A power transmission and transformation project multi-source heterogeneous construction cost data association method, device and medium

Abstract

The present application relates to a kind of power transmission engineering multi-source heterogeneous cost data association method, equipment and medium, belong to electric power engineering management technical field, comprising the following steps: the present application obtains the multi-source heterogeneous cost data of each power transmission engineering and constructs feature vector;Meteorological and geographic data are obtained to obtain meteorological comprehensive influence value and geographic suitability coefficient;Based on feature vector, geographic suitability coefficient and meteorological comprehensive influence value determine adjacency matrix, and the neighborhood aggregation of feature vector is obtained to obtain fusion embedding vector;Combined with geographic suitability coefficient and meteorological comprehensive influence value, obtain deep fusion feature vector;Based on deep fusion feature vector, using expert experience guide mechanism obtains mixed decision weight, according to this and fusion embedding vector calculate the correlation between each cost data.The present application connects similar engineering into graph by characteristic cosine similarity and neighborhood aggregation, so that engineering absorbs similar information, superposition commonality, suppresses error and random fluctuation caused by accidental factor.

Description

Technical Field

[0001] This invention relates to a method, equipment, and medium for associating multi-source heterogeneous cost data in power transmission and transformation projects, belonging to the field of power engineering management technology. Background Technology

[0002] Power transmission and transformation projects are a crucial component of power system construction. Their cost management spans the entire lifecycle of a project, from planning and design to construction and operation, directly impacting investment efficiency and resource allocation efficiency. With the continuous expansion of power project scale and the increasing complexity of construction environments, cost data exhibits significant multi-source heterogeneity: on the one hand, data sources encompass various types, including feasibility study reports, design documents, contract texts, tender documents, change orders, claims documents, equipment and material quotations, and market price survey reports; on the other hand, data types include both structured numerical data such as line length, number of towers, concrete volume, steel consumption, and voltage level, as well as unstructured textual data such as design specifications, contract terms, and geological survey reports. These data differ significantly in format, semantics, granularity, and quality, posing a severe challenge to the unified modeling and correlation analysis of cost information.

[0003] Currently, in the practice of power transmission and transformation project cost management, the correlation analysis between multi-source heterogeneous data mainly relies on manual experience and traditional statistical correlation analysis methods. These methods typically make strong assumptions about data distribution, making it difficult to effectively integrate textual semantic information with numerical engineering parameters, and even more difficult to flexibly handle nonlinear interaction relationships between multi-source data. At the same time, existing methods generally ignore the systematic effects of meteorological and geographical factors on the influencing factors of project costs, leading to distorted data correlation results under complex external conditions. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a method, equipment, and medium for associating multi-source heterogeneous structural price data in power transmission and transformation projects.

[0005] The technical solution of the present invention is as follows: On the one hand, the present invention provides a method for associating multi-source heterogeneous structural cost data in power transmission and transformation projects, comprising the following steps: Obtain multi-source heterogeneous structural cost data for each power transmission and transformation project, and construct feature vectors based on the multi-source heterogeneous structural cost data; Obtain meteorological and geographical data for each power transmission and transformation project, and based on the meteorological and geographical data, obtain the comprehensive meteorological impact value and geographical suitability coefficient for each power transmission and transformation project; The adjacency matrix is ​​determined based on eigenvectors, geographical suitability coefficients, and comprehensive meteorological impact values. Based on the adjacency matrix, a neighborhood aggregation operation is performed on the feature vectors of each power transmission and transformation project to obtain a fused embedding vector; Deep fusion feature vectors are obtained based on fused embedding vectors, geographic suitability coefficients, and meteorological comprehensive impact values. Hybrid decision weights are obtained by using an expert experience-guided mechanism based on deeply fused feature vectors; The correlation between various multi-source heterogeneous structural valence data is obtained by using fusion embedding vectors and hybrid decision weights, and the association between various multi-source heterogeneous structural valence data is established based on the correlation.

[0006] Preferably, the feature vector includes numerical feature vectors and text feature vectors; The numerical feature vector is constructed from standardized numerical multi-source heterogeneous valence data; The text feature vector is constructed from encoded textual multi-source heterogeneous valence data.

[0007] Preferably, the adjacency matrix is ​​determined based on the feature vector, geographical suitability coefficient, and comprehensive meteorological impact value. The specific steps are as follows: The numerical feature vector, text feature vector, geographical suitability coefficient, and meteorological comprehensive impact value are concatenated to obtain the heterogeneous structural valence feature vector; The neighbor set is determined by calculating the cosine similarity of the heterogeneous eigenvectors among the power transmission and transformation projects; The adjacency matrix is ​​determined based on the cosine similarity of the heterogeneous eigenvectors of each power transmission and transformation project in the neighbor set.

[0008] Preferably, a deep fusion feature vector is obtained based on the fused embedding vector, the geographical suitability coefficient, and the comprehensive meteorological impact value. The specific steps are as follows: A feature modulation factor is constructed based on the fusion embedding vector, geographical suitability coefficient, and meteorological comprehensive impact value; The fused embedding vector is concatenated with the feature modulation factor and then... The fully connected layer of the activation function is mapped to a deep fusion feature vector.

[0009] Preferably, the hybrid decision weights are obtained based on the deep fusion feature vectors using an expert experience-guided mechanism. The specific steps are as follows: The prior weights of experts are set according to the pre-configured expert professional title levels. Obtain the fusion feature benchmark vectors provided by each expert; The L2 norm of the difference between the deep fusion feature vector and the fusion feature baseline vector is negatively exponentially normalized and used as the hybrid decision weight.

[0010] Preferably, the correlation between various multi-source heterogeneous structural value data is obtained by using the Pearson correlation coefficient algorithm based on the fusion embedding vector and the hybrid decision weight.

[0011] Preferably, the method further includes taking the average of the L2 norms of the differences between the deep fusion feature vector and the fusion feature benchmark vector of each expert as the deviation judgment value of the power transmission and transformation project.

[0012] Preferably, a fully connected layer is used to obtain the cost prediction value based on the deviation judgment value and the deep fusion feature vector.

[0013] In another aspect, the present invention also provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the multi-source heterogeneous structure price data association method for power transmission and transformation projects as described in any embodiment of the present invention.

[0014] In another aspect, the present invention also provides a computer-readable storage medium for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the multi-source heterogeneous structure cost data association method for power transmission and transformation projects as described in any embodiment of the present invention.

[0015] The present invention has the following beneficial effects: 1. This invention connects similar projects into a graph structure by calculating the cosine similarity between the features of each project, and then performs neighborhood aggregation on the feature vector of each project based on this graph. This process is equivalent to allowing each project to absorb information from other projects with similar conditions, superimposing the common components of multiple similar samples, thereby effectively suppressing random fluctuations in individual projects caused by data entry errors, special accidental factors, etc. No learnable parameters are introduced during the aggregation process; the summation is based entirely on fixed similarity weights. This allows the model to work stably without a large number of training samples, making it very suitable for the reality that power transmission and transformation project cost datasets are usually small. After this graph propagation operation, the feature representations of each project become smoother and more continuous, the internal consistency between similar projects is enhanced, and the distinction between different types of projects is clearer, providing a higher-quality and less noisy feature foundation for subsequent correlation analysis.

[0016] 2. This invention converts text content into semantic embedding vectors through a pre-trained language model and further extracts semantic strength, enabling textual information and numerical features to be jointly modeled in the same mathematical space. This fusion method does not rely on subjective human interpretation and can objectively and consistently process all types of data. Therefore, it fully utilizes all known information during correlation analysis, avoids information fragmentation caused by differences in data formats, and makes the analysis results more comprehensive and realistically reflect the inherent relationships between various elements of engineering cost.

[0017] 3. This invention designs an expert experience-guided mechanism, assigning prior weights to experts based on their different professional titles, and collecting benchmark vectors of fusion features provided by each expert. These benchmark vectors represent the characteristic patterns that experts believe a typical project should have. The actual deep fusion features of each project are compared with the benchmark vectors of each expert to calculate the degree of deviation, and a hybrid decision weight is generated based on this. Projects that deviate significantly from the expert consensus are assigned lower weights in subsequent correlation calculations, thus avoiding excessive interference from a few outliers on the overall correlation conclusion.

[0018] 4. Power transmission and transformation projects typically span different terrains and climate zones. The external environment significantly impacts construction difficulty, material transportation, and schedule, thereby affecting the quantitative relationships between various cost data. This invention collects meteorological and geographical data for each project location and constructs a comprehensive meteorological impact value and a geographical suitability coefficient. These two coefficients quantify the comprehensive effect of environmental conditions on project implementation and participate as important features in subsequent adjacency matrix construction and feature fusion. This allows projects with similar environmental conditions to be associated more closely, while projects with significant environmental differences are appropriately separated. This enables the correlation analysis to adaptively adjust according to actual external conditions, significantly improving the real-world relevance and robustness of the analysis results. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0021] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0023] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0024] Example 1: This embodiment provides a method for associating multi-source heterogeneous cost data in power transmission and transformation projects, including the following steps: S1. Obtain multi-source heterogeneous structural cost data for each power transmission and transformation project, including numerical multi-source heterogeneous structural cost data and textual multi-source heterogeneous structural cost data.

[0025] In at least one embodiment, the numerical multi-source heterogeneous structure cost data includes one or more of the following: total line length, number of towers, volume of concrete, amount of steel, amount of conductors, number of insulator strings, number of hardware, cable length, number of cable heads, number of grounding electrodes, earthwork excavation volume, site leveling area, building area, road area, wall length, drainage trench length, fire pipeline length, number of equipment foundations, number of transformers, number of circuit breakers, number of disconnector switch groups, number of instrument transformers, number of surge arresters, number of cable supports, planned construction period, and voltage level. In at least one embodiment, text-based multi-source heterogeneous pricing data includes one or more of the following: contract texts and tender documents, feasibility studies and design documents, change orders and claims information, and supplier and market price information.

[0026] The contract text and tender documents include one or more of the following: tender document terms, contract agreement, bill of quantities description, and technical tender document; The feasibility study and design documents include one or more of the following: feasibility study report, preliminary design specification, construction drawing design specification, geological survey report, and hydro-meteorological report; The change orders and claim documents include one or more of the following: design change notice, engineering change order, claim intention notice, extension of construction period application, and change cost approval form; The supplier and market price information includes one or more of the following: equipment and material quotations, technical attachments to purchase contracts, market price research reports, and supplier qualification documents.

[0027] S2. Standardize the numerical multi-source heterogeneous tectonic valence data and construct numerical feature vectors based on the standardized numerical multi-source heterogeneous tectonic valence data.

[0028] The elements of the numerical feature vector are expressed by the following formula: ; In the formula, Indicates the first The first power transmission and transformation project A standardized numerical multi-source heterogeneous structural value dataset, Indicates the first The first power transmission and transformation project Numerical multi-source heterogeneous structural value data, Indicates the first The mean of a numerical multi-source heterogeneous valence data set. Indicates the first Standard deviation of numerical multi-source heterogeneous valence data This represents a preset constant used to prevent the denominator from being zero; The mean is expressed by the formula: ; In the formula, Indicates the number of power transmission and transformation projects; The standard deviation is expressed by the formula: .

[0029] S3. Encode the text-based multi-source heterogeneous valence data and construct text feature vectors based on the encoded text-based multi-source heterogeneous valence data. The specific steps are as follows: S301. Based on text-based multi-source heterogeneous valence data, a pre-trained language model is used to obtain semantic embedding vectors, expressed by the formula: ; In the formula, Indicates the first The first power transmission and transformation project A semantic embedding vector, This represents a pre-trained language model. Indicates the first The first power transmission and transformation project A text-based, multi-source, heterogeneous valence data set; S302. The semantic strength is obtained by projecting the semantic embedding vector, expressed by the formula: ; In the formula, Indicates the first The first power transmission and transformation project The semantic strength is indicated by a value closer to 1, signifying that the text contains more cost-driving factors; a value closer to 0 indicates a weaker correlation between the text content and cost formation, or a lack of explicit cost-related information in the text itself. express Activation function Indicates the first A semantic projection weight vector, Indicates the first A semantic projection bias; S303. Construct text feature vectors based on semantic strength.

[0030] In at least one embodiment, the pre-trained language model may be implemented using a Sentence-BERT model or a BERT model; S4. Obtain meteorological and geographical data for each power transmission and transformation project, and based on the meteorological and geographical data, obtain the comprehensive meteorological impact value and geographical suitability coefficient for each power transmission and transformation project.

[0031] In at least one embodiment, meteorological and geographical data are normalized to eliminate the influence of dimensions.

[0032] In at least one embodiment, the meteorological data includes one or more of temperature, wind speed, humidity, and rainfall.

[0033] In at least one embodiment, the geographic data includes terrain level ("1" represents plains, "2" represents hills, and "3" represents mountains), road transport distance, and altitude.

[0034] The comprehensive meteorological impact value is expressed by the formula: ; In the formula, Indicates the first The comprehensive meteorological impact value of a power transmission and transformation project. Indicates the quantity of meteorological data. Indicates the first The first-order influence regression coefficient of the meteorological data Numerical multi-source heterogeneous valence data The first in One meteorological data point, Indicates the first The meteorological data and the first The second-order interaction regression coefficients of the meteorological data. Numerical multi-source heterogeneous valence data The first in Meteorological data; First-order term Capture the independent marginal contributions of each meteorological data point, second-order term By capturing the synergistic effects between pairs of meteorological data, this embodiment can reflect the nonlinear amplification effect produced by combined meteorological conditions such as high temperature and humidity, strong winds and heavy rain, compared to the assumption that meteorological factors are completely independent in traditional cost estimation.

[0035] The geographical suitability coefficient is expressed by the formula: ; In the formula, Indicates the first Geographic suitability coefficient of a power transmission and transformation project This represents the bias term in the geographic coefficient regression model. , , Indicates the regression weights. Indicates the first Normalized altitude of a power transmission and transformation project Indicates the first Normalized topographic classification of a power transmission and transformation project Indicates the first Normalized road transport distance for each power transmission and transformation project.

[0036] Geographic data after linear weighted combination and The data is transformed and compressed into a comprehensive coefficient geographic data set with values ​​ranging from 0 to 1. Geographic data closer to 1 indicates more favorable geographical conditions, while data closer to 0 indicates more unfavorable geographical conditions.

[0037] S5. Determine the adjacency matrix based on numerical feature vectors, text feature vectors, geographical suitability coefficients, and comprehensive meteorological impact values. The specific steps are as follows: S501. Concatenate the numerical feature vector, textual feature vector, geographical suitability coefficient, and meteorological comprehensive impact value to obtain the heterogeneous structural value feature vector, expressed by the formula: ; In the formula, Indicates the first The heterogeneous structural eigenvectors of a power transmission and transformation project Indicates the first Numerical feature vectors of a power transmission and transformation project Indicates the first Text feature vectors of a power transmission and transformation project; S502. The neighbor set is determined by calculating the cosine similarity of the eigenvectors of heterogeneous structures among various power transmission and transformation projects, expressed by the formula: ; In the formula, Indicates the first The neighbor set of a power transmission and transformation project Indicates the first The heterogeneous structural eigenvectors of a power transmission and transformation project This indicates the calculation of cosine similarity. Indicates the adjacent cutoff threshold; S503. Determine the adjacency matrix based on the cosine similarity of the heterogeneous eigenvectors between the power transmission and transformation projects in the neighbor set, expressed by the formula: ; In the formula, Represents the adjacency matrix of the th Line 1 Column elements, Indicates the first The heterogeneous structural eigenvectors of a power transmission and transformation project.

[0038] S6. Based on the adjacency matrix, perform neighborhood aggregation on the numerical feature vectors and textual feature vectors of each power transmission and transformation project in the neighbor set to obtain the fused embedding vector, expressed by the formula: ; In the formula, Indicates the first Fusion embedding vectors for individual power transmission and transformation projects; The numerical and textual features of each power transmission and transformation project are weighted and summed according to the adjacency matrix, allowing the cost impact features of similar projects to be mutually transferred in the graph. After this operation, the cost impact features of each power transmission and transformation project are... By incorporating information from neighboring samples, the original features are essentially smoothed under graph constraints. From a signal processing perspective, this operation is equivalent to performing a graph-based low-pass filter on the feature matrix, preserving common components among samples, suppressing individual noise, and making the features influencing engineering costs more continuous among similar projects.

[0039] Furthermore, standard Graph Attention Networks (GATs) introduce learnable linear transformation matrices and attention parameters during neighborhood aggregation, resulting in a relatively large number of parameters. Power transmission and transformation engineering cost datasets are typically small, and a large number of learnable parameters can easily lead to overfitting. This embodiment discards all learnable parameters and only uses fixed similarity weights for weighted summation, fundamentally eliminating the risk of overfitting and enabling this embodiment to work stably even with small sample sizes.

[0040] S7. Based on the fused embedding vector, geographical suitability coefficient, and comprehensive meteorological impact value, a feature modulation factor is constructed, expressed by the formula: ; In the formula, Indicates the first The characteristic modulation factor of a power transmission and transformation project Represents the graph embedding weight vector; In areas with complex terrain and long transportation distances, engineering construction faces problems such as difficulties in material transportation, reduced efficiency of machinery, and increased costs of personnel adaptation. As a result, the cost of the same power transmission and transformation project is significantly higher than that in plain areas. The geographical suitability coefficient is used to determine the quality of geographical conditions, and the threshold value is adjusted accordingly. In areas with low or high temperatures, high wind speeds, and frequent rainfall, construction efficiency decreases, the effective construction period is shortened, and additional measures to expedite the work or seasonal construction costs are required. By quantifying and adjusting the comprehensive meteorological impact value, these environmental disturbances can be adaptively weighted or suppressed by the model. This aligns with the physical law that "the more severe the external environment, the higher the uncertainty of the cost, and the lower the model's dependence on such samples should be."

[0041] S8. Obtain the deep fusion feature vector based on the fusion embedding vector and the feature modulation factor, expressed by the formula: ; In the formula, Indicates the first Deeply fused feature vectors of power transmission and transformation projects express Activation function This represents the weight matrix of the fully connected layer. This represents the bias vector of the fully connected layer; It is the feature vector output after graph propagation, which contains numerical features (physical engineering quantities such as line length and number of towers) and textual features (semantic information such as design difficulty and contract risks), reflecting the ontological attributes of the project - that is, the scale and difficulty of the project itself. It is a feature modulation factor that integrates meteorological and geographical conditions, reflecting the degree of disturbance of the external environment in which the project is located—that is, how much additional impact the environment has on construction. By splicing, the two types of information are merged into the same vector, so that subsequent fully connected layers can simultaneously consider the "features of the project itself" and the "degree of modulation of the features by the environment" in the same calculation process.

[0042] S9. Based on the deep fusion feature vector, an expert experience-guided mechanism is used to obtain the bias judgment value and the hybrid decision weight, expressed by the formula: ; ; In the formula, Indicates the first Deviation judgment value for a power transmission and transformation project Indicates the number of review experts. Indicates the first A fusion feature benchmark vector provided by an expert. This indicates the calculation of the L2 norm. Indicates the first The mixed decision-making weights of individual experts Indicates the first The prior weights of each expert are pre-set according to their professional title level. Indicates the first Prior weights of experts, Indicates the first A fusion feature benchmark vector provided by an expert.

[0043] S10. Based on the deviation judgment value and the deep fusion feature vector, a fully connected layer is used to obtain the cost prediction value, which is expressed by the formula: ; In the formula, Indicates the first The projected cost of a power transmission and transformation project. This represents the weight vector of the fully connected layer. This represents the bias term of the fully connected layer. This represents the deviation regularization penalty coefficient.

[0044] S11. Obtain the correlation between various multi-source heterogeneous structural valence data based on the fusion embedding vector and the hybrid decision weight, and establish the association between various multi-source heterogeneous structural valence data according to the correlation.

[0045] The correlation is expressed by the formula: ; ; ; ; In the formula, Indicates the first stage of power transmission and transformation project The multi-source heterostructure valence data and the first Correlation of multi-source heterogeneous structural value data Indicates the first Normalized weights for individual power transmission and transformation projects Represents the fused embedding vector The dimensional elements, Represents the fused embedding vector The dimensional elements, Represents element The weighted average, Represents element The weighted average, Indicates the first A deep fusion feature vector of a power transmission and transformation project.

[0046] Each dimension directly corresponds to the multi-source heterogeneous structural valence data, therefore The values ​​can be directly traced back to specific engineering feature pairs, which fundamentally ensures the interpretability of the correlation analysis. By introducing mixed decision weights as sample weighting coefficients, the contributions of high-weight experts to the correlation degree are made greater, reflecting the design principle of "expert experience guiding feature analysis".

[0047] Example 2: This embodiment provides an electronic device that stores a computer program. When the computer program is executed by a processor, it implements the multi-source heterogeneous structure price data association method for power transmission and transformation projects as described in any embodiment of the present invention.

[0048] Example 3: This embodiment provides a computer-readable storage medium for storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the multi-source heterogeneous structure cost data association method for power transmission and transformation projects as described in any embodiment of the present invention.

[0049] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0050] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0051] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0052] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0053] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for associating multi-source heterogeneous structural cost data in power transmission and transformation projects, characterized in that, Includes the following steps: Obtain multi-source heterogeneous structural cost data for each power transmission and transformation project, and construct feature vectors based on the multi-source heterogeneous structural cost data; Obtain meteorological and geographical data for each power transmission and transformation project, and based on the meteorological and geographical data, obtain the comprehensive meteorological impact value and geographical suitability coefficient for each power transmission and transformation project; The adjacency matrix is ​​determined based on eigenvectors, geographical suitability coefficients, and comprehensive meteorological impact values. Based on the adjacency matrix, a neighborhood aggregation operation is performed on the feature vectors of each power transmission and transformation project to obtain a fused embedding vector; Deep fusion feature vectors are obtained based on fused embedding vectors, geographic suitability coefficients, and meteorological comprehensive impact values. Hybrid decision weights are obtained by using an expert experience-guided mechanism based on deeply fused feature vectors; The correlation between various multi-source heterogeneous structural valence data is obtained by using fusion embedding vectors and hybrid decision weights, and the association between various multi-source heterogeneous structural valence data is established based on the correlation.

2. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, The feature vectors include numerical feature vectors and text feature vectors; The numerical feature vector is constructed from standardized numerical multi-source heterogeneous valence data; The text feature vector is constructed from encoded textual multi-source heterogeneous valence data.

3. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, The adjacency matrix is ​​determined based on eigenvectors, geographical suitability coefficients, and comprehensive meteorological impact values. The specific steps are as follows: The numerical feature vector, text feature vector, geographical suitability coefficient, and meteorological comprehensive impact value are concatenated to obtain the heterogeneous structural valence feature vector; The neighbor set is determined by calculating the cosine similarity of the heterogeneous eigenvectors among the power transmission and transformation projects; The adjacency matrix is ​​determined based on the cosine similarity of the heterogeneous eigenvectors of each power transmission and transformation project in the neighbor set.

4. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, The deep fusion feature vector is obtained based on the fused embedding vector, geographical suitability coefficient, and meteorological comprehensive impact value. The specific steps are as follows: A feature modulation factor is constructed based on the fusion embedding vector, geographical suitability coefficient, and meteorological comprehensive impact value; The fused embedding vector is concatenated with the feature modulation factor and then... The fully connected layer of the activation function is mapped to a deep fusion feature vector.

5. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, Based on deep fusion feature vectors, an expert experience-guided mechanism is used to obtain hybrid decision weights. The specific steps are as follows: The prior weights of experts are set according to their professional title levels; Obtain the fusion feature benchmark vectors provided by each expert; The L2 norm of the difference between the deep fusion feature vector and the fusion feature baseline vector is negatively exponentially normalized and used as the hybrid decision weight.

6. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, The correlation between various multi-source heterogeneous structural valence data is obtained by using the Pearson correlation coefficient algorithm based on the fusion embedding vector and hybrid decision weights.

7. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 1, characterized in that, The method also includes taking the average of the L2 norms of the differences between the deep fusion feature vector and the fusion feature benchmark vector of each expert as the deviation judgment value for the power transmission and transformation project.

8. The method for associating multi-source heterogeneous structure cost data in power transmission and transformation projects according to claim 7, characterized in that, Cost prediction values ​​are obtained using a fully connected layer based on the deviation judgment value and the deep fusion feature vector.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the multi-source heterogeneous structure cost data association method for power transmission and transformation projects as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the method for associating multi-source heterogeneous cost data in power transmission and transformation projects as described in any one of claims 1 to 8.