Power grid project construction progress risk dynamic early warning method based on machine learning

Abstract

The invention discloses a power grid project construction progress risk dynamic early warning method based on machine learning. The method comprises the following steps: S1, estimating a reasonable construction period by using machine learning; s2, constructing a process network and setting an early warning threshold value; according to the method, on the basis of historical data and a machine learning algorithm, the scientificity of construction period prediction is improved, the completion time is re-estimated according to the actual progress every day, dynamic early warning is achieved, the actual background of a power grid project is fully combined in the abnormal value processing and threshold value setting links, and the project progress risk dynamic early warning is achieved. According to the system, general and serious two-level early warning is set, hierarchical response and management are facilitated, the system is in butt joint with an existing system of a state grid to improve the informatization level of project management, intelligent and refined management and control of the construction progress of the power grid project are achieved through combination of machine learning and dynamic monitoring, and the system has high practicability and popularization value.

Description

Technical Field

[0001] This invention relates to the field of engineering management technology, specifically to a dynamic early warning method for construction progress risks in power grid engineering based on machine learning. Background Technology

[0002] Power grid construction is a crucial link in power system development, encompassing core components such as transmission lines, distribution lines, substation construction, and facility upgrades. Construction plans must be comprehensively developed based on project scale, geographical conditions, and technical standards to ensure the stability, reliability, and security of the power grid system. Specific construction content includes transmission line construction, distribution line construction, substation construction, and facility upgrades. Power grid projects are characterized by large scale, high investment, and involvement of numerous departments. Their construction progress is easily affected by multiple risks, including external environment, technical factors, and management coordination. Dynamic early warning of construction progress risks in power grid projects refers to identifying various risk factors affecting construction progress during the project implementation process, conducting real-time monitoring and analysis of project progress status, promptly identifying potential delay risks, and issuing early warning signals, thereby supporting the project team's management process. Currently, in the process of dynamic early warning of schedule risks, traditional early warning methods mainly rely on experience to estimate the project duration. However, this approach leads to a significant discrepancy between the estimated duration and the actual situation. Furthermore, in the risk early warning process, the estimated duration cannot be correlated with actual construction progress data, resulting in low accuracy in the actual early warning process and further exacerbating project delays. Summary of the Invention

[0003] This invention provides a machine learning-based dynamic early warning method for construction progress risks in power grid projects. It can effectively solve the problem mentioned in the background that the traditional early warning method mainly relies on experience to estimate the construction period in the process of dynamic early warning of progress risks. This method leads to a large deviation between the estimated construction period and the actual situation. Furthermore, in the process of risk early warning based on the estimated construction period, it is impossible to correlate it with the actual construction progress data, resulting in low early warning accuracy and further aggravating the problem of project delays.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a machine learning-based dynamic early warning method for construction progress risks in power grid projects, comprising: Machine learning is used to estimate the reasonable construction period of a project to be built. Specifically, this includes collecting data on historical completed projects, handling missing values ​​and outliers, and using planned substation capacity, planned line length, total dynamic investment in the feasibility study, and the month of commencement as input features. A quantile regression model is used to predict the reasonable construction period and provide a construction period range. Based on the reasonable construction period, the planned start and end times of each process are collected using the construction schedule, the preceding processes of each process are identified, a process priority relationship network is constructed, and the planned start and completion times of the project are determined accordingly. A progress warning threshold is set based on the planned start time and planned completion time and the reasonable construction period. A warning threshold is set based on the difference between the planned start time and planned completion time and the reasonable construction period. Dynamic warning of progress risk is implemented in the construction process based on the relationship between the actual progress of the project and the warning threshold.

[0005] Preferably, the method for setting the early warning threshold includes: In identifying each process i For the immediate preceding process, the following formula is used for calculation: P i ={ p | p ∈(1,…, N ), S p ＜S i} (1); in, P i For process i The set of immediate preceding processes; Project time calculation specifically involves calculating the planned start date of the project to be constructed. B and planned completion time T Specifically, it is calculated using the following formula: (2); (3); Step S304: When setting the early warning threshold, the specific time should be based on the planned start time. B Planned completion time T and reasonable construction period R To determine the general early warning threshold for the schedule risk of projects under construction. J 1 and severe warning threshold J 2 Among them, the general warning threshold J 1 The difference between the planned start and completion times and the reasonable construction period constitutes a severe warning threshold. J 2 : Twice the general warning threshold; The specific formula is as follows: J 1 =min{ T,B+R} (4); J 2 =max{ T,B+R} (5).

[0006] Preferably, the data collection for estimating the reasonable construction period of the project to be constructed using machine learning is to collect relevant data from the construction schedule, including the number of processes N, the planned start time Si and the planned end time Fi of the i=1,...,N process; Identify and determine the preceding processes of each process based on the planned start time, forming a process priority relationship network.

[0007] Preferably, using machine learning to estimate the reasonable construction period of a project includes: During the data collection and screening phase, completed projects with the same voltage level and engineering type as the project to be constructed, within the preset time range and located in the same province are selected as training samples. The planned substation capacity, planned line length, dynamic total investment of the feasibility study, start month and actual construction period of the training samples are obtained. In the data preprocessing stage, missing values ​​in the training samples are counted. When the proportion of missing values ​​for a certain feature is less than 5%, samples containing that missing value are deleted. When the proportion of missing values ​​is greater than or equal to 5%, the K-nearest neighbor algorithm is used for filling. The isolated forest algorithm is combined with the business logic of power grid engineering to identify and process outliers, and the preprocessed training sample set is obtained. In the modeling and prediction phase, the preprocessed training sample set is divided into a training set and a validation set, with a sample size ratio of 7:3. The LightGBM quantile regression model is used for training, and the prediction results of the model at the preset high quantiles are used as the reasonable construction period of the project to be built.

[0008] Preferably, the modeling and prediction phase enables real-time progress monitoring and risk warning during construction, providing real-time alerts for progress risks during the construction process of the project. As required, the construction party needs to update the completion status of each process daily, as detailed below: r i ∈[0,1] ( i =1,…, N ),in, r i = 0 indicates the first i The process has not yet started. r i = 1 indicates the first i Each process has been completed. r i ∈ (0,1) indicates that the first ∈ (0,1) is the first ∈ (0,1). iThe process has started but not yet completed, and its completion rate is [percentage missing]. r i ; Specifically, the steps include the following: Data Update: Based on the daily progress reports from the construction unit, update the set of completed, started but not completed, and yet-to-be-started processes to form the current progress status of the project; Time estimation and schedule projection: Based on the updated process status, estimate the start and end times of each process, calculate the project's estimated completion time, and thereby project the schedule deviation relative to the planned start and completion times. Early warning judgment: The estimated completion time is compared with the preset general early warning threshold and severe early warning threshold. The early warning level is determined according to the range to which the deviation belongs, and the corresponding early warning information is output.

[0009] Preferably, the data update is as follows: if All r i = If the completion status is 1, it indicates that the project is completed and the alert should be stopped; otherwise, update the completion status of each process. r i ; At the same time, update the set of completed processes Ω1, the set of processes that have started but are not yet completed Ω2, and the set of processes that have not yet started Ω3 according to the following rules, as follows: (6); The start time of the update process in time estimation and schedule projection ( ) and end time ( ); In this case, all processes in set Ω1 have been completed, and their start and end times should be filled in according to the actual dates; For the process in set Ω2, fill in the actual start time and the end time according to the completion rate. r i estimate; The start and end times of the processes in set Ω3 are estimated based on the planned start and end times and the end times of the preceding processes, using the following formulas: (7); in, t Indicates the current time, i.e., the [number]th [time]. t sky.

[0010] Preferably, the updated project completion date for early warning judgment is determined by comparing the estimated completion time with the planned completion time, calculating the delay days, and issuing different levels of early warnings based on the relationship between the delay days and the early warning threshold. The specific formula is as follows: ; Warnings include the following situations: like T ＊ Less than or equal to the general warning threshold J 1 No warning; like T ＊ Given the general warning threshold J 1 With severe warning threshold J 2 In between, general early warning information is issued to the construction party; like T ＊ Greater than or equal to the severe warning threshold J 2 Issue a serious warning to the construction party and return to step S301.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention improves the scientific accuracy of project schedule prediction by relying on historical data and machine learning algorithms. It re-estimates the completion time daily based on the actual progress, enabling dynamic early warning. Furthermore, it fully integrates the actual background of power grid projects in the outlier handling and threshold setting stages, setting up two levels of early warning: general and severe, to facilitate hierarchical response and management. It can be integrated with the existing State Grid system to enhance the informatization level of project management. By combining machine learning with dynamic monitoring, it achieves intelligent and refined control over the construction progress of power grid projects, demonstrating strong practicality and promotional value. Furthermore, by limiting the voltage level, type, region, and time range of the samples, the accuracy of project schedule prediction is improved. By updating the network structure in real time through the process precedence matrix, it can adapt to complex construction logic and achieve closed-loop management of the entire process from historical data analysis to real-time risk monitoring, which can significantly improve the ability to predict the progress and risks of power grid projects. Attached Figure Description

[0012] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0013] In the attached diagram: Figure 1 This is a flowchart of the steps of the dynamic early warning method of the present invention. Detailed Implementation

[0014] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0015] Example: Figure 1 As shown, this invention provides a technical solution: a dynamic early warning method for the construction progress risk of power grid projects based on machine learning. Combining data mining, machine learning, and engineering management theories, it achieves intelligent early warning, aiming to monitor and warn of risks in real time during the construction progress of power grid projects. The method includes the following steps: Step S1: Use machine learning to estimate a reasonable construction period; Step S2: Construction of process network and setting of early warning threshold; Step S3: Dynamic early warning of progress risks for projects under construction.

[0016] Based on the above technical solution, step S1 is to predict the reasonable construction period of the project to be built based on historical data and give a confidence interval to provide a benchmark for subsequent early warning. Specifically, for projects of the same voltage level and type, recently completed projects of the same voltage level and type are used as training samples to ensure similarity between the samples and the projects to be built. Specifically, the voltage levels are 35kV, 110kV, and 220kV, and the project types are transmission and transformation projects, bay projects, and outgoing transmission projects. The confidence interval for the reasonable construction period of each project is predicted. The aim is to construct a high-quality training sample set to provide reliable input for the construction period prediction model. The specific steps include: Step S101, Data collection and filtering; Step S102, data preprocessing; Step S103, Modeling and Prediction.

[0017] Based on the above technical solution, in step S101, the input data includes the planned substation capacity, planned line length, total dynamic investment of the feasibility study and the month of commencement, and the output data is the actual construction period of the project. In order to improve the similarity between the training sample and the project to be built, the time range and geographical range of the sample should be appropriately limited, and completed projects in the same area in the near future should be selected first. Specifically, in terms of time frame, priority should be given to projects completed within the last 3 years to reduce the impact of changes in technical standards due to differences in time periods; in terms of geographical scope, projects within the same province should be limited to reduce the impact of differences in construction conditions due to geographical environment.

[0018] Based on the above technical solution, step S102, data preprocessing is to process the collected training sample data, specifically including missing value processing and outlier processing. Missing value handling involves systematically processing missing values ​​in the statistical training sample set. Specifically, this process involves counting the number and percentage of samples with missing values ​​for each feature, and then processing them accordingly based on the percentage of missing values. This includes the following: When the percentage of missing values ​​for a certain feature is less than 5%, samples containing that missing value can be directly deleted. When the percentage of missing values ​​is greater than or equal to 5%, the KNN imputation method is used for imputation. Before filling in the missing values, we first analyze the nature of the missing values ​​and distinguish between random missing values ​​and structural missing values. For structural missing values, we need to combine business logic to adopt a more reasonable approach to handle them, such as interpolation or deleting the entire column. Outlier handling involves identifying outliers in the training samples and processing them further. In this process, the isolated forest algorithm is used to initially identify and detect outliers and filter out outliers. After initially identifying outliers, further judgment is made in conjunction with the business logic of the power grid project. The causes of outliers are analyzed from multiple dimensions, including project scale, construction conditions, and construction environment. Based on the actual situation, it is decided whether to retain, correct, or delete the outliers. Outliers caused by special and reasonable reasons can be marked and retained to enrich the diversity of the sample.

[0019] Based on the above technical solution, in step S103, during the modeling and prediction process, the training samples after missing value processing and outlier processing are used as the training set. The LightGBM algorithm is used to perform quantile regression analysis, and the prediction result of the 95th percentile is used as the reasonable construction period of the project to be built. R ; In the specific modeling process, it is necessary to reasonably divide the training set and the validation set. In the specific data partitioning process, the training set accounts for 70% and the validation set accounts for 30% to ensure the generalization ability and prediction accuracy of the model. The output result is the reasonable construction period range of the project to be built, which serves as the benchmark for subsequent early warning.

[0020] Based on the above technical solution, step S2 refers to constructing a process priority relationship network for the project to be constructed and setting a progress warning threshold. The aim is to establish the process logical relationship of the project and set a warning threshold based on the planned start time, planned completion time and reasonable construction period to quantify the progress risk warning boundary. Specifically, addressing the lack of a project process priority relationship network in the online State Grid system, a feasible process priority relationship network is constructed using the construction schedule table. This lays the foundation for dynamic early warning of progress risks in step S3. Based on the planned start and completion times and reasonable construction periods, progress warning thresholds for the projects under construction are set. The specific steps are as follows: Step S201, Data Acquisition; Step S202: Construct the process network; Step S203, Project time calculation; Step S204: Set the warning threshold.

[0021] Based on the above technical solution, in step S301, relevant data, including the number of work processes, is collected from the construction schedule. N , No. i =1,…, N Planned start time for each process S i and the planned end time F i ; Step S302: Identify and determine the preceding processes of each process based on the planned start time to form a process priority relationship network; In the specific identification process, each step is identified. i For the immediate preceding process, the following formula is used for calculation: P i ={ p | p ∈(1,…, N ), S p ＜S i} (1); in, P i For process i The set of immediate preceding processes.

[0022] Based on the above technical solution, step S303, project time calculation specifically involves calculating the planned start time of the project to be constructed. B and planned completion time T Specifically, it is calculated using the following formula: (2); (3); Step S304: When setting the early warning threshold, the specific time should be based on the planned start time. B Planned completion time T and reasonable construction period R To determine the general early warning threshold for the schedule risk of projects under construction. J 1 and severe warning threshold J 2 Among them, the general warning threshold J 1 The difference between the planned start and completion times and the reasonable construction period constitutes a severe warning threshold. J 2 : Twice the general warning threshold; The specific formula is as follows: J 1 =min{ T,B+R} (4); J 2 =max{ T,B+R} (5).

[0023] Based on the above technical solution, step S3 mainly involves real-time progress monitoring and risk warning during the construction process. This includes providing real-time warnings of progress risks during the construction of the project. As required, the construction party needs to update the completion status of each process daily, as detailed below: r i ∈[0,1] ( i =1,…, N ),in, r i = 0 indicates the first i The process has not yet started. r i = 1 indicates the first i Each process has been completed. r i ∈ (0,1) indicates that the first ∈ (0,1) is the first ∈ (0,1). i The process has started but not yet completed, and its completion rate is [percentage missing]. r i ; Specifically, the steps include the following: Step S301, data update; Step S302, Time estimation and schedule forecasting; Step S303, early warning judgment.

[0024] Based on the above technical solution, step S301, the data update is as follows: if All r i = If the completion status is 1, it indicates that the project is completed and the alert should be stopped; otherwise, update the completion status of each process. r i ; At the same time, update the set of completed processes Ω1, the set of processes that have started but are not yet completed Ω2, and the set of processes that have not yet started Ω3 according to the following rules, as follows: (6); Step S302, update the start time of the process ( ) and end time ( ); In this case, all processes in set Ω1 have been completed, and their start and end times should be filled in according to the actual dates; For the process in set Ω2, fill in the actual start time and the end time according to the completion rate. r i estimate; The start and end times of the processes in set Ω3 are estimated based on the planned start and end times and the end times of the preceding processes, using the following formulas: (7); in, t Indicates the current time, i.e., the [number]th [time]. t sky; Step S303: Update the project completion date by comparing the estimated completion time with the planned completion time, calculating the number of days of delay, and issuing different levels of alerts based on the relationship between the number of days of delay and the alert threshold. The specific formula is as follows: ; Warnings include the following situations: like T ＊ Less than or equal to the general warning threshold J 1 No warning; like T ＊ Given the general warning threshold J 1 With severe warning threshold J 2 In between, general early warning information is issued to the construction party; like T ＊ Greater than or equal to the severe warning threshold J 2 Issue a serious warning to the construction party and return to step S301.

[0025] Furthermore, this method is applicable to power transmission and transformation projects, distribution line projects, and other power grid construction projects, aiming to improve the accuracy of project schedule prediction and risk identification capabilities through a data-driven approach.

[0026] First, before the project starts, the system obtains basic data of completed projects from the engineering management database. The collected data includes project type, construction location, substation capacity, line length, investment scale, start and completion time, etc. To ensure the timeliness and comparability of the data, the sample is selected from projects within the past three years and limited to the same voltage level and the same province to reduce the impact of regional and technical standard differences. The collected data is cleaned, filtered and deduplicated to form the basic dataset for model training.

[0027] Subsequently, the system preprocesses the data. For fields with few missing values, the system automatically deletes samples containing missing values. For fields with a high proportion of missing values, the system uses an imputation algorithm based on similar samples to complete the data. For outlier data, the system identifies outliers by combining statistical analysis with empirical judgment. If necessary, the system combines project background information to determine whether to retain the data. The preprocessed dataset is more complete and accurate, providing a reliable foundation for subsequent modeling.

[0028] Next, the system establishes a project duration prediction model based on historical data. The model is trained using machine learning algorithms and establishes a non-linear mapping relationship between project duration and project attributes through multi-dimensional input features. After model training, the system can automatically output a reasonable project duration range based on the input project parameters. This prediction result serves as a benchmark value for subsequent risk warnings and provides a reference for construction planning and progress control.

[0029] After obtaining a reasonable construction period, the system constructs a process relationship network based on the construction schedule. Each process node records its planned start and end times, and establishes preceding process relationships according to the construction sequence. Through this network structure, the system can identify the critical path, calculate the planned start and end times of the project, and provide logical support for risk monitoring.

[0030] During the progress warning phase, the system automatically receives daily feedback from the site regarding the completion of each work process. The construction unit must update the actual completion rate of each process at a prescribed frequency, and the system uses this information to assess the overall project progress. If the actual progress lags behind the reasonable progress predicted by the model, the system will automatically assess the possible number of delay days and generate a corresponding level of warning based on a set threshold. General warnings are used to remind project managers to pay attention to potential delays, while severe warnings will be synchronized to the management platform and require corrective measures.

[0031] The system interface displays the progress and risk level of each process in a visual format. Managers can view the current status of the project, key process risk points, and historical warning records on the platform. Through data accumulation and model optimization, the system can continuously improve the accuracy of schedule prediction and risk identification capabilities, realizing the transformation from static monitoring to dynamic intelligent early warning.

[0032] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0033] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0034] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0035] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0036] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0037] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A machine learning-based dynamic early warning method for construction progress risks in power grid engineering, characterized in that, include: Data from historically completed projects is collected, missing values ​​are handled, and outliers are processed. Using planned substation capacity, planned line length, dynamic total investment in feasibility studies, and the month of commencement as input features, a quantile regression model is used to predict the reasonable construction period and provide a construction period range. Based on the reasonable construction period, the planned start and end times of each process are collected using the construction schedule, the preceding processes of each process are identified, a process priority relationship network is constructed, and the planned start and completion times of the project are determined accordingly. A progress warning threshold is set based on the planned start time and planned completion time and the reasonable construction period. A warning threshold is set based on the difference between the planned start time and planned completion time and the reasonable construction period. Dynamic warning of progress risk is implemented in the construction process based on the relationship between the actual progress of the project and the warning threshold.

2. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 1, characterized in that: Methods for setting early warning thresholds include: In identifying each process i For the immediate preceding process, the following formula is used for calculation: P i =｛ p ∣ p ∈(1,…, N ), S p ＜S i ｝ （1）； in, P i For process i The set of immediate preceding processes; Project time calculation specifically involves calculating the planned start date of the project to be constructed. B and planned completion time T Specifically, it is calculated using the following formula: （2）； （3）； When setting early warning thresholds, the specific timing should be based on the planned start time. B Planned completion time T and reasonable construction period R To determine the general early warning threshold for the schedule risk of projects under construction. J 1 and severe warning threshold J 2 Among them, the general warning threshold J 1 The difference between the planned start and completion times and the reasonable construction period constitutes a severe warning threshold. J 2 : Twice the general warning threshold; The specific formula is as follows: J 1 =min{ T,B+R } (4); J 2 =max{ T,B+R } (5)。 3. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 2, characterized in that: The data collection for estimating the reasonable construction period of the project using machine learning involves collecting relevant data from the construction schedule, including the number of processes N, the planned start time Si and the planned end time Fi of the i=1,...,N process. Identify and determine the preceding processes of each process based on the planned start time, forming a process priority relationship network.

4. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 1, characterized in that: Data from historically completed projects is collected, missing values ​​are handled, and outliers are processed. Using planned substation capacity, planned line length, total dynamic investment from feasibility studies, and the month of commencement as input features, a quantile regression model is used to predict reasonable construction periods, including: Completed projects with the same voltage level and engineering type as the project to be constructed, within a preset time range and located in the same province are selected as training samples. The planned substation capacity, planned line length, dynamic total investment of the feasibility study, start month and actual construction period of the training samples are obtained. The missing values ​​in the training samples are statistically analyzed. When the proportion of missing values ​​for a certain feature is less than 5%, the samples containing the missing value are deleted. When the proportion of missing values ​​is greater than or equal to 5%, the filling method based on the K-nearest neighbor algorithm is used to fill the missing values. The isolated forest algorithm is combined with the business logic of the power grid project to identify and process outliers, and the preprocessed training sample set is obtained. The preprocessed training sample set is divided into a training set and a validation set, with a sample size ratio of 7:

3. The LightGBM quantile regression model is used for training, and the prediction results of the model at the preset high quantiles are used as the reasonable construction period of the project to be built.

5. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 4, characterized in that: The modeling and prediction phase enables real-time progress monitoring and risk warning during construction, providing real-time alerts for progress risks during the construction process. As required, the construction team needs to update the completion status of each work process daily, as detailed below: r i ∈[0,1] ( i =1,…, N ),in, r i = 0 indicates the first i The process has not yet started. r i = 1 indicates the first i Each process has been completed. r i ∈ (0,1) indicates that the first ∈ (0,1) is the first ∈ (0,1). i The process has started but not yet completed, and its completion rate is [percentage missing]. r i ; Specifically, the steps include the following: Data Update: Based on the daily progress reports from the construction unit, update the set of completed, started but not completed, and yet-to-be-started processes to form the current progress status of the project; Time estimation and schedule projection: Based on the updated process status, estimate the start and end times of each process, calculate the project's estimated completion time, and thereby project the schedule deviation relative to the planned start and completion times. Early warning judgment: The estimated completion time is compared with the preset general early warning threshold and severe early warning threshold. The early warning level is determined according to the range to which the deviation belongs, and the corresponding early warning information is output.

6. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 5, characterized in that: The data update is as follows: if All r i = If the completion status is 1, it indicates that the project is completed and the alert should be stopped; otherwise, update the completion status of each process. r i ; At the same time, update the set of completed processes Ω1, the set of processes that have started but are not yet completed Ω2, and the set of processes that have not yet started Ω3 according to the following rules, as follows: （6）； The start time of the update process for the time estimation and progress calculation ( ) and end time ( ); In this case, all processes in set Ω1 have been completed, and their start and end times should be filled in according to the actual dates; For the process in set Ω2, fill in the actual start time and the end time according to the completion rate. r i estimate; The start and end times of the processes in set Ω3 are estimated based on the planned start and end times and the end times of the preceding processes, using the following formulas: （7）； in, t Indicates the current time, i.e., the [number]th [time]. t sky.

7. The method for dynamic early warning of construction progress risks in power grid engineering based on machine learning according to claim 6, characterized in that: The updated project completion date for the early warning judgment is determined by comparing the estimated completion time with the planned completion time, calculating the delay days, and issuing different levels of early warnings based on the relationship between the delay days and the early warning threshold. The specific formula is as follows: ； Warnings include the following situations: like T ＊ Less than or equal to the general warning threshold J 1 No warning; like T ＊ Given the general warning threshold J 1 With severe warning threshold J 2 In between, general early warning information is issued to the construction party; like T ＊ Greater than or equal to the severe warning threshold J 2 Issue a serious warning to the construction party and return to step S301.

Images