Material supply decision-making method for cross-professional data sharing of power grid infrastructure

Abstract

The invention discloses a power grid infrastructure cross-professional data sharing material supply decision-making method. The method comprises the following steps: step 1, carrying out multi-source heterogeneous data integration on data; 2, data cleaning is conducted on the data, and semantic tags are added; 3, a knowledge graph is constructed, and entities comprise suppliers, materials and purchase orders; the relationship comprises supply and association; 4, extracting supplier historical data from the knowledge graph, and generating a supplier material portrait through the constructed neural network model; and 5, taking the generated supplier material portrait as input of a multi-objective optimization algorithm, and taking a Pareto optimal solution as a material supply decision result. According to the invention, the scientific, intelligent and efficient decision making is realized, the data fusion efficiency and accuracy are improved, the optimization effect and compliance of the purchase decision making are ensured, the purchase cost is effectively reduced, the material quality and service level are improved, and the delivery cycle is shortened.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of material management, and in particular to a material supply decision-making method for cross-professional data sharing of power grid infrastructure. BACKGROUND

[0002] In the process of power grid infrastructure, the scientificity and accuracy of material supply decision-making are crucial for the smooth progress, cost control and quality assurance of a project. However, the existing material supply decision-making methods have many problems. On the one hand, the data of systems such as ECP, ERP and e-infrastructure 2.0 are scattered and heterogeneous, data fusion is difficult, and it is difficult to obtain comprehensive and accurate information for decision analysis. On the other hand, the evaluation dimension of the supplier is single, and it is difficult to comprehensively consider factors such as material cost, quality, service and delivery timeliness, resulting in unscientific procurement decision-making, and problems such as high procurement cost, unstable material quality, unsatisfactory supplier service and delivery delay, which affect the overall benefit of the power grid infrastructure project. SUMMARY

[0003] In view of the deficiencies in the prior art, the application provides a material supply decision-making method for cross-professional data sharing of power grid infrastructure, to solve the technical problems of lack of accurate information for decision analysis and single evaluation dimension in the prior art.

[0004] The application provides a material supply decision-making method for cross-professional data sharing of power grid infrastructure, comprising the following steps:

[0005] Step 1: Obtain multiple system data and integrate the data from multiple sources and heterogeneous data;

[0006] Step 2: Data cleaning and adding semantic labels to the integrated data;

[0007] Step 3: Construct a knowledge graph, wherein the entities include suppliers, materials and purchase orders, and the relationships include supply and association;

[0008] Step 4: Extract supplier historical data from the knowledge graph, and generate a supplier material portrait through a constructed neural network model;

[0009] Step 5: Use the generated supplier material portrait as the input of a multi-objective optimization algorithm, and use the Pareto optimal solution as the material supply decision-making result.

[0010] Further, in step 1, the multiple systems include ECP, ERP and e-infrastructure 2.0 systems.

[0011] Further, in the step 1, the multi-source heterogeneous data integration is carried out by adopting the OWL ontology definition, and the integrated data comprises three kinds of entities, i.e., suppliers, materials and purchase orders, and twelve kinds of association relationships among the three kinds of entities.

[0012] Further, in the step 4, the loss function of the constructed neural network model is as follows:

[0013] Loss=β1*MSE 成本 +β2*MSE 质量 +β3*MSE 服务 +β4*MSE 时效 +R(w)

[0014] In the formula, β1, β2, β3 and β4 are weights of each target; MSE is a mean square error; and R(w) is a regularization term.

[0015] Further, a specific formula of the mean square error is as follows:

[0016]

[0017] In the formula, n is a sample quantity; y i is a real score; is a predicted score.

[0018] Further, a specific formula of the regularization term is as follows:

[0019]

[0020] In the formula, n is a sample quantity; α is a mixed proportion of L1 regularization and L2 regularization; λ is a total regularization strength; and w i is a weight parameter in the neural network.

[0021] Further, the step 4 further comprises adjusting each target weight in the loss function according to a period in which a project is located.

[0022] Further, in the step 5, the input of the multi-objective optimization algorithm further comprises the weights of each target.

[0023] Further, in the step 5, the multi-objective optimization algorithm further comprises: performing compliance verification by weight perception logic, and the multi-objective optimization algorithm outputs a solution set which simultaneously satisfies the strengthened constraint condition and the Pareto optimality.

[0024] Further, the step 5 further comprises: when a deviation occurs between a supplier evaluation and an actual performance, adjusting each target weight corresponding to the supplier.

[0025] The present application has the following beneficial effects:

[0026] The application adopts a unified data model, a knowledge graph, a neural network model, an improved NSGA-III algorithm and a closed-loop feedback mechanism, realizes scientific, intelligent and efficient power grid infrastructure material supply decision, improves data fusion efficiency and accuracy, makes supplier evaluation more comprehensive and in-depth, ensures optimization effect and compliance of procurement decision, effectively reduces procurement cost, improves material quality and service level, shortens delivery cycle, and provides strong guarantee for smooth progress of power grid infrastructure project.

[0027] The application realizes efficient fusion and accurate integration of multi-source heterogeneous data through a unified data model and data fusion technology, solves the problems of data dispersion, heterogeneity and semantic ambiguity, provides high-quality, comprehensive and accurate real-time updated data basis for subsequent decision analysis, improves data availability and reliability, and avoids decision errors caused by data problems.

[0028] The application can comprehensively and objectively reflect the comprehensive performance of suppliers in material cost, quality, service and delivery timeliness and other key dimensions through the constructed neural network for evaluation, breaks the limitation of traditional single dimension evaluation, makes procurement decision more scientific and reasonable, can effectively screen out suppliers with better comprehensive performance, and improves the overall quality and supply stability of procurement materials.

[0029] The application considers the differences in procurement demand of different stages of power grid infrastructure, introduces dynamic adjustment weight, can flexibly highlight the key points of attention in different procurement stages, such as timeliness in the early stage, makes the model training more targeted, better meets the actual procurement demand, realizes accurate guidance of multi-dimensional evaluation without modifying the network structure, improves the adaptability and practicality of the model, and enhances the scientificity and flexibility of the decision.

[0030] The application increases the consideration of weight in the multi-objective optimization algorithm in the evaluation, can more effectively search the Pareto optimal solution set in the multi-objective space of cost, quality and delivery cycle, improves the efficiency and accuracy of optimization search, provides decision makers with more high-quality and diversified procurement scheme selection, ensures that the procurement decision can realize comprehensive optimization in multiple key indicators, and improves the configuration efficiency and benefit of procurement resources.

[0031] The application adjusts the weight through the evaluation result feedback, forms a closed-loop decision mechanism, can strictly ensure that the procurement decision meets the compliance requirements such as the procurement specification of the State Grid while pursuing the optimization target, automatically intercepts suppliers who do not meet the qualification, avoids the risk of violation, realizes the dynamic balance of optimization and compliance, improves the reliability and credibility of the decision, and makes the whole procurement decision process more intelligent, efficient and stable. BRIEF DESCRIPTION OF DRAWINGS

[0032] The features and advantages of the present application will be appreciated upon consideration of the following detailed description of specific embodiments of the application given by way of example only, in conjunction with the accompanying drawings, of which:

[0033] Figure 1 is a flowchart of a specific embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0035] The present application will be further illustrated below with reference to specific embodiments. Those skilled in the art should understand that these embodiments are only used to illustrate the present application and not used to limit the scope of the present application. Various modifications of the present application are all within the scope of the appended claims.

[0036] As shown in Figure 1 , the present application provides a material supply decision method for power grid infrastructure cross-professional data sharing, comprising the following steps:

[0037] Step 1: Obtain multiple system data, adopt a unified data model defined by OWL ontology, integrate multiple source heterogeneous data of ECP, ERP and e-infrastructure 2.0 system, correlate and fuse data about suppliers, materials and projects in different systems, and the integrated data includes three types of entities, i.e. suppliers, materials and procurement orders, and twelve types of association relationships between the three types of entities, which enables data of different systems to be fused and interacted under a unified framework;

[0038] Step 2: Clean the integrated data to remove repeated, erroneous and incomplete data and improve data quality; add semantic labels to different data, for example, label data such as supplier name, material category and procurement price as corresponding semantic concepts, to facilitate subsequent data processing and analysis;

[0039] Step 3: Based on the labeled data, utilize graph database technology such as Neo4j to construct a knowledge graph, nodes in the graph represent entities such as suppliers, materials and procurement orders, and edges represent relationships between entities, such as relationships between suppliers and materials, and between procurement orders and materials, thereby converting scattered data into a structured knowledge network;

[0040] Step 4: Extracting supplier historical data from the knowledge graph to build a material procurement evaluation model using a neural network model.

[0041] Extract information related to suppliers and materials from the knowledge graph, including historical procurement prices of materials, quality inspection reports provided by suppliers, past service records, and delivery time records, as input data for the material procurement evaluation model.

[0042] The output layer of the material procurement evaluation model is set with four neurons corresponding to material cost, quality, service, and delivery timeliness, using a linear activation function to directly output the predicted score.

[0043] The material procurement evaluation model scores from four perspectives: cost, quality, service, and timeliness. To reduce computational resource consumption and speed up model iteration, a multi-objective weighted loss function is designed, where β is the weight of each objective. Since power grid infrastructure materials are purchased in batches, procurement needs differ at different stages. For example, during the early stages of a project, rapid progress is needed, so timeliness can be assigned a higher weight to make the model focus more on timeliness during training and reduce prediction errors for timeliness.

[0044] The final loss function is designed as:

[0045] Loss=β1*MSE 成本 +β2*MSE 质量 +β3*MSE 服务 +β4*MSE 时效 +R(w)

[0046] where β1, β2, β3, and β4 are the weights of each objective; MSE is the mean squared error; and R(w) is the regularization term.

[0047] The material procurement evaluation model uses the mean squared error (MSE) to calculate the error between the predicted score and the true score to guide model parameter updates and prevent overfitting during model training. The formula for the mean squared error is:

[0048]

[0049] where n is the number of samples; y i is the true score; is the predicted score.

[0050] The formula for the regularization term R(w) is:

[0051]

[0052] where n is the number of samples; α is the mixing ratio of L1 regularization and L2 regularization; λ is the total regularization strength; w iFor the weight parameters in the neural network.

[0053] In terms of optimization algorithm, Adam algorithm is selected to adaptively adjust the learning rate and accelerate the convergence speed of the model. During the training process, the real scores of each dimension are taken as labels, and the model prediction scores are compared to calculate the loss and update the model parameters by back propagation. The final supplier material portrait contains the cost, quality, service and delivery time of the material.

[0054] Through the constructed material procurement evaluation model, the supplier material portrait is generated;

[0055] Step 5: The generated supplier material portrait and the weight of each target are taken as the input of the multi-objective optimization algorithm NSGA-III. NSGA-III performs dimensionless unification on the four-dimensional score to eliminate the numerical deviation caused by the weight; at the same time, according to the target with high weight, high-density reference points are dynamically generated to guide the algorithm to preferentially explore the high-weight target area. In power grid material procurement, there are hard compliance constraints that cannot be compromised, such as supplier credit rating, material safety certification, and blacklist avoidance. While the multi-objective optimization algorithm NSGA-III pursues high-weight targets such as timeliness, it may output technically optimal but rule-breaking solutions. By adding weight-aware logic to NSGA-III for compliance checking, it ensures that high-weight target optimization does not break the compliance bottom line, and NSGA-III outputs a solution set that meets both the strengthened constraint condition and the Pareto optimality. After a period of decision-making, when there is a deviation between the supplier evaluation and the actual performance, adjust the weight of each target corresponding to the supplier, forming a closed-loop process of "model training-decision analysis".

[0056] Although the embodiments of the present application are described in conjunction with the drawings, various modifications and variations can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for material supply decision-making through cross-disciplinary data sharing in power grid infrastructure construction, characterized in that, Includes the following steps: Step 1: Acquire data from multiple systems and integrate the data from multiple heterogeneous sources; Step 2: Clean the integrated data and add semantic tags; Step 3: Construct a knowledge graph, where entities include: suppliers, materials, and purchase orders; relationships include: supply and association. Step 4: Extract historical supplier data from the knowledge graph and generate supplier material profiles using the constructed neural network model; Step 5: Use the generated supplier material profile as input to the multi-objective optimization algorithm, and use the Pareto optimal solution as the material supply decision result.

2. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 1, characterized in that, In step 1, multiple systems include: ECP, ERP, and e-infrastructure 2.0 system.

3. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 1 or 2, characterized in that, In step 1, multi-source heterogeneous data is integrated by using OWL ontology definition. The integrated data includes three types of entities: suppliers, materials and purchase orders, as well as twelve types of relationships between the three types of entities.

4. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 1, characterized in that, In step 4, the loss function of the constructed neural network model is: Loss=β1*MSE 成本 +β2*MSE 质量 +β3*MSE 服务 +β4*MSE 时效 +R(w) In the formula, β1, β2, β3 and β4 are the weights of each objective; MSE is the mean squared error; R(w) is the regularization term.

5. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 4, characterized in that, The specific formula for the mean square error is: In the formula, n is the sample size; y i Provides genuine ratings; For predicting scores.

6. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 4, characterized in that, The specific formula for the regularization term is: In the formula, n is the number of samples; α is the mixing ratio of L1 regularization and L2 regularization; λ is the total regularization intensity; w i These are the weight parameters in a neural network.

7. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 1 or 4, characterized in that, Step 4 also includes adjusting the weights of each objective in the loss function according to the stage of the project.

8. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 1, characterized in that, In step 5, the input to the multi-objective optimization algorithm also includes the weights of each objective.

9. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 8, characterized in that, In step 5, the multi-objective optimization algorithm further includes: performing compliance verification through weight-aware logic, and outputting a solution set that simultaneously satisfies the enhanced constraint conditions and Pareto optimality.

10. The material supply decision-making method for cross-disciplinary data sharing in power grid infrastructure as described in claim 8 or 9, characterized in that, Step 5 also includes: when there is a discrepancy between the supplier evaluation and the actual performance, adjusting the target weights corresponding to the supplier.

Images