A method for determining a mode partition of a power transmission line detection operation cost

Abstract

The present application relates to the technical field of power transmission line operation and maintenance management, and particularly relates to a power transmission line detection operation cost mode partition determination method, comprising the following steps: S1, collecting and processing historical project data, constructing a cost sequence, dividing a line section, extracting section characteristics and constructing a relationship graph; S2, dividing a sample group based on the cost sequence and the section characteristics, calculating a disturbance index, decomposing the cost into three types of modes and optimizing verification; S3, matching a to-be-detected project, determining three types of operation costs and identifying a cost dominant type; S4, performing mechanism partitioning in combination with the section characteristics and physical connection relationships, correcting the cost based on historical residuals and verifying the output result. The present application realizes fine determination of detection operation costs under the influence of different line conditions, operation characteristics and external disturbance factors, and improves the accuracy, adaptability and reliability of the cost determination result.

Description

Technical Field

[0001] This invention relates to the field of power transmission line operation and maintenance management technology, and in particular to a method for determining the modal partitioning of power transmission line inspection operation costs. Background Technology

[0002] With the continuous expansion of transmission line scale and the increasing complexity of line operating environments, transmission line inspection operations are gradually exhibiting characteristics of diversified inspection objects, differentiated operation areas, and complex cost influencing factors. The cost of transmission line inspection operations is affected not only by the line's inherent conditions such as line length, number of towers, number of inspection points, and inspection frequency, but also by external factors such as voltage level, terrain complexity, defect risk level, weather windows, relocation conditions, special crossing coordination, and the urgency of the operation. Therefore, it is necessary to rationally determine the cost of transmission line inspection operations by combining historical inspection project data and line section characteristics, providing a basis for inspection operation budget preparation, resource allocation, and operation and maintenance management.

[0003] Existing methods for determining the cost of transmission line inspection operations typically rely on historical average costs, quota standards, or manual experience for estimation. These methods primarily analyze the overall project cost, making it difficult to distinguish between stable costs, cost variations caused by the line's own conditions, and cost variations caused by external disturbances. Furthermore, different line sections differ in physical location, corridor environment, risk type, and work organization methods. Existing methods lack analysis of the cost formation mechanisms for each section, hindering differentiated cost management by region. In addition, external disturbances are sporadic and triggering; directly incorporating them into routine cost forecasting can easily lead to cost estimation biases, affecting the accuracy and adaptability of the cost determination results. Summary of the Invention

[0004] This invention provides a method for determining the modal partitioning of transmission line inspection operation costs. By structuring historical project cost data and combining line segment characteristics and segment correlations, the method performs multimodal decomposition of inspection operation costs, separately determining stable operation costs, endogenous driving operation costs, and exogenous disturbance operation costs. Furthermore, based on the segment cost formation mechanism, the method performs line mechanism partitioning and cost correction, thereby improving the precision, accuracy, and reliability of transmission line inspection operation cost determination.

[0005] A method for determining the modal zoning of transmission line inspection operation costs includes the following steps:

[0006] S1. Collect historical data of transmission line testing and inspection projects, clean the samples, unify the standards and build a hierarchical database, construct the cost sequence of historical projects, divide the line sections according to the line structure and the boundaries of testing operations, extract the workload characteristics, operation difficulty characteristics and external disturbance characteristics of the sections, form a section feature matrix, and construct a section relationship diagram through the adjacent connections, the same channel environment, the same risk type and the operation transfer relationship between sections.

[0007] S2, based on the historical project cost sequence and segment feature matrix, divides the same sample groups according to detection type, voltage level and operation scenario, calculates the endogenous driving strength, exogenous disturbance strength and exogenous disturbance triggering flag of the project, decomposes the historical project approved operation cost into stable operation cost mode, endogenous driving operation cost mode and exogenous disturbance operation cost mode, and performs constraint optimization solution and decomposition result verification for each type of operation cost mode;

[0008] S3. Match the project to be tested to the corresponding sample group of the same type, use the weighted robust statistical method to determine the stable operating cost, use the segment graph attention prediction network to determine the endogenous driving operating cost, use the disturbance factor triggered correction model to determine the exogenous disturbance operating cost, and determine the project cost dominant type according to the contribution rate of various operating cost modes.

[0009] S4 allocates stable operating costs and exogenous disturbance operating costs to each line segment, and calculates the initial operating cost and modal contribution rate of the segment based on the segment-level endogenous driving operating costs. The line mechanism is partitioned according to the segment cost dominance type, segment characteristics and physical connectivity. The historical modal residuals of various operating cost modes of historical projects are calculated, and each partition is corrected based on the historical modal residuals. The results are summarized to obtain the suggested operating costs and complete the cost consistency verification.

[0010] Optionally, S1 includes:

[0011] S11: Collect operational data from historical transmission line testing projects, including line parameters, testing types, testing methods, on-site operation records, testing reports, settlement data, and verified operational costs. Perform sample access verification on the collected operational data, and perform duplicate data merging and deduplication based on project number, line name, testing time, and verified operational cost. Perform unified conversion on operational data with inconsistent units of measurement. Identify abnormal operational data by combining unit line length cost, unit tower cost, and unit testing point cost. Abnormal operational data formed by crossing high-speed railways, long-distance relocation in mountainous areas, extreme weather windows, or emergency testing operations are retained and marked as disturbance samples. Correct or remove abnormal operational data formed by input errors, duplicate collection, or inconsistent cost standards. Build a layered database according to testing type, voltage level, and operation scenario.

[0012] S12, Based on historical project operation data, construct a historical project cost sequence using the verified operation cost of each historical transmission line testing and inspection project;

[0013] S13. Based on historical project samples and according to the actual structure of the transmission line and the boundaries of the inspection operation, the transmission line is divided into line sections, including tower sections, tension sections, special crossing sections, and passageway hidden danger sections. Workload characteristics, operation difficulty characteristics, and external disturbance characteristics are extracted for each line section. The workload characteristics include section length, number of towers, number of inspection points, and inspection frequency. The operation difficulty characteristics include voltage level, terrain passage complexity, and defect risk level. The external disturbance characteristics include the degree of weather window limitation, relocation distance or relocation difficulty, market price fluctuation level, special crossing coordination difficulty, and operation urgency. A section feature matrix is ​​constructed according to the line sequence.

[0014] S14. Using each line segment as a graph node, construct relationship edges based on the adjacent connection relationship, the same channel environment relationship, the same risk type relationship, and the operation transfer relationship between line segments. Calculate the comprehensive edge weight between segments based on the edge weights corresponding to each type of relationship edge to construct the segment relationship graph.

[0015] Optionally, S2 includes:

[0016] S21. Based on the historical project cost sequence and section feature matrix, the sample groups are divided into similar groups according to the detection type, voltage level and operation scenario. The verified operation cost of each historical project is decomposed into stable operation cost mode, endogenous driving operation cost mode, exogenous disturbance operation cost mode and decomposition residual. Among them, the stable operation cost mode is used to characterize the relatively stable operation cost in similar detection projects, the endogenous driving operation cost mode is used to characterize the operation cost change caused by the project's own factors, and the exogenous disturbance operation cost mode is used to characterize the operation cost change caused by external conditions.

[0017] S22, based on the obtained sample groups of the same type, summarize the project-level features of each historical project, including project-level endogenous features and project-level exogenous features, normalize each project-level feature, and calculate the endogenous driving strength, exogenous disturbance strength, and exogenous disturbance triggering flag of each historical project. The endogenous driving strength is used to characterize the degree of driving effect of the project's own factors on the operating cost, the exogenous disturbance strength is used to characterize the degree of influence of external disturbance factors on the operating cost, and the exogenous disturbance triggering flag is used to determine whether the exogenous disturbance operating cost mode is triggered.

[0018] S23. Based on various activity cost modes, endogenous driving strength, exogenous disturbance strength, and exogenous disturbance triggering identifiers, construct a constraint optimization objective that includes reconstruction error constraints, stability constraints, endogenous consistency constraints, and exogenous triggering constraints, and set non-negative constraints for various activity cost modes and triggering constraints for exogenous disturbance activity cost modes.

[0019] S24. Based on the constructed constraint optimization objective, determine the initial value of the stable operation cost mode, and update the stable operation cost mode, the endogenous driven operation cost mode, and the exogenous disturbance operation cost mode in sequence. Update the proportional coefficients corresponding to the endogenous driving strength and the exogenous disturbance strength until the preset convergence condition is met, and complete the constraint optimization solution of various operation cost modes.

[0020] S25. Based on the various activity cost modes obtained from the solution, calculate the decomposition residuals of each historical item, and perform residual verification according to the proportion of the decomposition residuals to the approved activity costs. Review the historical items whose residuals exceed the preset residual threshold, and retain the decomposition results for the historical items whose residuals do not exceed the preset residual threshold.

[0021] S26. Based on the various activity cost modes, endogenous driving strength, exogenous disturbance strength and exogenous disturbance triggering flags obtained by solving, perform a mode consistency check, verify the correspondence between various activity cost modes and project workload, operation difficulty and external disturbance factors, and determine that various activity cost modes of each historical project meet the cost formation mechanism.

[0022] S27 outputs the stable activity cost mode, endogenous driving activity cost mode, exogenous disturbance activity cost mode, decomposed residual, endogenous driving strength, exogenous disturbance strength, and exogenous disturbance trigger flag for each historical project.

[0023] Optionally, S3 includes:

[0024] S31. Based on the detection type, voltage level, and operating scenario of the test item, match the test item to the corresponding similar sample group, and extract the stable operating cost mode, endogenous driven operating cost mode, and exogenous disturbance operating cost mode obtained from the decomposition of the similar sample group as the basis for cost determination.

[0025] S32, select historical projects from the same sample group where exogenous disturbances have not been triggered, endogenous driving strength is within a preset quantile range and data is complete, determine sample weights based on the time difference, feature distance and data reliability between historical projects and projects to be tested, and determine the stable operating cost of projects to be tested using the weighted truncated median.

[0026] S33. Input the segment relationship graph of the project to be tested, as well as the workload and operation difficulty features of each segment, into the segment graph attention prediction network. Use adjacent connections, same channel environment, same risk type and operation transition relationship to perform graph encoding on the segment features, predict the segment-level endogenous driving operation cost, and sum the prediction results of each segment to obtain the endogenous driving operation cost of the project to be tested.

[0027] S34. Calculate the exogenous disturbance intensity based on the degree of weather window limitation, relocation distance or relocation difficulty, market price fluctuation level, special crossing coordination difficulty and operational urgency of the project under test, determine the exogenous disturbance triggering indicator based on the exogenous disturbance triggering threshold, and determine the exogenous disturbance operation cost of the project under test using the disturbance factor triggering correction model.

[0028] S35. Calculate the contribution rate of various activity cost modes based on the stable activity cost, endogenous driven activity cost and exogenous disturbance activity cost, and determine the project to be tested as stable baseline dominated, activity scale and risk condition dominated or exogenous disturbance sensitive based on the maximum contribution rate.

[0029] Optionally, S33 includes:

[0030] S331, Obtain the segment relationship diagram of the project to be tested. The segment relationship diagram includes a set of segment nodes, a set of segment relationship edges, and a segment feature matrix. Extract the segment length, number of towers, number of detection points, detection frequency, voltage level, terrain channel complexity, and defect risk level of each segment from the segment feature matrix, and construct the segment endogenous driving feature vector.

[0031] S332, the segment relationship edges are divided into adjacent connection edges, same channel environment edges, same risk type edges, and operation transfer edges. The attention scores between adjacent segments under each type of segment relationship edge are calculated, and the attention scores are normalized to obtain attention weights. The features of adjacent segments are weighted and aggregated according to the attention weights to obtain the endogenous driving graph encoding results of each segment.

[0032] S333: Input the endogenous driving graph encoding results of each segment into the prediction layer of the multilayer perceptron, and output non-negative segment-level endogenous driving operation costs through the Softplus activation function. Sum the segment-level endogenous driving operation costs to obtain the endogenous driving operation cost of the project to be tested.

[0033] S334, the segment graph attention prediction network is trained using the historical endogenous driven operation cost mode as the supervised label, and prediction loss, monotonic constraint loss and graph smoothing constraint are constructed. The total loss function is formed by combining the network parameter regularization term and the network parameters of the segment graph attention prediction network are updated through the total loss function.

[0034] Optionally, S34 includes:

[0035] S341: Obtain the degree of weather window limitation, transfer distance or transfer difficulty, market price fluctuation level, special crossing coordination difficulty and operation urgency of the project to be tested, construct the project-level exogenous disturbance feature vector, and normalize each exogenous disturbance feature.

[0036] S342, based on the determined weights of each exogenous disturbance characteristic, the normalized weather window restriction degree, transfer distance or transfer difficulty, market price fluctuation level, special crossing coordination difficulty and operational urgency are weighted and calculated to obtain the exogenous disturbance intensity of the project to be measured.

[0037] S343, compare the exogenous disturbance intensity with the exogenous disturbance trigger threshold to determine the exogenous disturbance trigger flag. When the exogenous disturbance intensity is lower than the exogenous disturbance trigger threshold, exogenous disturbance correction is not triggered. When the exogenous disturbance intensity reaches or exceeds the exogenous disturbance trigger threshold, exogenous disturbance correction is triggered.

[0038] S344, construct a three-segment disturbance correction function, and fit the correspondence between the historical exogenous disturbance operation cost mode and the exogenous disturbance intensity through the constrained nonlinear least squares method to determine the slight disturbance trigger threshold, strong disturbance trigger threshold, general disturbance correction intensity and strong disturbance correction intensity;

[0039] S345, calculate the disturbance correction value based on the exogenous disturbance intensity, and multiply the stable operating cost of the project under test by the disturbance correction value to obtain the exogenous disturbance operating cost of the project under test.

[0040] Optionally, S4 includes:

[0041] S41, the determined stable operation costs and exogenous disturbance operation costs are allocated to each line section. The allocation weight of stable operation costs is determined according to the section length, the number of detection points, and the basic operation allocation items. The allocation ratio of exogenous disturbance operation costs is determined according to the intensity of exogenous disturbances in the section. Combined with the section-level endogenous driving operation costs, the initial operation costs of each section and the contribution rate of various operation cost modes are calculated.

[0042] S42. Based on the contribution rates of various operation cost modes in each section and the preset difference threshold, determine the cost dominance type of each section, and verify the consistency of the correspondence between the cost dominance type of the section and the workload, operation difficulty, intensity of exogenous disturbance and project cost dominance type. For sections that do not meet the correspondence, readjust the corresponding modal contribution rate.

[0043] S43, combine the basic features of the segment, the segment map coding results and the segment cost-dominant type to form the segment mechanism partitioning features, calculate the mechanism difference distance between segments, use the physical adjacency connection relationship as a constraint, perform initial coarse division on physically continuous segments with similar cost formation mechanisms, and iteratively adjust the partition boundaries through partition cohesion, cross-partition association penalty and partition number penalty to obtain the line mechanism partitioning results;

[0044] S44, calculate the historical modal residuals between the various operation cost modes of historical projects and the corresponding backtest results, screen the matching historical modal residuals according to the detection type, voltage level, operation scenario, partition dominant type, channel complexity and defect risk level, determine the sub-modal correction coefficients of each line mechanism partition, and constrain the partition correction items according to the partition dominant type and the maximum correction ratio to obtain the corrected operation cost of each partition.

[0045] S45 summarizes the operation cost correction items of each line mechanism zone, and adds them to the stable operation cost, endogenous driving operation cost, and exogenous disturbance operation cost of the project under test to obtain the suggested operation cost. At the same time, it summarizes the corrected operation costs of each zone, performs consistency verification between the modal summation calibrator and the zone summary calibrator, and performs normalization calibration on the correction items of each zone when the deviation exceeds the preset threshold. It outputs the suggested operation cost, the composition of various operation costs, the line mechanism zone results, the operation cost correction items of each zone, the cost dominance type of each section, and the corresponding operation control classification.

[0046] Optionally, S43 includes:

[0047] S431: Obtain the standardized basic characteristics of each segment, the segment map coding results, the contribution rate of various operation cost modes, the intensity of exogenous disturbances and the dominant type of segment cost, construct the segment mechanism partitioning characteristics, and calculate the mechanism difference distance between any two segments.

[0048] S432, with physical adjacency as a constraint, scan each section along the actual sequence of the transmission line, divide the physically continuous sections with the same or similar cost-dominant type and mechanism difference distance less than the preset mechanism difference threshold into the same initial partition, set the partition boundary at the position where the mechanism difference distance reaches or exceeds the preset mechanism difference threshold, and obtain the initial partition result of the line.

[0049] S433, Based on the initial partitioning results of the line, with the goals of optimal partition cohesion, minimum cross-partition association penalty, and minimum number of partitions, the partitions to which the boundary segment belongs are iteratively adjusted, and the connectivity of each partition in terms of physical adjacency is maintained during the iteration process, and the partitioning results of the line mechanism are updated.

[0050] S434: Determine whether the relative change of the objective function corresponding to the circuit mechanism partitioning result is less than the preset change threshold or whether the number of iterations has reached the maximum number of iterations. If so, stop the iteration and output the circuit mechanism partitioning result. Otherwise, return to S433 to continue iterative optimization.

[0051] Optionally, S44 includes:

[0052] S441, Obtain the stable activity cost mode, endogenous driving activity cost mode and exogenous disturbance activity cost mode of the historical projects obtained from the decomposition, and backtest each historical project using the methods for determining stable activity cost, endogenous driving activity cost and exogenous disturbance activity cost, calculate the historical deviation of each type of activity cost mode, and calculate the historical relative residual of each type of activity cost mode;

[0053] S442, according to the detection type, voltage level, operation scenario, dominant type of partition, channel complexity and defect risk level, select various historical relative residual sets that match the current line mechanism partition, clean the outliers of the various historical relative residual sets, and remove historical relative residuals that are mainly random fluctuations;

[0054] S443, the sample weights are determined based on the time distance, feature similarity and data credibility of the matching historical items. The weighted median is used to calculate the stable mode correction coefficient, endogenous mode correction coefficient and exogenous mode correction coefficient of the current line mechanism partition.

[0055] S444, based on the partition dominance type of the current line mechanism partition, constrain the correction direction and correction magnitude of various modal correction coefficients, and calculate the partition correction item based on various modal correction coefficients and the corresponding partition operation cost mode, and apply a maximum correction ratio limit to the partition correction item to obtain the constrained partition correction item;

[0056] S445, add the initial operation cost of the current line mechanism partition to the partition correction term after the constraint to obtain the operation cost after the current line mechanism partition correction.

[0057] The beneficial effects of this invention are:

[0058] This invention collects historical transmission line inspection and testing project data, performs sample cleaning, standardization, and hierarchical database construction, and combines line segment division and segment feature extraction to construct a segment feature matrix containing segment workload characteristics, operation difficulty characteristics, and external disturbance characteristics. Furthermore, it utilizes multiple types of segment relationships to construct a segment relationship diagram, thereby establishing a correlation between historical cost data and the actual structure of the line and inspection operation conditions, thus improving the data integrity and structure of transmission line inspection operation cost analysis.

[0059] This invention decomposes the historical project-verified operating costs into stable operating cost modes, endogenously driven operating cost modes, and exogenously disturbed operating cost modes. It then combines the endogenous driving strength, exogenous disturbance strength, and exogenous disturbance triggering indicators for constrained optimization solutions and decomposition result verification to distinguish and identify different cost-forming factors. Simultaneously, it utilizes a segment graph attention prediction network, a disturbance factor triggering correction model, and a weighted robust statistical method to determine various operating costs, thereby improving the accuracy of cost determination under different line conditions, operating scales, and external disturbances.

[0060] This invention divides the line mechanism into zones based on the cost-dominant type, zone characteristics, and physical connectivity of the zone, and determines the zone correction coefficient by combining historical mode residuals. It then performs sub-mode correction and cost calibrator consistency calibration on each line mechanism zone, enabling the cost determination results to adapt to the differences in cost formation mechanisms of different line sections. This reduces cost deviations caused by differences in regional conditions, hidden work inputs, and mode prediction biases, thereby improving the rationality and reliability of the cost determination results for transmission line inspection operations. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in this invention or the prior art, 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 for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a schematic diagram of the determination method according to an embodiment of the present invention. Detailed Implementation

[0063] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0064] like Figure 1 As shown, a method for determining the modal zoning of transmission line inspection operation costs includes the following steps:

[0065] S1, Construct a cost sample and section relationship diagram for power transmission line testing and inspection projects:

[0066] This step establishes a unified sample base for historical transmission line testing projects. First, it collects data including line parameters, testing types, testing methods, field operation records, testing reports, settlement documents, and verified operating costs. Missing information, duplicate records, abnormal values, and inconsistent data are then cleaned and standardized. High-cost samples resulting from reasons such as relocation to mountainous areas, special crossings, extreme weather, or emergency testing are not simply discarded; instead, they are retained in the sample database with disturbance indicators to reflect specific cost formation scenarios in actual testing operations.

[0067] 1. Historical project data collection and sample preprocessing:

[0068] Collect basic data on historical transmission line testing projects. The data list includes: project name, project number, line name, voltage level, testing type, testing method, line length, number of towers, number of testing points, testing frequency, terrain conditions, defect risk, weather window, relocation distance, special crossing conditions, market price fluctuations, urgency of operation, submitted operation cost, verified operation cost, testing report, operation record, settlement data, and line drawings.

[0069] The business meaning of each dimension indicator is as follows:

[0070] Line length, number of poles and towers, number of inspection points, and inspection frequency: these represent the workload of the inspection operation;

[0071] Voltage level, terrain conditions, and defect risk: characterize the difficulty of the operation;

[0072] Weather window, relocation distance, special crossing situations, market price fluctuations, and operational urgency: characterize external disturbance conditions;

[0073] (1) Sample inclusion and standardization:

[0074] Historical samples should include at least the core fields such as line length, number of towers, voltage level, testing method, testing type, and verified operating cost; projects lacking core fields and unable to be supplemented through testing reports, line drawings, or settlement details will not be included in the valid sample database.

[0075] Duplicate data: Merge and remove duplicates using project number, line name, inspection time, and verified operating cost as the primary key;

[0076] Unit standardization: Data with inconsistent units of measurement are uniformly converted to the same measurement standard;

[0077] Outlier handling: Identify outlier samples by combining unit line length cost, unit tower cost, and unit inspection point cost;

[0078] If the anomaly originates from crossing high-speed railways, long-distance relocation in mountainous areas, extreme weather windows, or emergency testing operations, the sample should be retained and marked as a disturbance sample.

[0079] If the anomaly is caused by data entry error, duplicate data collection, or inconsistent cost standards, it should be corrected or removed.

[0080] (2) Layered database construction:

[0081] The database is stratified according to testing type, voltage level, and operating scenario to avoid mixing projects with significantly different cost formation mechanisms in the model:

[0082] Inspection types: UAV visible light inspection, infrared thermography, live-line inspection, power outage test, and passageway hazard detection;

[0083] Voltage level: Classified according to the actual voltage level of the transmission line;

[0084] Operation scenarios: regular sections in plains, complex sections in mountainous areas, special sections crossing certain terrains, and sections with high risks and potential hazards;

[0085] 2. Project-level cost sequence construction:

[0086] The first Each historical transmission line inspection and testing item is recorded as Its approved operating cost is recorded as All historical projects constitute the project cost series:

[0087] ;

[0088] in, This indicates the historical project-verified activity cost sequence. Indicates the first The approved operating costs for each project This represents the number of historical items that have entered the valid sample library. This cost sequence is used in step S2 to decompose the stable mode, endogenous driving mode, and exogenous perturbation mode.

[0089] 3. Line segment division and segment feature extraction:

[0090] (1) Segment division rules:

[0091] The section division does not adopt a simple equidistant division, but is determined based on the actual structure of the transmission line and the boundaries of the inspection work, according to tower sections, tension sections, crossing sections, and sections with potential hazards in the corridor:

[0092] Conventional continuous lines: using adjacent towers or continuous tower sections as the base section;

[0093] Tension / Corner Position: The section boundary is formed by the location of the tension tower, corner tower, or the change in operation mode;

[0094] Special crossing locations: crossing railways, highways, rivers, important power transmission channels, etc., forming a separate crossing section;

[0095] High-risk areas: Areas along the route with concentrated risks of tree obstruction, external damage, wildfires, icing, and geological disasters, forming separate sections with potential hazards.

[0096] Let the first The set of segments obtained from the division of each project is:

[0097] ;

[0098] in, Indicates the first A collection of segments for each project. Indicates the first The first project Each section Indicates the first The number of segments obtained from dividing each project.

[0099] (2) Definition and quantification of segment characteristics:

[0100] For each segment Workload characteristics, task difficulty characteristics, and external disturbance characteristics were extracted, and all qualitative indicators were quantified and mapped.

[0101]

[0102] Therefore, the section The comprehensive feature vector is represented as:

[0103] ;

[0104] in, Indicates the first The first project The comprehensive feature vector of each section covers three types of factors: workload, operational difficulty, and external disturbances, and is used to describe the main technical conditions for the formation of section costs.

[0105] (3) Segment feature matrix:

[0106] Arrange the comprehensive feature vectors of all sections within the same project in order of route to form a section feature matrix:

[0107] ;

[0108] in, Indicates the first The segment feature matrix for each project. Each row of the matrix corresponds to a route segment, and each column corresponds to a cost impact feature, preserving both the segment differences within the project and the spatial order of the segments within the route.

[0109] 4. Construction of the segment relationship diagram:

[0110] Treating each segment as a graph node and the engineering relationships between segments as graph edges, we obtain the first... Section relationship diagram of each project:

[0111] ;

[0112] in, Indicates the first Section relationship diagram of each project Represents the set of segment nodes. Represents the set of edges representing segment relationships. This represents the feature matrix of the segment nodes.

[0113] (1) Definition of four types of relation edges:

[0114] Section relationship edges include adjacent connection edges, edges in the same channel environment, edges of the same risk type, and operation transfer edges, each corresponding to different engineering association logic:

[0115] 1) Adjacent connecting edges: Describe the physical continuity of the line. Generated when two sections are connected end to end, the edge weight is positively correlated with the adjacent length, tower connection relationship, and work continuity.

[0116] 2) Same-channel environment edges: Describes the similarity relationship of the working environment. They are generated when the corridors are parallel, the towers are double-loop, the terrain is similar, or the spatial distance is less than a preset threshold. The edge weight is determined by the similarity of the channel conditions.

[0117] 3) Edges of the same risk type: Describes the risk transmission relationship. These edges are generated when two segments belong to the same risk type and the difference in risk level does not exceed a preset threshold. The edge weight is determined by the consistency of risk type and the similarity of risk levels.

[0118] 4) Operation Transfer Edge: Describes the relationship between operation movements. It is generated based on the distance of the entry path between sections, road accessibility, and the difficulty of organizing the transfer. The edge weight is positively correlated with the transfer distance and difficulty.

[0119] (2) Calculation of the overall edge weight:

[0120] For any two segments and Its overall edge weight is a weighted fusion of the four types of edge weights:

[0121] ;

[0122] in, For section and The combined edge weights between them , , , The weights are respectively for adjacent connections, same channel environment, same risk type, and operation transfer. Let be the weight coefficients of various relation edges, satisfying If there is no corresponding relationship between two segments, the weight of that sub-item is 0.

[0123] S2, Constrained Activity Cost Modal Decomposition:

[0124] To address the cost formation mechanism of transmission line inspection operations, three types of cost modes are customized with engineering meanings and constraints. This ensures that the decomposition results correspond to three different sources: stable operation input, line body and inspection workload-driven factors, and external disturbance-driven factors. This provides a hierarchical labeling basis for subsequent operation cost prediction, inspection operation scheme optimization, and operation and maintenance resource allocation. Specifically, a constraint decomposition method is used. First, the endogenous driving force and exogenous disturbance intensity are calculated within the same sample group. Then, through iterative solution, the sum of the three types of costs is made consistent with historically verified costs. Constraints are set to ensure that stable costs remain stable, endogenous costs match the operation scale and risk level, and exogenous costs are only formed when disturbing factors are triggered. For samples with large residuals, a backtracking verification method is used for correction.

[0125] 1. Basic definition of three-mode decomposition:

[0126] For the The approved operating cost of a historical transmission line testing and inspection project. Decomposed into:

[0127] ;

[0128] The engineering meanings of each component are as follows:

[0129] Stable operating cost mode: Characterizes the relatively stable operating costs incurred in similar testing projects, such as site organization, basic labor, basic equipment preparation, data processing, and test report preparation.

[0130] Endogenous driving operation cost mode: Characterizes the changes in operation costs caused by project-specific factors such as line length, number of towers, number of inspection points, inspection frequency, voltage level, terrain complexity, and defect risk level;

[0131] : Exogenous disturbance operation cost mode, which represents changes in operation cost caused by external conditions such as restricted weather windows, long-distance transition, market price fluctuations, special crossing coordination and emergency operations;

[0132] : Decomposition residual, which represents a small amount of random error that cannot be explained by the three types of operation cost modes.

[0133] 2. Division of similar sample groups:

[0134] Based on the layered sample library formed in step S1, similar sample groups are divided according to the dimension of "detection type + voltage level + operation scenario", so as to ensure that the cost formation mechanism of projects in the group is comparable, and the th similar sample group is denoted as .

[0135] The division rules are as follows:

[0136] Detection type: UAV visible light inspection, infrared temperature measurement, live detection, power outage test, channel hidden danger detection;

[0137] Voltage level: divided according to the actual voltage level of the transmission line;

[0138] Operation scenario: conventional plain section, complex mountain section, special crossing section, high-risk hidden danger section;

[0139] Sample size bottoming rule: if the number of valid samples in a group is insufficient, perform upward merging in the order of merging operation scenarios first and then merging adjacent voltage levels, but do not mix samples across detection types, so as to avoid mutual interference of operation cost formation mechanisms under different detection methods.

[0140] 3. Feature normalization and driving intensity index calculation:

[0141] (1) Project-level feature aggregation:

[0142] For each project, project-level features are obtained by aggregating the section feature matrix obtained in step S1:

[0143] Project-level endogenous features: total line length , total number of poles and towers , total number of detection points , detection frequency , voltage level , average terrain channel complexity , highest defect risk level ;

[0144] Project-level exogenous features: degree of restricted weather windows , transition distance / transition difficulty , market price fluctuation level Special cross-border coordination difficulties urgency of the task ;

[0145] To eliminate dimensional differences, minimum-maximum normalization is performed on all features within the same sample group:

[0146] ;

[0147] in, For normalized eigenvalues, , The minimum and maximum values ​​of this feature within the group. The denominator should be a very small positive number to avoid being zero.

[0148] (2) Indicators of endogenous driving force:

[0149] Instead of using a simple equal-weighted average, the endogenous driving force strength is assigned differentiated weights based on the contribution of various endogenous characteristics to historically determined operating costs:

[0150] ;

[0151] in, For the first The higher the value of the intrinsic driving force of a project, the stronger the cost-driven effect of workload and operational difficulty. For the weights of each endogenous feature, satisfying ;

[0152] The weights are determined by partial correlation analysis within the same sample group: under the control of other endogenous characteristics, the partial correlation coefficients between each characteristic and the approved operating cost are calculated, and the absolute values ​​are normalized and used as the weights.

[0153] Engineering adaptation principle: The number of inspection points, the complexity of terrain and passage, and the level of defect risk have a more direct impact on on-site operation organization, inspection time and safety control intensity. Their weight is usually higher than that of the line length alone. The specific value is determined by calculation based on historical data.

[0154] (3) Exogenous disturbance intensity index:

[0155] Similarly, the exogenous disturbance intensity index is calculated:

[0156] ;

[0157] in, For the first The intensity of exogenous disturbances for each project is indicated by a higher value, which represents a more significant impact of the external disturbances on operating costs. The weights of each exogenous perturbation feature satisfy the following conditions: The weights are determined by fitting historical data within the same sample group. The higher the contribution of the corresponding factor to the explanation of the determined operating cost, the greater the weight.

[0158] (4) Exogenous disturbance trigger flag:

[0159] Set an exogenous disturbance trigger flag to distinguish between normal items and disturbance items:

[0160] ;

[0161] in, This indicates that the project faces external conditions such as weather restrictions, special crossings, long-distance relocations, or emergency testing, requiring separate extraction of exogenous disturbance modes. This indicates that the project has no significant external disturbances, and exogenous disturbances are not a major source of cost. The threshold for triggering exogenous disturbances is determined based on the quantile of the intensity of exogenous disturbances in the same sample group or the identifier of historical disturbance samples.

[0162] 4. Constrained Modal Decomposition Optimization Objective:

[0163] In each similar sample group The core objectives of the internal execution constraint decomposition are: low fluctuation in stable modes, consistency between endogenous modes and driving strength, and activation of exogenous modes only upon triggering. The optimization objective function is:

[0164] ;

[0165] The meanings of each item are as follows:

[0166] The first item: Reconstruct the error term, minimizing the squared error between the verified cost and the sum of the three modes;

[0167] The second item: stability constraints. The variance of the stable modes within the group is used to constrain the stable modes from excessive fluctuations.

[0168] The third item: Endogenous consistency constraint, which forces the endogenous driving mode to correspond proportionally with the endogenous driving strength;

[0169] The fourth item: exogenous triggering constraint, which forces the exogenous disturbance mode to exist only when the trigger flag is activated, and is positively correlated with the disturbance intensity;

[0170] , : These are the proportional coefficients corresponding to the costs of endogenous driving force intensity and exogenous disturbance intensity, respectively;

[0171] , , The weighting coefficients for the three types of constraints are determined by the fitting error of historical samples and the stability of decomposition.

[0172] The optimization problem simultaneously satisfies the following hard constraints:

[0173] ;

[0174] The first set of constraints ensures that all three types of modes have a non-negative physical meaning of activity cost; the second set of constraints ensures that exogenous disturbance modes only occur when triggered by external conditions, avoiding misclassification of regular activity costs as external disturbances.

[0175] 5. Optimize problem solving:

[0176] (1) Determination of the initial center of stable modes:

[0177] Before solving, robust statistical methods are used to determine the initial values ​​of the stable modes to avoid getting trapped in local optima during iteration. (In similar sample groups) Among them, items with exogenous disturbance triggering indicators of 0 and endogenous driving strength in the 25%~75% quantile range within the group were selected to form a stable sample subset. .

[0178] For a stable subset of samples, weighted by sample confidence, calculate the weighted truncated median of the approved operating costs, which serves as the initial center of the stable mode:

[0179] ;

[0180] in, This represents the weighted truncated median. For the first The credibility weight of each sample is determined based on the completeness of the data, the completeness of the test report, the consistency of the settlement caliber, and the status of anomaly identification. This process can reduce the impact of abnormally high values, abnormally low values, and inconsistencies in early data caliber on the initial value of the stable mode.

[0181] (2) Alternating iterative constrained least squares solution:

[0182] The above optimization problem is solved using the alternating iterative constrained least squares method, and the specific process is as follows:

[0183] The first step is to fix the stable modes and solve for the endogenous and exogenous modes: Initialize to The optimization problem is simplified to about , , and The constrained linear regression problem is solved by combining the driving strength and triggering flag to obtain the initial endogenous and exogenous modes;

[0184] The second step is to fix the endogenous and exogenous modes and update the stable mode: fix the endogenous and exogenous mode results, and update the stable mode sequence under non-negativity constraints and variance constraints, so that the stable mode can explain the remaining operating costs as much as possible on the one hand, and maintain the low fluctuation characteristics within the group on the other hand, so as to avoid being skewed by individual high workload or strong perturbation samples.

[0185] The third step is to update the scaling factor: based on the fitting relationship between the current endogenous mode and the endogenous driving strength. Based on the current exogenous modality and Fitting relationship update The updated coefficients are used in the next round of modal iteration;

[0186] Step 4, Iterative Convergence Judgment: Repeat the above process alternately until the change in the objective function value between two adjacent iterations is less than the preset convergence threshold, or the maximum number of iterations is reached. The convergence condition is:

[0187] ;

[0188] in, For the first The objective function value of the next iteration. This is the convergence threshold. When the sample size is large or there are many similar sample groups, the Alternating Direction Multiplier Method (ADMM) can be used to accelerate the solution.

[0189] 6. Verification of decomposition results:

[0190] (1) Residual value verification:

[0191] After the decomposition is completed, calculate the decomposition residual for each item:

[0192] ;

[0193] like occupy If the proportion of residuals exceeds the preset residual threshold, the project is marked as a sample to be reviewed, and the process returns to step S1 to check its detection range, work records, special crossing markers, and work cost aggregation caliber; if the residual proportion does not exceed the threshold, the decomposition results are considered valid.

[0194] (2) Modal consistency check:

[0195] Verify the rationality of the decomposition results from an engineering mechanism perspective to avoid mathematical optimization that distorts business applications.

[0196] If the project's line length, number of towers, number of inspection points, and defect risk level are significantly higher, then the cost of the endogenous driving mode should be increased accordingly.

[0197] If the project involves weather window limitations, long-distance relocation, special cross-regional coordination, or emergency testing operations, then the exogenous disturbance mode should be triggered;

[0198] If the project is a routine planned inspection task and external disturbances are not significant, the operating cost should be mainly explained by the stable mode and the endogenous driving mode.

[0199] 7. Step Output:

[0200] After step S2 is completed, the following results are output for each historical item:

[0201] (1) Stable operating cost mode Endogenous activity-driven costing model Exogenous disturbance operation cost mode ;

[0202] (2) Decompose the residual ;

[0203] (3) Endogenous driving force intensity Intensity of exogenous disturbances Exogenous disturbance trigger flag .

[0204] S3, the three activity cost modes are determined respectively:

[0205] Separate methods are established for determining stable operating cost modes, endogenously driven operating cost modes, and exogenously disturbed operating cost modes. Specifically, stable operating costs are determined using robust statistical methods with similar samples, endogenously driven operating costs are determined using a segment graph attention prediction network, and exogenously disturbed operating costs are determined using a disturbance factor-triggered correction model. Stable operating costs are determined using a weighted robust statistical method, focusing on similar projects with complete data, weak disturbances, and moderate scale. Endogenously driven operating costs are determined using a segment graph attention prediction method. The segment characteristics and segment relationship diagram of the project to be tested are input into the model, which identifies segments with high workload, high terrain difficulty, high risk, and strong relocation correlation, and outputs segment-level and project-level endogenous costs. Exogenously disturbed operating costs are determined using a disturbance trigger correction method. Costs are not calculated separately if the trigger condition is not met; when the trigger condition is met, they are corrected according to the disturbance intensity. The final outputs are stable cost, endogenous cost, exogenous cost, and cost-dominant type.

[0206] 1. Matching of similar sample groups:

[0207] Let the current project to be measured be denoted as . According to the sample stratification rules in step S1, first... Match samples to the corresponding group of samples based on detection type, voltage level, and operating scenario. If multiple candidate sample groups exist, the selection is based on the similarity between the current project and historical sample groups in terms of detection type, voltage level, line size, operating scenario, and risk level.

[0208] ;

[0209] in, This indicates the group number of the sample group of the same type that best matches the current project; Indicates the current project With the The overall similarity between sample groups is calculated based on the weighted average of the matching degree of the classification dimension and the cosine similarity of the normalized features.

[0210] After matching is complete, extract the stable activity cost mode sequence obtained in step S2 from the sample group. Endogenous activity-driven costing modal sequence and exogenous disturbance operation cost mode sequence This serves as the statistical basis and training labels for this step.

[0211] 2. Determination of stable operating costs (weighted robust statistical method):

[0212] Stable operating costs mainly correspond to relatively stable operational inputs such as on-site organization, basic labor, basic equipment preparation, data processing, and test report preparation. This part is not suitable for direct fitting with complex prediction models to avoid individual sample noise being solidified into the results. Therefore, the weighted truncated median of similar samples is used to determine it.

[0213] In the same sample group In this study, historical items with exogenous disturbance triggering indicators of 0, endogenous driving strength within the 25th to 75th percentile range of the group, and data completeness meeting the requirements were selected to form a stable reference sample set. .

[0214] For each reference sample Calculate its weights:

[0215] ;

[0216] in, For the first The weight of each historical sample For the first The annual difference between a historical sample and the current project The weighted feature distance between the current project and historical samples in terms of detection type, voltage level, operating scenario, line size, and risk level. For the first The reliability of the data for each sample , These are the time decay coefficient and the similarity decay coefficient, respectively.

[0217] The current project's stable operating cost is determined to be:

[0218] ;

[0219] in, This represents the current stable operating cost of the project; This represents the weighted truncated median; This indicates that the result obtained in step S2 is the first... The method utilizes stable operating cost models from historical projects. It preserves relatively stable baseline operating cost levels across similar testing projects while mitigating the impact of early data variability, outliers, and outliers on the results.

[0220] 3. Determination of endogenously driven activity costs:

[0221] The intrinsically driven operating costs are mainly determined by the line's inherent conditions and the workload of inspections, including factors such as line length, number of towers, number of inspection points, inspection frequency, voltage level, terrain complexity, and defect risk level. Since different line sections are related by physical adjacency, shared corridor environment, similar risk types, and work relocation, a segment map attention prediction network is used for prediction.

[0222] (1) Input and graph structure definition:

[0223] According to step S1, the current project The segment relationship diagram is represented as follows:

[0224] ;

[0225] in, This diagram represents the segment relationship of the current project. Represents the set of segment nodes; Represents the set of edges representing segment relationships; This represents the segment feature matrix.

[0226] Endogenous-driven prediction uses only workload and task difficulty features for segment nodes. Its endogenous driving feature vector is:

[0227] ;

[0228] in, The length of the segment; The number of towers; This refers to the number of detection points; For detection frequency; The voltage level for the current project; For terrain and passage complexity; This represents the defect risk level.

[0229] (2) Multi-relationship graph attention encoding:

[0230] Section relationship edges include four categories: adjacent connecting edges, edges in the same channel environment, edges of the same risk type, and operation transfer edges. Regarding edge types... Section For section Attention score is calculated as follows:

[0231] ;

[0232] in, edge type lower section For section Attention score edge type The corresponding feature transformation matrix, For attention parameters, Embed vectors for edge types. This represents vector concatenation. It is a linear unit activation function with leakage correction;

[0233] Normalize the attention scores to obtain the attention weights:

[0234] ;

[0235] in, edge type lower section For section Attention weights; To be consistent with the section There exists the first The set of adjacent segments with class relationships.

[0236] Section The image encoding result is:

[0237] ;

[0238] in, For section The endogenous driving graph encoding results; It is a set of edge types, including adjacent connections, environments with the same channel, environments with the same risk type, and job transfers; It is a non-linear activation function.

[0239] (3) Section-level cost forecasting and project-level summary:

[0240] Inputting the segment map encoding results into the prediction layer yields the segment-level endogenous driving activity cost:

[0241] ;in, For section The predicted value of endogenous driving activity costs, For the prediction layer of a multilayer perceptron, Used to ensure that the cost of the output is non-negative;

[0242] The current project's endogenous activity cost is the sum of all segment-level endogenous activity costs:

[0243] ;

[0244] in, This represents the endogenous driving activity cost of the current project.

[0245] (4) Network training and multi-constraint loss function:

[0246] The segment graph attention prediction network uses the historical endogenous driving activity cost modality obtained in step S2. The data is used as a supervisory label for training. If historical projects can provide segment-level activity cost allocation data, then segment-level endogenous activity costs are used as training labels; if historical projects only have project-level endogenous activity cost modalities, then the summed segment outputs are used in conjunction with the project-level labels for training, and the predicted loss is:

[0247] ;

[0248] To ensure that the prediction results conform to the operational patterns of transmission line inspection and testing, a monotonic constraint loss is incorporated. This requires that the endogenous driving operation cost does not decrease unreasonably when the number of inspection points increases, the complexity of the terrain becomes more complex, and the defect risk level rises.

[0249] ;

[0250] in, This represents the total number of testing sites. This represents the average terrain passage complexity. This represents the highest defect risk level.

[0251] Simultaneously, graph smoothing constraints are added to prevent excessive jumps in prediction results for strongly correlated segments with similar features:

[0252] ;

[0253] in, For section and section Edge weights between them; For the first A set of segment relationships between historical projects.

[0254] The ultimate training objective is:

[0255] ;

[0256] in, To predict the total loss of the network driven by endogenous activity costs, For network parameters, , , These are the weighting coefficients for each loss term.

[0257] 4. Determination of exogenous disturbance operation costs:

[0258] Exogenous disturbance operation costs are mainly caused by factors such as limited weather windows, relocation distance, market price fluctuations, special cross-border coordination, and the urgency of operations. They are sporadic and trigger-based, and are not suitable for direct prediction as regular costs. Therefore, a disturbance factor trigger correction model is used to determine them.

[0259] For the current project Construct project-level exogenous perturbation feature vectors:

[0260] ;

[0261] in, The degree of weather window limitation; Transition distance or transition difficulty; This represents the level of market price fluctuations. This presents a unique challenge in terms of coordination. The urgency level of the task.

[0262] Based on the exogenous disturbance weights determined in step S2, calculate the exogenous disturbance intensity of the current project:

[0263] ;

[0264] in, The intensity of the exogenous disturbance; , , , , The characteristics of exogenous perturbations after normalization; to The weights are for each exogenous perturbation feature.

[0265] Calculate the disturbance trigger flag based on the exogenous disturbance intensity:

[0266] ;

[0267] in, Indicates whether the current project triggers exogenous disturbance activity costs; This is the threshold for triggering exogenous disturbances.

[0268] when When, the cost of exogenous disturbance operations is taken as 0 or close to 0; when When this occurs, an exogenous disturbance correction is triggered. The disturbance correction function adopts a three-segment piecewise form:

[0269] ;

[0270] in, This is the perturbation correction function. The threshold is set for triggering minor disturbances. This is the threshold for triggering strong disturbances. This is the general disturbance correction strength. For strong disturbance correction strength;

[0271] The above parameters are determined by fitting the correspondence between historical exogenous disturbance modes and disturbance intensity using a constrained nonlinear least squares method, with the constraint slope being non-negative and the threshold being within a reasonable quantile range.

[0272] The current project's exogenous disturbance activity cost is determined as follows:

[0273] ;

[0274] in, The cost of exogenous disturbances in the current project; This represents the stable operating cost of the current project. If the current project experiences no significant external disturbances, then... Take 0 or close to 0; if there are situations such as long-distance transfers across highways, railways, mountainous areas, extreme weather windows, or emergency testing, then... Increased costs are triggered by exogenous disturbances.

[0275] 5. Modal contribution rate and cost-dominant type identification:

[0276] After determining the costs of the three types of activities, calculate the contribution rates of the three modalities for the current project:

[0277] ;

[0278] ;

[0279] ;

[0280] in, , , The contribution rates of stable, endogenous-driven, and exogenous-disturbance operating cost modes are respectively. It is a very small positive number, used to avoid the denominator being zero;

[0281] Determine the project cost dominance type based on the maximum contribution rate: If At its maximum, the project belongs to the stable benchmark-driven type. The largest project is one dominated by operational scale and risk conditions. The project is the most sensitive to exogenous disturbances.

[0282] Cost-driven type serves as an important basis for line mechanism partitioning and partition correction in step S4.

[0283] S4, Circuit mechanism partitioning correction and cost caliber calibration:

[0284] This step differs from the traditional approach of dividing routes based on administrative sections, fixed lengths, or geographical areas. Instead, it uses the operational cost formation mechanism as the core basis while preserving the physical continuity requirement of inspection operations, thus aligning the operational cost control logic with the on-site inspection operation organization logic. First, project costs are allocated to each route segment. Stable operational costs are allocated based on route length, inspection points, and basic operational factors; exogenous disturbance operational costs are allocated based on weather constraints, relocation difficulty, special crossings, and urgency; endogenous driving operational costs are based on the segment prediction values ​​output by the model. Then, a mechanism-based zoning method is used, grouping physically continuous segments with similar cost formation mechanisms into the same zone based on segment continuity, feature similarity, risk type, disturbance intensity, and cost-dominant type. After zoning, a limited-amplitude correction method is used to compensate for implicit operational inputs, taking into account historical deviation patterns of similar zones. Finally, a dual-caliber consistency verification method is used to calculate both the total project cost and the zone summary cost. If the deviation between the two is within the allowable range, the final suggested activity cost is output; if the deviation exceeds the limit, the partition correction item is normalized to ensure that the total project cost is consistent with the sum of the partition costs.

[0285] Let the set of segments for the current project be:

[0286] ;

[0287] in, This represents the set of segments in the current project. Indicates any segment. This indicates the number of segments obtained by dividing the current project.

[0288] 1. Segmented allocation of project-level activity costs:

[0289] Since the endogenous driving operation cost is directly output by the segment-level result from the segment graph attention prediction network in step S3, This step only applies segmented allocation to stable operating costs and exogenous disturbance operating costs.

[0290] (1) Allocation of stable operating costs:

[0291] The allocation weights are constructed based on the section length, the number of detection points, and the basic work allocation items:

[0292] ;

[0293] in, For section The weighting of stable operating costs. For the length of the section line, The number of detection points in the section. , , These are the weighting coefficients for line length, detection points, and basic amortization items, respectively, satisfying... ;

[0294] The weights are determined based on the partial correlation between stable operating costs and various factors in similar projects: a 4:3:3 initial ratio can be used for routine inspection projects; for projects with high requirements for basic operation organization, such as live-line testing and power outage testing, the weight of the basic amortized items can be increased; the final values ​​are based on the calibration results of historical data from similar sample groups.

[0295] The section stabilization operation cost is:

[0296] ;

[0297] This allocation method satisfies the principle of total quantity conservation:

[0298] ;

[0299] (2) Allocation of exogenous disturbance operation costs:

[0300] The impact is allocated based on the intensity of exogenous disturbances in each segment. For each segment... Calculate the intensity of exogenous disturbances in the section:

[0301] ;

[0302] in, The intensity of exogenous disturbance in the section, , , These are the normalized segment weather window limitations, transition difficulties, and special crossing coordination difficulties, respectively. , These are the normalized project-level market price fluctuations and the urgency of the operation. Reuse the exogenous perturbation feature weights determined in step S2;

[0303] The cost of exogenous disturbance operation in the section is:

[0304] ;

[0305] in, The value should be a very small positive number to avoid a denominator of zero. If the project does not trigger significant exogenous disturbances, Since the cost is close to zero, the cost-sharing process will not artificially amplify the disturbance costs; sections that cross railways, are relocated in mountainous areas, or experience extreme weather will receive a higher cost-sharing ratio.

[0306] (3) Initial cost of the section and modal contribution rate:

[0307] Each segment generates three types of operating costs. , , The initial operating cost of the section is:

[0308] ;

[0309] Calculate the contribution rates of the three modalities at the segment level:

[0310] ;

[0311] ;

[0312] ;

[0313] 2. Identification and consistency verification of segment cost-dominant types:

[0314] (1) Dominant type determination rules:

[0315] Based on the contribution rates of three modalities, the dominant type of segment operation cost is identified. A difference threshold is set to avoid forced classification, and the threshold is used for discrimination. The same criteria are used as the project-level judgment threshold in step S3:

[0316] ;

[0317] in, , It is the second largest contribution rate among the three categories; the four categories correspond to the stable benchmark-dominated type, the operation scale and risk-dominated type, the exogenous disturbance sensitive type, and the mixed-dominated type, respectively.

[0318] (2) Consistency check:

[0319] After the initial determination of the dominant type, a triple check is performed to avoid business logic conflicts caused by simple numerical allocation:

[0320] 1) Feature matching verification: If the exogenous disturbance intensity of the section Below the minimum disturbance threshold However, if it is identified as sensitive to exogenous disturbances, its exogenous mode contribution rate is reduced and it is re-evaluated; if the workload, terrain complexity, and risk level of a section are significantly lower than the average of the same type but it is identified as endogenous dominant, the endogenous driving prediction results are retrospectively checked.

[0321] 2) Hierarchical logic verification: If the dominant type at the project level is in obvious conflict with the dominant type of the section that accounts for the highest proportion of the total line length / total cost, backtrack and verify the allocation weight of stable and exogenous costs, and recalculate the section contribution rate if necessary.

[0322] 3. Line-based partitioning with connectivity constraints:

[0323] (1) Segmentation mechanism partitioning characteristics:

[0324] Constructing segmentation features based on segmentation mechanism, and directly reusing the graph coding results from step S3:

[0325] ;

[0326] in, For the standardized basic features of the segment, The segment map encoding result output in step S3, This is a segment cost-dominant type.

[0327] The distance between the mechanisms of any two segments is:

[0328] ;

[0329] in For indicator functions, a fixed distance penalty is added when the dominant type is inconsistent; These are the weighting coefficients for each dimension.

[0330] (2) Two-phase graph constraint partitioning method:

[0331] The circuit mechanism is partitioned using a two-stage graph constraint partitioning method with physical connectivity constraints. The partitioning result is denoted as... , For the first One partition, This represents the number of partitions.

[0332] Phase 1, Initial Coarse Segmentation:

[0333] Based on physically adjacent connecting edges, all segments are scanned in the actual sequence of the line. Segments with the same or similar dominant type, a mechanism difference distance of less than a preset threshold, and physical continuity are assigned to the same initial partition. Partition boundaries are set where the difference exceeds the threshold. This stage ensures that each initial partition is physically continuous on the line, conforming to the segment organization logic of the inspection operation.

[0334] The second phase, iterative optimization:

[0335] Based on the initial coarse segmentation results, the assignment of boundary segments is iteratively adjusted with the objectives of maximizing mechanism cohesion and minimizing cross-partition association penalties. The objective function is:

[0336] ;

[0337] in:

[0338] First item: Partition cohesion constraint. For the first The central feature of each partition (the weighted average of the segment features within the partition) ensures that the cost formation mechanism of the same partition is similar;

[0339] The second item is the graph structure coherence constraint, which penalizes all relation edges that cross different partitions to prevent strongly associated segments from being cut off without a basis. The partition number to which the segment belongs. These are constraint coefficients;

[0340] The third item: partition count penalty, to prevent excessive partitions from causing result fragmentation. This is the penalty coefficient;

[0341] A hard constraint is enforced during the iteration process: The same partition must remain connected on physically adjacent edges, i.e.:

[0342] ;

[0343] Edges in the same channel environment or with the same risk type are only used for similarity identification and do not replace the connectivity constraints of physically adjacent edges. Number of partitions. The number of initial blocks, elbow rule, total line length, and number of segments are combined to determine the total number of segments.

[0344] The iteration stops when the relative change in the objective function value is less than a preset threshold, or when the maximum number of iterations is reached.

[0345] 4. Modal partitioning correction:

[0346] Partition correction is used to compensate for implicit systematic operational inputs that are not fully covered in modal decomposition and submodal prediction and are strongly correlated with the dominant type of the partition (such as safety protection for crossing sections, commuting support for mountainous sections, and retesting inputs for high-risk sections), and is not used to absorb random errors and sample noise.

[0347] (1) Calculation of historical submodal residuals:

[0348] For historical projects Calculate the historical deviations for the three modes respectively:

[0349] ;

[0350] in, , , The true values ​​of the historical modes obtained from step S2 decomposition are... , , This is the backtest result of the method in step S3.

[0351] Further calculation of the modal relative residuals:

[0352] ;

[0353] (2) Determination of zoning correction coefficients:

[0354] Based on the detection type, voltage level, operating scenario, dominant type of zone, channel complexity, and defect risk level, filter the samples corresponding to the current zone. The historical residual set is matched. Outlier cleaning is first performed on the residual set (samples outside 1.5 times the interquartile range are removed). If the median of the residuals after cleaning is close to 0 and has no stable direction, it is considered to be mainly random fluctuation and is not used for correction.

[0355] The partitioning correction coefficients for the three modes were determined using the weighted median:

[0356] ;

[0357] ;

[0358] ;

[0359] in, , , To match and clean the three types of historical relative residual sets, The weights for historical samples are determined by time distance, feature similarity, and data reliability.

[0360] (3) Business constraints and scope limitations:

[0361] Add business rule constraints to modify the direction and magnitude:

[0362] Stable benchmark-dominated partitioning: The upper limit of the correction magnitude is lower than that of other types, and the correction coefficient should not deviate significantly from 0;

[0363] Operation scale and risk-driven zoning: If the endogenous modal correction coefficient is significantly negative, backtrack and verify the endogenous driving prediction results to avoid unreasonably suppressing normal operation input;

[0364] Exogenous disturbance sensitive partition: Only when the partition has obvious disturbance characteristics and Positive correction is allowed when the exogenous mode correction coefficient is positive; otherwise, the correction coefficient is forced to be 0.

[0365] The partitioning correction term is the sum of the correction amounts for the three modalities:

[0366] ;

[0367] Set a maximum correction ratio constraint to prevent historical deviations from being excessively amplified:

[0368] ;

[0369] The modified terms after constraints are:

[0370] ;

[0371] in, This is a truncation function.

[0372] The activity cost after partition correction is:

[0373] ;

[0374] 5. Project activity cost summary and output:

[0375] After completing the line mechanism partitioning and partition correction, first calculate the total line partition correction amount for the current project:

[0376] ;

[0377] in, The total amount to be adjusted for the current project's line zones. For the first The operation cost adjustment item for each line mechanism partition after amplitude constraint. This represents the number of partitions for the current project's routing mechanism.

[0378] Based on this, the operating costs are determined by generating recommendations for current transmission line inspection and testing projects using modal summation.

[0379] ;

[0380] in, Determine the operating costs for the current recommended testing and inspection projects for transmission lines. To stabilize operating costs, To drive operating costs internally, For exogenous disturbance operating costs, The total amount of correction is adjusted for the line mechanism partitioning.

[0381] At the same time, the revised activity costs of each region are summarized to obtain the summarized results for each region:

[0382] ;

[0383] in, The operating costs are summarized by region according to the route mechanism. For the first The adjusted operating costs for each partition;

[0384] Perform dual-caliber consistency check: If and If the relative deviation does not exceed a preset threshold, the modal summation caliber is considered to be consistent with the partition summary caliber, and the output is... The activity cost is determined as the final recommendation; if the relative deviation between the two exceeds the preset threshold, the correction items of each partition are normalized and calibrated according to the modal summation result so that the final output maintains the same activity cost caliber.

[0385] After step S4 is completed, output the recommended activity cost for the current project. The document outlines the cost structure of three types of operations, the results of the line mechanism zoning, the cost adjustment items for each zone, the dominant cost type for each section, and the corresponding operation control classification. Specifically, stable baseline-dominated sections are used to implement standardized operation configurations; operation scale and risk-dominated sections are used to match testing personnel and equipment resources; and exogenous disturbance-sensitive sections are used to develop special operation plans and relocation organization schemes.

[0386] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0387] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining the modal partitioning of transmission line inspection operation costs, characterized in that, Includes the following steps: S1. Collect historical data of transmission line testing and inspection projects, clean the samples, unify the standards and build a hierarchical database, construct the cost sequence of historical projects, divide the line sections according to the line structure and the boundaries of testing operations, extract the workload characteristics, operation difficulty characteristics and external disturbance characteristics of the sections, form a section feature matrix, and construct a section relationship diagram through the adjacent connections, the same channel environment, the same risk type and the operation transfer relationship between sections. S2, based on the historical project cost sequence and segment feature matrix, divides the same sample groups according to detection type, voltage level and operation scenario, calculates the endogenous driving strength, exogenous disturbance strength and exogenous disturbance triggering flag of the project, decomposes the historical project approved operation cost into stable operation cost mode, endogenous driving operation cost mode and exogenous disturbance operation cost mode, and performs constraint optimization solution and decomposition result verification for each type of operation cost mode; S3. Match the project to be tested to the corresponding sample group of the same type, use the weighted robust statistical method to determine the stable operating cost, use the segment graph attention prediction network to determine the endogenous driving operating cost, use the disturbance factor triggered correction model to determine the exogenous disturbance operating cost, and determine the project cost dominant type according to the contribution rate of various operating cost modes. S4 allocates stable operating costs and exogenous disturbance operating costs to each line segment, and calculates the initial operating cost and modal contribution rate of the segment based on the segment-level endogenous driving operating costs. The line mechanism is partitioned according to the segment cost dominance type, segment characteristics and physical connectivity. The historical modal residuals of various operating cost modes of historical projects are calculated, and each partition is corrected based on the historical modal residuals. The results are summarized to obtain the suggested operating costs and complete the cost consistency verification.

2. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 1, characterized in that, S1 includes: S11: Collect operational data from historical transmission line testing projects, including line parameters, testing types, testing methods, on-site operation records, testing reports, settlement data, and verified operational costs. Perform sample access verification on the collected operational data, and perform duplicate data merging and deduplication based on project number, line name, testing time, and verified operational cost. Perform unified conversion on operational data with inconsistent units of measurement. Identify abnormal operational data by combining unit line length cost, unit tower cost, and unit testing point cost. Abnormal operational data formed by crossing high-speed railways, long-distance relocation in mountainous areas, extreme weather windows, or emergency testing operations are retained and marked as disturbance samples. Correct or remove abnormal operational data formed by input errors, duplicate collection, or inconsistent cost standards. Build a layered database according to testing type, voltage level, and operation scenario. S12, Based on historical project operation data, construct a historical project cost sequence using the verified operation cost of each historical transmission line testing and inspection project; S13. Based on historical project samples and according to the actual structure of the transmission line and the boundaries of the inspection operation, the transmission line is divided into line sections, including tower sections, tension sections, special crossing sections, and passageway hidden danger sections. Workload characteristics, operation difficulty characteristics, and external disturbance characteristics are extracted for each line section. The workload characteristics include section length, number of towers, number of inspection points, and inspection frequency. The operation difficulty characteristics include voltage level, terrain passage complexity, and defect risk level. The external disturbance characteristics include the degree of weather window limitation, relocation distance or relocation difficulty, market price fluctuation level, special crossing coordination difficulty, and operation urgency. A section feature matrix is ​​constructed according to the line sequence. S14. Using each line segment as a graph node, construct relationship edges based on the adjacent connection relationship, the same channel environment relationship, the same risk type relationship, and the operation transfer relationship between line segments. Calculate the comprehensive edge weight between segments based on the edge weights corresponding to each type of relationship edge to construct the segment relationship graph.

3. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 2, characterized in that, S2 includes: S21. Based on the historical project cost sequence and section feature matrix, the sample groups are divided into similar groups according to the detection type, voltage level and operation scenario. The verified operation cost of each historical project is decomposed into stable operation cost mode, endogenous driving operation cost mode, exogenous disturbance operation cost mode and decomposition residual. Among them, the stable operation cost mode is used to characterize the relatively stable operation cost in similar detection projects, the endogenous driving operation cost mode is used to characterize the operation cost change caused by the project's own factors, and the exogenous disturbance operation cost mode is used to characterize the operation cost change caused by external conditions. S22, based on the obtained sample groups of the same type, summarize the project-level features of each historical project, including project-level endogenous features and project-level exogenous features, normalize each project-level feature, and calculate the endogenous driving strength, exogenous disturbance strength, and exogenous disturbance triggering flag of each historical project. The endogenous driving strength is used to characterize the degree of driving effect of the project's own factors on the operating cost, the exogenous disturbance strength is used to characterize the degree of influence of external disturbance factors on the operating cost, and the exogenous disturbance triggering flag is used to determine whether the exogenous disturbance operating cost mode is triggered. S23. Based on various activity cost modes, endogenous driving strength, exogenous disturbance strength, and exogenous disturbance triggering identifiers, construct a constraint optimization objective that includes reconstruction error constraints, stability constraints, endogenous consistency constraints, and exogenous triggering constraints, and set non-negative constraints for various activity cost modes and triggering constraints for exogenous disturbance activity cost modes. S24. Based on the constructed constraint optimization objective, determine the initial value of the stable operation cost mode, and update the stable operation cost mode, the endogenous driven operation cost mode, and the exogenous disturbance operation cost mode in sequence. Update the proportional coefficients corresponding to the endogenous driving strength and the exogenous disturbance strength until the preset convergence condition is met, and complete the constraint optimization solution of various operation cost modes. S25. Based on the various activity cost modes obtained from the solution, calculate the decomposition residuals of each historical item, and perform residual verification according to the proportion of the decomposition residuals to the approved activity costs. Review the historical items whose residuals exceed the preset residual threshold, and retain the decomposition results for the historical items whose residuals do not exceed the preset residual threshold. S26. Based on the various activity cost modes, endogenous driving strength, exogenous disturbance strength and exogenous disturbance triggering flags obtained by solving, perform a mode consistency check, verify the correspondence between various activity cost modes and project workload, operation difficulty and external disturbance factors, and determine that various activity cost modes of each historical project meet the cost formation mechanism. S27 outputs the stable activity cost mode, endogenous driving activity cost mode, exogenous disturbance activity cost mode, decomposed residual, endogenous driving strength, exogenous disturbance strength, and exogenous disturbance trigger flag for each historical project.

4. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 3, characterized in that, S3 includes: S31. Based on the detection type, voltage level, and operating scenario of the test item, match the test item to the corresponding similar sample group, and extract the stable operating cost mode, endogenous driven operating cost mode, and exogenous disturbance operating cost mode obtained from the decomposition of the similar sample group as the basis for cost determination. S32, select historical projects from the same sample group where exogenous disturbances have not been triggered, endogenous driving strength is within a preset quantile range and data is complete, determine sample weights based on the time difference, feature distance and data reliability between historical projects and projects to be tested, and determine the stable operating cost of projects to be tested using the weighted truncated median. S33. Input the segment relationship graph of the project to be tested, as well as the workload and operation difficulty features of each segment, into the segment graph attention prediction network. Use adjacent connections, same channel environment, same risk type and operation transition relationship to perform graph encoding on the segment features, predict the segment-level endogenous driving operation cost, and sum the prediction results of each segment to obtain the endogenous driving operation cost of the project to be tested. S34. Calculate the exogenous disturbance intensity based on the degree of weather window limitation, relocation distance or relocation difficulty, market price fluctuation level, special crossing coordination difficulty and operational urgency of the project under test, determine the exogenous disturbance triggering indicator based on the exogenous disturbance triggering threshold, and determine the exogenous disturbance operation cost of the project under test using the disturbance factor triggering correction model. S35. Calculate the contribution rate of various activity cost modes based on the stable activity cost, endogenous driven activity cost and exogenous disturbance activity cost, and determine the project to be tested as stable baseline dominated, activity scale and risk condition dominated or exogenous disturbance sensitive based on the maximum contribution rate.

5. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 4, characterized in that, S33 includes: S331, Obtain the segment relationship diagram of the project to be tested. The segment relationship diagram includes a set of segment nodes, a set of segment relationship edges, and a segment feature matrix. Extract the segment length, number of towers, number of detection points, detection frequency, voltage level, terrain channel complexity, and defect risk level of each segment from the segment feature matrix, and construct the segment endogenous driving feature vector. S332, the segment relationship edges are divided into adjacent connection edges, same channel environment edges, same risk type edges, and operation transfer edges. The attention scores between adjacent segments under each type of segment relationship edge are calculated, and the attention scores are normalized to obtain attention weights. The features of adjacent segments are weighted and aggregated according to the attention weights to obtain the endogenous driving graph encoding results of each segment. S333: Input the endogenous driving graph encoding results of each segment into the prediction layer of the multilayer perceptron, and output non-negative segment-level endogenous driving operation costs through the Softplus activation function. Sum the segment-level endogenous driving operation costs to obtain the endogenous driving operation cost of the project to be tested. S334, the segment graph attention prediction network is trained using the historical endogenous driven operation cost mode as the supervised label, and prediction loss, monotonic constraint loss and graph smoothing constraint are constructed. The total loss function is formed by combining the network parameter regularization term and the network parameters of the segment graph attention prediction network are updated through the total loss function.

6. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 5, characterized in that, S34 includes: S341: Obtain the degree of weather window limitation, transfer distance or transfer difficulty, market price fluctuation level, special crossing coordination difficulty and operation urgency of the project to be tested, construct the project-level exogenous disturbance feature vector, and normalize each exogenous disturbance feature. S342, based on the determined weights of each exogenous disturbance characteristic, the normalized weather window restriction degree, transfer distance or transfer difficulty, market price fluctuation level, special crossing coordination difficulty and operational urgency are weighted and calculated to obtain the exogenous disturbance intensity of the project to be measured. S343, compare the exogenous disturbance intensity with the exogenous disturbance trigger threshold to determine the exogenous disturbance trigger flag. When the exogenous disturbance intensity is lower than the exogenous disturbance trigger threshold, exogenous disturbance correction is not triggered. When the exogenous disturbance intensity reaches or exceeds the exogenous disturbance trigger threshold, exogenous disturbance correction is triggered. S344, construct a three-segment disturbance correction function, and fit the correspondence between the historical exogenous disturbance operation cost mode and the exogenous disturbance intensity through the constrained nonlinear least squares method to determine the slight disturbance trigger threshold, strong disturbance trigger threshold, general disturbance correction intensity and strong disturbance correction intensity; S345, calculate the disturbance correction value based on the exogenous disturbance intensity, and multiply the stable operating cost of the project under test by the disturbance correction value to obtain the exogenous disturbance operating cost of the project under test.

7. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 6, characterized in that, S4 includes: S41, the determined stable operation costs and exogenous disturbance operation costs are allocated to each line section. The allocation weight of stable operation costs is determined according to the section length, the number of detection points, and the basic operation allocation items. The allocation ratio of exogenous disturbance operation costs is determined according to the intensity of exogenous disturbances in the section. Combined with the section-level endogenous driving operation costs, the initial operation costs of each section and the contribution rate of various operation cost modes are calculated. S42. Based on the contribution rates of various operation cost modes in each section and the preset difference threshold, determine the cost dominance type of each section, and verify the consistency of the correspondence between the cost dominance type of the section and the workload, operation difficulty, intensity of exogenous disturbance and project cost dominance type. For sections that do not meet the correspondence, readjust the corresponding modal contribution rate. S43, combine the basic features of the segment, the segment map coding results and the segment cost-dominant type to form the segment mechanism partitioning features, calculate the mechanism difference distance between segments, use the physical adjacency connection relationship as a constraint, perform initial coarse division on physically continuous segments with similar cost formation mechanisms, and iteratively adjust the partition boundaries through partition cohesion, cross-partition association penalty and partition number penalty to obtain the line mechanism partitioning results; S44, calculate the historical modal residuals between the various operation cost modes of historical projects and the corresponding backtest results, screen the matching historical modal residuals according to the detection type, voltage level, operation scenario, partition dominant type, channel complexity and defect risk level, determine the sub-modal correction coefficients of each line mechanism partition, and constrain the partition correction items according to the partition dominant type and the maximum correction ratio to obtain the corrected operation cost of each partition. S45 summarizes the operation cost correction items of each line mechanism zone, and adds them to the stable operation cost, endogenous driving operation cost, and exogenous disturbance operation cost of the project under test to obtain the suggested operation cost. At the same time, it summarizes the corrected operation costs of each zone, performs consistency verification between the modal summation calibrator and the zone summary calibrator, and performs normalization calibration on the correction items of each zone when the deviation exceeds the preset threshold. It outputs the suggested operation cost, the composition of various operation costs, the line mechanism zone results, the operation cost correction items of each zone, the cost dominance type of each section, and the corresponding operation control classification.

8. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 7, characterized in that, S43 includes: S431: Obtain the standardized basic characteristics of each segment, the segment map coding results, the contribution rate of various operation cost modes, the intensity of exogenous disturbances and the dominant type of segment cost, construct the segment mechanism partitioning characteristics, and calculate the mechanism difference distance between any two segments. S432, with physical adjacency as a constraint, scan each section along the actual sequence of the transmission line, divide the physically continuous sections with the same or similar cost-dominant type and mechanism difference distance less than the preset mechanism difference threshold into the same initial partition, set the partition boundary at the position where the mechanism difference distance reaches or exceeds the preset mechanism difference threshold, and obtain the initial partition result of the line. S433, Based on the initial partitioning results of the line, with the goals of optimal partition cohesion, minimum cross-partition association penalty, and minimum number of partitions, the partitions to which the boundary segment belongs are iteratively adjusted, and the connectivity of each partition in terms of physical adjacency is maintained during the iteration process, and the partitioning results of the line mechanism are updated. S434: Determine whether the relative change of the objective function corresponding to the circuit mechanism partitioning result is less than the preset change threshold or whether the number of iterations has reached the maximum number of iterations. If so, stop the iteration and output the circuit mechanism partitioning result. Otherwise, return to S433 to continue iterative optimization.

9. The method for determining the modal partitioning of transmission line inspection operation costs according to claim 8, characterized in that, S44 includes: S441, Obtain the stable activity cost mode, endogenous driving activity cost mode and exogenous disturbance activity cost mode of the historical projects obtained from the decomposition, and backtest each historical project using the methods for determining stable activity cost, endogenous driving activity cost and exogenous disturbance activity cost, calculate the historical deviation of each type of activity cost mode, and calculate the historical relative residual of each type of activity cost mode; S442, according to the detection type, voltage level, operation scenario, dominant type of partition, channel complexity and defect risk level, select various historical relative residual sets that match the current line mechanism partition, clean the outliers of the various historical relative residual sets, and remove historical relative residuals that are mainly random fluctuations; S443, the sample weights are determined based on the time distance, feature similarity and data credibility of the matching historical items. The weighted median is used to calculate the stable mode correction coefficient, endogenous mode correction coefficient and exogenous mode correction coefficient of the current line mechanism partition. S444, based on the partition dominance type of the current line mechanism partition, constrain the correction direction and correction magnitude of various modal correction coefficients, and calculate the partition correction item based on various modal correction coefficients and the corresponding partition operation cost mode, and apply a maximum correction ratio limit to the partition correction item to obtain the constrained partition correction item; S445, add the initial operation cost of the current line mechanism partition to the partition correction term after the constraint to obtain the operation cost after the current line mechanism partition correction.

Images