Disassembly sequence planning method and device for retired power batteries considering cascading failure

Abstract

The present invention discloses a method and device for planning the disassembly sequence of retired power batteries considering cascading failures, the method comprising: establishing a Petri net disassembly model of retired power batteries, the model capturing the disassembly priority relationship of the battery and the disassembly status of the components; in the Petri net disassembly model, using a Markov decision process to describe the disassembly sequence of retired power batteries; and using an improved Q learning algorithm to optimize the decision of the disassembly sequence. The present invention can well optimize and make decisions on the disassembly sequence planning under cascading failures.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and device for planning a disassembly sequence of retired power batteries taking cascading failure into consideration. Background Art

[0002] Disassembly operations are divided into manual disassembly and automated disassembly on disassembly lines. Since the disassembly process faces many risks and problems, such as excessive battery pack voltage, complex internal wiring harness arrangement, and the risk of electric shock and short circuit, disassembly lines are the safest, most efficient, and most economical operation method for large-scale retired power batteries. Automated disassembly on disassembly lines requires a process planning process for disassembly sequence planning.

[0003] Disassembly sequence planning (DSP) refers to the process of finding the optimal disassembly sequence for end-of-life (EOL) product parts. According to the required disassembly level, disassembly sequence planning can be divided into "complete disassembly" and "selective disassembly". The purpose of complete disassembly is to dismantle EOL products to the smallest component units, while selective disassembly refers to pre-specifying the target components to be recycled based on prior knowledge such as their recycling value. Disassembly sequence planning has been proven to be an NP-hard problem with special discreteness and subject to precedence constraints.

[0004] The particularity of disassembly sequence planning is that it is not a simple reverse assembly process. Due to the large number of EOL products and the different service environments, service cycles, maintenance and other conditions, the quality status after scrapping is also uneven. At present, there is no effective detection equipment and technical methods that can accurately identify the quality status of each part inside retired electromechanical products in advance, which makes process failures occur from time to time during the large-scale disassembly operation, and may cause cascading failures of subsequent processes. The failure behavior of the disassembly process can be caused by the internal factors such as fracture, deformation, wear, corrosion, etc. of the internal parts of the EOL product that are difficult to detect, or it may be induced by unpredictable external factors such as human operation errors and disassembly equipment failures during the disassembly process. The diversity and difficulty of predicting the causes of its generation bring great challenges to the modeling and optimization of the disassembly sequence planning problem. Therefore, the current research on power battery disassembly failure only focuses on the random failure of a certain task. However, practice shows that the failure of a certain disassembly process may directly cause the cascading failure of subsequent processes, thereby bringing great uncertainty to the disassembly sequence planning. At the same time, most of the existing optimization methods are used to solve simple disassembly sequence planning problems, and cannot optimize and make decisions on disassembly sequence planning under cascading failure conditions. Summary of the invention

[0005] The embodiments of the present invention provide a method and device for planning a disassembly sequence of retired power batteries taking into account cascading failures, which can well optimize and make decisions on the disassembly sequence planning under cascading failure conditions.

[0006] An embodiment of the present invention provides a method for planning a disassembly sequence of retired power batteries taking into account cascading failures, comprising:

[0007] A Petri net disassembly model for retired power batteries is established, which captures the disassembly priority relationship and component disassembly status of the battery;

[0008] In the Petri net disassembly model, the Markov decision process is used to describe the disassembly sequence of retired power batteries;

[0009] An improved Q-learning algorithm is used to optimize the decision of disassembly sequence.

[0010] As an improvement of the above solution, the Petri net disassembly model of retired power batteries is established, which captures the disassembly priority relationship of the battery and the disassembly status of the components, including:

[0011] Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N = (P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is a directed arc between the two;

[0012] In P, each place pi represents a specific disassembly state E;

[0013] The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered;

[0014] In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly;

[0015] Use the symbol M to represent the Petri net state;

[0016] Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

[0017] As an improvement of the above solution, in the Petri net disassembly model, a Markov decision process is used to describe the disassembly sequence of retired power batteries, including:

[0018] The reachability set R(M) of the Petri net is transformed into the state space S of the Markov decision process in the reinforcement learning environment; the state space S contains all possible marking configurations, and each marking M represents a specific disassembly state of the disassembly sequence;

[0019] The action space A of the Markov decision process is directly corresponded to the transition set T in the Petri net; the transition set T contains all possible disassembly actions of the disassembly sequence, and each transition represents a feasible disassembly action;

[0020] A reward function R is designed to evaluate the quality of the disassembly action represented by the selected transition after each disassembly action a is selected to execute the disassembly sequence.

[0021] As an improvement of the above solution, the improved Q learning algorithm is used to optimize the disassembly sequence decision, including:

[0022] Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state.

[0023] Define an ε-greedy strategy to decide what action the agent takes at each time step;

[0024] This is repeated for multiple iterations, where the agent performs the following steps in each iteration:

[0025] a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery;

[0026] b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at;

[0027] c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence;

[0028] d. Reward calculation: Calculate the immediate reward rt based on the new state;

[0029] e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence;

[0030] Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

[0031] Another embodiment of the present invention provides a disassembly sequence planning device for retired power batteries considering cascading failures, including:

[0032] A model building module is used to build a Petri net disassembly model of retired power batteries, which captures the disassembly priority relationship of the battery and the disassembly status of the components;

[0033] A description module is used to describe the disassembly sequence of retired power batteries using a Markov decision process in a Petri net disassembly model;

[0034] The optimization module is used to optimize the decision of the disassembly sequence by using an improved Q-learning algorithm.

[0035] As an improvement of the above solution, the model building module is specifically used for:

[0036] Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N = (P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is a directed arc between the two;

[0037] In P, each place pi represents a specific disassembly state E;

[0038] The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered;

[0039] In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly;

[0040] Use the symbol M to represent the Petri net state;

[0041] Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

[0042] As an improvement of the above solution, the description module is specifically used for:

[0043] The reachability set R(M) of the Petri net is transformed into the state space S of the Markov decision process in the reinforcement learning environment; the state space S contains all possible marking configurations, and each marking M represents a specific disassembly state of the disassembly sequence;

[0044] The action space A of the Markov decision process is directly corresponded to the transition set T in the Petri net; the transition set T contains all possible disassembly actions of the disassembly sequence, and each transition represents a feasible disassembly action;

[0045] A reward function R is designed to evaluate the quality of the disassembly action represented by the selected transition after each disassembly action a is selected to execute the disassembly sequence.

[0046] As an improvement of the above solution, the optimization module is specifically used for:

[0047] Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state.

[0048] Define an ε-greedy strategy to decide what action the agent takes at each time step;

[0049] This is repeated for multiple iterations, where the agent performs the following steps in each iteration:

[0050] a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery;

[0051] b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at;

[0052] c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence;

[0053] d. Reward calculation: Calculate the immediate reward rt based on the new state;

[0054] e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence;

[0055] Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

[0056] Another embodiment of the present invention provides a disassembly sequence planning device for retired power batteries considering cascading failures, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the disassembly sequence planning method for retired power batteries considering cascading failures as described in the above-mentioned embodiment of the invention when executing the computer program.

[0057] Another embodiment of the present invention provides a storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the disassembly sequence planning method for retired power batteries considering cascading failure as described in the above-mentioned embodiment of the invention.

[0058] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0059] First, a Petri net model is constructed to accurately characterize the priority and component status of battery disassembly, and then a Markov decision process is embedded in the model to dynamically analyze and guide the selection of the disassembly sequence to ensure that the decision takes into account the risk of cascading failure. On this basis, an improved Q-learning algorithm is used to continuously optimize the disassembly decision process, adapt the strategy to changes, reduce blind attempts, and improve disassembly efficiency and economy. The embodiments of the present invention can not only effectively deal with the uncertainty in the disassembly process and reduce the losses caused by cascading failures, but also through the improved Q-learning algorithm optimization, it can well optimize and make decisions on the disassembly sequence planning under cascading failure conditions, thereby promoting the sustainability and environmental friendliness of retired power battery processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of a method for planning a disassembly sequence of retired power batteries taking into account cascading failures provided by one embodiment of the present invention;

[0061] Figure 2 is a schematic diagram of a Petri structure model provided by an embodiment of the present invention;

[0062] Figure 3 It is a schematic diagram of a process of disassembly sequence decision based on Q-learning provided by an embodiment of the present invention;

[0063] Figure 4 is a schematic diagram of a disassembly sequence planning based on q-learning provided by an embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of a Markov decision process provided by an embodiment of the present invention;

[0065] Figure 6 It is a structural schematic diagram of a disassembly sequence planning device for retired power batteries taking into account cascading failures provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0067] See also Figure 1 , is a flow chart of a method for planning a disassembly sequence of retired power batteries considering cascading failures provided by an embodiment of the present invention. The method for planning a disassembly sequence of retired power batteries considering cascading failures comprises steps S10 to S12:

[0068] S10, establish a Petri net disassembly model of retired power batteries, which captures the disassembly priority relationship of the battery and the disassembly status of the components;

[0069] S11, in the Petri net disassembly model, the Markov decision process is used to describe the disassembly sequence of retired power batteries;

[0070] S12, an improved Q-learning algorithm is used to optimize the decision of the disassembly sequence.

[0071] In an embodiment of the present invention, a Petri net model is first constructed to accurately characterize the priority and component status of battery disassembly, and then a Markov decision process is embedded in the model to dynamically analyze and guide the selection of the disassembly sequence to ensure that the decision takes into account the risk of cascading failure. On this basis, an improved Q-learning algorithm is used to continuously optimize the disassembly decision process, adapt the strategy to changes, reduce blind attempts, and improve disassembly efficiency and economy. The embodiment of the present invention can not only effectively deal with the uncertainty in the disassembly process and reduce the losses caused by cascading failures, but also through the improved Q-learning algorithm optimization, it can well optimize and make decisions on the disassembly sequence planning under cascading failure conditions, thereby promoting the sustainability and environmental friendliness of retired power battery processing.

[0072] As one example, the Petri net disassembly model of retired power batteries is established, which captures the disassembly priority relationship and component disassembly status of the battery, including:

[0073] Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N = (P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is a directed arc between the two;

[0074] In P, each place pi represents a specific disassembly state E;

[0075] The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered;

[0076] In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly;

[0077] Use the symbol M to represent the Petri net state;

[0078] Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

[0079] In this embodiment, specifically, the disassembly sequence planning modeling process based on Petri net is as follows:

[0080] 1. Petri Net Model:

[0081] Petri net: It usually contains a place, which represents a subassembly; a transition, which represents a disassembly task; two directed arcs, one from the place to the transition, which represents the input function, and the other from the transition to the place, which represents the output function; and a token, which represents the state of the object in the place, that is, the state and properties of the subassembly. The disassembly Petri net can be regarded as an extension of the disassembly AND / OR diagram, and the disassembly information in the AND / OR diagram can be represented in the Petri net.

[0082] Petri net is a typical model for describing concurrent, asynchronous, distributed, and uncertain systems. It can intuitively represent the state, behavior, structure, and dynamics of the system, and support formal verification and analysis. Transition plays an important role in Petri nets. It represents the process of the system transitioning from one state to another at a specific time, reflecting the dynamic characteristics of the system, such as the order, conditions, and results of events, and describing multiple events that occur simultaneously or alternately. Therefore, this embodiment uses Petri nets to comprehensively model and represent the structural information of the power battery and the disassembly state of the components, providing a comprehensive and intuitive representation for the combination of various disassembly paths.

[0083] According to the basic elements and basic structure expression principle of Petri net, a disassembly Petri net (DPN) is established to accurately describe the characteristic information of the disassembly process of retired power batteries, and to realize the external representation of the multi-level logical relationship between the disassembly units and disassembly processes in the whole disassembly process, such as Figure 2As shown in the figure, the place represents the disassembly units at each level, the transition represents the disassembly process, the directed edge between the place and the transition represents the direction of the disassembly process route, the circle in the center of the place represents the token, and the place carrying the token is in an activated state, which means that the place can be currently disassembled, that is, the transition immediately following it can be selected for execution at this time.

[0084] 2. Petri net symbolic modeling:

[0085] Basic Petri Net (BPN) is a symbolic expression of Petri Net, and its structure can be expressed by a ternary function N = (P, T, F), where P = {p1, p2, …, pm} is a finite set of positions, T = {t1, t2, …, tn} is a finite set of transitions, and is a directed arc set based on places and transitions. For the ternary set (N, K, W), K is the capacity function of P, which refers to the maximum number of tokens that each place can accommodate; W is the weight function of F, which specifies the token transfer coefficient between places and transitions. BPN is a special case of Petri Net, and its capacity function K and weight function W are both constants 1. The place token of Petri Net is denoted as m(pi) = ki, where ki represents the number of tokens at position pi. For the entire network structure N, the distribution state of tokens on all places constitutes the label of Petri Net, denoted as M∈N. |P| In the tuple (N, M0), M0 represents the initial token of the Petri net N, and (N, M0) is called a labeled Petri net. i When triggered, the marking of the Petri net M will be updated to a new marking M', which is called the reachable marking of M. The position marking in M' can be calculated using formula (3-1). All can be calculated by emitting a transition sequence from 0 The markings that can be reached are called the reachable set of the Petri net, denoted by R(M).

[0086]

[0087] Among them, the change t i The input position set is denoted as ′t i , transition t i The output position set is denoted as t i ′. If a position p∈′t i \t i ′, which means that the position p only belongs to t i On the other hand, if p∈′t i ∩t i ′, which means that the library place p and transition t i A self-circulating structure is formed. iThe change of the label M to M' must satisfy equation (3-2). When the condition is met, transition ti is considered to be an enabled transition under the label M, represented by M(ti), and the process from M to M' is represented by M(ti)M′.

[0088]

[0089] The emission sequence in a Petri net is the sequence of transitions that can be emitted from a given initial marking in the net. In a given state, only enabled transitions can be chosen to be emitted. The emission sequence is a way to describe the behavior and evolution of a Petri net over time. For example, the emission sequence of a Petri net N and initial marking M0 is the sequence Then M0→M1∧...∧M n-1 →M n .

[0090] 3. Disassembly sequence planning based on Petri net

[0091] When modeling the disassembly activity based on Petri nets, the Petri net's place represents the disassembly state of the battery, the transition represents a specific disassembly operation, and the directed arc between the place and the transition represents the constraint relationship of the disassembly process. Therefore, the following characteristics are required to accurately represent the process of a Petri network model: (1) Petri net refers to the BPN defined above. (2) Petri net has a unique initial place node and a well-defined initial token. This initial place reflects the initial complete composition state of the retired power battery. (3) Self-loop structures should be avoided in Petri nets because the disassembly operation represented by the transition is irreversible. (4) Petri nets should have a clear direction, and each transition must satisfy the priority relationship of component disassembly. This will ensure that the disassembly operations are performed in the correct order. (5) The final position node of the Petri net is fixed, representing all the smallest structural units of the battery.

[0092] By combining the above requirements, Table 1 summarizes the disassembly states represented by each position in the Petri net, where the disassembly state {X} represents the remaining set of components to be disassembled. Table 3 lists the specific disassembly actions represented by the transition, where p n / {X} indicates that from position p n Remove component {X} from the components represented. The transition trigger sequence in the Petri net is a feasible solution to the disassembly sequence, and the disassembly sequence optimization is actually to find the optimal transition trigger sequence in the Petri net.

[0093] Table 3-1 Disassembly status represented by library

[0094]

[0095] Table 3-2 Disassembly actions represented by transitions

[0096]

[0097] As one example, in the Petri net disassembly model, a Markov decision process is used to describe the disassembly sequence of retired power batteries, including:

[0098] The reachability set R(M) of the Petri net is transformed into the state space S of the Markov decision process in the reinforcement learning environment; the state space S contains all possible marking configurations, and each marking M represents a specific disassembly state of the disassembly sequence;

[0099] The action space A of the Markov decision process is directly corresponded to the transition set T in the Petri net; the transition set T contains all possible disassembly actions of the disassembly sequence, and each transition represents a feasible disassembly action;

[0100] A reward function R is designed to evaluate the quality of the disassembly action represented by the selected transition after each disassembly action a is selected to execute the disassembly sequence.

[0101] In this embodiment, specifically:

[0102] In the Petri net model, the triggering of a transition will lead to a new mark, thus transforming the system to a new state. This dynamic behavior can be effectively described by the Markov process. Therefore, the disassembly sequence decision problem based on Petri nets can be solved by converting it into an equivalent Markov model and adopting reinforcement learning (RL) technology. To use the RL method to solve the disassembly sequence optimization problem, it is necessary to construct an MDP framework described by a four-element function (S, A, R, Φ). Among them, S represents the state space, A represents the action space, R represents the reward obtained by the action, and Φ represents the state transition. The specific design of S, A, and R is as follows:

[0103] (1) State space S

[0104] The reachability set R(M) of the Petri net simulating product decomposition constitutes the state set in the reinforcement learning environment, namely the state space S. In addition, each row in Q-Learing corresponds to the label M in R(M). The labeling of the library in the disassembly modeling based on Petri nets is expressed by formula (3-3), and the labeling at time t is defined by formula (3-4). For example, Figure 4The initial labeling of the Petri net shown in M0 = (1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) corresponds to the initial state S0 = M0 in reinforcement learning. Transition t1 is triggered, and we obtain the labeling M1 = (0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1) at the next moment, correspondingly labeled S1 = M1.

[0105]

[0106] M(t)=(m(p1), m(p2),..., m(p n )) t (3-4)

[0107] (2) Action Space A

[0108] For the disassembly model based on Petri network, it can be defined as a four-element function (P, T, F, M0), and the action space A is equal to the set of transitions T. Similarly, each column in the q table corresponds to a transition t (t∈T). During the training process, the reinforcement learning agent learns the optimal sequence strategy π, that is, the optimal transfer sequence of the Petri network.

[0109] (3) Reward function R

[0110] After the agent selects a transition action a, it will receive a certain reward from the environment. The reward serves as feedback to evaluate the quality of the disassembly action represented by the selected transition. The reward value reflects the effectiveness or desirability of the agent's behavior and is used to guide the learning process. In the context of disassembly sequence optimization, the immediate reward obtained after executing action a can be defined as formula (3-5).

[0111] R(a)=ΔV(a)+β·C(a)+b (3-5)

[0112] Where ΔV(a) represents the increment of the component recovery value obtained by performing disassembly operation a, and reflects the added value obtained by extracting components during disassembly, calculated by formula (3-6); β is the coefficient of C(a), C(a) is the cost required to perform process disassembly action a, and is a negative value representing the cost incurred in the disassembly action. b is a negative constant that distinguishes disassembly sequences with the same cumulative reward by the number of disassembly operations performed, so that disassembly sequences with fewer operations can obtain slightly larger cumulative rewards. The design size containing b is small to prevent the number of operations from masking cost factors and to encourage the agent to prioritize more efficient disassembly sequences.

[0113]

[0114] Among them, vMt and vMt+1 respectively represent the total recovery value of each part before and after the disassembly action a is executed; vp corresponds to the recovery value of the part state; m(pj) and m(pi) respectively represent the marks of each part before and after the disassembly action a is executed.

[0115] As one example, the improved Q-learning algorithm is used to optimize the disassembly sequence decision, including:

[0116] Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state.

[0117] Define an ε-greedy strategy to decide what action the agent takes at each time step;

[0118] This is repeated for multiple iterations, where the agent performs the following steps in each iteration:

[0119] a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery;

[0120] b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at;

[0121] c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence;

[0122] d. Reward calculation: Calculate the immediate reward rt based on the new state;

[0123] e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence;

[0124] Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

[0125] In this embodiment, the process of disassembling sequence decision based on Q-learning is as follows:

[0126] 1 Overall Framework

[0127] The disassembly state and feasible disassembly operation of the power battery correspond to the net mark and transition in the Petri net, respectively, and also serve as the input state and output operation of the RL. The required disassembly sequence corresponds to the transition trigger sequence in the Petri net and the strategy obtained by the RL. The detailed flow chart of this method is shown in Figure 3-4 The solid line in the figure represents the main process of the algorithm, and the dotted line represents the interaction process between the agent and the environment.

[0128] like Figure 3 As shown in the figure, the RL-based method involves two main components: the environment and the agent. The environment contains all the necessary information required to make disassembly decisions. This includes information related to the power battery disassembly decision, structural information represented by the Petri network model, and various hyperparameters used in the algorithm. On the other hand, the agent, as an abstract entity, is responsible for obtaining the optimal disassembly sequence. This includes interfacing with the environment, collecting relevant data, and navigating the optimization and decision-making stages through the Q table. Throughout the process, the improved Q learning method dynamically formulates and selects from a fluctuating action space based on different Petri net observation states. It fine-tunes the reward associated with the "stop" action and subsequently updates the Q table for the changing action space. Finally, in order to ensure superior convergence, the exploration rate is adjusted after each feedback optimization.

[0129] The disassembly sequence decision process based on Q-learning is as follows Figure 4 As shown. Q-Learing is a table-based reinforcement learning algorithm, where the rows of the q-table represent product states, the columns represent disassembly operations, and the corresponding q-values ​​Q(st, at) represent the expected rewards that can be obtained by performing the disassembly operation at in the current product disassembly state st. In each iteration, the agent selects the column index with the largest Q value in the current state as the execution action to be executed with a certain probability. The agent then receives a reward, updates the q-table, and transitions to the next state. This process is repeated until the product reaches the final disassembly state. The sequential combination of all actions taken by the agent forms the obtained strategy π. After server iterations, the obtained strategy π' represents the optimized disassembly sequence.

[0130] The q-learning algorithm is a reinforcement learning algorithm based on value iteration, which uses the action value function Q(s, a) instead of the state value function. The expression of the Q function is defined by equation (3-7), that is, the expected reward obtained when the agent starts from the disassembly state s, takes the disassembly action a and selects subsequent actions according to the strategy π until the disassembly process ends.

[0131] Q π (s, a) = E π [G t |S t =S,A t =a] (3-7)

[0132] When Q(s, a) converges to the optimal value Q*(s, a) after continuous iterations, the agent learns the optimal disassembly strategy π*. The iterative update process of Q(s, a) is shown in Equation (2-8).

[0133]

[0134] Where Q(st, at) represents the Q value of st, at at time t; Q new (s t , a t ) represents the updated Q value; The maximum Q value for executing at to enter the next state st+1; α is the learning rate, which indicates the degree of learning during one iteration; γ is the decay factor, which indicates the degree of decay of future rewards.

[0135] 2 Cascading failure model construction

[0136] Due to the long service life and serious aging of retired power batteries, the disassembly process may fail due to wear or damage of parts, or operational errors may occur when manually separating tightly bound parts. Currently, no research has shown what probability distribution the failure of retired power battery disassembly task conforms to. Therefore, this paper assumes that the occurrence of retired power battery disassembly behavior a conforms to random distribution, and its probability density function is shown in formula (3-9).

[0137]

[0138] The power battery has a multi-level nested structure of whole pack-module-cell, with complex series-parallel circuits between them, and each component unit is sealed by thick sealant. When a disassembly action fails, it is very likely to cause cascading failures in subsequent processes. The prediction of cascading failure probability is given by Bayesian posterior probability, as shown in formula (3-10).

[0139]

[0140] To optimize the disassembly sequence with uncertain goals, in addition to optimizing the original disassembly sequence, it is also necessary to determine the appropriate termination condition based on the recycling benefit. The new termination condition replaces the original termination condition of the q-learing agent, which includes disassembling all or reaching a fixed goal. To adapt to this change, we introduce a new "stop" action in the variable action space, denoted as A. Therefore, the action space of the selective disassembly MDP with uncertain goals becomes {A′∪STOP}, as Figure 5 shown.

[0141] It should be noted that in the disassembly sequence optimization, the "STOP" action does not have a direct disassembly meaning. Executing the "STOP" action does not cause the agent to enter the next state St+1. Instead, the execution of the "Stop" action means an early termination of the iterative process, allowing the agent to start the next iteration. Therefore, when updating the Q function of "STOP", its meaning is completely opposite to that of other actions in. Specifically, it is used to determine whether the action selection after time t will have a negative impact on the cumulative reward. If so, the agent will choose to terminate the strategy π at time t. Therefore, the Q value of the "Stop" action needs to be updated, which ensures that the agent learns to identify situations where further disassembly operations are no longer required.

[0142]

[0143] In this equation, represents the maximum q value of all actions a∈{A′|s t}, where {A′|s t} is the action space at time t, excluding the “STOP” action; is the reward obtained through the “STOP” action.

[0144] (1) Variable action space

[0145] When defining the valid actions in each step in a Petri net, it is important to consider the concept of "enabled transitions". Enabled transitions are transitions that can be triggered based on the current marking or state of the Petri net. By incorporating this concept, a mutable action space is designed.

[0146] The action space is obtained by updating the original action space A at time t and observing the state in the environment. The update method is described in equation (2-10), which determines the set of enabled transitions at each step. By only considering the enabled transitions in the action space, the agent focuses on selecting actions that may lead to meaningful changes in the labeling or state of the Petri net.

[0147]

[0148] By incorporating the concept of enabling transitions and updating the action space accordingly, the agent can learn to choose effective and meaningful actions while avoiding invalid transitions that do not change the labeling or state of the Petri net. This helps guide the agent's decision-making process and ensures that it focuses on actions that contribute to the decomposition process.

[0149] (2) Improved Greedy Strategy

[0150] When reinforcement learning faces the classic exploration-exploitation dilemma, the ε-greedy strategy is the most commonly used solution. The ε-greedy strategy selects the action with the largest Q value and probability (1-ε) at each action selection, and randomly selects from the action space with probability ε.

[0151] In the early stage of Q-value iteration, the ε-greedy strategy effectively avoids the agent period. The existence of ε has a negative impact on convergence, making the final output not necessarily the optimal solution. Therefore, it falls into the local optimum, but in the later stage of iteration, this paper makes the following modifications to the ε-greedy strategy:

[0152]

[0153] Where Δε represents the increment of random probability, episode represents the number of iterations, and k is a constant that should satisfy equation (2-13). In this way, a larger initial value of ε can be set in the early stages of the iteration to meet the exploration needs in the early stages of the iteration. As the number of iterations increases, the random probability decreases at a faster rate. Finally, 1-(ε-Δε) will be slightly greater than 1 in the later stages of the iteration, meeting the requirements of development and actual disassembly tasks. This method ensures that stable results are obtained.

[0154]

[0155] See also Figure 6 , is a schematic diagram of a disassembly sequence planning device for retired power batteries considering cascading failures provided by an embodiment of the present invention. The disassembly sequence planning device for retired power batteries considering cascading failures comprises:

[0156] A model building module 10 is used to build a Petri net disassembly model of retired power batteries, which captures the disassembly priority relationship of the battery and the disassembly status of the components;

[0157] A description module 11 is used to describe the disassembly sequence of retired power batteries by using a Markov decision process in a Petri net disassembly model;

[0158] The optimization module 12 is used to optimize the decision of the disassembly sequence by using an improved Q learning algorithm.

[0159] As an improvement of the above solution, the model building module is specifically used for:

[0160] Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N = (P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is a directed arc between the two;

[0161] In P, each place pi represents a specific disassembly state E;

[0162] The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered;

[0163] In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly;

[0164] Use the symbol M to represent the Petri net state;

[0165] Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

[0166] As an improvement of the above solution, the description module is specifically used for:

[0167] The reachability set R(M) of the Petri net is transformed into the state space S of the Markov decision process in the reinforcement learning environment; the state space S contains all possible marking configurations, and each marking M represents a specific disassembly state of the disassembly sequence;

[0168] The action space A of the Markov decision process is directly corresponded to the transition set T in the Petri net; the transition set T contains all possible disassembly actions of the disassembly sequence, and each transition represents a feasible disassembly action;

[0169] A reward function R is designed to evaluate the quality of the disassembly action represented by the selected transition after each disassembly action a is selected to execute the disassembly sequence.

[0170] As an improvement of the above solution, the optimization module is specifically used for:

[0171] Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state.

[0172] Define an ε-greedy strategy to decide what action the agent takes at each time step;

[0173] This is repeated for multiple iterations, where the agent performs the following steps in each iteration:

[0174] a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery;

[0175] b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at;

[0176] c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence;

[0177] d. Reward calculation: Calculate the immediate reward rt based on the new state;

[0178] e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence;

[0179] Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

[0180] An embodiment of the present invention provides a disassembly sequence planning device for retired power batteries considering cascading failures. The disassembly sequence planning device for retired power batteries considering cascading failures of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a disassembly sequence planning program for retired power batteries considering cascading failures. When the processor executes the computer program, the steps in each of the above-mentioned disassembly sequence planning method embodiments for retired power batteries considering cascading failures are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0181] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the disassembly sequence planning device for retired power batteries considering cascading failures.

[0182] The disassembly sequence planning device for retired power batteries considering cascading failures may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The disassembly sequence planning device for retired power batteries considering cascading failures may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the schematic diagram is only an example of a disassembly sequence planning device for retired power batteries considering cascading failures, and does not constitute a limitation on the disassembly sequence planning device for retired power batteries considering cascading failures. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the disassembly sequence planning device for retired power batteries considering cascading failures may also include input and output devices, network access devices, buses, etc.

[0183] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the disassembly sequence planning device for retired power batteries considering cascading failures, and uses various interfaces and lines to connect various parts of the disassembly sequence planning device for retired power batteries considering cascading failures.

[0184] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the disassembly sequence planning device for retired power batteries considering cascading failure by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0185] Wherein, if the module / unit integrated in the disassembly sequence planning device for retired power batteries considering cascading failure is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0186] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0187] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for disassembly sequence planning of retired power batteries considering cascading failures, characterized in that: include: A Petri net disassembly model for retired power batteries is established, which captures the disassembly priority relationship and component disassembly status of the battery; In the Petri net disassembly model, the Markov decision process is used to describe the disassembly sequence of retired power batteries; An improved Q-learning algorithm is used to optimize the decision of disassembly sequence; Wherein, in the Petri net disassembly model, the Markov decision process is used to describe the disassembly sequence of retired power batteries, including: The RL method is used to solve the disassembly sequence optimization problem and construct an MDP framework described by a four-element function (S, A, R, Φ), where S represents the state space, A represents the action space, R represents the reward obtained by the action, and Φ represents the state transition; the specific design of S, A, and R is as follows: State space S: Reachability Sets of Petri Nets Simulating Product Decomposition R(M) It constitutes the state set in the reinforcement learning environment, that is, the state space S. Each row in Q-Learing corresponds to R(M) Mark in M The library mark in the disassembly modeling based on Petri network is expressed by formula (3-3), and the mark at time 𝑡 is defined by formula (3-4); the initial mark of the Petri net M0 = (1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) corresponds to the initial state in reinforcement learning S0 = M0 , transition 𝑡1 is triggered, and at the next moment, it is labeled as 𝑀1=(0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1),correspondingly labeled as 𝑆1=𝑀1 ; ； Action space A: The Petri network-based disassembly model is defined as a quaternion function ( P, T, F, M0 ), action space A Equal to change T A set of; each column in the q table corresponds to a transition t , t ∈T ,During the training process, the reinforcement learning agent learns the optimal sequence strategy 𝜋, which is the optimal transition sequence of the Petri network; Reward function R: After the agent selects a transition action 𝑎, it will receive a certain reward from the environment. In the context of disassembly sequence optimization, the immediate reward obtained after executing action 𝑎 is defined as formula (3-5); ； in Indicates the execution of disassembly operation The increment of the component recycling value obtained, and reflects the added value obtained from extracting components during disassembly, calculated by formula (3-6); for The coefficient of Decomposing the execution process The required cost is a negative value representing the cost incurred in the disassembly action; is a negative constant that distinguishes disassembly sequences with the same cumulative reward by the number of disassembly operations performed, so that the disassembly sequence with fewer operations can obtain a slightly larger cumulative reward; is the coefficient; ； in, and Respectively represent disassembly actions Total recovery value of each part before and after execution; The recovery value of the corresponding part status; and Respectively represent disassembly actions Perform marking of the front and back parts.

2. The method for disassembly sequence planning of retired power batteries considering cascading failure according to claim 1, characterized in that: The Petri net disassembly model of retired power batteries is established, which captures the disassembly priority relationship and component disassembly status of the battery, including: Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N=(P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is the directed arc between the two; In P, each place pi represents a specific disassembly state E; The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered; In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly; Use the symbol M to represent the Petri net state; Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

3. The method for disassembly sequence planning of retired power batteries considering cascading failure according to claim 1, characterized in that: The improved Q-learning algorithm is used to optimize the disassembly sequence decision, including: Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state. Define an ε-greedy strategy to decide what action the agent takes at each time step; This is repeated for multiple iterations, where the agent performs the following steps in each iteration: a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery; b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at; c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence; d. Reward calculation: Calculate the immediate reward rt based on the new state; e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence; Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

4. A disassembly sequence planning device for retired power batteries considering cascading failures, characterized in that: include: A model building module is used to build a Petri net disassembly model of retired power batteries, which captures the disassembly priority relationship of the battery and the disassembly status of the components; A description module is used to describe the disassembly sequence of retired power batteries using a Markov decision process in a Petri net disassembly model; An optimization module, used to optimize the decision of the disassembly sequence using an improved Q-learning algorithm; Wherein, the description module is specifically used for: The RL method is used to solve the disassembly sequence optimization problem and construct an MDP framework described by a four-element function (S, A, R, Φ), where S represents the state space, A represents the action space, R represents the reward obtained by the action, and Φ represents the state transition; the specific design of S, A, and R is as follows: State space S: Reachability Sets of Petri Nets Simulating Product Decomposition R(M) It constitutes the state set in the reinforcement learning environment, that is, the state space S. Each row in Q-Learing corresponds to R(M) Mark in M The library mark in the disassembly modeling based on Petri network is expressed by formula (3-3), and the mark at time 𝑡 is defined by formula (3-4); the initial mark of the Petri net M0 = (1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) corresponds to the initial state in reinforcement learning S0 = M0 , transition 𝑡1 is triggered, and at the next moment, it is labeled as 𝑀1=(0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1),correspondingly labeled as 𝑆1=𝑀1 ; ； Action space A: The Petri network-based disassembly model is defined as a quaternion function ( P, T, F, M0 ), action space A Equal to change T A set of; each column in the q table corresponds to a transition t , t ∈T ,During the training process, the reinforcement learning agent learns the optimal sequence strategy 𝜋, which is the optimal transition sequence of the Petri network; Reward function R: After the agent selects a transition action 𝑎, it will receive a certain reward from the environment. In the context of disassembly sequence optimization, the immediate reward obtained after executing action 𝑎 is defined as formula (3-5); ； in Indicates the execution of disassembly operation The increment of the component recycling value obtained, and reflects the added value obtained from extracting components during disassembly, calculated by formula (3-6); for The coefficient of Decomposing the execution process The required cost is a negative value representing the cost incurred in the disassembly action; is a negative constant that distinguishes disassembly sequences with the same cumulative reward by the number of disassembly operations performed, so that the disassembly sequence with fewer operations can obtain a slightly larger cumulative reward; is the coefficient; ； in, and Respectively represent disassembly actions Total recovery value of each part before and after execution; The recovery value of the corresponding part status; and Respectively represent disassembly actions Perform marking of the front and back parts.

5. The disassembly sequence planning device for retired power batteries considering cascading failures as claimed in claim 4, characterized in that: The model building module is specifically used for: Define the Petri net model of retired power batteries, which is a basic Petri net structure, represented by a triple N=(P, T, F), where P is the set of disassembly states, T is the set of disassembly actions, and F is the directed arc between the two; In P, each place pi represents a specific disassembly state E; The transition ti in T corresponds to a specific disassembly operation, which means removing F from the current state; the input and output place sets and the input place set of the transition define the conditions under which the transition can be triggered; In the Petri net, the disassembly priority relationship is defined through the directed arc F to ensure that the transition can reflect the logical order and dependency of the disassembly; Use the symbol M to represent the Petri net state; Through the transition triggering of Petri nets, the model captures the battery disassembly process, and the trigger update mark of each transition represents the gradual disassembly state change of the battery component.

6. The disassembly sequence planning device for retired power batteries considering cascading failures as claimed in claim 4, characterized in that: The optimization module is specifically used for: Before Q learning begins, a Q table is first created. The rows of the table represent the disassembly status of the current retired power battery, which is the mark in the Petri net. The columns correspond to possible disassembly operations, which are transitions in the Petri net. The value of each cell stores the expected cumulative reward for performing a specific operation in the current state. Define an ε-greedy strategy to decide what action the agent takes at each time step; This is repeated for multiple iterations, where the agent performs the following steps in each iteration: a. State evaluation: Identify a set of feasible operations based on the current Petri net’s label - the disassembly state of the battery; b. Decision-making: Use the ε-greedy strategy to select an operation based on the current Q table and ε value to determine the disassembly action at; c. Execution: Trigger operation at, update the battery mark to the new state according to the Petri net rules, that is, the progress of the disassembly sequence; d. Reward calculation: Calculate the immediate reward rt based on the new state; e. Update the Q table: Use Q-learning to adjust the Q value of the current state action according to the actual reward rt and the maximum Q value of the next state, ensuring that the Q table gradually approaches the optimal disassembly sequence; Iterate until the predetermined stopping condition is reached, at which point the optimal path in the Q table is the learned disassembly sequence.

7. A disassembly sequence planning device for retired power batteries considering cascading failures, characterized in that: The invention comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for disassembly sequence planning of retired power batteries taking into account cascading failures as described in any one of claims 1 to 3 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the disassembly sequence planning method for retired power batteries considering cascading failures as described in any one of claims 1 to 3.

Images