolation or elimination method; detecting abnormal values ​​of the order parameters and the time parameters by using one of a mean method, a standard deviation method and a box plot method, and deleting the detected abnormal values; The first order data is processed by feature engineering to generate order feature data, and the order feature data is processed by a trained order demand prediction model to predict and output the battery demand parameters, wherein the order demand prediction model is a classification prediction model trained based on the CART algorithm, and is trained to predict and output the order demand corresponding to the first feature data based on the input first feature data.

8. A service system, comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps of the retired power battery matching and reorganization method according to any one of claims 1 to 7.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the retired power battery matching and reorganization method according to any one of claims 1 to 7 is implemented.

Images