Power engineering supervision scheme generation method and system based on multi-level knowledge graph

Abstract

This application relates to the field of power engineering supervision and management, and discloses a method and system for generating power engineering supervision schemes based on a multi-level knowledge graph. The method includes: acquiring the real-time data fluctuation amplitude corresponding to different levels of models used to generate power engineering supervision schemes; the different levels of models include project-level models, sub-project-level models, item-project-level models, and process-level models, with the time granularity of supervision adjustment for the supervision schemes corresponding to the project-level model, sub-project-level model, item-project-level model, and process-level model decreasing sequentially; when the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model is greater than a preset first fluctuation amplitude, an optimized supervision scheme is generated for the second supervision task corresponding to the second level model and the first supervision task corresponding to the first level model through a second correction module corresponding to the second level model. This application enables the engineering construction supervision scheme to adapt to dynamic changes during the construction process.

Description

Technical Field

[0001] This application relates to the field of power engineering supervision and management technology, and more specifically, to a method and system for generating power engineering supervision schemes based on multi-level knowledge graphs. Background Technology

[0002] In existing technologies, the development of supervision plans is typically based on engineering survey results, combined with construction drawings, relevant industry standards, and project contract requirements. In the early stages of developing a supervision plan, a professional supervision team needs to collect basic information such as the geological conditions of the project site, the surrounding environment, and the qualifications of the construction unit. Subsequently, a specific analysis is conducted on the structural characteristics of the project and the difficulties in construction techniques. During the development of the supervision plan, the key points of supervision at each stage of construction must be considered holistically, clearly defining quality control points, safety supervision nodes, and progress verification cycles. Simultaneously, the core needs of the construction unit, design unit, and construction unit must be integrated to form a complete plan system covering the supervision process, personnel allocation, testing methods, and emergency measures, providing guidance for the supervision work throughout the entire construction process.

[0003] Traditional construction supervision plans are formulated once and followed throughout the entire process, resulting in significant adjustment delays and difficulty in adapting to dynamic changes during construction. The construction phase is susceptible to multiple uncertainties, such as deviations in geological conditions and survey results, design optimizations, fluctuations in construction material performance, and sudden changes in the external environment, all requiring timely adjustments to the supervision plan. However, these adjustments are often delayed. This delay not only forces supervision into a passive follow-up situation, making it impossible to correct construction deviations in a timely manner, but may also lead to quality hazards, schedule delays, and even safety accidents due to mismatches between supervisory measures and actual working conditions, reducing the effectiveness of supervision and the overall construction benefits of the project. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for generating power engineering supervision schemes based on multi-level knowledge graphs, which solves the technical problem that engineering construction supervision schemes are difficult to adapt to dynamic changes in the construction process, and achieves the technical effect that engineering construction supervision schemes can adapt to dynamic changes in the construction process.

[0005] This application provides a method for generating power engineering supervision schemes based on a multi-level knowledge graph. The method includes: acquiring real-time data fluctuation amplitudes corresponding to different levels of models used to generate power engineering supervision schemes; the different levels of models include project-level models, sub-project-level models, item-project-level models, and process-level models; the time granularity of supervision adjustment for the supervision schemes corresponding to the project-level models, sub-project-level models, item-project-level models, and process-level models decreases sequentially; the project-level models, sub-project-level models, item-project-level models, and process-level models are used to generate supervision schemes for power engineering supervision tasks with different time granularities of supervision adjustment for different supervision schemes; and the project-level models, sub-project-level models, item-project-level models, and process-level models utilize historical supervision knowledge at different time granularities of supervision adjustment. The historical supervision knowledge graph obtained through graph training includes supervision scheme input data and supervision scheme output data at different supervision adjustment time granularities. The supervision scheme input data includes engineering parameters, environmental conditions, and specification requirements in the form of a knowledge graph, while the supervision scheme output data includes supervision inspection items, rectification requirements, and acceptance standards in the form of a knowledge graph. When the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, the second correction module corresponding to the second-level model generates an optimized supervision scheme for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model. Among them, the second-level model is a hierarchical model with a supervision adjustment time granularity one level larger than that of the first-level model, and the second supervision task is the superior supervision task of the first supervision task.

[0006] In one possible implementation, the second correction module is trained using the following method: acquiring second historical supervision data at the second supervision adjustment time granularity corresponding to the second-level model, and acquiring first historical supervision data at the first supervision adjustment time granularity corresponding to the first-level model during the same period of the second historical supervision model; determining first target historical supervision data in the first historical supervision data where the first fluctuation amplitude is greater than a preset first fluctuation amplitude; acquiring second target historical supervision data during the same period of the first target historical supervision data in the second historical supervision data; acquiring first supervision scheme correction data of the first-level model and second supervision scheme correction data of the second-level model corresponding to the first target historical supervision data and the second target historical supervision data; and training the second correction module based on the first target historical supervision data, the second target historical supervision data, the first supervision scheme correction data, and the second supervision scheme correction data.

[0007] In another possible implementation, the method further includes: obtaining a first fluctuation range corresponding to the first fluctuation amplitude of the real-time data corresponding to the first-level model; obtaining multiple second correction modules corresponding to different first fluctuation ranges of the real-time data; determining a target second correction module corresponding to the first fluctuation range among the multiple second correction modules corresponding to different first fluctuation ranges of the real-time data; generating an optimized supervision plan for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model through the target second correction module; wherein, the target second correction module is trained using first target historical supervision data, second target historical supervision data, first supervision plan correction data of the first-level model corresponding to the first target historical supervision data and second target historical supervision data, and second supervision plan correction data of the second-level model, the first target historical supervision data includes first historical supervision data whose real-time data fluctuation amplitude is within the first fluctuation range corresponding to the first historical supervision model, and the second target historical supervision data is second historical supervision data corresponding to the first target historical supervision data at the same time.

[0008] In another possible implementation, the method further includes: when the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, obtaining the first switching transition time period corresponding to the first-level model and the second-level model, and determining the number of times the first fluctuation amplitude within the first switching transition time period is greater than the preset first fluctuation amplitude; obtaining the preset number of times the first fluctuation exceeds the limit corresponding to the first switching transition time period; when the number of times the first fluctuation exceeds the limit is greater than or equal to the preset number of times the first fluctuation exceeds the limit, generating an optimized supervision plan for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model; when the number of times the first fluctuation exceeds the limit is less than the preset number of times the first fluctuation exceeds the limit, continuing to generate a basic supervision plan for the first supervision task through the first-level model.

[0009] In another possible implementation, the switching transition time period corresponding to different level models is determined by the following method: obtaining the second supervision adjustment time granularity corresponding to the second level model; obtaining the first fluctuation sensitivity coefficient of the first supervision task corresponding to the first level model and the second fluctuation sensitivity coefficient of the second supervision task corresponding to the second level model, where the first fluctuation sensitivity coefficient characterizes the speed at which the first supervision task responds to data fluctuations and the second fluctuation sensitivity coefficient characterizes the speed at which the second supervision task responds to data fluctuations; and determining the product of the second supervision adjustment time granularity, the first fluctuation sensitivity coefficient, and the second fluctuation sensitivity coefficient as the switching transition time period.

[0010] In another possible implementation, the method further includes: when the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, acquiring multiple first target historical supervision data with the same supervision adjustment time granularity as the first real-time supervision data and a first fluctuation amplitude greater than the preset first fluctuation amplitude, and acquiring multiple first target historical level models for generating supervision schemes based on the first target historical supervision data; determining multiple first similarities corresponding to the multiple first target historical supervision data and the first real-time supervision data, and determining the maximum first similarity among the multiple first similarities, and generating a first target supervision scheme for the first supervision task based on the first target historical level model corresponding to the maximum first similarity.

[0011] In another possible implementation, the method further includes: acquiring multiple second historical supervision data and multiple second historical hierarchical models, wherein the multiple second historical hierarchical models generate a second supervision plan for the second supervision task based on the multiple second historical supervision data; acquiring second real-time supervision data from the same period as the first real-time supervision data, determining multiple second similarities between the multiple second historical supervision data and the second real-time supervision data, and determining the maximum second similarity among the multiple second similarities; and generating a second target supervision plan for the second supervision task based on the second historical hierarchical model corresponding to the maximum second similarity.

[0012] In another possible implementation, the method further includes: obtaining the cause of the fluctuation exceeding the preset first fluctuation range when the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model exceeds the preset first fluctuation amplitude; wherein, the cause of the fluctuation exceeding the limit includes construction violation operation, equipment quality defect, sudden change in environmental conditions, and design scheme change; when the cause of the fluctuation exceeding the limit includes at least one of construction violation operation or equipment quality defect, an optimized supervision scheme is generated synchronously for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model; when the cause of the fluctuation exceeding the limit includes at least one of sudden change in environmental conditions or design scheme change, an optimized supervision scheme is generated for the first supervision task according to the first historical level model corresponding to the maximum first similarity, and an optimized supervision scheme is generated for the second supervision task according to the second historical level model corresponding to the maximum second similarity.

[0013] In another possible implementation, the method further includes: when the cause of the fluctuation exceeding the limit includes at least one of sudden changes in environmental conditions and changes in design schemes, as well as at least one of illegal construction operations and equipment quality defects, an optimized supervision scheme is generated simultaneously for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model.

[0014] This application also provides a power engineering supervision scheme generation system based on a multi-level knowledge graph, including units for implementing the above-described power engineering supervision scheme generation method based on a multi-level knowledge graph.

[0015] The beneficial effects of the embodiments in this application compared with the prior art are:

[0016] This application provides a method for generating power engineering supervision schemes based on a multi-level knowledge graph. The method includes: acquiring the real-time data fluctuation amplitudes corresponding to different level models used to generate the power engineering supervision scheme; when the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model is greater than a preset first fluctuation amplitude, generating an optimized supervision scheme for the second supervision task corresponding to the second level model and the first supervision task corresponding to the first level model through a second correction module corresponding to the second level model. The method in this application enhances the resilience of power engineering supervision in responding to sudden quality hazards, schedule delays, and other unexpected problems, optimizes the efficiency of supervision resource utilization, avoids waste of supervision resources caused by changes in construction load or fluctuations in the external environment, and reduces supervision work costs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the first method for generating power engineering supervision schemes based on multi-level knowledge graphs, provided in this application embodiment;

[0019] Figure 2 A schematic diagram illustrating the workflow of the first method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment;

[0020] Figure 3 A schematic diagram of the workflow of the second method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in the embodiments of this application;

[0021] Figure 4 A schematic diagram of the workflow of the third method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in the embodiments of this application;

[0022] Figure 5 A schematic diagram illustrating the workflow of the fourth method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment;

[0023] Figure 6 A schematic diagram illustrating the workflow of the fifth method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment;

[0024] Figure 7 A schematic diagram illustrating the workflow of the sixth method for generating power engineering supervision schemes based on multi-level knowledge graphs, provided in this application embodiment;

[0025] Figure 8 A schematic diagram of the workflow of the seventh method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in the embodiments of this application;

[0026] Figure 9 is a schematic diagram of the logical structure of a power engineering supervision scheme generation system based on a multi-level knowledge graph provided in an embodiment of this application. Detailed Implementation

[0027] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0028] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0029] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0030] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0031] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0032] The construction phase of a project is susceptible to multiple uncertainties. The lag in adjusting the supervision plan not only puts the supervision work in a passive follow-up predicament, making it impossible to correct construction deviations in a timely manner, but may also lead to quality hazards, schedule delays, or even safety accidents due to the mismatch between supervision measures and actual working conditions, thereby reducing the effectiveness of supervision work and the overall construction benefits of the project.

[0033] Based on the above reasons, this application provides a method for generating power engineering supervision schemes based on multi-level knowledge graphs. The method includes: acquiring real-time data fluctuation amplitudes corresponding to different levels of models used to generate power engineering supervision schemes; the different levels of models include project-level models, sub-project-level models, item-project-level models, and process-level models, with the supervision adjustment time granularity corresponding to the project-level model, sub-project-level model, item-project-level model, and process-level model decreasing sequentially; the project-level model, sub-project-level model, item-project-level model, and process-level model are used to generate supervision schemes for power engineering supervision tasks with different supervision adjustment time granularities; the project-level model, sub-project-level model, item-project-level model, and process-level model utilize historical supervision knowledge at different supervision adjustment time granularities. The historical supervision knowledge graph obtained through graph training includes supervision scheme input data and supervision scheme output data at different supervision adjustment time granularities. The supervision scheme input data includes engineering parameters, environmental conditions, and specification requirements in the form of a knowledge graph. The supervision scheme output data includes supervision inspection items, rectification requirements, and acceptance standards in the form of a knowledge graph. When the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, an optimized supervision scheme is generated by the second correction module corresponding to the second-level model for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model. The second-level model is a hierarchical model with a supervision adjustment time granularity one level larger than the first-level model, and the second supervision task is the superior supervision task of the first supervision task. The method in this embodiment enhances the resilience of power engineering supervision in dealing with sudden quality hazards, schedule delays, and other unexpected problems, optimizes the efficiency of supervision resource utilization, avoids waste of supervision resources caused by changes in construction load or external environmental fluctuations, and reduces supervision work costs.

[0034] In some scenarios, the power engineering supervision scheme generation method based on multi-level knowledge graph of this application embodiment can be applied to the generation of power engineering supervision schemes, which can improve the efficiency of power engineering supervision scheme formulation and improve the construction supervision effect.

[0035] The following section provides a detailed explanation of a power engineering supervision scheme generation method based on a multi-level knowledge graph, as provided in the embodiments of this application, using specific examples.

[0036] Figure 1 A flowchart illustrating the first method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application is shown below. Figure 1 As shown in the embodiment of this application, a method for generating power engineering supervision schemes based on multi-level knowledge graphs is provided, including S110 to S120. S110 to S120 will be described in detail below.

[0037] S110. Obtain the real-time data fluctuation amplitude corresponding to different levels of models used to generate power engineering supervision schemes. These different levels of models include project-level models, sub-project-level models, item-by-item-level models, and process-level models. The time granularity of supervision adjustments for these models decreases sequentially. These models are used to generate supervision schemes for power engineering supervision tasks with different time granularities of supervision adjustments. The project-level, sub-project-level, item-by-item-level, and process-level models are trained using historical supervision knowledge graphs with different time granularities of supervision adjustments. These historical knowledge graphs include supervision scheme input data and supervision scheme output data. The supervision scheme input data includes engineering parameters, environmental conditions, and specification requirements in the form of a knowledge graph. The supervision scheme output data includes supervision inspection items, rectification requirements, and acceptance standards in the form of a knowledge graph.

[0038] Figure 2 A schematic diagram of the workflow of the first power engineering supervision scheme generation method based on multi-level knowledge graph provided in the embodiments of this application is shown below. Figure 2 As shown, the power engineering supervision scheme generation method based on multi-level knowledge graph includes obtaining the real-time data fluctuation range corresponding to different level models used to generate power engineering supervision schemes. The different level models include project-level models, sub-project-level models, sub-item-level models, and process-level models. The supervision adjustment time granularity of the supervision schemes corresponding to these level models decreases sequentially, which can generate supervision schemes for power engineering supervision tasks with different supervision adjustment time granularities.

[0039] It should be noted that these hierarchical models are trained using historical supervision knowledge graphs with different time granularities for supervision adjustments. These historical supervision knowledge graphs include supervision scheme input data and supervision scheme output data. The supervision scheme input data includes engineering parameters, environmental conditions, and specification requirements in the form of knowledge graphs, while the supervision scheme output data includes supervision inspection items, rectification requirements, and acceptance standards in the form of knowledge graphs.

[0040] For example, in the process of power engineering supervision, the real-time data fluctuation range corresponding to each level of model can be continuously obtained. The project-level model can focus on the macro progress and quality indicator change trend of the entire project. The sub-project-level model can monitor the implementation of sub-projects such as substation construction and line laying. The sub-item-level model can track the detailed parameters of specific construction links. The process-level model can monitor the execution status of individual construction processes. The time granularity of supervision adjustment corresponding to these levels of models also decreases sequentially.

[0041] S120. When the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, an optimized supervision plan is generated for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model through the second correction module corresponding to the second-level model. Here, the second-level model is a hierarchical model with a larger time granularity for supervision adjustment than the first-level model, and the second supervision task is the superior supervision task of the first supervision task.

[0042] like Figure 2 As shown in this implementation, when the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, an optimized supervision plan can be generated for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model through the second correction module corresponding to the second-level model. The second-level model is a hierarchical model with a larger time granularity of supervision adjustment than the first-level model, and the second supervision task is the superior supervision task of the first supervision task.

[0043] For example, in power line construction supervision, if the process-level model detects that the real-time data fluctuation of the cable laying process exceeds the preset value, the corresponding correction module of the sub-project level model can simultaneously generate an optimized supervision plan for the line laying sub-project and the cable laying process. This hierarchical linkage mechanism can ensure the coordination and consistency of the supervision plan.

[0044] The beneficial effects of the above implementation method are as follows: By using a hierarchical structure with progressively decreasing time granularity for supervision adjustments—project-level, sub-project-level, item-level, and process-level models—data deviations in supervision can be detected in real time. When a data deviation at a certain level exceeds a preset threshold, a unified supervision plan for the corresponding two levels of supervision objects is generated through the correction module corresponding to the next higher-level model. Through dynamic model switching and targeted correction, the global control stability of the coarse-grained supervision model compensates for the unreliability of plan generation when data deviations are too large, achieving adaptive generation of supervision plans. This effectively avoids errors in supervision plans caused by real-time supervision data anomalies such as distorted quality inspection data and deviations in progress records, reducing the mismatch between supervision plans and actual project needs. By dynamically monitoring the real-time data deviation amplitude of models with different time granularities, adaptive adjustment of power engineering supervision plans is achieved, significantly improving the relevance of supervision plans and the overall efficiency of engineering supervision work.

[0045] The aforementioned implementation method also has the beneficial effects of enhancing the resilience of power engineering supervision in dealing with sudden quality problems, delays in progress, and other unexpected issues. It also optimizes the efficiency of supervision resource utilization, avoids the waste of supervision resources caused by changes in construction load or fluctuations in the external environment, reduces the cost of supervision work, and ensures the continuity and reliability of supervision plans, providing a more intelligent and flexible solution for generating plans for power engineering supervision.

[0046] In some implementations, the second correction module is trained through S210 to S220 in the above method. S210 to S220 will be explained in detail below.

[0047] S210. Obtain the second historical supervision data corresponding to the second level model at the second supervision adjustment time granularity, and obtain the first historical supervision data corresponding to the first level model at the same time period of the second historical supervision model at the first supervision adjustment time granularity.

[0048] Figure 3 A schematic diagram of the workflow for the second method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 3 As shown, the second correction module can be trained by acquiring the second historical supervision data corresponding to the second supervision adjustment time granularity of the second-level model. At the same time, it can acquire the first historical supervision data of the first supervision adjustment time granularity of the first-level model corresponding to the second historical supervision model at the same time. These data can cover the historical records of different supervision adjustment time granularity levels, providing a sufficient data foundation for subsequent training.

[0049] S210. Determine a first target historical supervision data point from the first historical supervision data point where the first fluctuation amplitude is greater than a preset first fluctuation amplitude. Obtain second target historical supervision data from the second historical supervision data point for the same period as the first target historical supervision data point. Obtain the first supervision scheme correction data of the first-level model and the second supervision scheme correction data of the second-level model corresponding to the first target historical supervision data and the second target historical supervision data. Train a second correction module based on the first target historical supervision data, the second target historical supervision data, the first supervision scheme correction data, and the second supervision scheme correction data.

[0050] In this implementation, a first target historical monitoring data with a first fluctuation amplitude greater than a preset first fluctuation amplitude can be determined from the first historical monitoring data. This filtering method can identify abnormal situations with large data fluctuations. By filtering out these key data, the correction module can be trained more effectively.

[0051] In this implementation, the second target historical supervision data can be further obtained from the second historical supervision data at the same time as the first target historical supervision data. This synchronous acquisition can ensure that the data at the two levels are consistent in the time dimension, which is convenient for subsequent comparative analysis and model training. Time-synchronized data helps to accurately reflect the data characteristics of different granular levels within the same supervision period.

[0052] In this implementation, the first supervision scheme correction data of the first-level model and the second supervision scheme correction data of the second-level model corresponding to the first target historical supervision data and the second target historical supervision data can be further obtained. These correction data record the adjustment records of the scheme during the historical supervision process, providing a reference for training the correction module.

[0053] In this implementation, a second correction module can be trained based on the historical supervision data of the first target, the historical supervision data of the second target, the correction data of the first supervision plan, and the correction data of the second supervision plan. The second correction module can be a neural network model based on deep learning. The model is trained by using sample historical supervision data and corresponding correction data, so that the model can learn the correction pattern when the data fluctuates greatly.

[0054] For example, in the process of power engineering supervision, when it is found that the supervision data of the first-level model has large fluctuations, the supervision plan can be adjusted through the second correction module. This training method enables the correction module to specifically handle scenarios with large data deviations, thereby improving the accuracy of the supervision plan.

[0055] The beneficial effects of the above implementation method are as follows: it obtains historical supervision data of the second supervision adjustment time granularity corresponding to the second-level model, and simultaneously obtains historical supervision data of the first supervision adjustment time granularity of the first-level model during the same period. Then, it filters out the first target historical supervision data containing the first data deviation exceeding the standard from these data, and matches the corresponding correction data. Finally, based on these target data and correction data that can cover scenarios with excessive deviation, a second correction module that can accurately deal with such deviation situations is trained. This ensures that the correction module can accurately match the scheme generation requirements of supervision data deviation exceeding the standard, avoiding the defect of insufficient adaptability of general correction methods to deviation scenarios, and greatly reducing problems such as mismatch between supervision schemes and actual project needs and supervision omissions caused by supervision data deviation.

[0056] In some implementations, the above method also includes S310 to S320, which are described in detail below.

[0057] S310. Obtain the first fluctuation range corresponding to the first fluctuation amplitude of the real-time data corresponding to the first level model. Obtain multiple second correction modules corresponding to different first fluctuation amplitude ranges of the real-time data.

[0058] Figure 4 A schematic diagram of the workflow for the third method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 4 As shown, in this implementation, the first fluctuation range corresponding to the first fluctuation amplitude of the real-time data corresponding to the first level model can be obtained first. The first fluctuation range represents the degree range of the real-time data deviating from the expected value. By dividing different fluctuation ranges, the degree of data deviation can be finely classified.

[0059] For example, in the process of power engineering supervision, when a fluctuation is detected in a certain supervision data, multiple fluctuation ranges can be divided according to the degree of deviation of the supervision data, and each range corresponds to a different level of data anomaly.

[0060] In this implementation, multiple second correction modules corresponding to different first fluctuation ranges of real-time data can be obtained. Each second correction module is specifically designed for data deviation within a specific fluctuation range. These correction modules constitute a correction module library for different degrees of deviation.

[0061] For example, in power line supervision, multiple second correction modules can be set up to address voltage fluctuations, corresponding to different amplitude ranges such as slight fluctuations, moderate fluctuations, and severe fluctuations.

[0062] S320. Among multiple second correction modules corresponding to different first fluctuation ranges in real-time data, determine the target second correction module corresponding to the first fluctuation range. Generate optimized supervision schemes for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model using the target second correction module. The target second correction module is trained using first target historical supervision data, second target historical supervision data, first supervision scheme correction data of the first-level model corresponding to the first target historical supervision data and the second target historical supervision data, and second supervision scheme correction data of the second-level model. The first target historical supervision data includes first historical supervision data corresponding to the first historical supervision model whose real-time data fluctuation range is within the first fluctuation range. The second target historical supervision data is the second historical supervision data corresponding to the first target historical supervision data at the same time.

[0063] In this implementation, a target second correction module corresponding to the first fluctuation range can be determined from multiple second correction modules corresponding to different first fluctuation ranges of real-time data. By matching the first fluctuation range of the current real-time data with the preset fluctuation range, the target correction module most suitable for the current deviation can be accurately selected from multiple second correction modules.

[0064] For example, when an abnormal fluctuation of a specific amplitude is detected in the current data of power equipment, a target second correction module that matches the fluctuation amplitude can be selected from multiple second correction modules for processing.

[0065] When determining the optimized supervision plan, the second target correction module can generate optimized supervision plans for the second supervision task corresponding to the second level model and the first supervision task corresponding to the first level model. The second target correction module can comprehensively consider the needs of supervision tasks at different levels and generate more accurate optimized supervision plans based on the current data fluctuation situation.

[0066] For example, in substation engineering supervision, the second target correction module can simultaneously handle supervision tasks at two levels: equipment operation status monitoring and construction quality supervision, generating a coordinated and unified optimized supervision plan.

[0067] In this implementation, the second target correction module can be trained using the historical supervision data of the first target, the historical supervision data of the second target, the first supervision scheme correction data of the first-level model corresponding to the historical supervision data of the first target and the second supervision scheme correction data of the second-level model. During the training process, historical supervision data and correction scheme data corresponding to specific fluctuation ranges are used to ensure that the correction module has targeted correction capabilities.

[0068] For example, the second correction module can be trained using a deep learning model, using historical monitoring data and corresponding successful correction schemes that occurred within the same fluctuation range as training samples.

[0069] In this implementation, the first target historical supervision data includes the first historical supervision data corresponding to the first historical supervision model whose real-time data fluctuation range is within the first fluctuation range. These historical data are specifically selected from past supervision cases that meet the current fluctuation range and have a high degree of scenario relevance.

[0070] For example, in power engineering quality supervision, the primary target historical supervision data can include monitoring data and supervision records of transformer temperature fluctuations within the same range throughout history.

[0071] In this implementation, the second target historical supervision data is the second historical supervision data corresponding to the first target historical supervision data at the same time. These data are synchronized with the first target historical supervision data in time, providing supervision information at different levels or in different dimensions within the same time period.

[0072] For example, in power engineering construction supervision, the second target historical supervision data may include construction environment data and safety monitoring data collected simultaneously with equipment operation data.

[0073] The beneficial effects of the above implementation method are that it can further accurately locate the specific range of deviation of the first real-time supervision data in the first-level model, and multiple second correction modules corresponding to different deviation ranges in the second-level model. Each second correction module is trained by selecting historical supervision models and high-granular data of the same period corresponding to the deviation range. Therefore, it can accurately match the target second correction module for the current deviation scenario from multiple correction modules. Finally, the supervision plan is generated through the correction module, avoiding the defect of poor adaptability of general correction modules to different deviation scenarios. This greatly improves the accuracy of supervision plan generation under data deviation scenarios and effectively reduces problems such as excessive or insufficient supervision intervention and engineering quality risks caused by improper deviation range adaptation.

[0074] In some implementations, the above method also includes S410 to S420, which are described in detail below.

[0075] S410. When the first fluctuation amplitude of the first real-time monitoring data corresponding to the first-level model is greater than the preset first fluctuation amplitude, obtain the first switching transition time period corresponding to the first-level model and the second-level model, and determine the number of times the first fluctuation amplitude is greater than the preset first fluctuation amplitude within the first switching transition time period. Obtain the preset number of times the first fluctuation exceeds the limit corresponding to the first switching transition time period.

[0076] Figure 5 A schematic diagram of the workflow for the fourth method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 5 As shown, the first fluctuation amplitude of the first real-time monitoring data corresponding to the first-level model can be obtained. When the first fluctuation amplitude is greater than the preset first fluctuation amplitude, the first switching transition time period corresponding to the first-level model and the second-level model can be obtained.

[0077] In this implementation, the first switching transition period is the time interval between the transition from the first-level model to the second-level model. During this period, the continuous changes in the supervision data can be observed.

[0078] In this implementation, the number of times the first fluctuation amplitude is greater than the preset first fluctuation amplitude during the first switching transition period can be determined. The number of times the first fluctuation exceeds the limit reflects the cumulative number of times the supervision data fluctuation exceeds the preset range during the first switching transition period. By counting this number, the persistence characteristics of the data fluctuation can be evaluated.

[0079] At the same time, the preset number of times the first fluctuation exceeds the limit corresponding to the first switching transition period can also be obtained. The preset number of times the first fluctuation exceeds the limit can be a reference standard set based on historical supervision data, which is an important basis for judging whether model switching is required.

[0080] S420. When the number of times the first fluctuation exceeds the limit is greater than or equal to the preset number of times the first fluctuation exceeds the limit, an optimized supervision plan is generated for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model. When the number of times the first fluctuation exceeds the limit is less than the preset number of times the first fluctuation exceeds the limit, a basic supervision plan is generated for the first supervision task through the first-level model.

[0081] In this implementation, when the number of times the first fluctuation exceeds the limit is greater than or equal to the preset number of times the first fluctuation exceeds the limit, the second correction module corresponding to the second level model can generate an optimized supervision plan for the second supervision task and the first supervision task. The second correction module can make fine adjustments to the supervision task and generate an optimized plan that is more in line with the actual engineering needs.

[0082] In this implementation, when the number of times the first fluctuation exceeds the limit is less than the preset number of times the first fluctuation exceeds the limit, a basic supervision plan can be generated for the first supervision task through the first-level model, which can avoid unnecessary adjustments caused by short-term data fluctuations.

[0083] For example, in the process of power engineering supervision, when the first fluctuation of the first real-time supervision data is found to exceed the preset range, the data fluctuation during the first switching transition period can be observed. If the number of times the first fluctuation exceeds the limit reaches the preset standard, the correction function of the second-level model is activated to generate an optimization plan; if the preset standard is not reached, a basic supervision plan is generated for the first supervision task through the first-level model.

[0084] The beneficial effect of the above implementation method is that when the deviation of the supervision data of the first-level model exceeds the preset threshold, the scheme generation model is not directly switched. Instead, the judgment is further combined with the switching transition period: the switching transition period corresponding to the two-level model is obtained, the number of deviations exceeding the limit during the switching transition period is counted and compared with the preset value. If the number of deviations exceeds the limit, the correction module of the high-granularity model is started to generate a scheme. If it does not meet the limit, the original model is used. This avoids frequent scheme adjustments caused by short-term small deviations in supervision data and ensures the stability and continuity of supervision work.

[0085] The beneficial effect of the above implementation method is that when there are persistent and significant deviations in the supervision data, the correction module of the higher-level model can be called in a timely manner to generate a solution. This fully leverages the advantages of the macro-level supervision model in grasping the overall trend of the project, and realizes the accuracy and efficiency of the power engineering supervision solution in dealing with complex data deviations, thereby improving the quality stability and resource utilization efficiency of the overall supervision work.

[0086] In some implementations, in S410 above, the switching transition time periods corresponding to different level models are determined by S411 to S412. S411 to S412 will be explained in detail below.

[0087] S411. Obtain the second supervision adjustment time granularity corresponding to the second-level model. Obtain the first fluctuation sensitivity coefficient of the first supervision task corresponding to the first-level model and the second fluctuation sensitivity coefficient of the second supervision task corresponding to the second-level model. The first fluctuation sensitivity coefficient represents the speed at which the first supervision task responds to data fluctuations, and the second fluctuation sensitivity coefficient represents the speed at which the second supervision task responds to data fluctuations.

[0088] Figure 6 A schematic diagram of the workflow for the fifth method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 6 As shown in this implementation, the second supervision adjustment time granularity corresponding to the second-level model can be obtained. The second supervision adjustment time granularity reflects the time interval characteristics of parameter adjustment of the second-level model during the generation of the supervision plan.

[0089] In this implementation, the first fluctuation sensitivity coefficient of the first supervision task corresponding to the first-level model and the second fluctuation sensitivity coefficient of the second supervision task corresponding to the second-level model can be obtained simultaneously. The first fluctuation sensitivity coefficient represents the speed at which the first supervision task responds to data fluctuations, and the second fluctuation sensitivity coefficient represents the speed at which the second supervision task responds to data fluctuations.

[0090] For example, in power engineering supervision, the first-level model may be responsible for the macro-monitoring of the overall project progress, and its corresponding first supervision task has relatively low requirements for responding to data fluctuations. The second-level model may be responsible for the quality inspection of specific construction links, and its corresponding second supervision task has higher requirements for responding to data fluctuations. The fluctuation sensitivity coefficient can be an empirical value assigned by analyzing historical supervision data, reflecting the sensitivity characteristics of different supervision tasks to data changes.

[0091] For example, the faster the first supervision task responds to data fluctuations, the smaller the first fluctuation sensitivity coefficient; the faster the second supervision task responds to data fluctuations, the smaller the second fluctuation sensitivity coefficient.

[0092] S412. Determine the product of the second supervisor's adjustment time granularity, the first fluctuation sensitivity coefficient, and the second fluctuation sensitivity coefficient as the switching transition time period.

[0093] In this implementation, the product of the second supervisor's adjustment time granularity, the first fluctuation sensitivity coefficient, and the second fluctuation sensitivity coefficient can be determined as the switching transition time period. This calculation method comprehensively considers the model's time characteristics and the task's data sensitivity, making the determination of the switching transition time period more scientific and reasonable. The switching transition time period obtained through product calculation can better adapt to the switching needs between different levels of models.

[0094] For example, in the process of generating a power engineering supervision plan, when the adjustment time granularity of the second-level model is large and the supervision tasks corresponding to the two-level models are not sensitive to data fluctuations, the calculated switching transition time period will be extended accordingly. This ensures that the influence of various factors is fully considered during the model switching process, and avoids deviations in the supervision plan due to improper switching time settings.

[0095] The beneficial effect of the above implementation method is that, when determining the transition period between the two-level models, the transition period is calculated by combining the time granularity characteristics of the second-level model and the sensitivity of the two-level supervision objects to data deviations. This ensures that the transition time is compatible with the characteristics of the supervision objects and the time granularity of the model in the actual supervision scenario, avoiding deviations in scheme judgment caused by unreasonable transition time settings, avoiding ignoring short-term fluctuations due to excessively fast transitions, and avoiding delays in handling issues caused by excessively slow transitions.

[0096] The beneficial effects of the above implementation method are also from a technical perspective. By combining the time granularity of the hierarchical model with the deviation sensitivity coefficient of the corresponding supervised object to determine the switching transition period, the scientific nature and pertinence of the determination of the switching transition period are significantly improved, and the problem of mismatch between the switching time and the actual supervision needs of the project is effectively solved.

[0097] The beneficial effects of the above implementation method are that it reduces the delay or misadjustment of the supervision plan generation caused by improper switching transition time period settings, thereby improving the stability and accuracy of the entire multi-level model supervision plan generation method, ensuring the reliability of engineering supervision work under different supervision time granularities, and reducing the risks to engineering quality and schedule caused by plan deviations.

[0098] In some implementations, the above method also includes S510 to S520, which are described in detail below.

[0099] S510. When the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model is greater than the preset first fluctuation amplitude, obtain multiple first target historical supervision data with the same supervision adjustment time granularity as the first real-time supervision data and a first fluctuation amplitude greater than the preset first fluctuation amplitude, and obtain multiple first target historical level models for generating supervision schemes based on the first target historical supervision data respectively.

[0100] Figure 7 A schematic diagram of the workflow for the sixth method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 7 As shown, in the process of power engineering supervision, when the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model is greater than the preset first fluctuation amplitude, multiple first target historical supervision data with the same supervision adjustment time granularity as the first real-time supervision data and a first fluctuation amplitude greater than the preset first fluctuation amplitude can be obtained.

[0101] At the same time, multiple first-target historical hierarchical models can be obtained to generate supervision plans based on the first-target historical supervision data. These first-target historical hierarchical models are used to generate supervision plans for specific fluctuation scenarios.

[0102] It should be noted that when obtaining multiple first-target historical hierarchical models, these models are trained using first-historical supervision data from different time periods. Using multiple first-target historical hierarchical models can further improve the adjustment effect of the engineering supervision plan.

[0103] For example, in the construction supervision of substations, when the fluctuation range of equipment installation progress in real-time supervision data exceeds the preset range, multiple historical supervision data with the same time granularity and fluctuation range exceeding the preset range can be filtered out from the historical database. These historical data are similar sudden progress deviation scenarios from different projects.

[0104] S520. Determine multiple first similarities corresponding to multiple first target historical supervision data and first real-time supervision data, and determine the maximum first similarity among multiple first similarities. Generate a first target supervision plan for the first supervision task based on the first target historical hierarchy model corresponding to the maximum first similarity.

[0105] In this implementation, multiple first similarities can be determined for multiple first target historical supervision data and multiple first real-time supervision data, and the maximum first similarity among multiple first similarities can be determined. Then, a first target supervision plan can be generated for the first supervision task based on the first target historical hierarchical model corresponding to the maximum first similarity. Through similarity matching, the historical model most similar to the current deviation scenario can be selected to generate the supervision plan.

[0106] For example, in the construction supervision of power transmission lines, by calculating the similarity between the current supervision data and multiple historical supervision data, we can find the historical cases most similar to the current sudden changes in geological conditions that cause fluctuations in construction progress. By selecting the hierarchical model corresponding to the historical case with the highest similarity, we can generate supervision adjustment plans for the current changes in geological conditions, including suggestions for adjusting construction methods and rescheduling progress.

[0107] The beneficial effect of the above implementation method is that by matching the current high deviation supervision scenario with similar historical deviation events, it can more accurately call the historical supervision scheme generation strategy that has been verified by engineering, instead of simply switching to a model with a larger time granularity. This enhances the pertinence and scenario adaptability of the supervision scheme. In particular, when dealing with sudden deviations, it can reduce the scheme error caused by model mismatch and improve the operability of the supervision scheme.

[0108] In some implementations, the above method also includes S530 to S540, which are described in detail below.

[0109] S530. Obtain multiple second historical supervision data and multiple second historical hierarchical models. The multiple second historical hierarchical models generate second supervision plans for the second supervision tasks based on the multiple second historical supervision data.

[0110] Figure 8 A schematic diagram of the workflow for the seventh method for generating power engineering supervision schemes based on multi-level knowledge graphs provided in this application embodiment is shown below. Figure 8 As shown, in the process of power engineering supervision, multiple second historical supervision data and multiple second historical hierarchical models can be obtained. These second historical hierarchical models are obtained by training the second supervision tasks with multiple second historical supervision data, and can generate corresponding second supervision schemes.

[0111] For example, in power engineering supervision, the second historical supervision data can include key indicators such as construction efficiency records, quality risk point statistics, and resource allocation status at different times during the seasonal construction phase. The second historical hierarchical model can learn supervision strategies for different time periods through this historical data, and then generate supervision plans for new second supervision tasks.

[0112] It should be noted that when acquiring multiple second-historical-level models, these models are trained using first-historical supervision data from different time periods. Using multiple second-historical-level models can further improve the adjustment effect of the engineering supervision plan.

[0113] S540. Obtain second real-time supervision data from the same period as the first real-time supervision data. Determine multiple second similarities between multiple second historical supervision data and the second real-time supervision data, and determine the maximum second similarity among the multiple second similarities. Generate a second target supervision plan for the second supervision task based on the second historical hierarchy model corresponding to the maximum second similarity.

[0114] In this implementation, second real-time supervision data can be obtained at the same time as the first real-time supervision data. The second real-time supervision data is the actual status and key parameters of the power engineering supervision in the current period with a larger adjustment granularity than the first real-time supervision data. By calculating multiple second similarities between multiple second historical supervision data and the second real-time supervision data, the matching degree between historical data and current real-time data can be evaluated. The second similarity can quantify the degree of similarity between the historical supervision data corresponding to the second historical level model and the current real-time supervision data in terms of time period characteristics.

[0115] For example, in power engineering supervision, the second real-time supervision data can include real-time indicators such as construction efficiency, quality risk points, and resource allocation status during the current seasonal construction phase. By calculating the similarity between these real-time indicators and the indicators of the same time period in each second historical supervision data, the historical data that best matches the characteristics of the current time period can be found.

[0116] In this implementation, the maximum second similarity among multiple second similarities can be determined. The second historical supervision data corresponding to the maximum second similarity is most similar to the current second real-time supervision data in terms of time period characteristics. Based on the second historical level model corresponding to the maximum second similarity, a second target supervision plan can be generated for the second supervision task. The second historical level model can generate a supervision plan suitable for the actual needs of the current time period based on the historical supervision patterns that best match the current time period.

[0117] The beneficial effect of the above implementation method is that power engineering supervision has significant time-related characteristics. For example, during different periods of seasonal construction in the same construction phase, key factors such as construction efficiency, quality risk points, and resource allocation status will show unique patterns. Ignoring the time-related attributes may lead to a disconnect between the selected historical model and the time-related characteristics of the current real-time scenario. By introducing second real-time supervision data and second historical supervision data from the same period for similarity comparison, the selected second historical supervision data can accurately map the engineering supervision patterns of the current period. The corresponding second historical hierarchical model is also more in line with the actual supervision needs of the current period in terms of scheme generation strategy, thereby making the supervision scheme more adaptable to the scenario and more practical.

[0118] In some implementations, the above method also includes S610 to S620, which are described in detail below.

[0119] S610. Obtain the reason for the first fluctuation amplitude exceeding the preset first fluctuation amplitude in the first real-time supervision data corresponding to the first-level model. Among them, the reasons for the fluctuation exceeding the limit include construction violations, equipment quality defects, sudden changes in environmental conditions, and design changes.

[0120] In this implementation, the reasons for the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model exceeding the preset first fluctuation amplitude can be obtained. The reasons for the fluctuation exceeding the limit include construction violations, equipment quality defects, sudden changes in environmental conditions, and design changes. These reasons for the fluctuation exceeding the limit can help identify the specific root causes of abnormal fluctuations in the supervision data.

[0121] For example, in the process of power engineering supervision, when the first fluctuation range of the first real-time supervision data exceeds the preset first fluctuation range, the specific cause of the fluctuation exceeding the limit can be determined by analyzing the on-site monitoring records, equipment test reports, environmental monitoring data and design change notices. This analysis helps to accurately locate the specific link in the problem.

[0122] S620. When the cause of the fluctuation exceeding the limit includes at least one of construction violation or equipment quality defects, an optimized supervision plan is generated simultaneously for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model. When the cause of the fluctuation exceeding the limit includes at least one of sudden changes in environmental conditions or design changes, an optimized supervision plan is generated for the first supervision task based on the first historical level model corresponding to the maximum first similarity, and an optimized supervision plan is generated for the second supervision task based on the second historical level model corresponding to the maximum second similarity.

[0123] In this implementation, when the cause of the fluctuation exceeding the limit includes at least one of the following: construction violation operation or equipment quality defect, the second correction module corresponding to the second level model can simultaneously generate an optimized supervision plan for the second supervision task and the first supervision task. The second correction module can comprehensively consider the correlation between supervision tasks with different adjustment time granularities to ensure that the generated optimized supervision plan has coordination and consistency.

[0124] For example, in substation construction supervision, if construction personnel are found to be operating in violation of regulations or if the transformer has quality defects, the second correction module can generate a unified optimization plan for both equipment installation supervision and construction process supervision, including measures such as strengthening on-site supervision and replacing unqualified equipment.

[0125] In this implementation, when the cause of the fluctuation exceeding the limit includes at least one of the sudden changes in environmental conditions or changes in the design scheme, an optimized supervision scheme can be generated for the first supervision task based on the first historical level model corresponding to the maximum first similarity, and an optimized supervision scheme can be generated for the second supervision task based on the second historical level model corresponding to the maximum second similarity, thus providing targeted solutions for different types of supervision tasks.

[0126] For example, when encountering sudden severe weather or design changes in transmission line supervision, a first-level historical model matching similar historical cases can be used to generate response plans for on-site construction supervision, while a second-level historical model can be used to develop corresponding adjustment plans for material acceptance supervision.

[0127] The beneficial effects of the above implementation method are as follows: Construction violations and substandard material quality are core issues at the level of project physical quality. These issues often affect supervised objects at different levels and time granularities. In this case, it is necessary to use a second-level model correction module with a larger time granularity and global coordination capabilities to uniformly generate solutions, so as to avoid problems such as inconsistent supervision standards and uncoordinated rectification measures caused by scattered solutions. Sudden changes in environmental conditions and design scheme changes are fluctuations at the management or design levels. These issues have relatively independent impacts on supervised objects at different levels. Historical hierarchical models based on similarity matching can accurately adapt to the solution generation needs of such scenarios. Therefore, using historical models to generate solutions is more flexible and adaptable, without the need to use global correction resources, reducing unnecessary waste of supervision solution correction resources.

[0128] In some implementations, the above method also includes: when the cause of the fluctuation exceeding the limit includes at least one of sudden changes in environmental conditions and changes in design schemes, as well as at least one of illegal construction operations and equipment quality defects, an optimized supervision scheme is generated simultaneously for the second supervision task and the first supervision task through the second correction module corresponding to the second-level model.

[0129] In this implementation, during the power engineering supervision process, when the cause of fluctuation exceeding the limit includes at least one of the following: sudden change in environmental conditions and change in design scheme, as well as at least one of the following: illegal construction operation and equipment quality defects, an optimized supervision scheme can be generated synchronously through the second correction module corresponding to the second-level model. This scenario of multiple causes superimposed reflects the complex deviation situation that may be encountered in power engineering supervision. These complex causes can be uniformly processed through the second correction module.

[0130] For example, the calculation process of the second correction module takes into account the mutual influence between different causes, and can perform synchronous optimization processing on the second supervision task and the first supervision task.

[0131] In this implementation, when generating an optimized supervision plan through the second correction module, it can ensure that core issues such as construction violations and equipment quality defects are addressed first, while also taking into account the impact of sudden changes in environmental conditions and design changes.

[0132] For example, in the supervision of substation construction projects, when seasonal weather changes cause sudden changes in the construction environment, and at the same time, violations of regulations by construction personnel and defects in the quality of equipment and materials are discovered, the second correction module can simultaneously generate an optimized supervision plan for these complex reasons. This plan can simultaneously adjust the supervision task arrangement for construction safety supervision and equipment quality inspection to ensure that core issues are addressed in a timely manner.

[0133] The beneficial effect of the above implementation method is that, for complex deviation scenarios with multiple superimposed causes, the global unified second correction module generates the supervision plan, which ensures that the core issues of construction violations and substandard material quality are addressed in a priority and comprehensive manner. At the same time, it avoids the confusion in the supervision plan caused by the interaction of construction violations, substandard material quality, sudden changes in environmental conditions and design changes, and significantly improves the adaptability and reliability of the power engineering supervision plan in complex deviation scenarios.

[0134] This application also provides a power engineering supervision scheme generation system based on a multi-level knowledge graph, including units for implementing the above-described power engineering supervision scheme generation method based on a multi-level knowledge graph.

[0135] Figure 9 A schematic diagram of the logical structure of a power engineering supervision scheme generation system based on a multi-level knowledge graph, as provided in this application embodiment, is shown below. Figure 9 As shown, the system 1 of this embodiment includes a processing unit 11, a storage unit 12, and a transceiver unit 13. The processing unit 11 is used to process data, the storage unit 12 is used to store data, and the transceiver unit 13 is used to send and receive data. The processing unit 11, the storage unit 12, and the transceiver unit 13 cooperate with each other to implement the above-described method. The beneficial effects of the embodiments of this application have been described in the above-described method and will not be repeated here.

[0136] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0138] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0139] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0140] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0141] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between apparatuses or units may be electrical, mechanical, or other forms.

[0142] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0143] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating power engineering supervision schemes based on multi-level knowledge graphs, characterized in that, The method includes: The system acquires real-time data fluctuation amplitudes corresponding to different levels of models used to generate power engineering supervision schemes. These different levels include project-level models, sub-project-level models, item-project-level models, and process-level models. The time granularity of the supervision adjustment schemes corresponding to these models decreases sequentially. The project-level, sub-project-level, item-project-level, and process-level models are used to generate supervision schemes for power engineering supervision tasks with different time granularities of supervision adjustment. These models are trained using historical supervision knowledge graphs with different time granularities of supervision adjustment. The historical supervision knowledge graphs include input data and output data for the supervision schemes. Input data includes engineering parameters, environmental conditions, and specification requirements in the form of a knowledge graph. Output data includes supervision inspection items, rectification requirements, and acceptance standards in the form of a knowledge graph. When the first fluctuation amplitude of the first real-time supervision data corresponding to the first-level model is greater than the preset first fluctuation amplitude, the second correction module corresponding to the second-level model generates an optimized supervision plan for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model. The second-level model is a hierarchical model with a larger time granularity of supervision adjustment than the first-level model, and the second supervision task is the superior supervision task of the first supervision task.

2. The method as described in claim 1, characterized in that, The second correction module is obtained through training using the following method: Obtain the second historical supervision data corresponding to the second-level model at the second supervision adjustment time granularity, and obtain the first historical supervision data corresponding to the first-level model at the same time period of the second historical supervision model at the first supervision adjustment time granularity. In the first historical supervision data, determine the first target historical supervision data whose first fluctuation range is greater than the preset first fluctuation range; obtain the second target historical supervision data in the second historical supervision data for the same period as the first target historical supervision data; obtain the first supervision scheme correction data of the first level model and the second supervision scheme correction data of the second level model corresponding to the first target historical supervision data and the second target historical supervision data; The second correction module is trained based on the historical supervision data of the first target, the historical supervision data of the second target, the correction data of the first supervision plan, and the correction data of the second supervision plan.

3. The method as described in claim 2, characterized in that, The method further includes: Obtain the first fluctuation range corresponding to the first fluctuation amplitude of the real-time data corresponding to the first-level model; obtain multiple second correction modules corresponding to different first fluctuation ranges of the real-time data. Among multiple second correction modules corresponding to different first fluctuation ranges of real-time data, a target second correction module corresponding to the first fluctuation range is determined. An optimized supervision plan is generated for the second supervision task corresponding to the second-level model and the first supervision task corresponding to the first-level model through the target second correction module. The target second correction module is trained using first target historical supervision data, second target historical supervision data, first supervision plan correction data of the first-level model corresponding to the first target historical supervision data and the second target historical supervision data, and second supervision plan correction data of the second-level model. The first target historical supervision data includes first historical supervision data corresponding to the real-time data fluctuation range within the first fluctuation range of the first historical supervision model, and the second target historical supervision data is the second historical supervision data corresponding to the first target historical supervision data at the same time period.

4. The method as described in claim 3, characterized in that, The method further includes: When the first fluctuation amplitude of the first real-time monitoring data corresponding to the first-level model is greater than the preset first fluctuation amplitude, the first switching transition time period corresponding to the first-level model and the second-level model is obtained, and the number of times the first fluctuation amplitude is greater than the preset first fluctuation amplitude within the first switching transition time period is determined; the preset number of times the first fluctuation exceeds the limit corresponding to the first switching transition time period is obtained. When the number of times the first fluctuation exceeds the limit is greater than or equal to the preset number of times the first fluctuation exceeds the limit, the second correction module corresponding to the second level model generates an optimized supervision plan for the second supervision task and the first supervision task; when the number of times the first fluctuation exceeds the limit is less than the preset number of times the first fluctuation exceeds the limit, the basic supervision plan continues to be generated for the first supervision task through the first level model.

5. The method as described in claim 4, characterized in that, The switching transition time periods corresponding to different model levels are determined using the following method: Obtain the second supervision adjustment time granularity corresponding to the second-level model; obtain the first fluctuation sensitivity coefficient of the first supervision task corresponding to the first-level model and the second fluctuation sensitivity coefficient of the second supervision task corresponding to the second-level model. The first fluctuation sensitivity coefficient represents the speed at which the first supervision task responds to data fluctuations, and the second fluctuation sensitivity coefficient represents the speed at which the second supervision task responds to data fluctuations. The product of the second supervisor's adjustment time granularity, the first fluctuation sensitivity coefficient, and the second fluctuation sensitivity coefficient is determined as the switching transition period.

6. The method according to claim 5, characterized in that, The method further includes: When the first fluctuation amplitude of the first real-time supervision data corresponding to the first level model is greater than the preset first fluctuation amplitude, multiple first target historical supervision data with the same supervision adjustment time granularity as the first real-time supervision data and a first fluctuation amplitude greater than the preset first fluctuation amplitude are obtained, and multiple first target historical level models are obtained for generating supervision schemes based on the first target historical supervision data respectively. Multiple first similarities are determined for each of the first target historical supervision data and the first real-time supervision data. The maximum first similarity among the multiple first similarities is determined. Based on the first target historical hierarchy model corresponding to the maximum first similarity, a first target supervision plan is generated for the first supervision task.

7. The method according to claim 6, characterized in that, The method further includes: Acquire multiple second-historical supervision data and multiple second-historical hierarchical models. The multiple second-historical hierarchical models generate second-supervision plans for the second-supervision tasks based on the multiple second-historical supervision data. Acquire second real-time supervision data from the same period as the first real-time supervision data, determine multiple second similarities between multiple second historical supervision data and the second real-time supervision data, and determine the maximum second similarity among the multiple second similarities; generate a second target supervision plan for the second supervision task based on the second historical hierarchical model corresponding to the maximum second similarity.

8. The method according to claim 7, characterized in that, The method further includes: Obtain the reasons for the first fluctuation amplitude exceeding the preset first fluctuation amplitude in the first real-time supervision data corresponding to the first level model; among which, the reasons for the fluctuation exceeding the limit include construction violations, equipment quality defects, sudden changes in environmental conditions, and design changes; When the cause of the fluctuation exceeding the limit includes at least one of the following: construction violation operation or equipment quality defect, the second correction module corresponding to the second level model generates an optimized supervision plan for the second supervision task and the first supervision task simultaneously; when the cause of the fluctuation exceeding the limit includes at least one of the following: sudden change in environmental conditions or design change, the optimized supervision plan is generated for the first supervision task based on the first historical level model corresponding to the maximum first similarity, and the optimized supervision plan is generated for the second supervision task based on the second historical level model corresponding to the maximum second similarity.

9. The method according to claim 8, characterized in that, The method further includes: When the cause of the fluctuation exceeding the limit includes at least one of sudden changes in environmental conditions and changes in design schemes, as well as at least one of illegal construction operations and equipment quality defects, the second correction module corresponding to the second-level model generates an optimized supervision scheme for the second supervision task and the first supervision task simultaneously.

10. A power engineering supervision scheme generation system based on a multi-level model, characterized in that, Includes units for implementing the method of any one of claims 1 to 9.

Images