Detired battery dynamic disassembly path decision-making method based on deep search

Abstract

The invention relates to the technical field of power battery disassembly and recovery, and discloses a decommissioned battery dynamic disassembly path decision-making method based on deep search. Constructing a disassembly path based on battery disassembly process steps, and constructing a graph matrix code according to the disassembly path; obtaining a machine state, and inputting the machine state and the graph matrix code into the constructed reinforcement learning model to obtain a disassembly weight; obtaining a first optimized disassembling path according to the disassembling weight in combination with deep search, performing disassembling according to the first optimized disassembling path, obtaining battery information, obtaining a second optimized disassembling path according to the battery pack information in combination with the reinforcement learning model and deep search, and distributing machines according to the second optimized disassembling path to complete disassembling of the retired battery. The problem that an existing disassembling path cannot be adjusted in real time is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power battery disassembly and recycling, and in particular to a method for dynamic disassembly path decision of retired batteries based on deep search. Background Art

[0002] With the rapid development of the global electric vehicle industry, the production and use of power batteries have surged, and the disassembly and recycling of retired batteries have become increasingly prominent. How to efficiently and safely disassemble retired batteries and maximize the recovery of valuable materials has become a technical problem that needs to be solved in the battery recycling industry.

[0003] At present, the process of power battery disassembly mainly relies on traditional path planning and scheduling methods. Although these methods perform well in certain fixed scenarios, they are obviously insufficient in dealing with the complexity and diversity of retired power batteries.

[0004] Several existing technical methods include static path planning methods, disassembly path selection methods based on optimization algorithms, frequent pattern mining methods, and scene matching power battery disassembly decision methods. Although the aforementioned technologies have solved some problems in the disassembly of retired batteries, they still have obvious limitations, such as limited matrix expression capabilities, insufficient complex path selection capabilities, insufficient ability to cope with battery structure diversity, and weak real-time adjustment capabilities. These issues limit the efficiency and resource utilization of power battery disassembly in terms of dealing with complex paths, dynamic adjustments, and real-time optimization. However, the market demand for efficient and intelligent disassembly technology is increasing, and traditional static scheduling methods and simple path planning can no longer meet the needs for real-time adjustment and complex path selection during the disassembly process. In order to improve disassembly efficiency, reduce disassembly costs, and maximize resource recovery rates, the market needs a disassembly method that can flexibly adjust and intelligently select paths in complex environments. Summary of the invention

[0005] The present invention provides a retired battery dynamic disassembly path decision method based on deep search to solve the problem that the existing disassembly path cannot be adjusted in real time.

[0006] In order to achieve the above object, the present invention is implemented by the following technical solutions: The present invention provides a method for dynamic disassembly path decision of retired batteries based on deep search, comprising the following steps: Step 1: Construct a disassembly path based on the battery disassembly process steps, and construct a graph matrix encoding according to the disassembly path; Step 2: Obtain the machine state, and input the machine state and graph matrix encoding into the built reinforcement learning model to obtain the disassembly weight; Step 3: Obtain a first optimized disassembly path according to the disassembly weight and in-depth search, and perform disassembly according to the first optimized disassembly path; Step 4: After inputting the obtained battery pack information of the retired battery into the reinforcement learning model again to obtain the disassembly weight, a second optimized disassembly path is obtained in combination with the deep search, and machines are allocated according to the second optimized disassembly path to complete the disassembly of the retired battery; The first optimized disassembly path and the second optimized disassembly path are both determined in the following manner: according to the disassembly weight combined with the depth search, the disassembly path is traversed under predetermined constraints to calculate the statistical value, and the first optimized disassembly path is determined based on the statistical value combined with the battery pack detection information.

[0007] Further, constructing a disassembly path based on the battery disassembly process steps includes: constructing a disassembly path according to the battery disassembly process steps in combination with a first predetermined rule; The first predetermined rule includes: numbering the main processes and key nodes of the battery disassembly process steps, setting the nodes of the paths that need to meet multiple process conditions at the same time as AND nodes, and setting the path nodes of whether to disassemble the battery after performance testing as OR nodes.

[0008] Further, constructing the graph matrix code according to the disassembly path includes: using the numbers of the nodes of the disassembly path as the rows and columns of the matrix code, and filling the rows and columns of the matrix code based on the node relationship of the disassembly path in combination with a second predetermined rule; The second predetermined rule includes: if the node relationship is an AND node or a connection relationship, it is defined as 1; if the node relationship is an OR node, it is defined as 2.

[0009] Further, the graph matrix code is constructed according to the disassembly path: a coding file is constructed according to the disassembly path, and a graph matrix code is constructed based on the coding file; The encoding file includes: an identifier and an array of several groups of side information representing the disassembly path; The identifier is set based on the node relationship, and the array is set based on the node number.

[0010] Furthermore, the reinforcement learning model constructs an incentive function based on average completion time, process waiting time, delay penalty, assembly and disassembly time, and station processing capacity; The activation function is expressed by the following formula: ; in, represents the activation function, represents the average completion time, Indicates the process waiting time, represents the delay penalty, Indicates assembly and disassembly time, Indicates the processing capacity of the workstation. The weight coefficients are based on the aforementioned parameters respectively; The average completion time is expressed by the following formula: ; in, Indicates The time required to complete the disassembly of a battery pack, Indicates the maximum allowed completion time; The process waiting time is expressed by the following formula: ; in, Indicates The waiting time for the disassembly process of each battery pack, Indicates the maximum allowed waiting time; The delay penalty is expressed by the following formula: ; in, Indicates the total delay time for disassembling the battery pack; The assembly and disassembly time is expressed by the following formula: ; in, Indicates the battery pack from The time required for disassembly on the machine, Indicates that the battery pack is delivered to The time required by the machine; The processing capacity of the workstation is expressed by the following formula: ; in, Indicates the maximum processing capacity of the station, that is, the time required to process a battery pack. Indicates The battery pack is in The first process is The time required for machine processing.

[0011] Further, the predetermined constraints are set based on the status of the machine and the battery, the process sequence, the single process, the assembly and disassembly time, and the detection station node; The state constraints of the machine and the battery are expressed by the following formula: ; in, Indicates that all machines and batteries are at the time It can be used at any time and is not bound to any machine; The process sequence constraint is expressed by the following formula: ; in, Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The process is The time required for machine processing; Indicates the battery pack from The time required to disassemble the machine, Indicates that the battery pack is delivered to The time required on the machine; The single process constraint is expressed by the following formula: ; in, Indicates The total number of processes for a battery pack, Indicates the number of machines in the workshop; Indicated in Moment, Battery Pack The process is Machine processing, , otherwise 0; ; The assembly and disassembly time constraints are expressed by the following formula: ; in, exist Moment, Battery Pack The process is Machine processing, , otherwise 0; The inspection station node constraints are expressed by the following formula: ; in, Indicates the processing capacity of the inspection station, that is, the number of batteries that can be inspected per unit time.

[0012] Furthermore, the statistical value is calculated by the following formula: ; in, Indicates traversal to Node, The first branch The statistics of nodes, Indicates that from The sum of all possible out-degrees triggered by the node, Indicates The first branch The in-degree of a node, Indicates The first branch The out-degree of a node, Indicates The first branch The in-degree of a node.

[0013] Furthermore, in the process of calculating the statistical value, if the statistical value of the node is 0, the node is defined as a JOIN node.

[0014] Further, the reinforcement learning model is a multi-graph convolutional network, and the multi-graph convolutional network is constructed based on MGCN units that extract connectivity and topological structures between different parts of retired batteries; The MGCN unit is constructed based on Conv2d k1 blocks and Conv2d k3 blocks.

[0015] Furthermore, the MGCN unit is a three-input single-output multi-source information fusion feature extraction unit, and the MGCN unit includes an input layer, a fusion extraction layer, a weighted layer, and an output layer; The input layer includes three Conv2d k1 blocks for extracting local features of three types of inputs respectively; The fusion extraction layer includes a fusion block and a Conv2d k3 block for extracting spatial relations; The weighted layer comprises a SE block; The output layer includes a Conv2d k1 block and an output block for performing linear transformation and dimensionality compression.

[0016] Further, the multi-graph convolutional network is constructed according to the MGCN unit for extracting connectivity and topological structures between different parts of the retired battery, including: constructing an action module and a value module according to the MGCN unit, and constructing a multi-graph convolutional network based on the action module and the value module; The action module extracts multi-scale features based on machine state and graph matrix encoding to obtain disassembly weights; The value module obtains the multi-scale features extracted by the action module and optimizes the disassembly weights in combination with the incentive function.

[0017] Furthermore, the action module sequentially includes an MGCN unit for extracting machine state and graph matrix encoding features, a detail extraction Conv2d k3 block for further extracting surface and internal detail features of the machine state and graph matrix encoding, an MGCN unit for strengthening the machine state and graph matrix encoding features, a detail strengthening Conv2d k3 block for further strengthening the surface and internal detail features of the machine state and graph matrix encoding, and a Conv2d k1 block for compressing and integrating all strengthened features.

[0018] Furthermore, the value module is connected to the detail extraction Conv2d k3 block and the detail enhancement Conv2d k3 block of the action module respectively, and the acquired multi-scale features are integrated and evaluated; The value module includes an acquisition unit for acquiring multi-scale features, an integration unit for integrating the multi-scale features, and an evaluation unit for evaluating the integrated multi-scale features; The acquisition unit includes a first channel, a second channel and a Conv2d k1 block, the first channel and the second channel are each composed of two Conv2d k3 blocks, and are both connected to the action module and the integration module, and the Conv2d k1 block is arranged between the first Conv2d k3 block in the direction where the first channel, the second channel and the action module are connected; The integration unit includes 1 Conv2d k1 block; The evaluation unit includes 1 fully connected block and one MGCN unit.

[0019] Beneficial effects: The present invention provides a retired battery dynamic disassembly path decision method based on deep search. Dynamic search is achieved by introducing deep search and reinforcement learning models. Compared with traditional static scheduling methods, it has better real-time and adaptability. In this process, secondary path planning is performed according to the acquired battery pack information of the retired battery, and disassembly path planning is better performed based on battery type, thereby achieving dynamic response to complex data and more efficient planning of circuit disassembly paths.

[0020] Furthermore, the incentive function set based on the industrial environment takes into account the detection result processing, resource utilization optimization, real-time response, adaptability and bottleneck effect. This design not only learns and optimizes in a better direction, but also can quickly adjust strategies in complex environments, reflecting the high efficiency and flexibility that existing scheduling methods do not have.

[0021] Furthermore, the multi-graph convolutional network enhances the ability to capture the complex relationships and interactions between different parts of the power battery by introducing the MGCN unit. Traditional methods may not be able to effectively integrate multi-scale information, but the present invention optimizes the multi-scale feature extraction path through the MGCN unit, which can accurately capture the different levels of details of the battery assembly during the disassembly process. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A disassembly path flow chart constructed based on three different battery disassembly processes in a dynamic disassembly path decision method for retired batteries according to an embodiment of the present invention; Figure 2 A schematic diagram of a matrix coding constructed according to an embodiment of the present invention; Figure 3 A schematic diagram of constructing a coding file based on a disassembly path according to an embodiment of the present invention; Figure 4 It is a disassembly process step flow chart of three different batteries according to an embodiment of the present invention after being optimized; Figure 5 Schematic diagram of the network structure of a multi-graph convolutional network according to an embodiment of the present invention, wherein A represents a first optimized disassembly path or a second optimized disassembly path; Figure 6 Schematic diagram of the network structure of the MGCN unit in the multi-graph convolutional network in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0025] The embodiment of the present invention provides a method for dynamic disassembly path decision of retired batteries based on deep search, comprising the following steps: Step 1: Construct a disassembly path based on the battery disassembly process steps, and construct a graph matrix encoding according to the disassembly path; See also Figure 1 In this embodiment, according to the disassembly process steps of three different batteries, the disassembly paths are constructed in combination with the first predetermined rule to obtain the following Figure 1 3 disassembly path flow chart shown in FIG.

[0026] like Figure 2 As shown, in this embodiment, directly by Figure 1 The disassembly path of battery 1 constructs a coding matrix, and the numbers of the nodes based on the disassembly path are used as the rows and columns of the matrix code. The rows and columns of the matrix code are filled based on the node relationship of the disassembly path combined with the second predetermined rule. The construction of the matrix code more effectively expresses the OR node, improves the decision-making of the node selection and the efficiency of the disassembly path.

[0027] like Figure 3 As shown, in other embodiments, a coding file can also be constructed based on the disassembly path of the above-mentioned battery 1, and then a graph matrix code can be further constructed based on the coding file. In the process of constructing the coding file, an identifier is set based on the node relationship, and then an array is set based on the node number. For example, "#OR" is used as an identifier in the first line of the file, and the subsequent connection node number "3 7" indicates that the connection relationship between node 3 and node 7 is an OR node, and each group of arrays recorded in the coding file is the edge information of the disassembly path. After the coding file is constructed, a matrix code is constructed according to the identifier and data in the coding file.

[0028] Among them, the first predetermined rule includes: numbering the main processes and key nodes of the battery disassembly process steps, setting the nodes of the paths that need to meet multiple process conditions at the same time as AND nodes, and setting the path nodes of whether to disassemble the battery after performance testing as OR nodes.

[0029] The second predetermined rule includes that if the node relationship is an AND node or a connection relationship, it is defined as 1, and if the node relationship is an OR node, it is defined as 2; Step 2: Obtain the machine state, and input the machine state and graph matrix encoding into the built reinforcement learning model to obtain the disassembly weight; The reinforcement learning model is a multi-graph convolutional network, which builds an incentive function based on average completion time, process waiting time, delay penalty, assembly and disassembly time, and station processing capacity. The machine state represents the number of machines and the time required for the machine to execute its corresponding process. Specifically, the activation function is expressed by the following formula: ; in, represents the activation function, represents the average completion time, Indicates the process waiting time, represents the delay penalty, Indicates assembly and disassembly time, Indicates the processing capacity of the workstation. Respectively represent weight coefficients based on the aforementioned parameters. In this embodiment, the weight coefficients of each parameter are 0.2; The average completion time is expressed by the following formula: ; in, Indicates The time required to complete the disassembly of a battery pack, Indicates the maximum allowed completion time; The process waiting time is expressed by the following formula: ; in, Indicates The waiting time for the disassembly process of each battery pack, Indicates the maximum allowed waiting time; The delay penalty is expressed by the following formula: ; in, Indicates the total delay time for disassembling the battery; The assembly and disassembly time is expressed by the following formula: ; in, Indicates the battery pack from The time required for disassembly on the machine, Indicates that the battery pack is delivered to The time required by the machine; The processing capacity of the workstation is expressed by the following formula: ; in, Indicates the maximum processing capacity of the workstation. Indicates The battery pack is in The first process is The time required for machine processing, Indicates the total number of battery packs to be disassembled.

[0030] Step 3: Obtain a first optimized disassembly path according to the disassembly weight and deep search, and perform disassembly according to the first optimized disassembly path; Specifically, according to the disassembly weight and the combination depth, the disassembly path is searched and traversed under the predetermined constraint to calculate the statistical value, and the first optimized disassembly path is determined based on the statistical value; The predefined constraints are based on the status of the machine and battery, the process sequence, the single process, the assembly and disassembly time, and the detection station node settings; The state constraints of the machine and battery are expressed by the following formula: ; in, Indicates that all machines and batteries are at the time It can be used at any time and is not bound to any machine; The process sequence constraint is expressed by the following formula: ; in, Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The process is The time required for machine processing; Indicates the battery pack from The time required to disassemble the machine, Indicates that the battery pack is delivered to The time required on the machine, this formula indicates that the battery process is performed in sequence; The single process constraint is expressed by the following formula: ; in, Indicates The total number of processes for a battery pack, Indicates the number of machines in the workshop; Indicates at time , No. Battery Pack The process is Machine processing, , otherwise 0; This formula indicates that the battery can only perform one process at any time; ; This formula indicates that each machine can only handle one process at any time; The assembly and disassembly time constraints are expressed by the following formula: ; in, exist Moment, Battery Pack The process is Machine processing, , otherwise 0; This formula indicates that each battery must consider the time constraint during assembly and disassembly; The inspection station node constraints are expressed by the following formula: ; in, Represents the total number of retired batteries, It represents the processing capacity of the inspection station, that is, the number of batteries that can be inspected per unit time; this formula means that the module and cell of each battery are the process bottleneck at the inspection station, and its processing capacity needs to be given priority.

[0031] Step 4: After inputting the reinforcement learning model again to obtain the disassembly weight based on the battery pack information of the retired battery, the second optimized disassembly path is obtained in combination with the deep search, and the machines are allocated according to the second optimized disassembly path to complete the disassembly of the retired battery.

[0032] See also Figure 4 After determining the first optimized disassembly path based on the above constraints, the optimized disassembly weight is obtained according to the module information and the cell information combined with the reinforcement learning model. Based on the optimized disassembly weight and combined with the deep search, the first optimized disassembly path is traversed under the predetermined constraints to calculate the statistical value. Based on the statistical value and the battery pack detection information, the second optimized disassembly path is determined.

[0033] Among them, the battery information includes module information and cell information, and the battery pack detection information includes the quality of the modules and cells, that is, the disassembly path is determined based on the statistical values ​​combined with the modules and cells disassembled from the battery pack.

[0034] The statistical value is calculated by the following formula: ; in, Indicates traversal to Node, The first branch The statistics of nodes, Indicates that from The sum of all possible out-degrees triggered by the node, Indicates The first branch The in-degree of a node, Indicates The first branch The out-degree of a node, Indicates The first branch The in-degree of a node.

[0035] See also Figure 1 , for the node 6 of battery 1, the out-degree of node 6 is -3. There are a total of Nodes, when When , the branch statistics are nodes 9 and 10, where the sum of the in-degree and out-degree of node 9 is equal to 0, and the in-degree of node 10 is 1. At this time, the statistical value is -2. When , the branch statistics are nodes 9, 10, 11 and 15, where the sum of the in-degree and out-degree of nodes 9, 10 and 11 is equal to 0, and the in-degree of node 15 is 3, so the statistical value is 0; In the process of calculating the statistical value, if the statistical value of the node is 0, the node is defined as a JOIN node. The JOIN node more intuitively represents the convergence point of the node, which is easier for operators to understand.

[0036] In this example, see Figure 5-6 ,The multi-graph convolutional network is also improved as follows: the multi-graph convolutional network is constructed according to the MGCN unit for extracting the connectivity and topological structure between different parts of retired batteries. Specifically, an action module and a value module are constructed according to the MGCN unit, and a multi-graph convolutional network is constructed based on the action module and the value module.

[0037] Among them, the action module includes, in sequence, an MGCN unit for extracting machine state and graph matrix encoding features, a detail extraction Conv2d k3 block for further extracting surface and internal detail features of machine state and graph matrix encoding, an MGCN unit for enhancing machine state and graph matrix encoding features, a detail enhancement Conv2d k3 block for further enhancing surface and internal detail features of machine state and graph matrix encoding, and a Conv2d k1 block for compressing and integrating all enhanced features.

[0038] The value module is connected to the detail extraction Conv2d k3 block and detail enhancement Conv2d k3 block of the action module respectively, and the acquired multi-scale features are integrated and evaluated; The value module includes an acquisition unit for acquiring multi-scale features, an integration unit for integrating the multi-scale features, and an evaluation unit for evaluating the integrated multi-scale features; The acquisition unit includes a first channel, a second channel and a Conv2d k1 block. The first channel and the second channel are both composed of two Conv2d k3 blocks, and are both connected to the action module and the integration module. The Conv2d k1 block is arranged between the first Conv2d k3 block in the direction where the first channel, the second channel and the action module are connected; The integration unit includes 1 Conv2d k1 block; The evaluation unit consists of 1 fully connected block and one MGCN unit.

[0039] The MGCN unit is built based on the Conv2d k1 block and the Conv2d k3 block.

[0040] Among them, the MGCN unit is a multi-source information fusion feature extraction unit with three inputs and one output. The MGCN unit includes an input layer, a fusion extraction layer, a weighted layer, and an output layer; The input layer includes three Conv2d k1 blocks that extract local features of three types of inputs respectively; The fusion extraction layer includes the fusion block and the Conv2d k3 block that extracts the spatial relationship; The weighted layer includes SE blocks; The output layer includes a Conv2d k1 block and an output block for linear transformation and dimension compression.

[0041] Finally, in this embodiment, the multi-graph convolutional network in the reinforcement learning model is combined with a dynamic disassembly path decision method for retired batteries based on deep search of the present invention, and manual scheduling, genetic algorithm and particle swarm optimization algorithm are used to make decisions on the disassembly path. For specific decision results, please see Table 1; Table 1: Average completion time, process waiting time, delay penalty and overall goal

[0042] As can be seen from Table 1, the multi-graph convolutional network is superior to the traditional manual scheduling, genetic algorithm and particle swarm optimization algorithm in the dynamic scheduling of retired battery disassembly. In terms of average completion time, process waiting time, delay penalty and other indicators, the multi-graph convolutional network using the method for dynamic disassembly path decision of retired batteries based on deep search of the present invention performs best, especially reducing the average completion time to 3287 seconds, the process waiting time to 482 seconds, and the delay penalty to 65 seconds, and the comprehensive total target value reaches 3834 seconds. Each value is the lowest among all methods, which clearly shows that the method of the present invention has excellent real-time scheduling optimization capabilities.

[0043] Compared with the static optimization genetic algorithm and particle swarm optimization algorithm, the multi-graph convolutional network of the retired battery dynamic disassembly path decision method based on deep search in the present invention can adjust the priority at each moment in the disassembly process by dynamically monitoring the status of battery modules and cells, and flexibly respond to changing working conditions. This dynamic and adaptive scheduling method significantly improves resource utilization and on-time completion rate, making the disassembly process more efficient and smooth, demonstrating its superiority in complex and changeable environments.

[0044] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A dynamic disassembly path decision method for retired batteries based on deep search, characterized in that: The steps include: Step 1: Construct a disassembly path based on the battery disassembly process steps, and construct a graph matrix encoding according to the disassembly path; Step 2: Obtain the machine state, and input the machine state and graph matrix encoding into the built reinforcement learning model to obtain the disassembly weight; Step 3: Obtain a first optimized disassembly path according to the disassembly weight and in-depth search, and perform disassembly according to the first optimized disassembly path; Step 4: After inputting the obtained battery pack information of the retired battery into the reinforcement learning model again to obtain the disassembly weight, a second optimized disassembly path is obtained in combination with the deep search, and machines are allocated according to the second optimized disassembly path to complete the disassembly of the retired battery; The first optimized disassembly path and the second optimized disassembly path are both determined in the following manner: according to the disassembly weight combined with the depth search, the disassembly path is traversed under the predetermined constraints to calculate the statistical value, and the first optimized disassembly path is determined based on the statistical value combined with the battery pack detection information.

2. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 1 is characterized in that: The constructing a disassembly path based on the battery disassembly process steps includes: constructing a disassembly path according to the battery disassembly process steps in combination with a first predetermined rule; The first predetermined rule includes: numbering the main processes and key nodes of the battery disassembly process steps, setting the nodes of the paths that need to meet multiple process conditions at the same time as AND nodes, and setting the path nodes of whether to disassemble the battery after performance testing as OR nodes.

3. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 2 is characterized in that: The constructing the graph matrix code according to the disassembly path includes: using the numbers of the nodes of the disassembly path as the rows and columns of the matrix code, and filling the rows and columns of the matrix code based on the node relationship of the disassembly path in combination with a second predetermined rule; The second predetermined rule includes: if the node relationship is an AND node or a connection relationship, it is defined as 1; if the node relationship is an OR node, it is defined as 2.

4. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 2 is characterized in that: The constructing of the graph matrix code according to the disassembly path: constructing a code file according to the disassembly path, and constructing the graph matrix code based on the code file; The encoding file includes: an identifier and an array of several groups of side information representing the disassembly path; The identifier is set based on the node relationship, and the array is set based on the node number.

5. The method for dynamic disassembly path decision of retired batteries based on deep search according to any one of claims 1 to 4, characterized in that: The reinforcement learning model constructs an incentive function based on average completion time, process waiting time, delay penalty, assembly and disassembly time, and station processing capacity; The activation function is expressed by the following formula: ； in, represents the activation function, represents the average completion time, Indicates the process waiting time, represents the delay penalty, Indicates assembly and disassembly time, Indicates the processing capacity of the workstation. The weight coefficients are based on the aforementioned parameters respectively; The average completion time is expressed by the following formula: ； in, Indicates The time required to complete the disassembly of a battery pack, Indicates the maximum allowed completion time; The process waiting time is expressed by the following formula: ； in, Indicates The waiting time for the disassembly process of each battery pack, Indicates the maximum allowed waiting time; The delay penalty is expressed by the following formula: ； in, Indicates the total delay time for disassembling the battery pack; The assembly and disassembly time is expressed by the following formula: ； in, Indicates the battery pack from The time required for disassembly on the machine, Indicates that the battery pack is delivered to The time required by the machine; The processing capacity of the workstation is expressed by the following formula: ； in, Indicates the maximum processing capacity of the station, that is, the time required to process a battery pack. Indicates The battery pack is in The process is The time required for machine processing, Indicates the total number of battery packs to be disassembled.

6. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 5 is characterized in that: The predetermined constraints are based on the status of the machine and the battery, the process sequence, the single process, the assembly and disassembly time, and the detection station node setting; The state constraints of the machine and the battery are expressed by the following formula: ； in, Indicates that all devices and battery packs are at the time It can be used at any time and is not bound to any machine; The process sequence constraint is expressed by the following formula: ； in, Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The time required for each process; Indicates The battery pack is in The process is The time required for machine processing; Indicates the battery pack from The time required to disassemble the machine, Indicates that the battery pack is delivered to The time required on the machine; The single process constraint is expressed by the following formula: ； in, Indicates The total number of processes for a battery pack, Indicates the number of machines in the workshop; Indicated in Moment, Battery Pack The process is Machine processing, , otherwise 0; ； The assembly and disassembly time constraint is expressed by the following formula: ； in, exist Moment, Battery Pack The process is Machine processing, , otherwise 0; The detection station node constraint is expressed by the following formula: ； in, Indicates the processing capacity of the inspection station, that is, the number of batteries that can be inspected per unit time.

7. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 1 is characterized in that: The statistical value is calculated by the following formula: ； in, Indicates traversal to Node, The first branch The statistics of nodes, Indicates that from The sum of all possible out-degrees triggered by the node, Indicates The first branch The in-degree of a node, Indicates The first branch The out-degree of a node, Indicates The first branch The in-degree of a node.

8. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 1, characterized in that: During the process of calculating the statistical value, if the statistical value of the node is 0, the node is defined as a JOIN node.

9. The method for dynamic disassembly path decision of retired batteries based on deep search according to any one of claims 1 to 4, characterized in that: The reinforcement learning model is a multi-graph convolutional network, which is constructed based on MGCN units that extract connectivity and topological structures between different parts of retired batteries; The MGCN unit is constructed based on Conv2d k1 blocks and Conv2d k3 blocks.

10. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 9, characterized in that: The MGCN unit is a multi-source information fusion feature extraction unit with three inputs and one output, and the MGCN unit includes an input layer, a fusion extraction layer, a weighted layer and an output layer; The input layer includes three Conv2d k1 blocks for extracting local features of three types of inputs respectively; The fusion extraction layer includes a fusion block and a Conv2d k3 block for extracting spatial relations; The weighted layer comprises a SE block; The output layer includes a Conv2d k1 block and an output block for performing linear transformation and dimensionality compression.

11. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 10, characterized in that: The multi-graph convolutional network is constructed according to the MGCN unit for extracting connectivity and topological structures between different parts of the retired battery, including: constructing an action module and a value module according to the MGCN unit, and constructing a multi-graph convolutional network based on the action module and the value module; The action module extracts multi-scale features based on machine state and graph matrix encoding to obtain disassembly weights; The value module obtains the multi-scale features extracted by the action module and optimizes the disassembly weights in combination with the incentive function.

12. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 11, characterized in that: The action module sequentially includes an MGCN unit for extracting machine state and graph matrix encoding features, a detail extraction Conv2d k3 block for further extracting machine state and graph matrix encoding surface and internal detail features, an MGCN unit for strengthening machine state and graph matrix encoding features, a detail strengthening Conv2d k3 block for further strengthening machine state and graph matrix encoding surface and internal detail features, and a Conv2d k1 block for compressing and integrating all strengthened features.

13. The method for dynamic disassembly path decision of retired batteries based on deep search according to claim 12, characterized in that: The value module is connected to the detail extraction Conv2d k3 block and the detail enhancement Conv2d k3 block of the action module respectively, and integrates and evaluates the acquired multi-scale features; The value module includes an acquisition unit for acquiring multi-scale features, an integration unit for integrating the multi-scale features, and an evaluation unit for evaluating the integrated multi-scale features; The acquisition unit includes a first channel, a second channel and a Conv2d k1 block, the first channel and the second channel are each composed of two Conv2d k3 blocks, and are both connected to the action module and the integration module, and the Conv2d k1 block is arranged between the first Conv2d k3 block in the direction where the first channel, the second channel and the action module are connected; The integration unit includes 1 Conv2d k1 block; The evaluation unit includes 1 fully connected block and one MGCN unit.

Images