Power material supply chain network equilibrium control method and device based on value function

Abstract

The application discloses a power material supply chain network equilibrium control method and device based on a value function, relates to the technical field of supply chain network equilibrium control, and comprises the following steps: according to local value functions of nodes, key neighborhood nodes are screened through influence among the nodes and sparse communication is carried out, and a local equilibrium control strategy is generated; nodes at the same level are combined into a subnetwork, coordination optimization is carried out in the subnetwork based on the local equilibrium control strategy, a subnetwork equilibrium control strategy is generated, and is summarized to a cross-level coordination layer, global state information is constructed, a coordination target is determined, and after relaxation iteration smoothing processing, multi-level coordination control instructions are output; the local strategy of each node is dynamically adjusted, and the execution result is fed back to each subnetwork; and the application solves the problems that in the existing power material supply chain management, local optimization leads to global imbalance, cooperation between levels is insufficient, and when facing supply and demand fluctuations and sudden disturbances, real-time response capability is poor and control stability is low.

Description

Technical Field

[0001] This invention relates to the field of supply chain network equilibrium control technology, and more specifically, to a method and apparatus for equilibrium control of power material supply chain networks based on value functions. Background Technology

[0002] With the continuous expansion of power systems and the increasing complexity of the power material supply chain, the production, storage, transportation, and distribution of power materials involve multiple participating nodes, including manufacturers, warehousing centers, logistics nodes, and construction units, forming a multi-level, multi-functional, and multi-constrained supply chain network. In traditional power material dispatching and management, common methods mainly rely on centralized planning and experience-based management, including manual dispatching, static strategies based on inventory limits, and single-level transportation optimization. These methods can play a certain role in small-scale or simple topology structures, but as the supply chain network expands and the interdependence between nodes increases, existing technologies are gradually revealing the following shortcomings: Existing technologies typically optimize individual nodes or links, lacking a comprehensive consideration of the overall resource distribution, demand matching, and inter-node collaboration benefits of the supply chain network. This can easily lead to situations where local strategies are optimal but the global network is unbalanced. For example, in inventory management or transportation scheduling, a single warehouse node may schedule according to its own inventory level, resulting in cross-level resource conflicts or transportation route competition, thereby reducing the overall operational efficiency of the supply chain.

[0003] Existing control strategies are typically based on static parameters or periodic updates, lacking real-time response capabilities. When faced with unforeseen events (such as material shortages, transportation delays, or node failures), existing methods struggle to adjust strategies quickly, leading to instability in the supply chain network and reduced robustness and reliability. Furthermore, the power supply chain exhibits a distinct hierarchical structure. Existing methods often neglect inter-level collaborative optimization, resulting in a lack of coordination in resource allocation and transportation scheduling among different levels of nodes, leading to frequent instances of inventory buildup, transportation bottlenecks, or material shortages. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a power material supply chain network equilibrium control method and apparatus based on value functions. By constructing a multi-level equilibrium control method based on value function-driven "local equilibrium control - sub-network collaborative optimization - cross-level coordination control", the present invention solves the problems of global imbalance caused by local optimization, insufficient inter-level coordination, poor real-time response capability and low control stability in the existing power material supply chain management.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Firstly, this application provides a power supply chain network equilibrium control method based on value functions. This method includes: generating local equilibrium control strategies for each node by selecting key neighboring nodes based on their local value functions and performing sparse communication through inter-node influence; forming sub-networks from nodes at the same level and coordinating and optimizing them within the sub-networks based on the local equilibrium control strategies to generate sub-network equilibrium control strategies; aggregating the equilibrium control strategies of each sub-network to a cross-level coordination layer, constructing global state information and determining coordination objectives, and outputting multi-level coordination control instructions after relaxation and iterative smoothing processing; dynamically adjusting the local strategies of each node according to the multi-level coordination control instructions and feeding back the execution results to each sub-network.

[0006] In one embodiment, the local value function is established based on the local state variables of each node, which include at least one of inventory level, material flow status, and resource capacity.

[0007] In one embodiment, based on the local value function of each node, key neighboring nodes are selected through influence among nodes and sparse communication is performed to generate local equilibrium control strategies for each node. This includes: each node exchanging information on local state variables with the selected key neighbors and fusing the exchanged information with its own state variables to form a local approximate representation; calculating the current optimal local strategy for each node based on the local approximate representation and the local value function; and iteratively updating each node according to the best response update rule: while keeping the strategies of key neighboring nodes fixed, resolving the strategy that maximizes its own local value function to obtain the next round of iterative strategy; repeating the iteration until the difference between two adjacent rounds of strategy is less than a preset convergence threshold, and outputting the final local equilibrium control strategy.

[0008] In one embodiment, each node exchanges local state variable information with the selected key neighborhoods, including: each node initializing its set of neighboring nodes, wherein the neighboring nodes are directly connected nodes with material flow relationships; calculating the comprehensive influence of each neighboring node on the current node's local strategy, wherein the comprehensive influence is determined by a weighted average of material flow, network topology centrality, and historical strategy contribution; sorting the neighboring nodes in descending order according to the comprehensive influence, selecting the node with the largest influence as the key neighborhood, and each node only exchanges state information with the key neighborhood to obtain neighborhood information.

[0009] In one embodiment, nodes at the same level are grouped into sub-networks. Within each sub-network, coordination and optimization are performed based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy. Specifically, this includes: dividing nodes at the same level that meet preset partitioning conditions into the same sub-network; within each sub-network, each node shares its own local state variables to form an overall state representation of the sub-network; based on the overall state representation of the sub-network, constructing an overall value function for the sub-network; and, under the condition that the constraints of each node are met, jointly adjusting the local equilibrium control strategies of each node within the sub-network through distributed optimization, and outputting the sub-network equilibrium control strategy after convergence.

[0010] In one embodiment, the equilibrium control strategies of each sub-network are aggregated to a cross-level coordination layer, global state information is constructed, and coordination objectives are determined. After relaxation and iterative smoothing, multi-level coordination control instructions are output. This includes: aggregating the equilibrium control strategies of each sub-network to a cross-level coordination layer, performing normalization processing, and constructing cross-level global state information; determining cross-level coordination objectives based on the cross-level global state information; identifying resource gaps and resource conflicts in each sub-network strategy according to the cross-level coordination objectives, and correcting the corresponding resource allocation parameters, task execution order parameters, and resource call priority parameters to generate candidate cross-level adjustment strategies; introducing a relaxation and iterative mechanism to smooth the candidate cross-level adjustment strategies, generating cross-level update strategies and distributing them to each sub-network, which then updates its equilibrium control strategy accordingly; repeating the aggregation, correction, and relaxation and iterative steps until convergence, and outputting multi-level coordination control instructions.

[0011] In one embodiment, a relaxation iteration mechanism is introduced to smooth the candidate cross-level adjustment strategy and generate a cross-level update strategy, including: obtaining the executed control strategy of the previous scheduling cycle as the initial control strategy; calculating the candidate update strategy based on the candidate cross-level adjustment strategy and the current system state; determining the relaxation coefficient according to the iteration convergence, and generating the smoothed cross-level update strategy using the initial control strategy, the candidate update strategy and the relaxation coefficient.

[0012] In one embodiment, the local strategies of each node are dynamically adjusted according to the multi-level coordination control instructions, and the execution results are fed back to each sub-network. This includes: acquiring the state change information of each node in real time; generating local strategy adjustment suggestions based on the state change information; each node dynamically updating and executing the local strategy according to the adjustment suggestions; and feeding back the execution results to the local sub-network and the upper-level coordination layer for the next round of strategy iteration.

[0013] Secondly, this application provides a power supply chain network equilibrium control device based on value functions. The device includes: a strategy generation module, used to generate local equilibrium control strategies for each node by filtering key neighboring nodes based on the local value functions of each node and performing sparse communication between nodes; a sub-network strategy generation module, used to form sub-networks from nodes at the same level, and perform coordination optimization within the sub-networks based on the local equilibrium control strategies to generate sub-network equilibrium control strategies; a global coordination module, used to aggregate the equilibrium control strategies of each sub-network to a cross-level coordination layer, construct global state information and determine coordination objectives, and output multi-level coordination control instructions after relaxation iterative smoothing processing; and a multi-level closed-loop iteration module, used to dynamically adjust the local strategies of each node according to the multi-level coordination control instructions and feed back the execution results to each sub-network.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the aforementioned value function-based power supply chain network equilibrium control method.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: Unlike conventional power supply chain control methods that employ centralized global optimization or hierarchical scheduling based on fixed topologies, this invention is driven by local value functions of nodes. It filters key neighboring nodes based on inter-node influence and implements sparse communication, retaining only interactions that significantly impact local decisions, thereby reducing the scale of information exchange and communication burden. Furthermore, it constructs sub-networks within the same level and performs coordinated optimization. At a cross-level coordination layer, it aggregates the balanced control strategies of each sub-network to form global state information. Through relaxed iterative smoothing under coordination objective constraints, it generates multi-level coordinated control commands, achieving a unification of local decision-making, hierarchical collaboration, and global coordination. Compared to conventional solutions, this invention organically combines value function-driven local autonomy, sub-network collaborative optimization, and cross-level coordination mechanisms. While ensuring the overall balance of the supply chain network, it effectively reduces computational complexity and communication overhead, improving the real-time response capability, control stability, and adaptability to resource supply and demand fluctuations and local disturbances in large-scale power supply chain networks. It possesses good scalability and engineering application value. Attached Figure Description

[0016] Figure 1 A schematic diagram of the power material supply chain network equilibrium control method based on value function provided in this application embodiment.

[0017] Figure 2A schematic diagram of the structure of a power material supply chain network equilibrium control device based on a value function provided in this application embodiment.

[0018] Figure 3 This is a structural diagram of an electronic device provided in an embodiment of this application.

[0019] Figure 4 This is a schematic diagram of key neighborhood information exchange in a distributed Nash equilibrium provided in an embodiment of this application.

[0020] Figure 5 This is a schematic diagram of the sub-network partitioning of nodes at the same level provided in an embodiment of this application.

[0021] Figure 6 A stability comparison diagram of the relaxation iteration mechanism provided in the embodiments of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] Reference Figure 1 As shown in the diagram, the power supply chain network equilibrium control method based on value function provided by this invention includes the following steps: S1 divides the nodes in the power supply chain network into manufacturing, warehousing and transportation layers according to their functions and levels, and establishes local value functions for each node to evaluate the long-term benefits of material scheduling, inventory management and transportation allocation at that node.

[0024] In this example, the nodes in the power supply chain network are divided according to function and hierarchy, and a local value function is established for each node, including: S11, Identify all participating nodes in the power material supply chain network, including manufacturers, warehousing centers, logistics and transportation nodes and construction units, and establish a network topology to clarify the material flow relationship between each node; S12, based on the main functions of the nodes in the supply chain, the identified nodes are divided into manufacturing layer, warehousing layer and transportation layer. The main functions of nodes in the supply chain include node production capacity, inventory capacity, transportation capacity, and main operational functions; the manufacturing layer is the node responsible for material production or procurement; the warehousing layer is the node responsible for material storage and inventory management; and the transportation layer is the node responsible for material transportation and distribution.

[0025] S13. Based on the classification of node functions, determine the hierarchical position of the node in the supply chain network and establish hierarchical relationships for subsequent multi-level coordination and control. The manufacturing layer is located at the upper level, the warehousing layer is located at the middle level, and the transportation layer is located at the lower level to form a top-down network structure. S14, Define local state variables for each node, including inventory level, material arrival status, transportation capacity, and production capacity, to describe the dynamic state of the node during network operation. Among them, the state variables are obtained through sensor data and ERP system interfaces, and are updated in real time or periodically according to a preset update cycle to accurately describe the dynamic state of nodes during network operation.

[0026] S15, based on the local state variables of the nodes, establish a local value function for each node.

[0027] The local value function Defined as a cost-minimizing function, the specific calculation formula is as follows:

[0028] In the formula, Let i be the state variable of node i. Let i be the state variable of the node adjacent to node i. For the node's own cost, For the timeliness of supplies, Due to resource constraints, For the sake of collaborative benefits, , , , These are the weighting coefficients, and all data have been normalized.

[0029] The node's own costs can be obtained by statistically analyzing operational data such as production costs, warehousing costs, transportation costs, and labor costs during the node's historical operating cycle through ERP systems, financial systems, or scheduling management systems. Material timeliness can be obtained by analyzing the actual time spent on materials from production and warehousing to transportation completion, such as recording transportation time, order response time, and arrival time through a logistics tracking system, and using average delay or overtime rate as timeliness indicators. Resource constraints can be determined based on the node's actual resource capabilities, including the rated capacity of production equipment, warehousing capacity, vehicle transportation capacity, and personnel scheduling capacity, and related data can be read through equipment ledgers, warehousing management systems, and transportation scheduling systems. Collaboration benefits can be obtained based on the historical collaborative scheduling results between nodes, such as statistically analyzing the success rate of joint transportation between nodes, the proportion of material sharing, the number of cross-node inventory adjustments, or the efficiency of collaborative task completion, and calculating the increase in revenue after node collaboration through historical operating data. Weighting coefficients can be obtained using empirical assignment methods, analytic hierarchy process (AHP), or normalized proportional allocation methods. For example, managers can pre-set initial weights based on the importance of cost, timeliness, constraints, and collaboration in actual business, ensuring that the sum of all weighting coefficients is 1.

[0030] In this embodiment, in order to eliminate the influence of different dimensions among parameters such as node cost, material timeliness, resource constraints and cooperation benefits in the local value function, the extremum normalization method (Min-Max normalization) commonly used in this field can be used to process each parameter.

[0031] S2, based on the local value function, adopts a distributed Nash equilibrium algorithm. By filtering key neighborhoods based on influence among nodes and conducting sparse communication, it generates local equilibrium control strategies for each node, including material allocation, inventory replenishment plans, and transportation route selection, so that each node can achieve the value function optimization objective in its own local environment.

[0032] In this embodiment, based on the local value function, a distributed Nash equilibrium algorithm is used. This involves filtering key neighborhoods based on influence among nodes and performing sparse communication to generate a local equilibrium control strategy for each node, including: S21, exchange information between the local state variables of each node and the key neighboring nodes to obtain neighborhood information; S22, the neighborhood information is fused with its own local state variables to form a local approximate representation of the overall state of the network by the node. The local information fusion can be carried out by weighted average or Laplace matrix method to reflect the importance of the neighborhood state. Among them, local approximation refers to the local estimate of the global network state formed by the fusion of the node's own state information and key neighborhood information, so as to approximately reflect the overall operating status under sparse communication conditions and support the optimization of local value functions.

[0033] S23, based on local approximation representation, each node independently calculates its optimal local strategy based on the local value function, including material allocation, inventory replenishment plan and transportation route selection; wherein, the node only updates its strategy based on the state of the key neighborhood, reducing the computational and communication burden; S24, each node takes the optimal local policy as the current iteration value and receives the latest policy information from the key neighboring nodes. It then adjusts its own policy according to the preset best response update rule. Among them, each node, while keeping the policies of other key neighboring nodes fixed, re-solves for the policy that minimizes its local value function to obtain the next round of iteration policy. S25, in each iteration, the nodes follow... The optimal response rule update strategy, where Let i be the local strategy for the (k+1)th iteration. This is the set of strategies for the key neighborhood nodes in the k-th iteration. The optimal response mapping for node i is used to find the optimal strategy for node i based on its local value function, given the strategies of key neighboring nodes. S26, when all nodes satisfy... Convergence is determined at the time of convergence, and the final local equilibrium control strategy is output. in, Let i be the local strategy for the k-th iteration. The vector norm, used to characterize the degree of difference between two adjacent iteration strategies, can be represented by the Euclidean norm. A preset convergence threshold is used to limit the allowable policy change error. When the policy change does not exceed this threshold, the node policy is considered to have stabilized. The local equilibrium control strategy refers to the resource allocation, inventory replenishment plan, and transportation route selection scheme obtained by each node through distributed Nash equilibrium iteration to optimize the local value function based on its own local state information and key neighboring node information, so as to achieve the optimal benefit of the node in the current local environment.

[0034] Furthermore, in S21, the local state variables of each node are exchanged with key neighboring nodes to obtain neighborhood information, such as... Figure 4 As shown, it includes: S211, based on the local state variables of the nodes, each node initializes its set of neighboring nodes, including all directly connected nodes with material flow relationships, for subsequent key neighbor filtering; S212, For each node in the neighboring node set, calculate its comprehensive influence on the local strategy of the current node; The specific formula for calculating the overall influence is as follows:

[0035] In the formula, Let represent the influence of node j on node i. For the flow of materials, Topological centrality is calculated by dividing node j by its degree centrality or betweenness centrality in the local network topology, and by the sum of the centralities of all its neighboring nodes. Contribution to historical strategy: the proportion of times node j has positively impacted the policy update of node i over a certain number of iterations (i.e., increased the value function value of node i). , , The weighting coefficients are α + β + γ = 1. Let be the total material flow between node i and all its neighboring nodes.

[0036] The formula normalizes and weights the real-time resource flow, network topology centrality, and historical policy contribution to obtain the comprehensive influence of each neighboring node on its local policy. This indicator is used for key neighbor selection, ensuring that distributed equilibrium control can accurately capture the influence of major neighbors even in the case of communication sparsity.

[0037] S213, based on the comprehensive influence, sort the set of neighboring nodes in descending order, and select the k neighboring nodes with the largest comprehensive influence as key neighbors, where k is an adjustable parameter used to control the degree of communication sparsity. Nodes only exchange information with key neighbors and ignore other neighboring nodes to achieve communication sparsity. S214, each node stores the list of key neighboring nodes and their status information in a local data structure, and exchanges status information only with key neighbors during the distributed iteration process through a predefined communication interface or protocol to obtain neighborhood information.

[0038] It should be noted that by combining local value functions, distributed Nash equilibrium algorithms, and influence-based key neighborhood selection mechanisms, a sparsity communication strategy is introduced into the distributed optimization process. By comprehensively considering resource flow, topological centrality, and historical policy contributions, the importance of neighborhood nodes is quantitatively evaluated, and only key neighborhoods are selected for information exchange and policy updates. This breaks through the traditional distributed control approach that relies on full neighborhood communication or centralized global information exchange. At the same time, through local approximate global state representation and distributed iteration mechanisms, nodes can achieve coordinated optimization of local policies and overall network equilibrium under limited communication conditions.

[0039] This scheme not only effectively reduces communication bandwidth usage and computational complexity, and improves real-time response capability and distributed convergence efficiency in large-scale power material supply chain networks, but also reduces resource conflicts and scheduling delays caused by information redundancy, making the generated material allocation, inventory replenishment plan and transportation route selection more reasonable.

[0040] S3. Nodes at the same level are grouped into sub-networks. Within the sub-networks, coordination and optimization are performed based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy.

[0041] In this embodiment, nodes at the same level are grouped into sub-networks. Within each sub-network, coordination and optimization are performed based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy, including: S31, within the same level, nodes that meet the preset partitioning conditions are divided into the same sub-network, forming several sub-networks at the same level, such as... Figure 5 As shown; The preset division conditions include: there is a direct or indirect material interaction relationship between nodes; the geographical distance between any two nodes does not exceed a preset distance threshold; and the number of nodes in the sub-network does not exceed a preset maximum number of nodes.

[0042] In one embodiment, the preset distance threshold setting method includes: taking the 85th percentile value based on the statistical distribution of historical material delivery radius to ensure that most nodes with material interaction relationships can be assigned to the same sub-network; the preset maximum number of nodes setting method includes: determining the number of nodes allowed to be included in a single sub-network based on the total number of nodes at the same level and the expected number of sub-networks, and using this number of nodes as the preset maximum number of nodes.

[0043] S32, within each sub-network, each node collects its own local state variables and shares the local state variables in vector form with other nodes in the sub-network through the internal communication interface of the sub-network, forming a sub-network state information set; S33, within the sub-network, a sub-network Laplace matrix is ​​constructed based on the connection relationships between nodes. Each node fuses the sub-network state information set based on the Laplace matrix to generate a sub-network overall state representation that reflects the overall resource distribution of the sub-network. S34. Based on the overall state representation of the sub-network, construct the overall value function of the sub-network, wherein the overall value function of the sub-network is a weighted sum of the local value functions of each node in the sub-network; S35, under the condition of satisfying the constraints of each node, the overall value function of the sub-network is optimized by a distributed optimization algorithm, and the local equilibrium control strategies of each node in the sub-network are jointly adjusted until convergence is determined. Then, the sub-network equilibrium control strategy is output and used as the input of the upper-level multi-level coordination or global equilibrium control to realize the hierarchical distributed equilibrium control of the power material supply chain network.

[0044] Subnetwork equilibrium control strategy refers to a material scheduling, inventory management and transportation control scheme that achieves overall resource allocation balance and optimal collaboration among nodes within the same level of subnetwork, based on the sharing of state information and joint optimization of the overall value function of the subnetwork.

[0045] The convergence determination process involves calculating the policy difference between two adjacent iterations in each iteration. The subnetwork's joint optimization process is considered converged when the policy difference for all nodes within the subnetwork is less than a preset convergence threshold. Constraints include node resource constraints, inventory constraints, and transportation constraints. Node resource constraints limit the amount of materials produced, processed, or allocated by each node during the joint optimization process to ensure the executability of the control strategy. Inventory constraints limit the node's inventory level to remain between the safety stock lower limit and the maximum inventory capacity upper limit to prevent material shortages or inventory backlogs and ensure inventory stability. Transportation constraints restrict the transportation routes and volumes selected by nodes to comply with existing transportation network conditions and capacity limitations to avoid generating unexecutable transportation plans and reduce transportation resource conflicts.

[0046] Joint adjustments are achieved through a distributed optimization algorithm: Specifically, each node first participates in subnetwork optimization by using its current material demand, inventory status, and feasible transportation routes as decision variables; then, information exchange and state fusion are carried out within the subnetwork, enabling each node to obtain state information reflecting the overall resource distribution of the subnetwork; based on this, each node iteratively updates its own material allocation, inventory replenishment plan, and transportation route selection while fixing the strategies of other nodes, and gradually eliminates resource conflicts and path competition among nodes through multiple rounds of distributed iteration, so that the decisions of each node converge under mutual coordination, thereby achieving a joint optimization result with more balanced material allocation, more reasonable inventory levels, and more efficient transportation routes within the subnetwork.

[0047] It should be noted that by dividing nodes at the same level into subnetworks and carrying out coordination and joint optimization within the subnetworks, nodes shift from making individual local decisions to collaborative decisions that take into account the overall state of the subnetwork, thereby effectively improving the overall balance and resource utilization efficiency of the supply chain operation. By sharing and integrating node state information within the subnetworks, the resource distribution and constraints (node ​​resource constraints, inventory constraints, and transportation constraints) within the subnetworks can be fully reflected, avoiding resource conflicts and path competition caused by information asymmetry. At the same time, by using distributed optimization algorithms to jointly adjust material allocation, inventory replenishment plans, and transportation route selection at the subnetwork level, not only is the computational and communication pressure of centralized optimization reduced, but the strategy convergence speed is also accelerated, enabling the subnetwork to form a stable equilibrium control strategy and serving as a reliable basis for multi-level coordination at higher levels.

[0048] S4 aggregates the equilibrium control strategies of each sub-network to the cross-level coordination layer for multi-level collaborative control. Through iterative optimization and strategy adjustment, it generates multi-level coordinated control instructions, enabling the entire supply chain network to achieve the global network equilibrium goal while satisfying the local equilibrium of each layer.

[0049] In this embodiment, the equalization control strategies of each sub-network are aggregated into a cross-level coordination layer for multi-level collaborative control. Through iterative optimization and strategy adjustment, multi-level coordinated control instructions are generated, including: S41, based on the aggregated sub-network equilibrium control strategy, the cross-level coordination layer organizes and normalizes the strategy information from different levels and different sub-networks to construct cross-level global state information that reflects the multi-level resource distribution, demand matching relationship and constraints. The cross-level global state information is used to describe the resource supply and demand relationship and its coupling characteristics among the manufacturing layer, storage layer and transportation layer. Normalization processing refers to unifying the dimensions, standardizing the data format and coordinating conflict constraints of the strategy information from different levels and sub-networks to form cross-level global state information that can be comprehensively analyzed.

[0050] S42, Determine the cross-level coordination objective based on cross-level global status information; The cross-level coordination objective is used to coordinate the flow of materials, inventory structure and transportation resource allocation between different levels, under the premise of satisfying the local equilibrium constraints of each sub-network, so as to reduce cross-level resource conflicts and improve overall operational efficiency.

[0051] S43, based on the cross-level coordination objective, parameters are extracted from the equilibrium control strategies corresponding to each sub-network to obtain resource demand information and resource allocation information corresponding to each equilibrium control strategy; the resource demand information and resource allocation information of different level sub-networks are compared item by item to identify resource gaps where the resource demand exceeds the resource allocation and resource conflicts where multiple sub-networks have competing demands for the same resource; combined with the resource remaining amount, resource capacity limit and scheduling constraints in the cross-level global state information, the resource gaps and resource conflicts are screened to determine cross-level adjustment objects that meet the preset judgment conditions; for each cross-level adjustment object, the resource allocation parameters, task execution order parameters and resource call priority parameters in the corresponding equilibrium control strategy are modified to generate candidate cross-level adjustment strategies; The preset determination conditions include: the same resource is requested simultaneously by two or more sub-networks and the total number of requests exceeds the allocatable amount of the resource; or the difference between the resource demand and the resource allocation exceeds a preset difference threshold.

[0052] In one implementation, the process of modifying the resource allocation parameters, task execution order parameters, and resource call priority parameters in the corresponding balanced control strategy is as follows: The rules for correcting the task execution order parameters are as follows: Based on the resource demand intensity and urgency of tasks in each sub-network, the task execution order is rearranged; whereby the resource demand intensity is determined by the ratio of the amount of resources required by the task to the amount of currently available resources, and the urgency is determined by the reciprocal of the time difference between the task deadline and the current time; the task execution order priority is calculated by weighted summation of resource demand intensity and urgency, and the task execution order is rearranged from high to low priority to generate the adjusted task execution order parameters; The rules for correcting the resource allocation priority parameter are as follows: First, for each sub-network under the same resource type, obtain the resource demand and the currently allocated resource amount, and calculate the resource gap ratio. The specific method is as follows: The difference between the resource demand and the allocated resources is used as the gap value. This gap value is then compared to the resource demand to obtain the resource gap ratio. Subsequently, the resource gap ratios of all sub-networks are compared, and a gap ratio threshold and half of that threshold are pre-set as dividing criteria. When the resource gap ratio of a sub-network is greater than or equal to the threshold, the sub-network is identified as a high-gap sub-network, and its resource allocation priority is increased by adding a preset fixed adjustment amount to its original priority value. When the resource gap ratio of a sub-network is less than or equal to half of the threshold, the sub-network is identified as a low-gap sub-network, and its resource allocation priority is decreased by reducing its original priority value by a preset fixed adjustment amount. When the resource gap ratio is between half and the threshold, its resource allocation priority remains unchanged. When multiple sub-networks compete for the same resource, they are sorted from high to low according to the adjusted resource allocation priority values, and the resource allocation order is determined accordingly. The preset fixed adjustment amount setting method includes: setting the preset fixed adjustment amount to 10% of the upper limit of the commonly used value range of the priority value (e.g., 1 to 10 points).

[0053] The rules for correcting resource allocation parameters are as follows: Under the premise of satisfying the total system resource constraints, the resource allocation of each sub-network is adjusted. First, a basic allocation ratio is calculated based on the resource demand of each sub-network, that is, the resource demand of each sub-network is compared with the total resource demand of all sub-networks to obtain its proportion in the overall demand. Then, the total available system resources are allocated to each sub-network according to the above proportions to obtain the basic allocated resource amount for each sub-network. When resource conflicts exist, the resource gap ratio is recalculated for each sub-network and adjusted accordingly: for sub-networks with a high gap ratio, a certain percentage of resource increment is added to their basic allocated resource amount; for sub-networks with a low gap ratio, a certain percentage of resource allocation is reduced accordingly. The resource increment or reduction is determined according to a pre-set adjustment coefficient and the degree of change in the gap ratio, so that sub-networks with higher gap ratios receive more resource increments, and sub-networks with lower gap ratios receive less resource allocation. Finally, the adjusted resource allocation results of all sub-networks are uniformly corrected to ensure that the sum of the resource allocation of each sub-network is consistent with the total available system resources, thereby satisfying the overall resource constraints.

[0054] S44. After obtaining the candidate cross-level adjustment strategy, a relaxation iteration mechanism is introduced to smooth the candidate cross-level adjustment strategy and generate the cross-level update strategy after relaxation iteration. S45, the cross-level update strategy is issued as a coordination instruction to the corresponding sub-network. After receiving the coordination instruction, each sub-network, while maintaining its original internal optimization framework, re-jointly optimizes its local material allocation, inventory replenishment plan and transportation route selection, and updates the sub-network balance control strategy. S46, the updated sub-network equilibrium control strategy is summarized again to the cross-level coordination layer as the historical strategy input for the next round of relaxation iteration. Through multiple rounds of relaxation iteration, the strategy relationship between the cross-level and sub-network is gradually adjusted. When the difference norm of the cross-level update strategy in two adjacent rounds of iteration is less than the preset convergence threshold, and the change amplitude of the equilibrium control strategy in two adjacent rounds of all sub-networks is less than the threshold, the multi-level distributed equilibrium control process is determined to have converged, and the final multi-level coordination control command is output.

[0055] The multi-level coordination control instructions are used to uniformly guide each level, sub-network and its internal nodes to perform material allocation, inventory management and transportation scheduling operations, thereby achieving global balanced operation of the power material supply chain network while ensuring local balance at each level.

[0056] Furthermore, after obtaining candidate cross-level adjustment strategies, a relaxation iteration mechanism is introduced to smooth the candidate cross-level adjustment strategies, generating a relaxed iteration-adjusted cross-level update strategy, including: S441, Obtain the control policies that have been executed at each level in the previous scheduling cycle, and use them as the initial control policies for the current relaxation iteration process; The initial control strategy includes material allocation, inventory replenishment plan, and transportation route selection, which serve as the baseline state for subsequent iterative updates.

[0057] S442, Based on the candidate cross-level adjustment strategy, and combined with the current system state and constraints, a distributed optimization algorithm is used to calculate the candidate update strategy. The candidate update strategy is the smoothing strategy for this iteration obtained by quantizing and solving the candidate cross-level adjustment strategy. Based on the candidate cross-level adjustment strategy, the cross-level coordination layer uses it as the initial correction direction. Combining the resource distribution, demand gap, and inventory level in the current system state, as well as the constraints between levels (such as material supply capacity, transportation capacity limits, and inventory capacity constraints), the cross-level coordination objective is decomposed into local sub-problems that can be independently optimized by each sub-network. Through a distributed optimization algorithm, the coordination layer issues correction parameters to each sub-network. Each sub-network calculates the optimal response strategy in parallel based on its own local value function and current state and returns it to the coordination layer. The coordination layer summarizes the optimization results of each sub-network, performs iterative updates and verifications according to global consistency constraints, and repeats the above process until convergence, thereby obtaining the candidate update strategy for this round of iteration. The specific calculation formula for the candidate update strategy is as follows:

[0058] In the formula, These are the candidate update strategies calculated in this iteration. This is a vector of decision variables for candidate cross-level adjustment strategies, including adjustable parameters such as material allocation, inventory replenishment, and transportation route selection. The constraint decision set for node i reflects various constraints under the current system state, such as maximum inventory capacity, maximum transportation capacity, material supply and demand balance constraints, and node resource limits. Let be the local value function of node i, which measures the value of node i given its own policy. Other node strategies The local value function value under the given conditions.

[0059] S443, The relaxation coefficient is determined by the historical iteration convergence and system stability. The relaxation coefficient is used to control the weighting ratio of historical policies and candidate policies to prevent excessive policy updates from causing oscillations and to balance convergence speed and stability. The relaxation coefficient is calculated using the following formula:

[0060] In the formula, Let be the relaxation coefficient in the k-th iteration. This represents the upper limit of the relaxation coefficient. These are the preset convergence adjustment parameters. Let i be the historical strategy of node i in the k-th iteration. Let be the historical strategy of node i in the (k-1)th iteration.

[0061] S444 generates the update control strategy after relaxation iteration based on the initial control strategy, candidate update strategies, and relaxation coefficients in a weighted combination manner. This strategy serves as the output of this round of relaxation iteration and is distributed to each sub-network for coordinated optimization.

[0062] like Figure 6 The figure shows a comparison of policy fluctuation curves using the relaxation iteration mechanism and the traditional direct update method. As can be seen from the figure, the traditional full update method exhibits significant oscillations during the iteration process, while the policy change curve is smoother and converges stably in a shorter time after adopting the dynamic relaxation coefficient of this invention. Simulation results show that, under the same network size, the global policy fluctuation amplitude is reduced by approximately 37% after adopting the relaxation iteration mechanism, and the system stability is significantly improved.

[0063] It should be noted that the advantages of introducing the relaxation iteration mechanism are as follows: by introducing a dynamic weighted smoothing process between historical strategies and candidate update strategies, the cross-level coordination strategy will not cause system oscillations, resource conflicts, or strategy instability due to a one-time large adjustment, thereby improving the stability and convergence accuracy of multi-level distributed equilibrium control in complex dynamic environments; at the same time, the mechanism can adaptively adjust the relaxation coefficient according to the historical iteration convergence and system stability, ensuring that the global coordination goal is gradually approached while taking into account the convergence speed and control smoothness, so that the resource allocation and scheduling relationship between the manufacturing layer, storage layer, and transportation layer can be gradually coordinated. Its innovation lies in the fact that it not only applies the relaxation iteration mechanism to the cross-level distributed equilibrium control process, but also combines cross-level global state information, supply and demand matching gap identification, and distributed optimization algorithms to dynamically decompose candidate cross-level adjustment strategies into multiple sub-network local optimization problems. Then, it adaptively generates relaxation coefficients through historical strategy feedback to achieve closed-loop iterative control of "cross-level coordination - local optimization - smooth feedback". Compared with the traditional coordination method of fixed step size or direct full update, it can effectively reduce the risk of strategy oscillation in large-scale power material supply chain networks and improve the real-time performance and global equilibrium capability of multi-level collaborative optimization.

[0064] S5 monitors the status changes of each node in real time, including changes in inventory, transportation status, and supply and demand. Based on multi-level coordination control instructions, it dynamically adjusts the local strategies of each node and feeds back the actual execution results to each level of sub-network, realizing a closed-loop cycle of multi-level distributed control to ensure that the supply chain network continuously approaches global equilibrium under different operating states.

[0065] In this embodiment, the state change information of each node is monitored in real time, and the local strategies of each node are dynamically adjusted according to multi-level coordinated control commands. The actual execution results are fed back to each level of sub-network to realize a closed-loop cycle of multi-level distributed control, including: S51 acquires real-time status change information of each node in the supply chain network and archives it according to categories, including inventory changes, transportation status changes and supply and demand fluctuations. S52 performs data preprocessing based on different types of state changes, such as missing value completion, outlier filtering, and time series smoothing, to obtain processed real-time state information. S53, based on real-time status information, maps each node or sub-network to the corresponding decision module in the current multi-level coordinated control command, and calculates the optimal local policy adjustment suggestion for the node in the current state through the policy mapping algorithm; Based on real-time status information, each node or sub-network first classifies and preprocesses the collected data on inventory levels, transportation status, and supply and demand changes. Then, through a strategy mapping algorithm, these dynamic states are mapped to the corresponding decision modules in the multi-level coordinated control instructions, that is, determining which local control rules or optimization objectives need to be activated or adjusted under the current network operating state. On this basis, the node or sub-network, combined with its own local constraints and the status information of key neighborhoods or sub-networks, uses distributed optimization algorithms (such as local Nash equilibrium, gradient descent, or ADMM) to calculate the optimal local strategy adjustment suggestions for material allocation, inventory replenishment plans, and transportation route selection. This ensures that in a dynamic environment, the node's decision-making can respond to system state changes in real time, while maintaining coordination with the sub-network and the global multi-level equilibrium strategy, achieving dynamic and adaptive local strategy optimization.

[0066] S54, each node dynamically updates its own local strategy and executes it based on the optimal local strategy adjustment suggestion, combined with local constraints and key neighborhood / sub-network information; After obtaining the optimal local strategy adjustment suggestion, each node first matches and corrects the suggestion with its own local constraints, including inventory upper and lower limits, transportation capacity, and resource availability restrictions, to ensure that the strategy is physically and logically feasible. Subsequently, the node combines the real-time status information of key neighborhoods or its subnetwork to evaluate the impact of the strategy adjustment on the neighborhood and the subnetwork as a whole, and smoothly corrects the strategy through distributed coordination or relaxation iteration methods to avoid network oscillations or conflicts caused by local adjustments. Finally, the node implements the corrected strategy in material allocation, inventory replenishment, and transportation route execution, and feeds back the execution results to the local subnetwork and the upper coordination layer in real time to achieve dynamic closed-loop updates, so that the local strategy remains consistent with the global multi-level equilibrium control within the constraints and is continuously optimized.

[0067] S55 collects execution results in real time, including actual material shipment volume, inventory changes and transportation completion status, and feeds the execution results back to the local sub-network and the upper-level coordination layer for the next round of multi-level equilibrium iteration; S56, Each node and sub-network updates its local state information and recalculates the multi-level equilibrium strategy based on the actual execution results. Through continuous closed-loop iteration, the strategies of nodes, sub-networks and the global network continuously approach the optimal equilibrium in the dynamic environment until the change in strategy is lower than the preset convergence threshold, thus completing the closed-loop control convergence determination. S57. After the closed-loop iteration converges, each sub-network outputs the final local equilibrium strategy. The upper coordination layer integrates the sub-network equilibrium strategies to form a global multi-level coordination control command. This strategy ensures the local equilibrium of each layer while realizing the global equilibrium and dynamic robust control of the supply chain network under different operating conditions.

[0068] It should be noted that by monitoring the inventory, transportation status, and supply and demand changes of each node in real time, dynamic perception of the supply chain network can be achieved. Combined with multi-level coordinated control commands and strategy mapping algorithms, the optimal local strategy can be adaptively calculated based on the real-time status of nodes and sub-networks. Dynamic and coordinated strategy updates are achieved through relaxation iteration and smooth correction. At the same time, the actual execution results are fed back to each level to achieve closed-loop iteration, so that nodes, sub-networks, and the global network continuously approach the optimal equilibrium in a dynamic environment. This improves the operational efficiency, resource utilization, and robustness of the supply chain network, ensures the synchronous optimization of local equilibrium and global equilibrium at each level, and realizes dynamic adaptive distributed control and global collaborative management.

[0069] Reference Figure 2 As shown in the diagram, the power supply chain network equilibrium control device based on value function provided by this invention includes a strategy generation module, a sub-network strategy generation module, a global coordination module, and a multi-level closed-loop iteration module, with connections between the modules: The strategy generation module is used to generate local equilibrium control strategies for each node by filtering key neighborhoods through the influence between nodes and performing sparse communication based on the local value function of each node. The sub-network strategy generation module is used to form sub-networks from nodes at the same level, perform coordination and optimization within the sub-networks, and generate sub-network equilibrium control strategies. The global coordination module is used to aggregate the equalization control strategies of each sub-network to the cross-level coordination layer, construct global state information and determine coordination objectives, and output multi-level coordination control instructions after relaxation iterative smoothing processing. The multi-level closed-loop iterative module is used to dynamically adjust the local strategies of each node according to the multi-level coordinated control instructions, and to feed back the execution results to each sub-network.

[0070] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0071] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the value function-based power supply chain network equilibrium control method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0072] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here. The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0073] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0074] The operating system 221 manages and controls the various hardware devices and computer programs 222 on the electronic device 20 to enable the processor 21 to perform calculations and processing on the massive amounts of data 223 in the memory 22. The operating system can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including a computer program capable of performing the value function-based power supply chain network equilibrium control method disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include data received by the electronic device from external devices, as well as data collected by its own input / output interface 25.

[0075] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0076] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0077] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0078] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0079] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power material supply chain network equilibrium control method based on value function, characterized in that, include: Based on the local value function of each node, key neighboring nodes are selected through the influence between nodes and sparse communication is carried out to generate a local equilibrium control strategy for each node. Nodes at the same level are grouped into sub-networks, and coordination and optimization are performed within the sub-networks based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy. The equalization control strategies of each sub-network are aggregated into the cross-level coordination layer, global state information is constructed and coordination objectives are determined, and multi-level coordination control instructions are output after relaxation iterative smoothing. The local strategies of each node are dynamically adjusted according to the multi-level coordination control instructions, and the execution results are fed back to each sub-network.

2. The power material supply chain network equilibrium control method based on value function according to claim 1, characterized in that, The local value function is established based on the local state variables of each node, which include at least one of inventory level, material circulation status and resource capacity.

3. The power material supply chain network equilibrium control method based on value function according to claim 1, characterized in that, The process involves selecting key neighboring nodes based on the local value function of each node, filtering them through the influence between nodes, and performing sparse communication to generate a local equilibrium control strategy for each node, including: Each node exchanges information on its local state variables with the selected key neighborhoods, and merges the exchanged information with its own state variables to form a local approximate representation. Based on the local approximation representation and the local value function, calculate the current optimal local strategy for each node; Each node iteratively updates according to the best response update rule: while keeping the policies of key neighboring nodes fixed, the policy that maximizes its own local value function is re-solved to obtain the next round of iterative policy; Repeat the iteration until the difference between two adjacent rounds of strategy is less than the preset convergence threshold, and output the final local equilibrium control strategy.

4. The power material supply chain network equilibrium control method based on value function according to claim 3, characterized in that, Each node exchanges information about its local state variables with the selected key neighborhood, including: Each node initializes its set of neighboring nodes, which are nodes that are directly connected and have a material flow relationship. Calculate the comprehensive influence of each neighboring node on the local strategy of the current node. The comprehensive influence is determined by a weighted average of three factors: resource flow, network topology centrality, and historical strategy contribution. The neighboring nodes are sorted in descending order according to their comprehensive influence, and the node with the greatest influence is selected as the key neighbor. Each node only exchanges state information with the key neighbor to obtain the neighborhood information.

5. The power material supply chain network equilibrium control method based on value function according to claim 1, characterized in that, The process of forming sub-networks from nodes at the same level, and then coordinating and optimizing within these sub-networks based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy, specifically includes: Nodes that meet the preset partitioning conditions within the same level are divided into the same sub-network; Within each subnetwork, each node shares its own local state variables, forming the overall state representation of the subnetwork; Based on the overall state representation of the subnetwork, an overall value function of the subnetwork is constructed. Under the condition of satisfying the constraints of each node, the local equilibrium control strategy of each node in the subnetwork is jointly adjusted through distributed optimization. After convergence, the equilibrium control strategy of the subnetwork is output.

6. The power material supply chain network equilibrium control method based on value function according to claim 1, characterized in that, The process involves aggregating the equilibrium control strategies of each sub-network to a cross-level coordination layer, constructing global state information, determining coordination objectives, and outputting multi-level coordination control instructions after relaxation and iterative smoothing processing. These instructions include: The equilibrium control strategies of each sub-network are aggregated into the cross-level coordination layer, normalized, and cross-level global state information is constructed. The cross-level coordination objective is determined based on the aforementioned cross-level global state information; Based on the cross-level coordination objectives, resource gaps and resource conflicts in each sub-network strategy are identified, and the corresponding resource allocation parameters, task execution order parameters, and resource call priority parameters are corrected to generate candidate cross-level adjustment strategies. A relaxation iteration mechanism is introduced to smooth the candidate cross-level adjustment strategies, generate cross-level update strategies and distribute them to each sub-network, and the sub-network updates its equilibrium control strategy accordingly. Repeat the summarization, correction, and relaxation iteration steps until convergence, and output multi-level coordinated control commands.

7. The power material supply chain network equilibrium control method based on value function according to claim 6, characterized in that, The introduction of a relaxation iteration mechanism to smooth the candidate cross-level adjustment strategies and generate cross-level update strategies includes: Obtain the executed control policy from the previous scheduling cycle as the initial control policy; Based on the candidate cross-level adjustment strategies and the current system state, calculate the candidate update strategies; The relaxation coefficients are determined based on the iterative convergence, and a smoothed cross-level update strategy is generated using the initial control strategy, candidate update strategy, and relaxation coefficients.

8. The power material supply chain network equilibrium control method based on value function according to claim 1, characterized in that, The local strategies of each node are dynamically adjusted according to the multi-level coordination control instructions, and the execution results are fed back to each sub-network, including: Real-time acquisition of status change information for each node; Based on the state change information, local strategy adjustment suggestions are generated; Each node dynamically updates and executes its local strategy based on the aforementioned adjustment recommendations; The execution results are fed back to the local sub-network and the upper-level coordination layer for use in the next round of policy iteration.

9. An apparatus for using the value function-based power supply chain network equilibrium control method as described in any one of claims 1-8, characterized in that, include: The strategy generation module is used to generate local equilibrium control strategies for each node by filtering key neighboring nodes through the influence between nodes and performing sparse communication based on the local value function of each node. The sub-network strategy generation module is used to form sub-networks from nodes at the same level, and to perform coordination and optimization within the sub-networks based on the local equilibrium control strategy to generate a sub-network equilibrium control strategy. The global coordination module is used to aggregate the equalization control strategies of each sub-network to the cross-level coordination layer, construct global state information and determine coordination objectives, and output multi-level coordination control instructions after relaxation iterative smoothing processing. The multi-level closed-loop iterative module is used to dynamically adjust the local strategies of each node according to the multi-level coordinated control instructions, and to feed back the execution results to each sub-network.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the value function-based power supply chain network equilibrium control method as described in any one of claims 1 to 8.

Images