Carbon emission optimization-oriented muck disposal method and system

Abstract

The invention relates to the technical field of green construction scheduling, in particular to a muck disposal method and system oriented to carbon emission optimization. The method comprises the following steps: obtaining muck disposal data, and extracting muck fixed disposal data and muck non-fixed disposal data; performing carbon factor decoupling calculation on the fixed disposal data to obtain fixed carbon emission data; constructing a disposal element graph based on the non-fixed disposal data, and calculating an edge floating value to generate a non-fixed disposal graph; executing disturbance perception propagation, and constructing a disturbance perception graph; path carbon emission accumulative reasoning is carried out based on the disturbance perception graph, and path carbon emission expected data are obtained; calculating a path carbon emission probability to obtain non-fixed carbon emission data; and finally, fusing the fixed and non-fixed carbon emission data, carrying out carbon load optimization scheduling, and outputting a carbon load optimization result. According to the invention, low-carbon path evaluation and intelligent scheduling under a multi-source disposal behavior are realized, and the precision and execution efficiency of carbon emission control are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of green construction scheduling, and particularly to a muck disposal method and system for carbon emission optimization. Background Art

[0002] With the acceleration of the urbanization construction process, a large amount of muck generated in construction, demolition and foundation projects has become an important link in urban logistics scheduling and environmental management. Traditional muck disposal methods mainly rely on manual scheduling methods based on fixed transportation routes and dumping points, lacking the comprehensive perception and intelligent optimization capabilities for dynamic changes in transportation routes, fluctuations in operating conditions, and carbon emission loads, which easily lead to problems such as unreasonable vehicle scheduling, redundant path repetition, congestion during transportation peaks, and excessive overall carbon emission levels. At the same time, most of the existing carbon emission assessment methods are only based on static distance or total amount accounting, while the actual operating conditions involve more multi-source elements, resulting in insufficient carbon emission accuracy and being unable to support the needs of green construction and low-carbon scheduling. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a muck disposal method and system for carbon emission optimization to solve at least one of the above technical problems.

[0004] The present application provides a muck disposal method for carbon emission optimization, including the following steps: Step S1: Obtain muck disposal data; perform extraction of fixed muck disposal and extraction of non-fixed muck disposal on the muck disposal data to obtain fixed muck disposal data and non-fixed muck disposal data respectively; Step S2: Perform decoupling calculation of carbon factors based on the fixed muck disposal data to obtain fixed carbon emission data; construct a disposal element map for the non-fixed muck disposal data to obtain disposal element map data; Step S3: Calculate the edge floating value for the disposal element map data to obtain non-fixed disposal map data; perform disturbance perception propagation on the non-fixed disposal map data to obtain disturbance perception map data; perform cumulative reasoning of path carbon emissions on the disturbance perception map data to obtain expected path carbon emission data; calculate the carbon emission probability based on the expected path carbon emission data to obtain non-fixed carbon emission data; Step S4: Perform carbon load optimization scheduling based on the fixed carbon emission data and the non-fixed carbon emission data to obtain optimized muck carbon load data.

[0005] In the present invention, the construction waste disposal data is classified and processed according to fixed and non-fixed characteristics, effectively improving the pertinence of data modeling and the flexibility of scheduling. For the fixed disposal part, the carbon factor decoupling calculation method is adopted, which can accurately analyze the emission contributions in various transportation and operation links. For the non-fixed disposal part, by constructing a disposal element map and introducing the edge floating value and disturbance perception propagation mechanism, it can dynamically reflect the impacts of factors such as traffic, path selection, and environmental changes on the carbon emission path, and realize the intelligent prediction of carbon emissions in uncertain scenarios through path carbon emission expectation reasoning and probability estimation. Finally, through fusing the fixed and non-fixed carbon emission data for carbon load optimal scheduling, it can achieve global fine control of carbon emissions and reasonable allocation of scheduling resources, significantly different from the existing methods that cannot handle path disturbances and multi-variable carbon source evaluation, and having higher carbon emission response ability and scheduling accuracy.

[0006] Preferably, step S1 is specifically as follows: Obtain the construction waste disposal data; Construct a disposal behavior map according to the construction waste disposal data to obtain disposal behavior map data; Extract the time series features from the disposal behavior map data to obtain time series feature data; Divide the disposal behavior map data according to the time series feature data to obtain construction waste fixed disposal data and construction waste non-fixed disposal data respectively.

[0007] In the present invention, the disposal behavior map expresses the operation processes such as construction waste generation, transportation, and dumping in the form of a graph structure with node and behavior relationships, which not only clearly depicts the relevance between different operation links but also lays a foundation for path reasoning and disturbance propagation. On this basis, the system further extracts the time series features in the map, including dynamic information such as operation frequency, scheduling period, and path stability, so as to realize the temporal analysis of the construction waste disposal behavior. By dividing the above map data according to the time series features, the system can accurately distinguish the fixed operations (such as daily fixed dumping) and non-fixed operations (such as temporary scheduling and path change) in the disposal behavior, which is significantly different from the traditional data processing method divided according to static rules.

[0008] Preferably, the carbon factor decoupling calculation is specifically as follows: Extract the fixed disposal working condition factors according to the construction waste fixed disposal data to obtain fixed disposal working condition factor data; Perform matrix decomposition calculation on the fixed disposal working condition factor data to obtain fixed disposal decomposition data; Estimate the unit carbon emissions according to the fixed disposal decomposition data to obtain fixed carbon emission data.

[0009] In the present invention, the system extracts key operating condition factors affecting carbon emissions from the fixed disposal data of construction waste, such as transportation distance, vehicle load, operation frequency, energy consumption level, etc., and constructs a multi-dimensional dataset of operating condition factors. The matrix decomposition method is used to structurally decouple the above multi-variable factors, and the complex coupling relationship is transformed into a set of basic feature factor matrices with linear interpretability, so as to realize the quantitative splitting of the carbon source contribution of different disposal paths. This method not only solves the problem that it is difficult to independently model the cross-influence of multiple factors in traditional methods, but also can dynamically adjust the unit carbon emission valuation according to different operating condition weights. Through the estimation of unit carbon emissions based on the decomposition results, more accurate and controllable fixed carbon emission data can be obtained.

[0010] Preferably, the construction of the disposal element graph is specifically as follows: Extract the disposal attributes from the non-fixed disposal data of construction waste to obtain disposal attribute data; Construct a graph structure according to the disposal attribute data to obtain disposal attribute graph data; Perform disposal node partitioning on the disposal attribute graph data to obtain disposal partition graph data; Perform disposal evolution on the disposal partition graph data to obtain disposal evolution graph data.

[0011] In the present invention, the system extracts key disposal attributes from the non-fixed disposal data of construction waste, including disposal location, path selection, dumping method, operation frequency, etc., and constructs a disposal attribute dataset; based on these attributes, a graph structure model is constructed, and various disposal units are used as nodes in the graph, and different attribute similarities or operation reachabilities are used as edges to obtain a complete disposal attribute graph. This graph can not only intuitively reflect the potential substitutable relationship between different disposal points, but also has good structural computability. By clustering or regionalizing the nodes in the graph structure, a disposal partition graph is formed to realize the optimization modeling of zoning based on geography or function. On the basis of the partition graph, time series or behavior-driven evolution rules are introduced to generate a disposal evolution graph, which is used to express the dynamic evolution process of different regional disposal strategies in the construction waste scheduling process.

[0012] Preferably, the calculation of the edge floating value is specifically as follows: Embed perturbation factors into the disposal element graph data to obtain graph embedding data; Perform transfer inertia analysis on the graph embedding data to obtain transfer inertia edge data; Calculate the edge floating value for the transfer inertia edge data to obtain edge floating value data; Adjust the edge weights of the disposal element graph data according to the edge floating value data to obtain non-fixed disposal graph data.

[0013] In the present invention, the system embeds external disturbance factors such as traffic congestion, meteorological conditions, equipment availability, and construction restrictions into the disposal element map data to generate graph embedding data, thereby enhancing the expression ability of the graph structure for real external environment fluctuations. By analyzing the temporal behavior of the transfer relationship between nodes, quantifying the transfer inertia of each path in historical scheduling, constructing transfer inertia edge data, and characterizing the stability of path selection and the path jump trend. Based on this inertia data, the system further calculates the edge floating value, that is, the fluctuation degree of attributes such as carbon emissions, cost, or time of each edge under the influence of disturbances. Using the floating value to adjust the edge weights in the original graph, generating dynamic adaptive non-fixed disposal graph data, providing a highly reliable structural basis for disturbance perception propagation and path carbon emission reasoning.

[0014] Preferably, the disturbance perception propagation specifically is: Extract the disturbance source nodes from the non-fixed disposal graph data to obtain the disturbance source node data; Extract the disturbance attributes according to the disturbance source node data to obtain the disturbance attribute data; Extract the disturbance coupling characteristics from the non-fixed disposal graph data according to the disturbance attribute data to obtain the disturbance coupling characteristic data; Construct a disturbance weight graph according to the disturbance coupling characteristic data to obtain the disturbance weight graph data; Use the disturbance weight graph data to perform Gaussian diffusion on the non-fixed disposal graph data to obtain the preliminary disturbance graph data; Suppress the disturbance coefficient of the preliminary disturbance graph data to obtain the disturbance perception graph data.

[0015] In the present invention, the system extracts disturbance source nodes (such as abnormal traffic nodes, construction conflict areas, meteorological risk areas, etc.) from the non-fixed disposal graph and analyzes their disturbance attributes (influence intensity, influence range, duration, etc.) to form disturbance attribute data. The system extracts disturbance coupling characteristics according to the interaction relationship between the disturbance attributes and the graph structure, thereby revealing the propagation potential and influence path of the disturbance factors in the graph structure. Using the disturbance coupling characteristics as the weight basis, a disturbance weight graph is constructed to realize the structural mapping of path sensitivity. The system uses the Gaussian diffusion mechanism to perform propagation calculations on this weight graph, simulating the process of layer-by-layer transmission of the disturbance influence in the path network, generating the preliminary disturbance graph data, and truly reflecting the carbon emission structure changes caused by the disturbance. Through the disturbance coefficient suppression mechanism, the propagation result is constrained and converged to generate the disturbance perception graph data to filter out the noise effect brought by low-intensity disturbances.

[0016] Preferably, the disturbance coupling characteristic extraction specifically is: Perform disturbance attribute projection on the non-fixed disposal graph data according to the disturbance attribute data to obtain the disturbance attribute projection data; Perform perturbation coupling calculation on the perturbed attribute projection data to obtain perturbed coupling data; Perform cross-node connection processing based on the perturbed coupling data to obtain cross-node connection data; Perform perturbation propagation intensity calculation on the perturbed coupling data according to the cross-node connection data to obtain perturbed coupling characteristic data.

[0017] In the present invention, the system projects the perturbed attribute data (such as traffic resistance level, meteorological anomaly category, construction operation fluctuation, etc.) onto a non-fixed disposal graph to obtain perturbed attribute projection data, so that the perturbed information has spatial correspondence on the graph structure. Based on the node attribute similarity and edge connection relationship, the system performs perturbation coupling calculation on the projection data to quantify the response degree of the perturbation factor between adjacent path nodes and obtain perturbed coupling data. On this basis, the system further identifies the indirect influence paths between different region or partition nodes, constructs cross-node connection data, and thus expands the propagation boundary of the perturbed influence. According to the cross-node connection situation, the system performs propagation intensity modeling on the perturbed coupling data to obtain perturbed coupling characteristic data, comprehensively expressing the dynamic diffusion ability and node sensitivity of the perturbation factor in the graph structure.

[0018] Preferably, the specific process of cumulative reasoning of path carbon emissions is as follows: Extract the perturbed path set from the perturbed perception graph data to obtain perturbed path set data; Perform path carbon factor mapping according to the perturbed path set data to obtain path carbon mapping data; Perform graph attention network reasoning according to the path carbon mapping data to obtain path carbon emission expectation data.

[0019] In the present invention, the system extracts the path set affected by the perturbation factor in the perturbed perception graph to form perturbed path set data, ensuring that the model focuses on the key traffic routes within the perturbation propagation range. The system performs carbon factor mapping on each path in the path set according to the path structure and historical scheduling characteristics, converts factors such as path length, traffic status, transportation type, operation frequency, etc. into a structured carbon emission characteristic expression, and forms path carbon mapping data. Based on the graph attention network, dynamic weighting is performed through the attention mechanism between nodes, highlighting the weight contributions of the perturbation-sensitive paths and high-emission paths in the reasoning process, so as to realize the expectation reasoning of path carbon emissions.

[0020] Preferably, the specific process of carbon emission probability calculation is as follows: Perform path carbon emission distribution fitting according to the path carbon emission expectation data to obtain path carbon emission distribution data; Perform perturbed perception weight fusion on the path carbon emission distribution data to obtain path carbon emission probability data; Perform non-fixed disposal carbon emission probability estimation on the path carbon emission probability data to obtain non-fixed carbon emission data.

[0021] In the present invention, the system fits the carbon emission distribution of each candidate path based on the expected data of path carbon emissions, generates path carbon emission distribution data, thereby expanding from a static expected value to a dynamic probability density expression, and depicting the change trend of carbon emissions under different perturbation conditions. The system introduces a perturbation perception weight, takes factors such as the perturbation influence intensity and the path historical sensitivity as weighted parameters, and embeds them into the carbon emission distribution model to achieve a responsive adjustment of the path carbon emission probability under different perturbation situations. Based on the fused probability data, the system completes the carbon emission probability estimation in the non-fixed disposal scenario, obtains the non-fixed carbon emission data, and provides highly reliable input for scheduling optimization and low-carbon strategy formulation.

[0022] Preferably, step S4 is specifically as follows: Perform carbon load modeling on the fixed carbon emission data and the non-fixed carbon emission data to obtain a carbon load model; Set scheduling variables for the carbon load model to obtain a carbon load scheduling model; Set constraint conditions according to the carbon load scheduling model to obtain a carbon load constraint model; Perform optimized solution of the carbon load for the carbon load constraint model to obtain optimized data of the muck carbon load.

[0023] In the present invention, the system incorporates the fixed carbon emission data and the non-fixed carbon emission probability data into the carbon load modeling process, constructs a carbon load model, and realizes the unified quantitative expression of the total carbon emissions at the scheduling level. The system sets scheduling variables according to factors such as the configuration of transport vehicles, operation cycles, and path accessibility to form a carbon load scheduling model, thereby enabling schedulability for multiple paths, multiple resources, and multiple stages. Combining external limit conditions such as urban management regulations, emission caps, traffic control, and operation time windows, a carbon load constraint model is constructed to ensure that the optimized solution has practical feasibility and policy compliance. By introducing linear programming, integer programming, or multi-objective evolutionary algorithms to optimize and solve the carbon load constraint model, a muck scheduling plan that meets the goal of minimizing carbon emissions is output, forming optimized data of the muck carbon load.

[0024] Preferably, the present application also provides a muck disposal system for carbon emission optimization, which is used to execute the muck disposal method for carbon emission optimization as described above. The muck disposal system for carbon emission optimization includes: A muck disposal data parsing module, which is used to obtain muck disposal data; perform extraction of muck fixed disposal and extraction of muck non-fixed disposal on the muck disposal data to respectively obtain muck fixed disposal data and muck non-fixed disposal data; A carbon factor modeling and structure diagram construction module, which is used to perform decoupling calculation of carbon factors according to the muck fixed disposal data to obtain fixed carbon emission data; construct a disposal element diagram for the muck non-fixed disposal data to obtain disposal element diagram data; A non-fixed path carbon emission reasoning module is used to calculate the edge floating value of the disposal element map data to obtain non-fixed disposal map data; perform disturbance-aware propagation on the non-fixed disposal map data to obtain disturbance-aware map data; perform path carbon emission cumulative reasoning on the disturbance-aware map data to obtain path carbon emission expectation data; calculate the carbon emission probability according to the path carbon emission expectation data to obtain non-fixed carbon emission data; A carbon load optimization scheduling module is used to perform carbon load optimization scheduling according to the fixed carbon emission data and the non-fixed carbon emission data to obtain optimized muck carbon load data.

[0025] The beneficial effects of the present invention are as follows: A muck disposal method for carbon emission optimization proposed by the present invention realizes refined evaluation and global optimization control of carbon emissions during the muck scheduling process based on the differential structure modeling and dynamic response mechanism of fixed and non-fixed disposal paths, and has significant green and low-carbon scheduling advantages. By constructing a behavior graph and performing temporal feature analysis on the muck disposal data, the system can accurately distinguish the stability types of disposal tasks, form dual-track processing paths of fixed and non-fixed, and improve the adaptability and hierarchical response ability of the scheduling strategy. In terms of the fixed path, through carbon factor decoupling and matrix decomposition, fine estimation of unit carbon emissions under multi-dimensional working conditions is realized; in terms of the non-fixed path, the system constructs a disposal element map and integrates edge floating value calculation and disturbance-aware propagation to truly reflect the diffusion process of environmental disturbances on path carbon emission changes, and completes the expectation reasoning and probability modeling of path carbon emissions with the help of a graph attention network, having the technical advantages of strong dynamics and accurate response. By constructing a carbon load scheduling optimization model and introducing multi-variable constraints and disturbance adaptation mechanisms, a muck scheduling scheme with minimized global carbon emissions is generated. Compared with traditional static scheduling methods, the present invention can not only dynamically perceive the carbon emission risks under working conditions, but also effectively coordinate the contradictions among path selection, resource allocation, and emission load, significantly improving the intelligent and scientific level of carbon emission reduction work during urban construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Other features, purposes, and advantages of the present application will become more obvious by reading the detailed description of the non-restrictive embodiments with reference to the following drawings: Figure 1 Shows a flowchart of the steps of a muck disposal method for carbon emission optimization according to an embodiment; Figure 2 Shows a flowchart of the steps of a muck disposal data parsing method according to an embodiment; Figure 3 Shows a flowchart of the steps of a carbon factor decoupling calculation method according to an embodiment; Figure 4 Shows a flowchart of the steps of an edge floating value calculation method according to an embodiment; Figure 5 The flowchart of the steps of a carbon load optimization scheduling method according to an embodiment is shown. Detailed implementation manners

[0027] The technical method of the present invention patent will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0028] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures denote the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor means and / or microcontroller means.

[0029] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0030] Please refer to Figures 1 to 5 , the present application provides a method for muck disposal oriented to carbon emission optimization, including the following steps: Step S1: Obtain muck disposal data; perform extraction of fixed muck disposal and extraction of non-fixed muck disposal on the muck disposal data to obtain fixed muck disposal data and non-fixed muck disposal data respectively; In one embodiment, the system obtains construction waste disposal data, which can be sourced from various heterogeneous acquisition channels such as the construction site scheduling system, GPS devices of transport vehicles, in-vehicle terminals, weighing systems, and historical operation records. The original construction waste disposal data includes, but is not limited to, the following field information: task number, loading time, transportation route trajectory, dumping location, vehicle type, operation frequency, etc. Based on the above data, the system classifies and identifies construction waste disposal behaviors. Specifically, the system sets a determination rule for fixed disposal routes: if a specific disposal route or dumping point appears with a frequency of 70% or more in no less than ten past operation records, then this route is marked as a fixed disposal route. Such routes usually exhibit characteristics such as stable scheduling behavior and concentrated dumping points. The remaining routes that do not meet this condition are marked as non-fixed disposal routes, which are characterized by high scheduling dynamics and large fluctuations in dumping point selection. The system divides the extraction results into two categories: construction waste fixed disposal data, denoted as , representing the set of all disposal records determined to be fixed routes; and construction waste non-fixed disposal data, denoted as , representing the set of all disposal records belonging to non-fixed routes.

[0031] Step S2: Perform carbon factor decoupling calculation based on the construction waste fixed disposal data to obtain fixed carbon emission data; construct a disposal factor map for the construction waste non-fixed disposal data to obtain disposal factor map data; In one embodiment, the system extracts key operating condition factor dimensions from the fixed route operation data, including but not limited to the following five items: transportation distance (denoted as L), vehicle type (V), loading weight (W), road condition level (R), and dumping operation time (T). The system constructs a factor matrix M with each fixed disposal task as a row, and the matrix dimension is m rows and k columns, where m represents the number of fixed operation tasks and k is the number of factor dimensions. Each row represents the factor vector corresponding to one operation. The system uses the non-negative matrix factorization (NMF) method to construct a model for factor decomposition of the above operating condition factor matrix M. Specifically, the matrix M is decomposed into the product of two non-negative matrices: one is the factor intensity matrix F, representing the proportion of each operation on each latent factor; the other is the factor carbon emission contribution matrix C, representing the contribution value of each operating condition factor to unit carbon emission. Based on this decomposition structure, the system further calculates the carbon emissions of each fixed route. Specifically, it multiplies the factor intensity vector corresponding to the operation and the factor carbon emission weight vector by dimension and then sums them to obtain the estimated value of unit carbon emission for this operation. Output the fixed carbon emission data set, denoted as , which represents the set of carbon emission values corresponding to all fixed disposal tasks. For the non-fixed path part, the system constructs a disposal element graph structure based on attribute perception. Path attribute information is extracted from non-fixed scheduling records, including the dumping point number, estimated transportation mileage, passable time period, traffic sensitivity level, and operation reliability score, etc. Each potential dumping point is regarded as a node in the graph, forming a node set. The system constructs the edge connection relationship in the graph structure based on the following rules, including that if the number of times two dumping points co-occur in the historical scheduling path is not less than two, or the similarity between them in the path space reaches or exceeds 0.8 (such as path coincidence rate or traffic feature similarity), then an edge is established between them. The attributes of each edge include transfer probability, path time consumption estimation, disturbance sensitivity, etc., which are used for propagation modeling. The system uses the Louvain modular clustering algorithm to perform node partitioning operations on the graph structure, forming multiple local structure clustering regions. Each partition represents a subset of dumping nodes with redundancy, substitutability, or similar scheduling characteristics in terms of scheduling, providing structural support for the deployment of path disturbance propagation and scheduling switching strategies. The system further introduces a time series marking mechanism within each partition to model the scheduling order between nodes, recording characteristic information such as the probability of switching from the main path to other paths, switching time window, and frequency, and constructing a disposal evolution graph.

[0032] Step S3: Calculate the edge floating value for the disposal element graph data to obtain non-fixed disposal graph data; perform disturbance-aware propagation on the non-fixed disposal graph data to obtain disturbance-aware graph data; perform cumulative inference of path carbon emissions on the disturbance-aware graph data to obtain path carbon emission expectation data; calculate the carbon emission probability based on the path carbon emission expectation data to obtain non-fixed carbon emission data; In one embodiment, the system embeds the set of external disturbance factors into the edge structure of the disposal factor graph. The disturbance factors include traffic delay index, meteorological impact level, equipment status availability, construction impact weight, etc. For any edge in the graph, its disturbance impact value can be obtained through weighted linear combination. Specifically, the above four types of disturbance factors are weighted and summed with preset weights to obtain the total disturbance impact value corresponding to each edge. The system conducts frequency statistical analysis on the historical scheduling transfer sequence of each edge and calculates the scheduling inertia coefficient of the edge, that is, the degree of continuity of the path in historical scheduling. Combining the disturbance impact value and transfer inertia, the system defines the edge floating value as the product of the disturbance intensity and the scheduling instability degree, which is used to characterize the fluctuation risk of the path under the current disturbance conditions. The system corrects the original edge weight according to the floating value. Specifically, the edge floating value is directly added to the original weight to obtain the edge weight after disturbance response, which is used to construct the non-fixed disposal graph structure. The system performs disturbance-aware propagation modeling. Extract the disturbance source nodes. The definition method is as follows: If the floating value of an edge is significantly higher than the statistical mean of the floating values of all edges, that is, greater than the mean plus one and a half times the standard deviation, then the nodes at both ends of the edge are marked as disturbance source nodes. Subsequently, a disturbance propagation graph is constructed with these disturbance sources as the center, and the disturbance propagation intensity between any two nodes in the graph is calculated. The propagation process adopts the Gaussian diffusion modeling method, and the propagation weight between nodes is assigned through the Gaussian kernel function. This function takes the path distance between nodes as a variable, and the closer the distance, the greater the propagation intensity. After the diffusion propagation is completed, the system performs disturbance coefficient suppression on the preliminary propagation results, that is, performs a threshold lower truncation operation on the nodes with low disturbance response, so as to weaken the influence of weak disturbance noise and obtain the disturbance-aware graph data. Based on the disturbance-aware graph, the system performs path carbon emission cumulative inference. The system extracts the set of disturbance propagation paths in the disturbance-aware graph. This set consists of all paths starting from the disturbance source nodes, with a step size not exceeding three hops, and finally reaching the effective dumping points. For each path, the system extracts its carbon emission factor characteristics, including the sum of the edge weights in the path, the average value of the node disturbance response values in the path, the product of the estimated path time consumption and the preset carbon factor table, etc. in multiple dimensions, which constitute the carbon emission feature vector of the path. The system constructs a graph attention neural network model, uses the above path factor characteristics as input, learns the dependency relationship between nodes through the attention mechanism, embeds and expresses the whole path, and outputs the expected carbon emission value of each path, which is used to characterize the potential carbon emission level of the path under the current disturbance conditions. The system performs carbon emission probability modeling according to the path expected carbon emission results. The system conducts fitting analysis on the carbon emission fluctuation of each path based on historical data, and selects the probability distribution model with the smallest fitting error (such as normal distribution, gamma distribution, etc.) as the carbon emission probability distribution model of the path. Introduce the disturbance-aware weight mechanism, fuse factors such as disturbance response value, scheduling frequency, and path stability, and construct the path sampling priority weight.The system performs weighted integration of the carbon emission probability density function of each path with its perturbation weight to form a path selection probability value, and further fuses and integrates the carbon emission distributions of all paths to obtain the overall carbon emission expectation distribution under the current non-fixed path set.

[0033] Step S4: Perform carbon load optimization scheduling based on the fixed carbon emission data and non-fixed carbon emission data to obtain the optimized muck carbon load data.

[0034] In one embodiment, the system defines the variable set in the scheduling model. Set the variable set X, which consists of multiple variable elements. Each variable represents the decision state of the i-th muck transportation task being assigned to the j-th feasible transportation path, where the variable value of 1 indicates assignment and the value of 0 indicates non-assignment. The carbon emission corresponding to each task path consists of two parts. One is the fixed carbon emission value of this path, representing the carbon emission data under stable working conditions of the path; the other is the non-fixed carbon emission expectation value, representing the estimated carbon emission load of the path in a dynamic perturbation environment. The system takes minimizing the global carbon load as the optimization objective function. Specifically, it traverses all combinations of tasks and their optional paths, calculates the total carbon emission of the task path combination, and minimizes the sum of the carbon emissions of all combinations. The objective function is expressed as minimizing the sum of the fixed carbon emission value and the non-fixed carbon emission expectation value corresponding to each task and path combination. The system establishes the constraint conditions of the scheduling model, mainly including the following categories. There is the uniqueness constraint of task assignment to ensure that each muck transportation task can only be assigned to one path, that is, for any task, the sum of the scheduling variables in all its optional paths should be 1. There is the path capacity limit constraint. Within a unit time window, the total transportation load of each path shall not exceed the maximum transportation capacity supported by this path, and the transportation load can be estimated according to the load data corresponding to the task. Optional additional constraint conditions include, but are not limited to, the time window constraint: some paths are unavailable during certain time periods; the carbon emission upper limit, setting that the total carbon emission value of the system shall not exceed the set upper limit threshold. At the solution level, the system uses mathematical programming methods such as linear programming (LP) and mixed integer programming (MIP) to jointly solve the above objective function and constraint model. For scheduling requirements with a large number of tasks or involving multi-objective optimization scenarios, non-dominated sorting genetic algorithm NSGA-II or other evolutionary intelligent optimization algorithms can also be used. The system outputs the optimized muck carbon load data as the decision basis for scheduling execution. The optimization results include core parameters such as the specific path number to which each transportation task is assigned, the corresponding carbon emission estimation value, the task execution duration, and the transportation weight.

[0035] Preferably, step S1 is specifically: Step S11: Obtain the muck disposal data; In one embodiment, muck disposal data is collected from the following platforms / systems. For example, from the construction site management platform, which provides task information of the current muck operation, including but not limited to task list fields such as task number, designated dumping point location, and task execution time; from the transport vehicle system, which records the sequence of spatial trajectory points, vehicle operation duration, and loading weight corresponding to a unit task during muck transportation through integrated on-vehicle GPS and weighing terminal devices; and from the urban muck supervision system, which provides supervision data related to the muck dispatching link, including management information such as the inbound record, outbound record, and dispatching batch number of muck vehicles at the dumping station.

[0036] Step S12: Construct a disposal behavior graph based on the muck disposal data to obtain disposal behavior graph data; In one embodiment, each "construction site", "path segment", and "dumping point" is regarded as a node in the graph; the node attributes include type (site / path / dump), operation frequency, and operation time period. If there is a path transfer between two nodes in the dispatching history (for example, site_A→dump_B, that is, from construction site A to dumping point B), a directed edge is established; the edge attributes include the number of occurrences, average transportation time, and operation time window. Use NetworkX or a custom graph class to construct a directed weighted graph; the node set is defined as , and the edge set is ; the weight represents the historical frequency or dispatching cost.

[0037] Step S13: Extract temporal features from the disposal behavior graph data to obtain temporal feature data; In one embodiment, the timestamp is standardized to a daily cycle (such as divided by hours). The dispatching concentration of a node (such as a dumping point) within a time window is statistically calculated. The temporal feature indicators include the dispatching frequency distribution entropy , where is the dispatching frequency distribution entropy of node , is the time window index (such as a time segment in hours), is the dispatching probability of node within the time window ; the dispatching periodicity , where is the variance of the node operation time interval, is the variance function, is the sequence of adjacent dispatching time intervals of node ; the job path repeatability, that is, the repeatability of the path pattern with a certain node as the end point, such as the proportion of repeated paths in all historical paths with node as the end point, which is quantified by the path frequency distribution entropy.

[0038] ​​​​Step S14: Divide the disposal behavior map data according to the timing feature data to obtain the fixed disposal data of construction waste and the non-fixed disposal data of construction waste respectively.

[0039] In one embodiment, if a certain node meets one of the following two conditions, it is regarded as a fixed disposal point. For example, condition A: the path repeatability ≥ 0.75 and the dumping path has been continuously used in the past 7 days; condition B: the variance of the scheduling period ≤ 1800 seconds and the scheduling frequency ≥ 5 times per day on average. Non-fixed disposal node: The dumping points that do not meet any of the above criteria are regarded as non-fixed disposal points. Output the fixed path data set and output the non-fixed path data set.

[0040] Preferably, the decoupling calculation of the carbon factor is specifically as follows: Step S21: Extract the fixed disposal working condition factors according to the fixed disposal data of construction waste to obtain the fixed disposal working condition factor data; In one embodiment, the system extracts the working condition parameters strongly related to carbon emissions in the fixed disposal data of construction waste and constructs structured factor data for subsequent carbon emission modeling calculations. The fixed disposal data set of construction waste contains multiple operation records, and each operation record includes, but is not limited to, the following information fields: path number, operation time period, transportation vehicle identification, dumping point number, etc. The system extracts the following working condition factors from each fixed disposal record of construction waste and uniformly standardizes them into numerical vector representations, including transportation distance (unit: kilometers), calculating the geographical distance between the operation starting point and the corresponding dumping point according to the GPS trajectory data of the transportation vehicle. Average vehicle speed (unit: kilometers per hour), calculating the average speed value of the whole journey based on the time series data of the trajectory speed. Vehicle tonnage (unit: tons), retrieving the rated load capacity of the vehicle from the vehicle ledger system as a static parameter. Actual load (unit: tons), the load data recorded in real time by the on-vehicle weighing equipment, reflecting the actual material weight of each transportation. Operation duration (unit: minutes), calculating the operation duration by recording the start time and end time of the operation task. Road condition resistance level (value range: [0,1]), obtaining the unit resistance coefficient through conversion by combining the road grade information provided by the road network platform and the traffic management index data. Dumping operation type (a discrete variable), indicating the specific operation type of this disposal behavior, such as "direct dumping" or "transfer to storage". This field is discretely numericalized using the one-hot encoding method for easy combination with other factors for modeling.

[0041] Step S22: Perform matrix decomposition calculation on the fixed disposal working condition factor data to obtain the fixed disposal decomposition data; In one embodiment, the factor matrix M is decomposed into a factor intensity matrix and a carbon emission weight matrix to achieve factor decoupling. The non-negative matrix factorization (NMF) algorithm is used to meet the physical constraints of environmental variables (such as emissions ≥ 0). , is the operating condition matrix (already standardized), is the task factor intensity matrix (indicating the response degree of each sample under different factor clusters), is the factor carbon emission contribution weight matrix (the weight of each latent factor on each original factor), and set the decomposition factor dimension (adjustable); is the error of the metric matrix decomposition. The system selects the KL divergence as the optimization objective and uses the KL divergence as the error metric, , where is the KL divergence, is the row index in the matrix, indicating the th task sample, is the column index in the matrix, indicating the th operating condition factor, is the original matrix in the row and column element value, is the element value of the in the row and column.

[0042] Step S23: Estimate the unit carbon emission based on the fixed disposal decomposition data to obtain the fixed carbon emission data.

[0043] In one embodiment, the decomposed factor matrix is converted into the unit carbon emission valuation (kgCO2 / ton·km) of each operation task. The unit carbon emission of each task is estimated as where is the intensity of the th record under the th latent factor; is the carbon emission contribution weight of the th factor (extracted from the factor matrix or selected as a representative value from the dominant field of each implicit factor); can be calibrated as the mean value of the real carbon emission sample or mapped using a carbon factor library (such as the diesel emission coefficient in the IPCC guidelines, etc.). The output unit carbon emission data can be standardized to the unit / task; or standardized to / ton·km.

[0044] Preferably, the construction of the disposal element map is specifically as follows: Extract the disposal attributes from the non-fixed disposal data of the construction waste to obtain the disposal attribute data; In one embodiment, let the non-fixed disposal data set of the construction waste be denoted as , where each record For a single independent non-fixed scheduling task, the record should at least include core information such as the tipping point identifier, path trajectory, task time, operation type, etc. The system regards each non-fixed tipping point as a graph node and extracts the following key attributes from the task historical data, including the geographical location of the tipping point, which represents the longitude and latitude coordinates of the tipping point. The average scheduling frequency represents the number of operation tasks received at this tipping point per unit time (such as per day or per week), reflecting the usage density of this point. The number of reachable paths counts the number of paths that successfully reach this tipping point from other paths in the historical scheduling data, indicating its network connectivity or accessibility. The operation window represents the available operation time range for this tipping point each day, for example, from 08:00 to 18:00, reflecting its feasible operation time interval. The operation mode type characterizes the disposal operation modes supported by the current tipping point, including but not limited to "landfill", "transfer", "diversion", "temporary storage", etc. This field is a discrete variable and can be processed using encoding or one-hot encoding methods. The environmental protection level or risk level introduces data labels from the construction waste supervision platform or environmental protection system to measure the rating of this tipping point in terms of ecological impact, such as "preferred recommendation", "restricted area", "high risk", etc. The historical average carbon emission, if there are historical carbon emission estimation records for this tipping point, calculates the average carbon emission per unit in its historical operations.

[0045] Construct a graph structure based on the disposal attribute data to obtain the disposal attribute graph data; In one embodiment, set the node set , that is, each non-fixed tipping point is a graph node. Establish a graph , where the edge satisfies the following connection criteria, including if the Euclidean distance between nodes , then establish the edge ; if there exists a task sequence , that is, in a certain operation path and appear simultaneously ≥2 times, it is regarded as a connected edge ; calculate the attribute similarity for each pair of nodes: , if , then it is regarded as a strongly correlated edge and given a high weight, such as directly setting as the edge similarity weight; generate the edge weight by combining three types of rules: , where is the geographical edge weight coefficient (taking the value of 0.3), is the edge established based on the geographical distance (geographical edge), is the historical edge weight coefficient (taking the value of 0.3), is the edge based on the historical task co-occurrence (historical edge), is the attribute similarity edge weight coefficient (taking the value of 0.4), For nodes and of the attribute similarity, output the graph structure data G = (V, E, W).

[0046] Dispose of the disposal attribute graph data to perform node partitioning to obtain disposal partition graph data; In one embodiment, the dumping point graph is divided into multiple sub-graph clusters with functional or scheduling redundancy as the basic unit for scheduling optimization. Perform unsupervised graph partitioning (applicable to weighted undirected graphs), and each node is assigned to a certain partition The mapping relationship is expressed as: where is the number of partitions, takes values from 1 to... which is adaptively determined by the internal structure of the graph, that is, when the modularity optimization reaches convergence, the algorithm automatically outputs the partition structure and the corresponding set of cluster numbers; for each partition its corresponding sub-graph structure can be extracted , is the set of nodes in the partition sub-graph, is the set of edges in the partition sub-graph, where , is the set of all nodes in the original graph, is the set of all edges in the original graph, and satisfies the global node coverage relationship .

[0047] Dispose of the disposal partition graph data for disposal evolution to obtain disposal evolution graph data.

[0048] In one embodiment, each time window can be set to a fixed period, such as 7 days as a window period. The system statistically analyzes the change trend of the attribute values of each dumping point node within each time window to construct a time evolution trajectory. If and at time there is a job transfer or attribute state transfer (such as an increase in scheduling frequency, dumping volume transfer, etc.), then a time-series evolution edge is established. The edge weight is calculated as where is the relative transfer probability or transfer weight of node migrating to node at time , is the number of job transfers from node to node within the time window , is the index of all target nodes having a transfer relationship with node , is the time window Inner slave node Migrate to all other nodes The cumulative number of transfers.

[0049] Preferably, the edge floating value calculation is specifically as follows: Step S31: Embed the perturbation factor into the disposal factor graph data to obtain graph embedding data; In one embodiment, the input is an original graph structure, represented as a triple where represents the set of nodes in the graph, represents the set of edges, represents the original weight value of each edge. The system simultaneously accesses a group of external perturbation data sources that are dynamically updated by time slices. This data includes but is not limited to the following four types of indicators, including the traffic passing index, represented as , with a value range of 0 to 1, used to describe the real-time traffic congestion degree of the path connected by edge ; the meteorological interference level, represented as , usually from a meteorological radar or monitoring platform, and can be quantified into indicators such as rainfall level and visibility level; the equipment status availability index, represented as , used to reflect the schedulability of construction machinery or work teams on this path; the road network closure rate or construction interruption level, represented as , measuring the degree of influence of the current path by factors such as construction enclosures and road closures. The system performs weighted aggregation processing on the above perturbation factors to generate the perturbation factor influence value corresponding to each edge. The specific calculation method is as follows , where is the perturbation factor influence value, is the traffic passing index weight coefficient, with a value of 0.25, is the traffic passing index, is the meteorological interference level weight coefficient, with a value of 0.25, is the meteorological interference level, is the equipment availability index weight coefficient, with a value of 0.25, is the equipment status availability index, is the construction interruption level weight coefficient, with a value of 0.25, is the road network closure rate / construction interruption level. The system generates a graph embedding data structure, where each edge simultaneously retains its original scheduling weight and the newly embedded perturbation factor score value.

[0050] Step S32: Perform transfer inertia analysis on the graph embedding data to obtain transfer inertia edge data; In one embodiment, the system reads the historical scheduling path sequence data of each edge in the graph structure. This data is sourced from the operation records of the construction unit or the urban management platform, which record the specific paths adopted in each construction waste transportation task. The system counts the usage of each edge in consecutive tasks and extracts two core indicators from it, including the consecutive usage count, which represents the maximum number of times the edge is reused in consecutive scheduling tasks, denoted as the consecutive usage count of the edge; and the cumulative usage count, which represents the total number of times the edge appears in all historical tasks, i.e., the total scheduling usage frequency of this path. Based on the above two indicators, the system defines the transfer inertia value of the edge, which is the tendency degree when a certain edge is continuously selected by the scheduling system or operators. The specific calculation method of the inertia value is: divide the consecutive usage count by the cumulative usage count. The value range of the inertia value is between 0 and 1, and the higher the value, the more likely the edge is to be repeatedly and continuously selected in history. For example, if an edge is used 10 times in history and the maximum consecutive usage reaches 8 times, then its inertia value is 0.8. The edges with inertia values close to 1 can be regarded as high-stability paths, while the edges with values lower than a certain threshold (such as 0.3) are scheduling paths with higher uncertainty.

[0051] Step S33: Calculate the edge floating value for the transfer inertia edge data to obtain the edge floating value data; In one embodiment, the system sets the calculation method of the edge floating value as: multiply the perturbation factor score value of the edge by its transfer instability degree (i.e., one minus the transfer inertia value) to form the edge floating value. That is, the edge floating value is equal to the perturbation intensity multiplied by the path scheduling instability. This calculation method ensures the following response behaviors. For example, if an edge has a high perturbation intensity and a low historical inertia, that is, it is severely perturbed currently and has an unstable historical scheduling, then the floating value of this edge will increase significantly, indicating its high volatility in the current environment; if an edge is strongly perturbed but shows a high inertia in history, that is, the system has a long-term preference for using this path, then the floating value of this edge is suppressed, reflecting its structural stability; if an edge has a low current perturbation impact but also a low historical inertia, then its floating value is at a medium level, indicating its potential uncertainty. Through the above calculation, the system obtains the edge floating value corresponding to each edge and outputs a structured result to form a set of floating value data.

[0052] Step S34: Adjust the edge weights of the disposal element graph data according to the edge floating value data to obtain the non-fixed disposal graph data.

[0053] In one embodiment, the system takes each edge as a unit and receives its original weight value and floating value as inputs. The original edge weight represents the basic scheduling cost or passing tendency of the path in a static environment, while the floating value represents the fluctuation risk or uncertainty of the path under the current perturbation conditions. To incorporate the perturbation perception factor into the scheduling graph model, the system sets the following edge weight update rule: multiply the floating value of the edge by a preset floating adjustment coefficient, and add the resulting value to the original edge weight to obtain a new edge weight value. The floating adjustment coefficient is a tuning parameter used to control the degree of influence of the perturbation on the overall graph structure. The value range of this coefficient is from 0.5 to 2.0. The larger the value, the more sensitive it is to edge floating, and the system dynamically adjusts the path cost according to the perturbation; conversely, if the coefficient is smaller, it means that the system adopts a more conservative reaction strategy to the perturbation fluctuation. In a specific scheduling scenario, this coefficient can be dynamically set by the policy setting module to adapt to different job requirements or regional rules. If the original graph structure uses the path cost representation method, that is, the larger the edge weight, the less optimal the path, then the weight increase will suppress the priority of this edge in subsequent path selection, playing a role in weakening the scheduling tendency. Therefore, by integrating the perturbation floating value into the edge weight, the system can directly guide the path scheduling to shift towards areas with low perturbation and low uncertainty at the graph structure level. The system outputs a non-fixed disposal graph structure based on the updated edge weight set, denoted as , which inherits the original node and edge topological structure and updates the edge weights to the results after perturbation response.

[0054] Preferably, the perturbation perception propagation is specifically as follows: Extract the perturbation source node from the non-fixed disposal graph data to obtain the perturbation source node data; In one embodiment, select the nodes that meet the following conditions as perturbation source nodes from all nodes , such as any perturbation index value of the node exceeding the local statistical upper bound , where is the perturbation value (such as the traffic delay level), is the average perturbation value of the node, is the standard deviation; or there have been abnormal record events such as task failure or dumping interruption within the previous 3 hours of the node.

[0055] Extract the perturbation attributes according to the perturbation source node data to obtain the perturbation attribute data; In one embodiment, for each disturbance source node, the system collects and constructs the following four types of disturbance attribute indicators from multiple dimensions, including traffic disturbance intensity, which represents the degree of traffic anomaly in the road network where the node is located, with a value range of 0 to 1. The higher the value, the more serious the congestion or delay. The data is sourced from the real-time state analysis of the traffic road network; meteorological impact level, which represents the intensity of the impact of adverse weather on the node, including the level scores of extreme meteorological events such as heavy rain, heavy snow, typhoon, etc. The higher the value, the greater the impact of the weather on the scheduling; geographical blockage rate, which represents the degree of closure of the road or area where the node is located, with a numerical range of 0 to 1, where 1 represents complete closure and no traffic conditions; average floating value of surrounding nodes, which represents the scheduling volatility of the adjacent area of the node. Based on the edge weight floating value in the previous stage (such as scheduling frequency or cost change), neighborhood weighted averaging is performed to reflect the overall dynamic stability of the area where the node is located. For each disturbance source node, the system summarizes the above disturbance attribute items to form a standardized disturbance attribute vector.

[0056] Extract the disturbance coupling characteristics from the non-fixed disposal graph data according to the disturbance attribute data to obtain the disturbance coupling characteristic data; In one embodiment, the disturbance attribute is mapped into the edge set E of the graph structure, and the following coupling index is used for calculation , where is the shortest path distance between points in the graph; is the node attribute or spatial similarity (such as 1 for the same partition); is the disturbance intensity of the disturbance source, and the coupling characteristic data matrix is output, indicating the disturbance coupling degree between each pair of nodes. The calculation results of the disturbance coupling strength between all node pairs are organized into a two-dimensional matrix, called the disturbance coupling characteristic matrix, whose structure is as follows. The matrix dimension is , where represents the number of nodes in the graph; each matrix element represents the coupling strength of the disturbance spreading from node to node .

[0057] Construct a disturbance weight graph according to the disturbance coupling characteristic data to obtain the disturbance weight graph data; In one embodiment, a graph is constructed based on the coupling characteristic matrix, where is the set of nodes in the graph, which is consistent with the nodes in the non-fixed disposal graph, is the set of edges in the graph, which only contains the node pairs with significant disturbance coupling strength. The edge weight , that is, the edge weight comes from the value in the coupling matrix in the previous stage ; only is retained (such as ), representing the attenuation weight path when the perturbation propagates from the source node to other nodes.

[0058] Perform Gaussian diffusion on the non-fixed disposal graph data using the perturbation weight graph data to obtain preliminary perturbation graph data; In one embodiment, the graph Gaussian diffusion formula is adopted to propagate the perturbation influence on the perturbation weight graph as follows where is the perturbation response value vector of each node in the th diffusion iteration, is the diffusion rate control coefficient, and its numerical range is set between 0.7 and 0.9, which is used to adjust the balance degree between the historical perturbation value and the initial perturbation, is the perturbation weight graph 's normalized adjacency matrix, and is where is the node degree matrix, is the edge weight matrix, is the perturbation response value vector of each node in the th diffusion iteration, is the initial perturbation state, which is non-zero only at the perturbation source node, and the remaining nodes are assigned zero. Iteratively propagate 3 - 5 times, or until the perturbation change amount ; obtain the perturbation response value of each node (i.e., the preliminary perturbation graph node value).

[0059] Perform perturbation coefficient suppression on the preliminary perturbation graph data to obtain perturbation-aware graph data.

[0060] In one embodiment, set the perturbation intensity threshold , and only retain the nodes whose perturbation values are greater than or equal to this threshold; for each node , its perturbation value is processed as follows: where represents the perturbation response value after threshold clipping. The system performs a maximum normalization operation on the clipped perturbation values, which is defined as follows: where is the normalized perturbation value of node , is the maximum value among all the clipped perturbation values. Sort all the nodes in descending order according to the normalized perturbation value ; only retain the nodes whose perturbation values rank in the top 30%, and set the perturbation values of the remaining nodes to 0.

[0061] Preferably, the perturbation coupling feature extraction is specifically as follows: Perform perturbation attribute projection on the non-fixed disposal graph data according to the perturbation attribute data to obtain perturbation attribute projection data; In one embodiment, the node perturbation vector is projected onto its connected edges, and the edge perturbation attribute is defined as: , representing the average perturbation intensity of the two end nodes of edge , where is the perturbation attribute projection vector of edge , is the perturbation factor vector of node , is the perturbation factor vector of node , and is the discriminative expression of the two end nodes of edge ; for heterogeneous factors, a weighted sum can be formed to obtain the edge perturbation influence value: , where is the perturbation influence intensity score of edge , is the weight coefficient of the traffic passing factor, with a value of 0.3, is the traffic congestion index on edge , is the weight coefficient of the meteorological interference factor, with a value of 0.2, is the meteorological influence level on edge , is the weight coefficient of the road network closure factor, with a value of 0.2, is the closure or construction interference level on edge , is the weight coefficient of the path fluctuation factor, with a value of 0.3, is the floating value of edge , and the output perturbation attribute projection data is a graph structure with perturbation attributes , is the set of nodes in the graph, is the set of edges in the graph, and the edge weight is .

[0062] Perform perturbation coupling calculation on the perturbation attribute projection data to obtain the perturbation coupling data; In one embodiment, calculate the perturbation coupling strength between any two nodes, that is, in the perturbation scenario, whether the change of one node has a coupling effect on another node. Define the perturbation coupling degree of node as a combination of three items, including the cosine similarity of the perturbation factor vectors , where is the perturbation factor vector of node , is the perturbation factor vector of node , whether there is a path from to , and the shortest path distance , where is the graph structure dependency of the node pair , is the node to the shortest path distance; The calculation formula for the coupling degree is: , is the node and the perturbation coupling degree, is the node and the cosine similarity of the perturbation factors, is the graph structure dependency of the node pair , and the perturbation coupling matrix is output.

[0063] Cross-node connection processing is performed based on the perturbation coupling data to obtain cross-node connection data; In one embodiment, a path sequence database is extracted from the scheduling task history; For any node pair , if any of the following conditions is met, it is considered that there is a "cross-node connection", such as the frequency of co-occurrence in the same scheduling path ≥ 2; There is a transfer path with a path length ≤ 3 between them; There are ≥ 1 common neighbors within 2 adjacent hops. The connection coefficient is given, where is the cross-node connection coefficient, indicating the indirect propagation channel strength between the nodes and , is the co-occurrence weight coefficient of the scheduling path, controlling the weight ratio of the frequency of the common path in the overall connection strength, with a value of 0.4, is the node path co-occurrence function, indicating the frequency of co-occurrence of the nodes , in the historical scheduling path, is the graph path distance weight coefficient, controlling the influence of the transfer distance in the graph on the connection strength, with a value of 0.3, is the shortest graph path distance, that is, the shortest number of hops in the graph between the nodes and , is the neighbor overlap weight coefficient, used to control the contribution of neighbor similarity to the connection strength, with a value of 0.3, is the adjacency Jaccard similarity, that is, the Jaccard similarity (the ratio of the number of common neighbors to the total number of neighbors) between the adjacency node sets of the nodes and , and the cross-node connection matrix is output, indicating the strength of the indirect propagation channel.

[0064] Calculate the perturbation propagation intensity of the perturbation coupling data based on the cross-node connection data to obtain the perturbation coupling characteristic data.

[0065] In one embodiment, for the perturbation propagation ability the calculation is as follows: , where is the perturbation propagation intensity, indicating the actual perturbation conduction ability of node to node considering the cross-node connection, is the basic perturbation coupling intensity, indicating the basic coupling value of node and in the graph structure based on the propagation path and similarity, which is calculated from the perturbation coupling matrix, is the cross-node propagation enhancement factor, used to control the improvement amplitude of the overall propagation intensity by the cross-node connection, and its value range is from 0.5 to 1, is the cross-node connection intensity, indicating the degree of non-local structure connection between node pairs, such as shared resources, cross-region dependencies, collaborative scheduling relationships, etc.; the value range is [0,1], if , it means only direct coupling is considered.

[0066] Preferably, the specific path carbon emission accumulation reasoning is as follows: Extract the perturbation path set from the perturbation perception graph data to obtain the perturbation path set data; In one embodiment, the system defines the starting point set S of the path. Nodes with a perturbation response value not lower than 0.7 are screened out among all nodes and added to the starting point set as high-perturbation response nodes. The system defines the ending point set T of the path. The ending point set can include two types of nodes: one is all dumping points in the non-fixed scheduling graph, representing the job termination position; the other is nodes that often serve as transportation endpoints in the historical scheduling path, indicating output nodes that the system highly depends on during actual operation. During the path generation process, for each starting point node, the system uses a graph structure search algorithm (such as depth-first search DFS or improved Beam Search) to expand the path. The system sets the following path screening constraints: the number of nodes in the path does not exceed 5, that is, at most 4 edges are included, to control the path complexity and propagation error; the perturbation weight of any edge in the path must not be lower than the set threshold; the average perturbation response value of all nodes in the path must not be lower than 0.3. After screening, the system outputs the path set that meets the conditions, that is, the perturbation path set data. This path set is composed of multiple node sequences, and each path represents a feasible propagation channel from the high-perturbation starting point to the potential ending point and serves as the input basis for path behavior modeling under the perturbation environment.

[0067] Perform path carbon factor mapping based on the perturbation path set data to obtain the path carbon mapping data; In one embodiment, the system represents each path as a set of numerical feature vectors, mainly including the following seven types of elements. It includes the mean of node perturbation responses, which represents the average of the perturbation response values of all nodes on the path and reflects the overall degree of perturbation of the path, denoted as the path perturbation mean; the path length (number of hops), which is the number of nodes in the path minus one and represents the actual number of steps the path takes in the graph; the total sum of cumulative edge weights, which represents the sum of the perturbation weights of all edges in the path and reflects the overall propagation resistance or perturbation cost of the path; the average edge floating value, which represents the average of the floating values corresponding to each edge in the path and can be used to measure the scheduling volatility of the path; the total sum of historical scheduling frequencies, which represents the cumulative frequency of the path being called in historical muck transportation tasks and reflects the actual feasibility and scheduling preference of the path; the total path distance (in kilometers), which is estimated based on the geographical coordinates or GPS trajectory data between nodes in the path and reflects the physical transportation distance of the path; the vehicle type and load mapping feature. If the task information recorded in the system has been associated with a specific vehicle, then the path can also load features such as the vehicle type (such as dump trucks, tippers, etc.) used in the task and the corresponding load tonnage, etc., to reflect the capacity load factor of carbon emissions.

[0068] Perform graph attention network inference based on the path carbon mapping data to obtain the expected path carbon emission data.

[0069] In one embodiment, the system converts the path data into a graph structure form, uses each node in the path as a node element of the graph, and constructs a local path subgraph. Each node is attached with its corresponding carbon factor feature vector, which is generated in the previous step and includes multi-dimensional input information such as perturbation mean, path length, cumulative edge weight, scheduling frequency, etc. The connection method of the edges adopts sequential directed connection, that is, the nodes are connected according to the physical order of the path to construct a directed graph structure to simulate the actual traveling direction of the path during the scheduling process. In the process of graph attention modeling, the system constructs an attention coefficient between each pair of connected nodes in the graph, representing the importance weight of the adjacent node to the current node. The calculation of the attention coefficient is usually based on the following mechanism: the feature vectors of the two nodes are respectively linearly transformed, then concatenated, and then input into a feed-forward attention function with a LeakyReLU activation function to measure their relative influence. The attention weights of all adjacent nodes are then normalized and used to weighted aggregate the information of the neighbor nodes. The system updates the feature representation of each node through the above mechanism, that is, the new representation of each node is the weighted sum of the features of all its adjacent nodes, and is output after passing through a non-linear activation function. This process can be stacked in multiple layers to enhance the model's ability to express the graph structure. After all node features are updated, the system aggregates all the node embedding vectors in the entire path to generate an overall representation vector of the path. The path vector is input into a multi-layer perceptron regression model to output the expected value of unit carbon emissions of the current path under the given perturbation conditions, with the unit of kilograms of carbon dioxide ( )

[0070] Preferably, the calculation of the carbon emission probability is specifically as follows: Fit the path carbon emission distribution according to the path carbon emission expectation data to obtain the path carbon emission distribution data; In one embodiment, the input path expected carbon emission data set , each path 's expected carbon emission value and the perturbation coefficient (source of variance) are used to construct a probability density function, where takes values from 1... ; if there is no real variance sample, the default variance can be set to 5% - 15% of the expected value, that is , where is the standard deviation of the path carbon emission distribution, is the variance coefficient (perturbation range adjustment coefficient), with a range of 5% - 15% (i.e., 0.05 - 0.15), is the path carbon emission expected value, and the normal distribution is used for fitting (Gamma distribution, etc. are also supported): , where is the carbon emission probability density function of path , is the carbon emission value variable, is the pi constant, is the standard deviation of the path carbon emission distribution, is the natural exponential function.

[0071] Perform perturbation-aware weight fusion on the path carbon emission distribution data to obtain path carbon emission probability data; In one embodiment, a perturbation-aware weight is assigned to each path carbon emission distribution to reflect the sampling priority of the path in the current perturbation environment. Each path is already associated with a perturbation propagation graph node value (such as the average perturbation response intensity ); Weight fusion , where is the path perturbation-aware fusion weight, is the perturbation response intensity weight coefficient, with a value of 0.5, is the path average perturbation response intensity, indicating the path The average response value of the nodes in the associated perturbation propagation graph (such as the node activation value or the average propagation weight), is the stability penalty term weight coefficient, with a value of 0.3, is the path stability index, is the task usage frequency suppression term weight coefficient, with a value of 0.2, is the path usage frequency.

[0072] Perform non-fixed disposal carbon emission probability estimation on the path carbon emission probability data to obtain non-fixed carbon emission data.

[0073] In one embodiment, if only the average level of overall carbon emissions is of concern, the system can use the weighted expected value method to calculate the weighted sum of the expected carbon emissions of all paths to obtain the expected carbon emissions of the overall non-fixed disposal paths. The specific calculation method is as follows: Weighted sum of the expected carbon emission values of each path according to their perturbation-aware weights: where represents the overall expected carbon emissions under the current non-fixed path network, is the perturbation-aware weight of the th path, is the expected carbon emission of the th path, is the non-fixed disposal path index

[0074] In one embodiment, if it is necessary to obtain the overall probability characteristics (such as distribution form, confidence interval, extreme risk, etc.) of the non-fixed path network in terms of carbon emissions, a weighted mixing method of path distribution can be adopted to construct the overall carbon emission probability density function. The specific definition is as follows: all path carbon emission density functions are weighted and added according to the disturbance perception weight to obtain the overall mixed density function: , where represents the carbon emission probability density function of the overall non-fixed disposal path; is the carbon emission density function of the th path (such as normal distribution or Gamma distribution), is the non-fixed disposal path index, is the path weight; is the carbon emission value variable.

[0075] Preferably, step S4 is specifically as follows: Step S41: Perform carbon load modeling on the fixed carbon emission data and the non-fixed carbon emission data to obtain a carbon load model; In one embodiment, the system defines a set of scheduling tasks, representing the total set of current muck transportation tasks to be executed. Each task needs to select a path from its candidate path set to perform the transportation scheduling operation. The candidate path set may include fixed paths and non-fixed paths, and the system constructs a corresponding path option list for each task. Based on all task and path combinations, the system further constructs a carbon emission coefficient matrix. This matrix is a two-dimensional structure, where the row index corresponds to the scheduling task and the column index corresponds to the optional path. Each element in the matrix represents the unit carbon emission value resulting from assigning a certain task to a certain path, with the unit of kilograms of carbon dioxide (kg CO2). For fixed paths, the system directly uses its known unit carbon emission value as the corresponding value of this path in the carbon emission matrix; for non-fixed paths, the system uses the expected carbon emission value inferred through the disturbance perception model as the predicted carbon emission index of this path. The system outputs a carbon load model, which includes three types of data structure information, the scheduling task identifier, the candidate path set corresponding to each task, and the unit carbon emission index matrix corresponding to the path scheduling.

[0076] Step S42: Set the scheduling variables for the carbon load model to obtain a carbon load scheduling model; In one embodiment, the system defines a scheduling variable for the mapping relationship between each scheduling task and its candidate paths. This variable is a binary decision variable used to represent the allocation status between tasks and paths. Suppose the i-th task has k candidate paths. For the j-th path among them, the system sets the value-taking rule of the scheduling variable as follows: when the i-th task is assigned to its j-th candidate path, the value of this variable is 1; if the task does not select this path, the value of this variable is 0. All the above scheduling variables can be combined to form a task-path binary decision matrix. The rows of the matrix represent task numbers, and the columns represent the numbers of selectable paths. The whole matrix is used to represent the allocation structure of the entire scheduling scheme in terms of path selection. On this basis, the system constructs an objective function for carbon load optimization. The goal is to minimize the total carbon emissions generated by all tasks under their path allocations. Specifically, for each scheduling task, the system traverses all its candidate paths and multiplies the scheduling variable indicating whether it is selected by the unit carbon emission value of the corresponding path. Summing up the products of all task-path combinations, the total carbon emission value as a whole is obtained. Therefore, the optimization objective function is to find a combination of the values of a set of scheduling variables among all feasible task-path allocation schemes, so that the total carbon emissions after scheduling reach the minimum.

[0077] Step S43: Set constraint conditions according to the carbon load scheduling model to obtain a carbon load constraint model; In one embodiment, based on the established carbon load scheduling variable model, the system sets the constraint conditions to be followed during the scheduling optimization process to form a complete carbon load scheduling optimization constraint model. The constraint conditions are one of the following: Task uniqueness constraint. The system ensures that each scheduling task can only be assigned to a unique candidate path. For any task, the system sums up the scheduling variable values of all its optional paths, and requires that the sum value be 1, that is, each task is only assigned once during scheduling, without repetition or omission. Vehicle load limit constraint (applicable to scenarios with vehicle resource capacity limitations). Within a unit time window, the system totals the loads of all scheduled tasks, and requires that the total weight of all tasks transported during this time period does not exceed the maximum total load capacity of the schedulable vehicle resources. Dumping point capacity limit constraint. A maximum receiving capacity per unit time is set for each dumping point. During the scheduling process, the system determines whether all paths lead to a certain dumping point. If so, the scheduling variable of the task belonging to this path is multiplied by its transport weight and then accumulated, and the sum is required not to exceed the maximum receiving capacity of this dumping point. Whether a path leads to a specific dumping point is determined by an indicator function, which takes a value of 1 when the path leads to this point, and 0 otherwise. Scheduling time window limit constraint. For the time periods allowed for different paths to pass, the system sets time window limitations. For example, for some paths, there are restrictions such as being closed during traffic peaks and prohibited from passing during construction periods. If the preset scheduling time period of a task is not within the path-allowed time window range, the scheduling variable of this path is forced to be set to 0, indicating that it cannot be selected. Carbon emission quota or carbon limit constraint. The system can set an upper limit on the total carbon emissions during the scheduling period as the carbon emission control target. When the carbon emissions of all scheduling tasks under the selected paths are accumulated, the sum must not exceed the set maximum carbon emission threshold.

[0078] Step S44: Perform carbon load optimization solution on the carbon load constraint model to obtain the optimized data of the muck carbon load.

[0079] In one embodiment, the carbon load optimization problem is formulated as a mixed-integer linear programming solution, and the objective is set to where is the minimization operator, is the task number index, is the optional path number index, is the path assignment variable. If task i is assigned to path j, then ; otherwise , is the carbon emission estimate for a single task on a single path, indicating the estimated carbon emission value (unit: when task is assigned to path ). The constraints include but are not limited to that each task is assigned only once; path schedulability limitations; upper limit of path carbon load; policy rules such as preferentially scheduling fixed paths. When the task scale is medium and the model structure is standardized, a mathematical programming solver such as the open-source PuLP or commercial high-performance solvers Gurobi and CPLEX is used, which has the advantages of fast solving speed and stable convergence; when the task scale is large or there are intertwined multi-objective and multi-constraint situations, an evolutionary algorithm (such as the non-dominated sorting genetic algorithm NSGA-II) or a heuristic optimization method with local search (such as simulated annealing and neighborhood perturbation) is used to search for an approximate global solution. Optimized data of the carbon load of the muck is output.

[0080] Preferably, the present application also provides a muck disposal system for carbon emission optimization, which is used to execute the muck disposal method for carbon emission optimization as described above. The muck disposal system for carbon emission optimization includes: A muck disposal data parsing module, which is used to obtain muck disposal data; perform extraction of fixed muck disposal and extraction of non-fixed muck disposal on the muck disposal data to obtain fixed muck disposal data and non-fixed muck disposal data respectively; A carbon factor modeling and structure diagram construction module, which is used to perform carbon factor decoupling calculation according to the fixed muck disposal data to obtain fixed carbon emission data; construct a disposal element diagram for the non-fixed muck disposal data to obtain disposal element diagram data; A non-fixed path carbon emission inference module, which is used to calculate the edge floating value of the disposal element diagram data to obtain non-fixed disposal diagram data; perform perturbation perception propagation on the non-fixed disposal diagram data to obtain perturbation perception diagram data; perform path carbon emission cumulative inference on the perturbation perception diagram data to obtain path carbon emission expectation data; calculate the carbon emission probability according to the path carbon emission expectation data to obtain non-fixed carbon emission data; A carbon load optimization scheduling module, which is used to perform carbon load optimization scheduling according to the fixed carbon emission data and the non-fixed carbon emission data to obtain optimized data of the muck carbon load.

[0081] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended application documents rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be included in the present invention.

[0082] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A muck disposal method for carbon emission optimization, characterized in that It includes the following steps: Step S1: Obtain muck disposal data; perform muck fixed-disposal extraction and muck non-fixed-disposal extraction on the muck disposal data to obtain muck fixed-disposal data and muck non-fixed-disposal data respectively; Step S2: Perform carbon factor decoupling calculation based on the muck fixed-disposal data to obtain fixed carbon emission data; construct a disposal factor map for the muck non-fixed-disposal data to obtain disposal factor map data; Step S3: Calculate the edge floating value for the disposal factor map data to obtain non-fixed disposal map data; perform disturbance perception propagation on the non-fixed disposal map data to obtain disturbance perception map data; perform path carbon emission cumulative reasoning on the disturbance perception map data to obtain path carbon emission expectation data; calculate the carbon emission probability based on the path carbon emission expectation data to obtain non-fixed carbon emission data; Step S4: Perform carbon load optimization scheduling based on the fixed carbon emission data and the non-fixed carbon emission data to obtain optimized muck carbon load data.

2. The method according to claim 1, characterized in that, Specifically, Step S1 is as follows: Obtain muck disposal data; Construct a disposal behavior map based on the muck disposal data to obtain disposal behavior map data; Extract time-series features from the disposal behavior map data to obtain time-series feature data; Divide the disposal behavior map data according to the time-series feature data to obtain muck fixed-disposal data and muck non-fixed-disposal data respectively.

3. The method according to claim 1, wherein Among them, the carbon factor decoupling calculation is specifically as follows: Extract fixed-disposal working condition factors from the muck fixed-disposal data to obtain fixed-disposal working condition factor data; Perform matrix decomposition calculation on the fixed-disposal working condition factor data to obtain fixed-disposal decomposition data; Estimate the unit carbon emission based on the fixed-disposal decomposition data to obtain fixed carbon emission data.

4. The method according to claim 1, wherein Among them, the construction of the disposal factor map is specifically as follows: Extract disposal attributes from the muck non-fixed-disposal data to obtain disposal attribute data; Construct a graph structure based on the disposal attribute data to obtain disposal attribute graph data; Partition the disposal nodes of the disposal attribute graph data to obtain disposal partition graph data; Perform disposal evolution on the disposal partition graph data to obtain disposal evolution graph data.

5. The method according to claim 1, wherein Among them, the calculation of the edge floating value is specifically as follows: Embed disturbance factors into the disposal factor map data to obtain graph embedding data; Analyze the transfer inertia of the graph embedding data to obtain transfer inertia edge data; Calculate the edge floating value for the transfer inertia edge data to obtain edge floating value data; Adjust the edge weights of the disposal factor map data according to the edge floating value data to obtain non-fixed disposal map data.

6. The method according to claim 1, characterized in that Among them, the disturbance perception propagation is specifically as follows: Extract disturbance source nodes from the non-fixed disposal map data to obtain disturbance source node data; Extract disturbance attributes based on the disturbance source node data to obtain disturbance attribute data; Extract disturbance coupling features from the non-fixed disposal map data according to the disturbance attribute data to obtain disturbance coupling feature data; Construct a disturbance weight graph based on the disturbance coupling feature data to obtain disturbance weight graph data; Perform Gaussian diffusion on the non-fixed disposal map data using the disturbance weight graph data to obtain preliminary disturbance map data; Suppress the disturbance coefficient of the preliminary disturbance map data to obtain disturbance perception map data.

7. The method according to claim 1, characterized in that, Among them, the path carbon emission cumulative reasoning is specifically as follows: Extract the disturbance path set from the disturbance perception graph data to obtain the disturbance path set data; Perform path carbon factor mapping based on the disturbance path set data to obtain path carbon mapping data; Perform graph attention network reasoning based on the path carbon mapping data to obtain path carbon emission expectation data.

8. The method according to claim 1, characterized in that, Among them, the calculation of carbon emission probability is specifically as follows: Perform path carbon emission distribution fitting based on the path carbon emission expectation data to obtain path carbon emission distribution data; Perform disturbance perception weight fusion on the path carbon emission distribution data to obtain path carbon emission probability data; Perform non-fixed disposal carbon emission probability estimation on the path carbon emission probability data to obtain non-fixed carbon emission data.

9. The method according to claim 1, wherein Step S4 is specifically as follows: Perform carbon load modeling on the fixed carbon emission data and the non-fixed carbon emission data to obtain a carbon load model; Set scheduling variables for the carbon load model to obtain a carbon load scheduling model; Set constraint conditions according to the carbon load scheduling model to obtain a carbon load constraint model; Perform carbon load optimization solution on the carbon load constraint model to obtain optimized data for the carbon load of construction waste.

10. A muck disposal system for carbon emission optimization, characterized in that, For implementing the construction waste disposal method for carbon emission optimization as described in claim 1, the construction waste disposal system for carbon emission optimization includes: A construction waste disposal data analysis module, configured to obtain construction waste disposal data; perform extraction of fixed construction waste disposal and non-fixed construction waste disposal on the construction waste disposal data to respectively obtain fixed construction waste disposal data and non-fixed construction waste disposal data; A carbon factor modeling and structure diagram construction module, configured to perform carbon factor decoupling calculation based on the fixed construction waste disposal data to obtain fixed carbon emission data; construct a disposal factor diagram for the non-fixed construction waste disposal data to obtain disposal factor diagram data; A non-fixed path carbon emission reasoning module, configured to calculate the edge floating value of the disposal factor diagram data to obtain non-fixed disposal diagram data; perform disturbance perception propagation on the non-fixed disposal diagram data to obtain disturbance perception graph data; perform path carbon emission cumulative reasoning on the disturbance perception graph data to obtain path carbon emission expectation data; calculate the carbon emission probability according to the path carbon emission expectation data to obtain non-fixed carbon emission data; A carbon load optimization scheduling module, configured to perform carbon load optimization scheduling based on the fixed carbon emission data and the non-fixed carbon emission data to obtain optimized data for the carbon load of construction waste.

Images