Coordinated optimization method and system for classification disposal and green disassembly of power grid waste materials

Abstract

The application discloses a power grid waste material classification disposal and green disassembly collaborative optimization method and system, relates to the technical field of material classification disassembly, and comprises the following steps: obtaining multi-source characteristic data of power grid waste materials to be disposed; constructing a residual value prediction model based on the multi-source characteristic data, and outputting a component residual value mapping atlas; analyzing structure information of target materials, combining the residual value mapping atlas to generate a disassembly path decoupling graph; establishing a double-objective optimization scheduling model, solving the scheduling model, and generating a green disassembly operation sequence meeting a first constraint condition; collecting real-time feedback data when the disassembly operation sequence is executed, and dynamically correcting residual value prediction model parameters and scheduling model constraint weights. The application constructs a disassembly path decoupling graph and a disassembly state transition graph, realizes collaborative representation of component values in equipment and disassembly topologies, and effectively solves the problems of decoupling of residual value evaluation and path planning and lack of greenness evaluation in traditional methods.

Description

Technical Field

[0001] This invention relates to the field of materials classification and dismantling technology, and more specifically, to a collaborative optimization method and system for the classification, disposal and green dismantling of waste materials in power grids. Background Technology

[0002] With the accelerating pace of power equipment upgrades, a large amount of obsolete power grid materials, such as transformers, circuit breakers, cables, and control cabinets, are gradually being decommissioned. How to efficiently and environmentally classify, dispose of, and recycle these materials has become a crucial issue for improving the sustainable operation and maintenance capabilities and circular economy level of the power grid. Current technologies typically involve three stages in the disposal of obsolete power equipment: residual value assessment, dismantling route planning, and recycling scheduling. However, these stages are often modeled independently, lacking effective coordination and information coupling.

[0003] Specifically, current residual value assessment methods mostly rely on static equipment properties or single material composition for coarse-grained estimations, making it difficult to characterize the residual value transfer and recyclability potential between components in complex structures. Meanwhile, green dismantling path planning primarily relies on structural diagrams or topological constraints for path searching, ignoring residual value differences between components and environmental impact factors such as pollution diffusion and carbon emission costs. These methods fail to jointly model residual value assessment and dismantling paths, making it difficult to achieve overall optimization aimed at maximizing residual value, thus limiting the effectiveness of green dismantling and resource recycling.

[0004] The above-disclosed technical solutions have at least the following technical problems: the existing technology suffers from model decoupling between residual value assessment of waste materials and green dismantling path, making it impossible to achieve collaborative optimization scheduling based on maximizing residual value.

[0005] To address the above problems, this invention proposes a solution. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a collaborative optimization method for the classification and disposal of waste materials in power grids and green dismantling. This method integrates residual value prediction modeling with multi-source feature data and structural semantic decoupling graph construction, and combines reinforcement learning algorithms to achieve dual-objective scheduling optimization. This addresses the problems of decoupling between residual value assessment and green dismantling path model, lack of environmental friendliness and dynamic adaptability in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] On the one hand, the collaborative optimization method for the classification and disposal of waste materials in the power grid and green dismantling includes the following steps: acquiring multi-source characteristic data of waste materials to be disposed of; constructing a residual value prediction model based on the multi-source characteristic data and outputting a component residual value mapping map; analyzing the structural information of the target materials and generating a dismantling path decoupling map in combination with the residual value mapping map; establishing a dual-objective optimization scheduling model that maximizes residual value benefits and minimizes environmental impact, and solving the scheduling model to generate a green dismantling operation sequence that satisfies the first constraint condition; collecting real-time feedback data when executing the dismantling operation sequence and dynamically correcting the residual value prediction model parameters and scheduling model constraint weights.

[0009] In a preferred embodiment, the step of constructing a residual value prediction model based on multi-source feature data and outputting a component residual value mapping map specifically involves: structurally encoding the multi-source feature data to construct a component map; training the component map using a graph neural network model to output the structural residual value estimate of the component under structural connection and disassembly constraints; predicting the trend of recyclable material value changes over time using a time-series regression model based on historical sequence data of component operating conditions, and outputting a material residual value estimate; fusing the structural residual value estimate and the material residual value estimate, and introducing pollution risk factors and carbon emission factors for green correction to generate a final residual value score; and generating a component residual value mapping map based on the final residual value score.

[0010] In a preferred embodiment, the estimated structural residual value and the estimated material residual value are combined and green-corrected by incorporating pollution risk factors and carbon emission factors to generate a final residual value score. Specifically, the estimated structural residual value and the estimated material residual value for each component are received respectively; pollution risk factors and carbon emission factors are set, and a green penalty weight is activated when a component meets any of the following conditions: the pollution level exceeds a preset threshold, it contains hazardous materials, or the energy consumption intensity of its dismantling path exceeds the standard; the estimated structural residual value and the estimated material residual value are synthesized into a basic residual value score through a weighted fusion function, and dynamically adjusted according to the green penalty weight to generate the final residual value score; the dynamic adjustment satisfies that the basic residual value score is weighted by the weighted sum of the pollution risk factors and the carbon emission factors.

[0011] In a preferred embodiment, the step of parsing the structural information of the target material and generating a dismantling path decoupling diagram in combination with the residual value mapping map specifically involves parsing the structural configuration information into a structural semantic diagram.

[0012] Based on the residual mapping graph, the residual scores of each component are embedded into the structural semantic graph as initial node values ​​to generate a residual propagation subgraph. The structural embedding vector of the structural semantic graph and the residual embedding vector of the residual propagation subgraph are extracted by graph neural network. Vector fusion is performed by embedding alignment strategy based on cosine similarity to output component embedding vector. Based on the component embedding vector, a decoupling graph of disassembly path is constructed. The nodes in the graph represent the component disassembly state, and the edge weights are determined by the distance metric of the component embedding vector.

[0013] In a preferred embodiment, the step of extracting the structural embedding vector from the structural semantic graph and the residual embedding vector from the residual propagation subgraph using a graph neural network, and then performing vector fusion using a cosine similarity-based embedding alignment strategy to output a component embedding vector, specifically involves: inputting the structural semantic graph into a first graph neural network to generate node-level structural embedding vectors; simultaneously inputting the residual propagation subgraph into a second graph neural network to generate node-level residual embedding vectors; using the residual embedding vectors as query vectors and the corresponding structural embedding vectors as key and value vectors, calculating the attention weights of nodes to their structural neighbor nodes through an attention mechanism; weighting and aggregating the embedding features of structural neighbor nodes based on the attention weights to generate residual-guided structural awareness vectors; and inputting the structural awareness vectors and the original residual embedding vectors into a gating fusion unit to dynamically adjust their contribution ratio, outputting the final component embedding vector.

[0014] In a preferred embodiment, the dual-objective optimization scheduling model is specifically constructed as follows: A disassembly state transition diagram is constructed based on the disassembly path decoupling diagram; a state space and action space are defined based on the disassembly state transition diagram, where disassembly states are represented as combinations of component disassembly states, and actions are represented as component disassembly operations; a set of feasible state transition paths is generated by combining structural blocking dependencies and disassembly direction constraints; based on the set of feasible state transition paths, and combined with the first constraint condition, a dual-objective optimization scheduling model is constructed with maximizing cumulative residual value as the positive objective and minimizing the cumulative pollution factor and carbon emission factor of the disassembly path as the negative objective.

[0015] In a preferred embodiment, the step of constructing a disassembly state transition graph based on the disassembly path decoupling graph specifically involves: parsing the entity connection relationships, nested dependency relationships, and contamination propagation path information between components in the disassembly path decoupling graph; defining the expression method for disassembly states, representing each disassembly state as a state vector containing all component disassembly situations, forming an enumerable set of state spaces; defining state transition actions based on the set of state spaces, and combining the structural blocking edges and contamination path edges in the decoupling graph to filter out the set of legal actions in the current state, forming a set of state-action pairs; constructing state transition relationships for each state-action pair based on the set of state-action pairs to obtain the new state to which the specified disassembly operation is performed, and attaching multi-dimensional weight information to the transition edge; and constructing a complete disassembly state transition graph based on the new state after the transition.

[0016] In a preferred embodiment, solving the scheduling model to generate a green dismantling operation sequence that satisfies the first constraint involves: transforming the bi-objective optimization scheduling model into a single optimization index function using a weighted normalization method; solving the single optimization index function using a reinforcement learning algorithm to obtain an optimal dismantling action strategy set; and generating a dismantling action sequence step by step from the initial state based on the optimal dismantling action strategy set to generate a green dismantling operation sequence that satisfies the first constraint.

[0017] On the other hand, the collaborative optimization system for the classification, disposal, and green dismantling of waste materials in the power grid includes the following modules: a multi-source data acquisition module for acquiring multi-source characteristic data of waste materials to be disposed of; a residual value prediction and modeling module for constructing a residual value prediction model based on multi-source characteristic data and outputting a component residual value mapping map; a dismantling path generation module for parsing the structural information of the target materials and generating a dismantling path decoupling diagram based on the residual value mapping map; a green optimization scheduling module for establishing a dual-objective optimization scheduling model that maximizes residual value benefits and minimizes environmental impact, solving the scheduling model, and generating a green dismantling operation sequence that satisfies the first constraint condition; and a model dynamic correction module for collecting real-time feedback data when executing the dismantling operation sequence and dynamically correcting the residual value prediction model parameters and scheduling model constraint weights.

[0018] The technical effects and advantages of the synergistic optimization method for the classification, disposal, and green dismantling of waste materials in power grids according to this invention are as follows:

[0019] 1. This invention introduces a graph neural network and heterogeneous graph fusion strategy to unify the modeling of structural configuration information and component residual value scoring, constructing a dismantling path decoupling graph and a dismantling state transition graph, thus achieving a collaborative representation of component value and dismantling topology in structurally complex equipment. By introducing pollution risk and carbon emission factors into residual value prediction, the model's ability to perceive green indicators is significantly enhanced, ensuring that the generated dismantling path not only maximizes economic value but also considers environmental protection goals, effectively solving the problems of decoupling residual value assessment and path planning, and the lack of green evaluation in traditional methods.

[0020] 2. This invention constructs a reinforcement learning-driven dual-objective scheduling model and collects feedback data (such as actual pollution emissions, recycling rates, and energy consumption) during the entity dismantling process. This data is then used to dynamically adjust the weights in the residual value scoring model and path graph, improving the model's adaptability and generalization ability to different equipment types and actual operating conditions. This method supports continuous learning and rolling optimization across batches and devices, ensuring that the green dismantling operation sequence always matches the actual execution effect, thus improving the intelligent scheduling and green decision-making level of the entire system in multi-source heterogeneous power grid waste material scenarios. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the synergistic optimization method for the classification, disposal, and green dismantling of waste materials in the power grid according to the present invention.

[0022] Figure 2 This is a schematic diagram of the structure of the collaborative optimization system for the classification, disposal and green dismantling of waste materials in the power grid according to the present invention. Detailed Implementation

[0023] 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.

[0024] Example 1, Figure 1 This invention presents a synergistic optimization method for the classification, disposal, and green dismantling of waste materials in power grids, comprising the following steps:

[0025] S1, acquire multi-source characteristic data of the waste power grid materials to be disposed of; the multi-source characteristic data includes equipment type, years of operation, maintenance records, material composition ratio, pollution level and internal structural configuration information of the equipment;

[0026] In this embodiment, the structural configuration information is the component hierarchy, connection relationship, and disassembly direction information obtained from the analysis of equipment CAD drawings, digital twin models, or 3D scanning models, specifically:

[0027] After obtaining the CAD drawings and digital twin model of the equipment, the structural tomography algorithm is used to analyze the equipment model layer by layer to identify the constituent units, geometric features and interconnection methods of its internal components.

[0028] Based on the above analysis results, a component hierarchy structure tree is constructed, where each node represents a detachable component unit. Node attributes include component number, material type, expected mass, size parameters, and the level it belongs to. The parent-child relationship between nodes represents the assembly nesting relationship, and the horizontal edge represents the connection method between components (such as bolts, welding, snap-fit, etc.).

[0029] Based on the component hierarchy structure tree, combined with the equipment maintenance manual and manufacturing standard library, the disassembly directionality (e.g., "can only be pulled out from the top") and disassembly sequence dependency constraints (e.g., "component B must be disassembled after component A") of each connection method are deduced, and weight indicators such as estimated disassembly force, required tool level, and disassembly time cost are assigned.

[0030] The final structure configuration information tensor is formed, whose core fields include component ID, hierarchy depth, connection type, adjacent component ID, disassembly direction, expected energy consumption value and pollution release risk level, which are used to support subsequent residual value modeling and path graph construction.

[0031] S2, construct a residual value prediction model based on multi-source feature data, and output a component residual value mapping map;

[0032] The residual value prediction model is used to output the estimated recycling value, expected recyclability, and remanufacturing potential of each component unit under disassembly conditions.

[0033] In this embodiment, the construction of the residual prediction model based on multi-source feature data and the output of the component residual mapping map are specifically as follows:

[0034] Structured encoding is performed on multi-source feature data to construct a component map containing component physical attributes, operating condition parameters, and structural configuration information;

[0035] A graph neural network model is used to train the component map to obtain the estimated structural residual values ​​of each component under structural connection and disassembly constraints;

[0036] Based on historical sequence data of component operating conditions, a time-series regression model is constructed to predict the trend of recyclable material value changes over time and output the estimated residual value of the material.

[0037] The structural residual value estimate is integrated with the material residual value estimate to construct the final residual value score. During the integration process, pollution risk factors and carbon emission factors are introduced to make green corrections to the residual values ​​of high-pollution and high-energy-consumption components.

[0038] The component residual value mapping map is generated based on the final residual value score, which is used to guide the subsequent dismantling path optimization and recycling priority ranking.

[0039] The method involves fusing structural residual value estimates with material residual value estimates to construct a final residual value score. During the fusion process, pollution risk factors and carbon emission factors are introduced to perform green corrections on the residual values ​​of high-pollution and high-energy-consumption components. Specifically:

[0040] For each component, structural residual value and material residual value are obtained separately. The structural residual value is output by a graph neural network model, reflecting the disassembly difficulty and structural coupling effect of the component in the equipment structure. The material residual value is output by a time series regression model, reflecting the residual value and economic trend of recyclable materials.

[0041] Pollution risk factors and carbon emission factors are set, and green penalty weights are introduced for components with high pollution levels, containing hazardous materials, or with high energy consumption dismantling paths.

[0042] A weighted fusion function is used to synthesize the structural residual value estimate and the material residual value estimate. During the fusion process, the residual value score is adjusted for green purposes based on the component pollution risk factor and carbon emission factor to form the final residual value score.

[0043] The final residual value score is used to determine the priority dismantling order and environmental value weight in the subsequent dismantling path decoupling diagram.

[0044] The formula for calculating the final residual value score is as follows:

[0045]

[0046]

[0047]

[0048] in, For the final residual value score, To integrate the weighting coefficients, This is the estimated residual value of the structure. This is the reference selling price after the component is remanufactured. The integrity score is calculated based on sample statistics. The probability of successful disassembly (obtained based on historical maintenance sample analysis). To reduce dismantling costs, This is an estimate of the residual value of the material. Let the mass of material k in component i be , To estimate the recyclability, Let K be the current unit price. To reduce processing costs, , The preset green correction weight coefficient satisfies , As a pollution risk factor, This is the carbon emission intensity factor.

[0049] The pollution risk factor is based on a multi-dimensional scoring and normalization of the types and contents of harmful substances contained in the components, pollution accumulation during operation, and regulatory constraints; the carbon emission intensity factor is calculated based on the energy consumption of component dismantling operations and the unit carbon emission coefficient of the contained materials. The two together serve as a green correction coefficient to impose environmental constraints on the final residual value score.

[0050] S3, analyze the structural information of the target material and generate a dismantling path decoupling diagram by combining the residual value mapping map;

[0051] The nodes in the disassembly path decoupling diagram represent component units, and the edges represent disassembly connections. It also includes disassembly energy consumption, pollution risk and carbon emission factors to express green disassembly constraints.

[0052] In this embodiment, the step of analyzing the structural information of the target material and generating a dismantling path decoupling diagram by combining the residual value mapping map specifically involves:

[0053] The structural configuration information is parsed into a structural semantic graph that includes entity connection edges (representing physical connection relationships), constraint surfaces (representing detachable directional constraints), and nested regions (representing nesting or envelope relationships between components);

[0054] Based on the residual value mapping graph, the residual value scores of each component are embedded as the initial values ​​of the nodes into the structural semantic graph to form a residual value propagation subgraph;

[0055] Based on the semantic inconsistency of graph structure, the structural semantic graph and the residual propagation subgraph are embedded and encoded separately. The structural embedding vector and the residual embedding vector are obtained through graph neural network, and a heterogeneous graph embedding alignment strategy is used for fusion learning to obtain the component embedding vector.

[0056] Based on component embedding vectors, information from the structure graph and residual graph is fused to construct a decoupling graph for the disassembly path.

[0057] The process involves obtaining structural embedding vectors and residual embedding vectors through a graph neural network, and then using a heterogeneous graph embedding alignment strategy for fusion learning to obtain component embedding vectors. Specifically:

[0058] Embedding vectors are obtained by performing embedding learning on the structural semantic graph and the residual propagation subgraph using graph neural networks. With residual embedding vector Among them: structural embedding vector focuses on the deconstruction topology features of components in entity connections and nesting levels; residual embedding vector reflects the propagation, reinforcement or blocking patterns of component value in the network;

[0059] Semantic alignment is performed between the structural embedding vector and the residual embedding vector, and the residual embedding vector is... as query vector The corresponding structural embedding vector As keys and values, and calculate the attention weights of a node to its structural neighbor nodes;

[0060] Based on attention-weighted aggregation structure embedding neighbor features, a residual-guided structure-aware vector is obtained. ;

[0061] Structure-aware vectors The component embedding vector is generated by gating fusion with the residual embedding vector.

[0062] The attention weights are specifically as follows:

[0063]

[0064] The structure-aware vector is specifically:

[0065]

[0066] The component embedding vector is specifically:

[0067]

[0068] in, For attention weights, Let i be the transpose of the query matrix for node i. The preset attention mapping matrix, The structural embedding vector for node j. Let be the set of neighboring nodes of node i in the structural semantic graph. For structure-aware vectors For activation function, Embed vectors for components. , , These are learnable parameters, obtained through the backpropagation algorithm.

[0069] The semantic alignment of the structure embedding vector and the residual embedding vector also includes the following strategies:

[0070] Residual-guided attention mechanism: Guided by residual embedding vectors, it guides structural embedding to focus on structural neighbors with high residual paths;

[0071] Semantic contrast loss: Construct positive and negative sample pairs to make the embedding vectors of the same component converge in the shared space under different semantic graphs, while maintaining the distinction between different components;

[0072] Structure preservation constraint: Preserve the topological reachability features of the structure graph during the fusion process to prevent the loss of value-driven structural information.

[0073] S4. Establish a dual-objective optimization scheduling model that maximizes residual value and minimizes environmental impact, and solve the scheduling model to generate a green dismantling operation sequence that satisfies the first constraint condition, which is dismantling order, tool selection, and priority of hazardous components.

[0074] The operation sequence includes disassembly steps, operation parameters, residual value retention strategy and high-risk module warning information, which are used to guide the physical disassembly process.

[0075] In this embodiment, a dual-objective optimization scheduling model is established to maximize residual value and minimize environmental impact. The scheduling model is then solved to generate a green dismantling operation sequence that satisfies the first constraint condition, specifically:

[0076] Based on the dismantling path decoupling graph, a dismantling state transition graph is constructed. The nodes in the graph are regarded as dismantling operation units, the edges represent the reachability dependencies between operations, and the pollution impact factors and carbon emission path weights are mapped to edge weight functions to characterize the environmental impact path in the dismantling process.

[0077] Based on the disassembly state transition diagram, the state space and action space are defined. The disassembly state is represented as a combination of component disassembly states, and the action is represented as a disassembly operation of a certain component. By combining structural blocking dependencies and disassembly direction constraints, a set of feasible state transition paths is generated.

[0078] Based on the set of feasible state transition paths and combined with the first constraint, a dual-objective optimization scheduling model is constructed with the positive objective of maximizing the cumulative residual value benefit, where the residual value comes from the aforementioned residual value scoring map; and the negative objective of minimizing the cumulative pollution factor and carbon emission factor of the decomposed path.

[0079] The dual-objective optimization scheduling model is as follows:

[0080]

[0081]

[0082] in, To break down the operation sequence path, , This is the edge weight adjustment coefficient for the preset environmental loss factor. The pollution impact factors during the disassembly process from component i to component j. This represents the carbon emissions corresponding to this state transition. This represents the score for component i in the residual score map. This is to decompose the state transition edge from node i to j in the path.

[0083] The first constraint is as follows:

[0084] Based on the disassembly sequence constraint, operations that cannot be disassembled temporarily due to unresolved physical barriers are dynamically eliminated;

[0085] Based on tool selection constraints, a tool-component adaptation diagram is established to limit the tool selection space for each operation;

[0086] Based on the priority of dangerous components, a penalty coefficient is introduced in the scheduling of high-risk components to delay or prioritize their processing, thereby improving safety.

[0087] The construction of the disassembly state transition diagram based on the disassembly path decoupling diagram is as follows:

[0088] Based on the entity connection relationship, nested dependency relationship and pollution propagation path information between the components in the disassembly path decoupling diagram, the expression method of disassembly state is defined. Each disassembly state is represented as a state vector containing the disassembly status of all components, where each component indicates whether the corresponding component has been disassembled, forming an enumerable set of state spaces.

[0089] Based on the state space set, state transition is defined as "performing a disassembly operation on a component that meets the disassembly preconditions in the current state". Combining the structural blocking edges and pollution path edges in the decoupling graph, the set of legal actions in the current state is selected, that is, component disassembly operations that meet the conditions of no physical blocking, no nested dependencies that have not been resolved, and available tools that match, forming a set of state-action pairs.

[0090] Based on the set of state-action pairs, a state transition relationship is constructed for each state-action pair to obtain the new state to which the operation is transitioned after the specified disassembly operation is performed. Multidimensional weight information is attached to the transition edge, including the residual value released by the operation, the amount of pollution emissions caused, the carbon emission cost, and whether high-risk components are involved.

[0091] Based on the new state after the transition, a complete decomposition state transition graph is constructed, with all reachable states as nodes in the graph, state-action transitions as directed edges, and edge weights representing the comprehensive evaluation of each decomposition operation in terms of green value and environmental impact.

[0092] The solution yields a green dismantling operation sequence, specifically:

[0093] The bi-objective optimization scheduling model is transformed into a single optimization index function by using a weighted normalization method;

[0094] The optimal set of action strategies is obtained by solving a single optimization index function based on reinforcement learning algorithm.

[0095] Based on the optimal set of disassembly action strategies, the disassembly action sequence is generated step by step from the initial state, and the green disassembly operation sequence is obtained by parsing.

[0096] S5 collects real-time feedback data when executing the dismantling operation sequence, and dynamically corrects the residual value prediction model parameters and scheduling model constraint weights.

[0097] The feedback data includes indicators such as actual recyclability, pollutant leakage, and energy consumption deviation, enabling rolling iteration of the scheduling model and adaptation and migration of multiple batches of materials.

[0098] In this embodiment, real-time feedback data is collected during the execution of the dismantling operation sequence to dynamically correct the residual value prediction model parameters and scheduling model constraint weights, specifically as follows:

[0099] Based on feedback data, a deviation evaluation mechanism is established between the output results of the residual value prediction model and the model. The structural residual value prediction error and the material residual value prediction error are calculated separately. The actual observed values ​​of pollution risk factors and carbon emission factors are introduced to perform gradient correction on the relevant parameters in the fusion residual value scoring model in order to achieve adaptive updating of the model.

[0100] Based on the actual path and green cost recorded during entity execution, the residual value release, pollution propagation probability and energy consumption factor in the edge weights of the original decomposed state transition graph are updated to construct a feedback-corrected state transition graph, and on this basis, the objective function and constraint terms of the optimization scheduling model are dynamically adjusted.

[0101] The revised residual value prediction model and the optimized scheduling model are redeployed into the dismantling control system, and a new green dismantling operation sequence is output based on the updated model.

[0102] Example 2, Figure 2 The present invention provides a collaborative optimization system for the classification, disposal, and green dismantling of waste materials in power grids, comprising the following modules:

[0103] Multi-source data acquisition module: used to acquire multi-source characteristic data of power grid waste materials to be disposed of;

[0104] The residual value prediction modeling module is used to build a residual value prediction model based on multi-source feature data and output a component residual value mapping map.

[0105] Disassembly path generation module: used to parse the structural information of the target material and generate a disassembly path decoupling diagram by combining the residual value mapping map;

[0106] Green Optimization Scheduling Module: Used to establish a dual-objective optimization scheduling model that maximizes residual value and minimizes environmental impact, solve the scheduling model, and generate a green dismantling operation sequence that satisfies the first constraint condition;

[0107] Model dynamic correction module: used to collect real-time feedback data when executing the disassembly operation sequence, and dynamically correct the residual prediction model parameters and scheduling model constraint weights.

[0108] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0109] 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.

[0110] 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.

[0111] 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.

[0112] 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.

[0113] 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 synergistic optimization method for the classification, disposal, and green dismantling of waste materials in power grids, characterized in that... Includes the following steps: Acquire multi-source characteristic data of power grid waste materials to be disposed of; A residual value prediction model is constructed based on multi-source feature data, and a component residual value mapping map is output. Specifically, the multi-source feature data is structured and encoded to construct a component map; the component map is trained using a graph neural network model to output the structural residual value estimate of the component under structural connection and disassembly constraints; based on the historical sequence data of component operating conditions, the trend of recyclable material value change over time is predicted using a time-series regression model, and the material residual value estimate is output. The residual value estimates of the structure and materials are combined, and pollution risk factors and carbon emission factors are introduced for green correction to generate the final residual value score. Generate a component residual value mapping map based on the final residual value score; Analyze the structural information of the target material and generate a dismantling path decoupling diagram by combining the residual value mapping map; A dual-objective optimization scheduling model is established to maximize residual value and minimize environmental impact. The scheduling model is solved to generate a green dismantling operation sequence that satisfies the first constraint condition. Real-time feedback data is collected during the execution of the dismantling operation sequence to dynamically adjust the residual value prediction model parameters and scheduling model constraint weights.

2. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 1, characterized in that, The estimated residual value of the fused structure and the estimated residual value of the materials are combined, and pollution risk factors and carbon emission factors are introduced for green correction to generate the final residual value score, specifically: The structural residual value estimate and material residual value estimate for each component are received separately. Set pollution risk factors and carbon emission factors. When a component meets any of the following conditions, the green penalty weight will be activated: the pollution level exceeds the preset threshold, it contains hazardous materials, or the energy intensity of the dismantling path exceeds the standard. The structural residual value estimate and the material residual value estimate are synthesized into a basic residual value score by a weighted fusion function, and dynamically adjusted according to the green penalty weight to generate the final residual value score; the dynamic adjustment satisfies that the basic residual value score is reduced in weight by the weighted sum of pollution risk factors and carbon emission factors.

3. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 2, characterized in that, The process of analyzing the structural information of the target material and generating a dismantling path decoupling diagram by combining it with the residual value mapping map is as follows: The structural configuration information is parsed into a structural semantic graph; Based on the residual value mapping graph, the residual value scores of each component are embedded as the initial values ​​of the nodes into the structural semantic graph to generate a residual value propagation subgraph; The structural embedding vector of the structural semantic graph and the residual embedding vector of the residual propagation subgraph are extracted by graph neural network. Vector fusion is performed by embedding alignment strategy based on cosine similarity to output component embedding vector. A decoupling graph for the decomposition path is constructed based on the component embedding vector. In the graph, nodes represent the component decomposition state, and the edge weights are determined by the distance metric of the component embedding vector.

4. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 3, characterized in that, The structural embedding vector of the structural semantic graph and the residual embedding vector of the residual propagation subgraph are extracted separately through a graph neural network. Vector fusion is then performed using an embedding alignment strategy based on cosine similarity to output component embedding vectors. Specifically: The structural semantic graph is input into the first graph neural network to generate node-level structural embedding vectors; at the same time, the residual propagation subgraph is input into the second graph neural network to generate node-level residual embedding vectors. Using the residual embedding vector as the query vector and the corresponding structural embedding vector as the key vector and value vector, the attention weight of a node to its structural neighbor nodes is calculated through an attention mechanism. The embedded features of structural neighbor nodes are weighted and aggregated based on attention weights to generate a residual-guided structural perception vector. The structure-aware vector and the original residual embedding vector are input into the gating fusion unit, and the contribution ratio of the two is dynamically adjusted to output the final component embedding vector.

5. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 4, characterized in that, The specific construction method of the dual-objective optimization scheduling model is as follows: Based on the decoupling path diagram, a decomposition state transition diagram is constructed. Based on the disassembly state transition diagram, the state space and action space are defined. The disassembly state is represented as a combination of component disassembly states, and the action is represented as the component disassembly operation. Combining structural blocking dependencies and disassembly direction constraints, a set of feasible state transition paths is generated. Based on the set of feasible state transition paths and combined with the first constraint, a dual-objective optimization scheduling model is constructed with the positive objective of maximizing cumulative residual value revenue and the negative objective of minimizing the cumulative pollution factor and carbon emission factor of the decomposed path.

6. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 5, characterized in that, The construction of the disassembly state transition diagram based on the disassembly path decoupling diagram is as follows: Analyze and decouple the entity connection relationships, nested dependency relationships, and pollution propagation path information between the components in the decoupling path diagram; Define the way to express the disassembly state, and represent each disassembly state as a state vector containing the disassembly of all components, forming an enumerable set of state spaces; Based on the state space set, state transition actions are defined. Combining the structural blocking edges and contaminated path edges in the decoupling graph, the set of legal actions in the current state is selected to form a set of state-action pairs. Based on the set of state-action pairs, a state transition relationship is constructed for each state-action pair to obtain the new state to which the state is transitioned after performing a specified decomposition operation, and multi-dimensional weight information is attached to the transition edge. Construct a complete decomposed state transition diagram based on the new state after the transition.

7. The synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials according to claim 6, characterized in that, The process of solving the scheduling model to generate a green dismantling operation sequence that satisfies the first constraint is as follows: The bi-objective optimization scheduling model is transformed into a single optimization index function by using a weighted normalization method; The optimal set of action strategies is obtained by solving a single optimization index function based on reinforcement learning algorithm. Based on the optimal set of disassembly action strategies, a disassembly action sequence is generated step by step from the initial state, resulting in a green disassembly operation sequence that satisfies the first constraint condition.

8. A system using the synergistic optimization method for the classification, disposal, and green dismantling of power grid waste materials as described in any one of claims 1-7, characterized in that, Includes the following modules: Multi-source data acquisition module: used to acquire multi-source characteristic data of power grid waste materials to be disposed of; The residual value prediction modeling module is used to build a residual value prediction model based on multi-source feature data and output a component residual value mapping map. Disassembly path generation module: used to parse the structural information of the target material and generate a disassembly path decoupling diagram by combining the residual value mapping map; Green Optimization Scheduling Module: Used to establish a dual-objective optimization scheduling model that maximizes residual value and minimizes environmental impact, solve the scheduling model, and generate a green dismantling operation sequence that satisfies the first constraint condition; Model dynamic correction module: used to collect real-time feedback data when executing the disassembly operation sequence, and dynamically correct the residual prediction model parameters and scheduling model constraint weights.

Images