Training data acquisition method and system for power transmission line fault prediction model

Abstract

The invention relates to the technical field of meteorological data processing, in particular to a training data acquisition method and system for a power transmission line fault prediction model. The method comprises the following steps: constructing a multi-source heterogeneous original data pool containing severe convection meteorological data, power transmission line operation and fault data and geographic space data; performing time-space reference unified processing and grid division on the data; mapping the line paragraph to a grid unit, identifying a severe convection event and dividing an influence area through a density clustering algorithm based on a multi-parameter collaborative threshold value, calculating a grid exposure degree, and combining a strength grade weight and a terrain correction weight to calculate a space correlation strength; setting a differentiation time association window to calculate a time factor according to the severe convection dominant type, calculating association credibility based on the space association strength and the time factor, and screening high-credibility association results; and performing attribution labeling on the fault information to obtain a training data set. The problem of low training data quality is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological data processing, in particular to a training data acquisition method and system for a power transmission line fault prediction model. BACKGROUND

[0002] With the development of power systems towards high voltage, large capacity and wide coverage, the operation safety of power transmission lines is increasingly affected by strong convective weather. In order to reduce the harm of strong convective weather to power transmission lines, the industry generally realizes early warning of fault risk by building a fault rate prediction model, and high-quality training data is the core prerequisite for ensuring the accuracy and reliability of the prediction model.

[0003] In the prior art, the influencing factors of line faults under strong convective weather involve meteorological, line operation, geographical space and other multi-dimensional data, and the data format and storage standard are not unified, lacking effective integration mechanism, the data cannot be directly associated, resulting in incomplete data coverage and information fragmentation. In addition, the existing method lacks a precise spatio-temporal matching mechanism, making it difficult to accurately establish the association between strong convective events, line sections and fault time, and unable to effectively eliminate pseudo-associated data, ultimately resulting in low quality of training data, affecting the accuracy and reliability of the prediction model. SUMMARY

[0004] The present application provides a training data acquisition method and system for a power transmission line fault prediction model to solve the problem of low quality of training data in the prior art.

[0005] To achieve the above purpose, the present application provides the following technical solution: The present application provides a training data acquisition method and system for a power transmission line fault prediction model to solve the problem of low quality of training data in the prior art. S100: building a multi-source heterogeneous original data pool, the multi-source heterogeneous original data pool comprising strong convective weather data, power transmission line operation and fault data and geographical space data; S200: performing spatio-temporal reference unified processing on the data in the multi-source heterogeneous original data pool, and performing grid division based on a preset spatio-temporal granularity to obtain spatio-temporal grid data; S300: splitting the power transmission line into line sections according to the latitude and longitude of the tower and mapping them to the corresponding grid cells of the spatio-temporal grid data; identifying strong convective events and dividing the influence area, and assigning an intensity level weight to each influence area; calculating the grid exposure degree of the line section, and calculating the spatial correlation strength in combination with the intensity level weight and the terrain correction weight extracted based on the geographical space data; S400: calculating a time factor according to the time relationship between the fault time and the strong convective occurrence time; calculating the correlation confidence by weighted summation based on the spatial correlation strength and the time factor, and screening the correlation results with a confidence higher than a preset threshold as the correlation matching data; S500: Attributing annotation is made to the power transmission line fault information in the correlation matching data, and a training data set is obtained, and the annotation content includes the corresponding relationship of the fault type, the strong convection inducement type and the characteristic parameter.

[0006] As a preferred technical solution of the present application, the identification of the strong convection event and the division of the influence area include: Based on the radar reflectivity factor, satellite cloud top brightness temperature and ground automatic station observation data, a multi-parameter cooperative threshold is set; By the density clustering algorithm, the adjacent grids meeting the multi-parameter cooperative threshold are aggregated, and the strong convection event is identified; According to the outward decreasing intensity distribution of the core area of the strong convection event, the core influence area, the significant influence area and the peripheral influence area are divided from inside to outside, and the corresponding intensity level weight is given to each influence area.

[0007] As a preferred technical solution of the present application, the calculation of the grid exposure degree of the line section includes: The power transmission line is split into basic line sections according to the tower coordinates, and according to the key geographical features in the line corridor, the line sections are subdivided again to generate a plurality of fine line sections; The fine line sections are spatially cut with the spatio-temporal gridding data to obtain sub-line sections completely falling into a single grid unit; The weighted length is calculated according to the length of each sub-line section and the attribute parameters of the line; The weighted lengths of all sub-line sections falling into the same grid are summed up, and then divided by the grid side length to obtain the grid exposure degree of the grid.

[0008] As a preferred technical solution of the present application, the terrain correction weight extracted based on geographic spatial data is obtained by the following way: The slope, slope direction and relative altitude parameters of the grid are extracted from the digital elevation model data; The weight coefficients of the slope, slope direction and relative altitude parameters are fitted based on historical fault data; The slope, slope direction and relative altitude parameters are weighted and summed up according to the corresponding weight coefficients, and then normalized to obtain the terrain correction weight.

[0009] As a preferred technical solution of the present application, the calculation of the spatial correlation strength includes: The line corridor environment weight is obtained, and the line corridor environment weight is obtained by weighted summation based on the threat degree of trees, the threat degree of buildings and the terrain channel index in the line corridor; The strength level weight, the terrain correction weight and the line corridor environment weight are multiplied to obtain the spatial comprehensive weight of the grid; The grid exposure of each grid through which the line section passes is multiplied by the spatial synthesis weight of the corresponding grid, and then accumulated to obtain the spatial correlation strength of the line section.

[0010] As a preferred technical solution of the present application, the calculation of the time factor according to the time relationship between the fault time and the time when the strong convection occurs comprises: Setting a differentiated time correlation window according to the dominant type of the strong convection event; Calculating the time difference between the line fault time and the strong convection peak time; Determining the time factor according to the inverse ratio of the time difference to the total length of the time correlation window.

[0011] As a preferred technical solution of the present application, the attribution labeling comprises: Labeling the fault type based on the operation and maintenance records, fault recording waveforms or on-site investigation results in the power transmission line fault data; Labeling the strong convection inducement type in combination with the dominant features of the strong convection event in the correlation matching data; Extracting the meteorological parameter peak corresponding to the strong convection inducement type as the feature parameter.

[0012] The present application also proposes a training data acquisition system for a power transmission line fault prediction model, comprising: A data pool construction module for constructing a multi-source heterogeneous original data pool, the multi-source heterogeneous original data pool comprising strong convection meteorological data, power transmission line operation and fault data and geographic spatial data; A data preprocessing module for performing time-space benchmark unified processing on the data in the multi-source heterogeneous original data pool, and performing grid division based on a preset time-space granularity to obtain time-space grid data; A spatial correlation matching module for splitting the power transmission line according to the tower longitude and latitude into line sections and mapping them to the corresponding grid units of the time-space grid data; identifying strong convection events and dividing the influence areas, and assigning intensity level weights to each influence area; calculating the grid exposure of the line section, and calculating the spatial correlation strength in combination with the intensity level weight and the terrain correction weight extracted based on the geographic spatial data; A time correlation screening module for calculating a time factor according to the time relationship between the fault time and the time when the strong convection occurs; calculating the correlation confidence by weighted summation based on the spatial correlation strength and the time factor, and screening the correlation results with a confidence higher than a preset threshold as the correlation matching data; An attribution labeling module for attributing labeling of the power transmission line fault information in the correlation matching data to obtain a training data set, the labeling content comprising the corresponding relationship of the fault type, the strong convection inducement type and the feature parameter thereof.

[0013] The application further provides a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the training data acquisition method of the power transmission line fault prediction model when executing the computer program.

[0014] The application further provides a readable storage medium, characterized in that the readable storage medium stores a computer program, and the computer program implements the training data acquisition method of the power transmission line fault prediction model when executed by a processor.

[0015] The application has the following beneficial effects: 1. The application realizes multi-dimensional quantitative matching of line sections and strong convective events by dividing the strong convective influence area and assigning intensity level weights, combining grid exposure, terrain correction weights and line corridor environment weights to calculate spatial correlation strength.

[0016] 2. The application realizes accurate matching and screening in time and space dimensions by setting differentiated time correlation windows according to the dominant type of strong convection and calculating time factors, combining spatial correlation strength and strong convection peak intensity to calculate correlation reliability. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, which together with the embodiments of the application, serve to explain the application, and do not constitute a limitation on the application. Figure 1 is a flowchart of the training data acquisition method of the power transmission line fault prediction model of the application; Figure 2 is a structural schematic diagram of the training data acquisition system of the power transmission line fault prediction model of the application. DETAILED DESCRIPTION

[0018] The preferred embodiments of the application will be described below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and do not limit the application.

[0019] Embodiment one: as shown in the figure, the training data acquisition method of the power transmission line fault prediction model of the application includes: Figure 1 S100: constructing a multi-source heterogeneous original data pool, the multi-source heterogeneous original data pool including strong convective weather data, power transmission line operation and fault data and geographic spatial data; S200: constructing a strong convective event database, the strong convective event database including strong convective event information and strong convective event influence area information; Specifically, in the present embodiment, constructing a multi-source heterogeneous raw data pool is to integrate key information from different sources with different data types and formats into a unified initial data set to support the training of a strong convective weather under transmission line fault rate prediction model. The multi-source heterogeneous raw data pool includes strong convective weather data, transmission line operation and fault data, and geographic spatial data.

[0020] The strong convective weather data mainly covers radar monitoring data, satellite remote sensing data, and ground automatic station measured data from the meteorological department, which are collectively used to accurately capture the occurrence period, impact range, and intensity characteristics of strong convective weather. Specifically, the radar monitoring data includes radar reflectivity factor (used to represent precipitation intensity and cloud cluster structure), Doppler radial wind speed data, etc.; satellite remote sensing data includes satellite cloud top brightness temperature data (used to identify the vertical development intensity of deep convective clouds), infrared and visible light channel images, etc.; ground automatic station measured data includes real-time observation values of ground wind speed, precipitation, air temperature, air pressure, and other conventional meteorological elements. In addition, it also includes lightning location system recorded cloud-to-ground lightning data, including time, location, intensity, and other information of lightning occurrence.

[0021] The transmission line operation and fault data come from the power grid dispatching and operation management system, which directly relates to the running state of the transmission line and the specific situation of the fault occurrence. Specifically, the transmission line basic information includes line number, voltage level, line direction, tower number, tower longitude and latitude coordinates, tower type, conductor type, insulator type, and other static parameters; the line operation data includes line load, current, voltage, power, and other real-time operation parameters; the fault record data includes fault occurrence time (accurate to minutes or seconds), fault line number, fault type (such as tripping, wire breakage, flashover, etc.), fault recording waveform data, protection action information, and post-fault site investigation report, etc.

[0022] The geographic spatial data comes from the geographic information service platform, mainly including digital elevation model data, high-resolution remote sensing image data, and line corridor special environment data. These data are used to analyze the influence of geographic environment on the intensity of strong convective weather and the exposure risk of the line under different geographic environments. Specifically, the digital elevation model data provides elevation information of the regional terrain, including elevation, slope, slope direction, and other terrain parameters; the high-resolution remote sensing image data is used to identify the land cover type, including vegetation distribution, building location, water body distribution, etc.; the line corridor special environment data includes tree height and distribution, building height and distance, terrain channel characteristics, and other environment information directly related to line safety within the line corridor.

[0023] By integrating the above three types of data into a multi-source heterogeneous raw data pool, it is ensured that the subsequent training data acquisition process can be based on all key influencing factors, thereby guaranteeing the comprehensive coverage and effectiveness of the final training data set.

[0024] S200: performing space-time reference unified processing on the data in the multi-source heterogeneous raw data pool, and performing grid division based on a preset space-time granularity to obtain space-time gridded data; Specifically, first, the time reference is unified, and the UTC time of the weather radar data, the local time of the power transmission line fault data, and the device recording time of the lightning location data are uniformly converted into Coordinated Universal Time or Local Standard Time. For data with different sampling frequencies, time interpolation or resampling processing is performed to make the time resolution consistent.

[0025] Subsequently, the space reference is unified, and the Gauss-Kruger projection adopted by the weather grid, the WGS-84 longitude and latitude of the power transmission line tower, and the local coordinate system of the geographic terrain data are converted into the WGS-84 coordinate system. The position deviation caused by different projection methods is corrected through coordinate conversion algorithm.

[0026] According to the influence range, change speed of severe convective weather, and spatial distribution characteristics of the power transmission line, a unified time interval and space unit size are preset. The time granularity can be selected as 5 minutes or 10 minutes as a time slice, and the space granularity can be selected as 1 km x 1 km or 2 km x 2 km as a grid unit.

[0027] The data that have undergone space-time reference unification are mapped to the corresponding space-time grid according to the preset granularity, so that each grid unit contains all weather data, line data, and geographic data within the time period and the space range, and space-time gridded data is obtained.

[0028] S300: splitting the power transmission line into line segments according to the longitude and latitude of the tower and mapping them to the corresponding grid units of the space-time gridded data; identifying severe convective events and dividing the influence area, and assigning a strength level weight to each influence area; calculating the grid exposure degree of the line segment, and calculating the spatial correlation strength in combination with the strength level weight and the terrain correction weight extracted based on geographic spatial data; Specifically, the power transmission line is split into line segments according to the longitude and latitude of the tower. Specifically, the longitude and latitude coordinates of two adjacent towers are taken as the starting point and the ending point, and the entire line is split into several segments of tower-to-tower line segments.

[0029] According to the longitude and latitude range of each line segment, the corresponding grid unit in the space-time gridded data is determined. By comparing the longitude and latitude of the starting point, ending point, and key nodes in the line segment with the spatial boundaries of each grid unit, it is determined whether the line segment falls completely or partially into the grid, and the accurate spatial correspondence between the line segment and the space-time grid is realized.

[0030] Further, the identifying the strong convective event and dividing the influence area comprises: setting a multi-parameter collaborative threshold based on radar reflectivity factor, satellite cloud top brightness temperature and ground automatic station observation data; identifying the strong convective event by density clustering algorithm to aggregate adjacent grids meeting the multi-parameter collaborative threshold; dividing the core influence area, the significant influence area and the peripheral influence area from inside to outside according to the outwardly decreasing intensity distribution of the core area of the strong convective event, and giving corresponding intensity level weights to each influence area.

[0031] Specifically, the multi-parameter collaborative threshold is set based on radar reflectivity factor, satellite cloud top brightness temperature and ground automatic station observation data.

[0032] The radar reflectivity factor is used to capture the concentration and size of the precipitation particles in the strong convective cloud cluster, and the threshold such as Z≥45dBZ indicates the strong thunderstorm core; the satellite cloud top brightness temperature is used to identify the vertical development intensity of the deep convective cloud, and the threshold such as TBB≤-50℃ represents the strong convective cloud cluster; the ground automatic station data is used to verify the actual influence of the strong convective on the ground, and the threshold such as maximum wind speed V≥17.2m / s and 1-hour rainfall R≥20mm. The three form a screening standard in collaboration, and only when the grid meets multiple parameter thresholds at the same time, it is determined as a strong convective influence grid, avoiding misjudgment of non-strong convective weather by a single parameter.

[0033] The adjacent grids meeting the multi-parameter collaborative threshold are aggregated by a density clustering algorithm (such as DBSCAN algorithm). Specifically, the spatial neighborhood radius and the minimum cluster point number are set, and the grids that are continuous in space and meet the intensity are classified into the same group, thereby identifying the single-field independent strong convective event and distinguishing different sessions and discontinuous strong convective.

[0034] The core influence area, the significant influence area and the peripheral influence area are divided from inside to outside according to the outwardly decreasing intensity distribution of the core area of the strong convective event. Specifically: the core influence area is the area with the highest intensity of the strong convective and the greatest threat to the line, such as the grid with radar reflectivity factor≥50dBZ; the significant influence area is the area with the second highest intensity but still with obvious threat, such as the grid with radar reflectivity factor between 45dBZ and 50dBZ; the peripheral influence area is the area with weaker intensity and lower threat, such as the grid with radar reflectivity factor between 40dBZ and 45dBZ.

[0035] Corresponding intensity level weights are given to each influence area. For example, the core influence area weight is 1.0, the significant influence area weight is 0.7, and the peripheral influence area weight is 0.4. The specific value of the weight can be obtained based on statistical analysis of historical failure data.

[0036] Further, the calculation of the grid exposure degree of the line section includes: The power transmission line is split into basic line segments according to the tower coordinates, and secondary subdivision is performed according to the key geographical features in the line corridor to generate a plurality of fine line segments; The fine line segments are spatially cut with the spatio-temporal grid data to obtain sub-line segments that completely fall into a single grid cell; The weighted length is calculated according to the length of each sub-line segment and the attribute parameters of the line to which it belongs; The weighted lengths of all sub-line segments falling into the same grid are summed up and divided by the grid side length to obtain the grid exposure degree of the grid.

[0037] Specifically, after the power transmission line is split into basic line segments according to the tower coordinates, secondary subdivision is performed according to the key geographical features in the line corridor to generate a plurality of fine line segments.

[0038] The key geographical features include: terrain transition features such as transition zones between plains and mountains, valley entrance and exit positions; environmental mutation features such as boundaries between dense forest areas and open farmland, and boundaries between towns and suburbs; and strong convective sensitive terrain features such as narrow passages between two mountains and historical thunderstorm high-frequency areas.

[0039] Through GIS spatial overlay analysis, the identified key geographical features are spatially matched with the basic line segments to determine the specific coordinate positions of the features on the line segments. Taking the feature coordinates as the center, each extending a predetermined distance (such as 50 meters) to the ends of the line segment as the subdivision endpoints, the basic line segments are split into a plurality of fine line segments, ensuring that each fine line segment only covers a single type of key geographical feature.

[0040] The fine line segments are spatially cut with the spatio-temporal grid data. Using GIS spatial analysis tools, the intersection coordinates of the fine line segments and the grid boundaries are calculated, and the fine line segments that may span multiple grids are split into shorter sub-line segments, ensuring that each sub-line segment completely falls into a single grid cell.

[0041] The weighted length is calculated according to the length of each sub-line segment and the attribute parameters of the line to which it belongs. The attribute parameters of the line are used to represent the importance of the power transmission line, including but not limited to line voltage level, line load level, and power supply area importance, etc. For example, according to the line voltage level, the weight of 1000kV ultra-high voltage line is 1.2, the weight of 500kV super-high voltage line is 1.0, and the weight of 220kV high voltage line is 0.8; according to the line load level, the weight of special class load line is 1.2, the weight of first class load line is 1.0, and the weight of second class load line is 0.8. The weighted length calculation formula of each sub-line segment is: weighted length = actual length of sub-line segment x weight coefficient of the line to which it belongs.

[0042] The weighted lengths of all sub-line segments falling into the same grid are summed to obtain the total weighted exposure length of the line in the grid. The total weighted exposure length is divided by the grid side length to obtain the grid exposure degree of the grid. For example, if the grid side length is 1 km and the total weighted length of the sub-line segments in the grid is 800 m, then the grid exposure degree of the grid is 0.8.

[0043] Further, the terrain correction weight extracted based on the geographic spatial data is obtained by the following method: The slope, aspect and relative elevation parameters of the grid are extracted from the digital elevation model data; The weight coefficients of the slope, aspect and relative elevation parameters are fitted based on historical fault data; The slope, aspect and relative elevation parameters are weighted and summed according to the corresponding weight coefficients, and normalized to obtain the terrain correction weight.

[0044] Specifically, the slope, aspect and relative elevation parameters of the grid are extracted from the digital elevation model (DEM) data. The slope is obtained by calculating the gradient change rate of the elevation in the grid, reflecting the steepness of the terrain; the aspect is obtained by calculating the direction of the elevation gradient, reflecting the slope direction; the relative elevation is the difference between the elevation in the grid and the average elevation of the surrounding area, reflecting the protrusion degree of the terrain.

[0045] The weight coefficients of the slope, aspect and relative elevation parameters are fitted based on historical fault data. By statistically analyzing the line fault rate corresponding to different terrain parameters under historical severe convective weather, the weight coefficients of each parameter are fitted by regression analysis method .

[0046] The slope, aspect and relative elevation parameters are weighted and summed according to the corresponding weight coefficients: ; The above initial value is normalized by dividing by the maximum initial value in the region to obtain the terrain correction weight in the range of 0~1.

[0047] Further, the calculation of the spatial correlation strength includes: Obtaining a line corridor environment weight, the line corridor environment weight being obtained by weighted summation based on the threat degree of trees, the threat degree of buildings and the terrain channel index in the line corridor; Multiplying the intensity level weight, the terrain correction weight and the line corridor environment weight to obtain the spatial comprehensive weight of the grid; Multiplying the grid exposure degree of each grid through which the line section passes and the spatial comprehensive weight of the corresponding grid and then accumulating to obtain the spatial correlation strength of the line section.

[0048] Specifically, the environmental weight of the route corridor is obtained by weighted summation of the threat level of trees, the threat level of buildings, and the terrain access index within the route corridor.

[0049] Identify trees along the route corridor and extract the height of each tree. and horizontal distance to the line The threat level of a single tree is calculated using the following nonlinear function: ; in, Tree height, The horizontal distance from the tree to the line. Set the high threshold for the line safety tree (e.g., 15m). The safe distance threshold (e.g., 30m). This is an adjustment factor for the influence of tree height (e.g., 1.5). This is a distance-related adjustment factor (e.g., 2.0). It is a natural exponential function.

[0050] The total threat value is obtained by summing the threat levels of all trees in the corridor. It is then normalized based on the maximum total threat value of the same type of corridor in history to obtain the forest impact weight between 0 and 1.

[0051] Building threat level calculation involves identifying buildings within the route corridor and extracting the height and horizontal distance of each building to the route. A threshold function is set based on the distance and height; buildings with a height exceeding a certain threshold and a distance to the route less than a safe distance are assigned a higher threat weight, while those with a lower threshold are assigned a lower weight. The weights are then accumulated and normalized to obtain the building impact weights.

[0052] The terrain access index is calculated based on digital elevation model data. Cross sections are taken at fixed intervals (e.g., 500m) along the route, and the width W, relative height H, slope S, and terrain continuity C parameters of each cross section are extracted. First, each parameter is normalized. ; ; ; in, , These represent the maximum and minimum cross-sectional widths within the region. This represents the maximum relative height between the two sides. The maximum slope on both sides is determined by statistically analyzing historical data. Regarding the width... Inverse normalization is used, so that the smaller the width (the narrower the terrain), the larger the normalization value. C is the terrain continuity parameter, which is already normalized (value 0-1).

[0053] The terrain channel index is calculated by a weighted sum formula: are weight coefficients of each parameter, which can be determined according to statistical analysis of the correlation between line faults and terrain features under historical severe convective weather, for example The higher the index, the more likely the terrain forms a channel effect. The index is normalized and mapped to a weight range of 1.0-1.5 to obtain the terrain channel effect weight.

[0054] By corresponding weight coefficients to the forest influence weight, the building influence weight and the terrain channel effect weight, the line corridor environment weight is calculated by a weighted sum formula, wherein satisfies and is obtained by fitting historical data.

[0055] The spatial comprehensive weight of the grid is obtained by multiplying the intensity level weight, the terrain correction weight and the line corridor environment weight.

[0056] The spatial correlation contribution value of each grid to the line paragraph is obtained by multiplying the grid exposure of each grid through which the line paragraph passes and the spatial comprehensive weight of the corresponding grid. The spatial correlation strength of the line paragraph is obtained by adding the contribution values of all the passing grids: is the spatial correlation strength of the line paragraph,

[0057] The spatial correlation strength quantifies the comprehensive degree of the line paragraph affected by severe convection in the spatial dimension, which is one of the key innovative points of the present application to realize precise spatiotemporal matching.

[0058] S400: calculating a time factor according to the time relationship between the fault time and the severe convective occurrence time; calculating an association credibility by weighted sum based on the spatial correlation strength and the time factor, and screening the association result with a credibility higher than a preset threshold as the association matching data; Further, the calculation of the time factor according to the time relationship between the fault time and the severe convective occurrence time comprises: ​​​​​​​​​​​​​​​According to the dominant type of the severe convective event, a differentiated time correlation window is set; The time difference between the line fault time and the severe convective peak time is calculated; According to the inverse ratio of the time difference to the total duration of the time correlation window, the time factor is determined.

[0059] Specifically, according to the dominant type of the severe convective event, a differentiated time correlation window is set. Different types of severe convective weather have significant differences in the mechanism and time characteristics of their effects on power transmission lines, so different time windows need to be set.

[0060] Lightning-dominant type: Lightning has an instantaneous impact on the line, so a narrow time window is set, such as the fault time ± 5 minutes; Wind-dominant type: Wind needs to accumulate to cause line dancing and tower stress deformation, so a wide time window is set, such as 20 minutes before the fault time to 10 minutes after the fault time; Short-duration heavy precipitation-dominant type: Heavy precipitation needs to penetrate and accumulate to cause tower foundation instability or geological disasters, so a wider time window is set, such as 30 minutes before the fault time to 20 minutes after the fault time.

[0061] The dominant type of the severe convective event is determined according to the peak value of each meteorological parameter in the event. For example, if the lightning location system records the highest flash density, it is determined to be a lightning-dominant type; if the ground automatic station records the most prominent maximum wind speed, it is determined to be a wind-dominant type.

[0062] Extract the fault time of the power transmission line This time is obtained from the fault record data, accurate to the minute or second level.

[0063] Extract the peak time of the severe convective event This time is the time when each meteorological parameter (such as radar reflectivity factor, ground wind speed, and precipitation intensity) of the severe convective event reaches the maximum value.

[0064] Calculate the time difference between the fault time and the severe convective peak time: ; According to the inverse ratio of the time difference to the total duration of the time correlation window, the time factor is determined. Specifically, the calculation formula of the time factor is: ; ; Where, is the time difference between the fault time and the severe convective peak time, is the total duration of the time correlation window (unit: minutes). When , the time factor , indicating that the time correlation is very weak; when , the time factor , indicating that the time correlation is the strongest.

[0065] For example, for lightning-dominated severe convective events, the time window is 30 minutes, if the time difference between the fault time and the peak time of the severe convective event is 30 minutes, then the time factor .

[0066] Based on the time factor , the spatial correlation strength , and the ratio of the peak intensity of the severe convective event to the threshold value , the correlation confidence between the severe convective event and the line fault is calculated by a weighted sum formula: ; wherein: is the correlation confidence; is the time factor, calculated by step 1; is the spatial correlation strength, calculated by step S300; is the ratio of the peak intensity of the severe convective event to the threshold value, and the calculation formula is , wherein is the peak value of the dominant parameter in the severe convective event (such as the maximum value of the radar reflectivity factor), is the threshold value of the parameter (such as 45 dBZ); , , are weight coefficients, which are obtained by regression fitting of the labeled data of severe convective induced faults in historical fault samples.

[0067] Set the confidence threshold , and only keep the correlation results with a confidence higher than the threshold as the correlation matching data. The setting of the confidence threshold needs to balance the accuracy and recall rate, and the optimal value is determined through historical data verification. Specifically, calculate the correlation confidence of the historical fault samples according to the above method, draw the accuracy-recall rate curve (PR curve) under different thresholds, and select the threshold corresponding to the maximum F1-Score as . For example, through verification, it is determined that .

[0068] For each line fault record, calculate the correlation confidence with all candidate severe convective events. If the correlation confidence of a severe convective event with the fault is , the correlation record is kept; if If the correlation record is removed, the strong convection event is considered to be insufficiently associated with the fault, and may be a false association (such as a line fault caused by equipment aging, but mistakenly associated with a strong convection event at a remote place at the same time).

[0069] After the credibility screening, the obtained correlation matching data contains the following information: line segment information including line number, tower number, and segment latitude and longitude range; strong convection event information including event number, dominant type, affected grid set, peak time, and peak intensity; fault information including fault time, fault type preliminary record; and correlation parameters including spatial correlation strength , time factor , and correlation credibility .

[0070] The correlation matching data realizes accurate correspondence of the transmission line, strong convection weather condition, and fault event in the time and space dimensions, and provides a high-quality correlation basis for subsequent fault attribution labeling.

[0071] By introducing the quantitative calculation of the time factor and the correlation credibility, the false correlation is effectively removed, and the accuracy and reliability of the training data are ensured.

[0072] S500: Attributing labeling is performed on the transmission line fault information in the correlation matching data to obtain a training data set, and the labeling content includes a corresponding relationship of a fault type, a strong convection cause type, and a characteristic parameter.

[0073] Further, the attribution labeling includes: Based on the operation and maintenance record, the fault recording waveform, or the field investigation result in the transmission line fault data, the fault type is labeled. In combination with the dominant features of the strong convection event in the correlation matching data, the strong convection cause type is labeled. The peak value of the meteorological parameter corresponding to the strong convection cause type is extracted as the characteristic parameter.

[0074] Specifically, based on the operation and maintenance record, the fault recording waveform, or the field investigation result in the transmission line fault data, the fault type is labeled.

[0075] Extract the protection action information, trip record, equipment state change and other operation and maintenance record data at the time of fault occurrence from the power grid dispatching system to preliminarily judge the fault type. For example, if the record shows "line protection tripping, reclosing success", it is determined to be a transient fault, and if the record shows "line protection tripping, reclosing unsuccessful", it is determined to be a permanent fault. At the same time, analyze the fault recording waveform characteristics. If the waveform shows a transient large current impact and the duration is extremely short (millisecond level), it is determined to be a lightning strike fault, and if the waveform shows a slow rise in current and then a sudden change, it may be a conductor breakage fault. Combined with the direct observation of equipment damage during on-site investigation, if discharge marks are found on the surface of the insulator, it is determined to be a flashover fault, and if the tower is found to be tilted or collapsed, it is determined to be a tower structure fault.

[0076] According to the above basis, the fault type is classified into line tripping (including transient tripping and permanent tripping), conductor breakage (including single strand breakage and multi-strand breakage), insulator flashover (including porcelain insulator flashover and composite insulator flashover), tower tilt or collapse, arrester damage and other fault types, ensuring that each fault record has a clear and unique fault type label.

[0077] According to the peak value of each meteorological parameter in the strong convective event and the corresponding relationship of the fault type, the key inducement of the fault is determined. If the radar reflectivity factor reaches the strong thunderstorm standard (such as Z≥50dBZ) in the strong convective event, and the lightning location system records ground flash data, the flash position is within the influence range of the fault line segment (such as the distance from the line is less than 500 meters), and the fault recording shows a transient large current impact feature, then the inducement type is determined to be lightning. If the maximum wind speed recorded by the ground automatic station reaches the strong wind standard (such as V≥25m / s) in the strong convective event, and the fault type is conductor breakage, tower tilt or conductor dancing induced tripping, and on-site investigation finds wind-induced damage marks, then the inducement type is determined to be strong wind. If the 1-hour precipitation reaches the heavy rain standard (such as R≥50mm) in the strong convective event, and the fault type is tower tilt or foundation settlement, and on-site investigation finds soil saturation or geological landslide signs, then the inducement type is determined to be short-time heavy rainfall. If the weather record shows that there is hail weather in the strong convective event, and the fault type is insulator damage or conductor damage, then the inducement type is determined to be hail.

[0078] According to the above determination basis, the strong convective inducement type is divided into lightning (including direct lightning and induced lightning), strong wind (including thunderstorm wind and downburst), short-time heavy rainfall (including local heavy rain and persistent heavy rainfall), hail (including small hail and large hail) and composite inducement (multiple strong convective factors acting together, such as lightning + strong wind, short-time heavy rainfall + strong wind, etc.).

[0079] For lightning inducement, extract the peak value of radar reflectivity factor (unit: dBZ), the peak value of ground flash density (unit: times / km2 / hour), peak lightning intensity (unit: kA), cloud top brightness temperature minimum (unit: °C). For the wind gust cause, the maximum wind speed peak (unit: m / s), wind speed duration (unit: minutes), radar reflectivity factor peak , 10-minute average wind speed (unit: m / s). For the short-time heavy precipitation cause, the 1-hour precipitation peak (unit: mm), 10-minute precipitation peak (unit: mm), precipitation duration (unit: minutes), radar reflectivity factor peak . For the hail cause, the hail diameter (unit: cm), cloud top brightness temperature minimum , radar reflectivity factor peak , hail duration (unit: minutes).

[0080] The characteristic parameters are extracted from the meteorological data of the space-time grid where the corresponding severe convective event of the associated matching data is located. Specifically, for the radar data, the maximum reflectivity factor value of the grid in the time slice corresponding to the failure time is extracted, for the ground automatic station data, the maximum wind speed or maximum precipitation recorded by the station in the time window before and after the failure time is extracted, and for the lightning location data, the number of lightning strikes and the maximum lightning intensity in the grid range within the time window before and after the failure time are counted.

[0081] After the above attribution labeling, each associated matching data is converted into a complete training sample. Each sample contains two parts of input features and labeled labels, wherein the input features include line section attributes (line number, voltage level, tower type, conductor type, insulator type, facility importance weight), geographical environment features (grid exposure, terrain correction weight, line corridor environment weight), severe convective event features (dominant type, impact area classification, intensity level weight, peak time), meteorological parameter features (radar reflectivity factor peak, cloud top brightness temperature, maximum wind speed, precipitation, lightning density, etc.), and space-time association features (spatial association strength, time factor, association credibility), and the labeled labels include failure type label, severe convective cause type label and characteristic parameter value.

[0082] All training samples are summarized to form a training data set, which can be directly used to train the prediction model of the fault rate of the power transmission line under the severe convective weather. Each sample in the data set has a clear "input feature-fault result-cause attribution" corresponding relationship, which ensures that the model can learn the internal correlation between the severe convective weather conditions, geographical environmental factors and the line fault. Through the systematic attribution labeling method, the original multi-source heterogeneous data is converted into high-quality structured training samples, providing reliable data support for the fault prediction model.

[0083] Embodiment two: In order to improve the accuracy of the fault prediction model of the power transmission line under the severe convective weather, a power grid company needs to obtain high-quality training data. For the high-occurrence season of severe convective weather in June to August 2025, the system for obtaining training data of the fault prediction model of the power transmission line is adopted, which comprises: A data pool construction module is used to construct a multi-source heterogeneous original data pool, which comprises severe convective weather data, power transmission line operation and fault data and geographical space data. A data preprocessing module is used to perform time-space benchmark unified processing on the data in the multi-source heterogeneous original data pool, and grid division based on a preset time-space granularity is performed to obtain time-space grid data. A spatial correlation matching module is used to split the power transmission line into line paragraphs according to the tower latitude and longitude and map them to the corresponding grid units of the time-space grid data; identify the severe convective event and divide the influence area, and assign a strength grade weight to each influence area; calculate the grid exposure degree of the line paragraph, and calculate the spatial correlation strength in combination with the strength grade weight and the terrain correction weight extracted based on the geographical space data. A time correlation screening module is used to calculate a time factor according to the time relationship between the fault time and the severe convective occurrence time; calculate the correlation confidence by weighted summation based on the spatial correlation strength and the time factor, and screen the correlation results with a confidence higher than a preset threshold as the correlation matching data. An attribution labeling module is used to attribute label the power transmission line fault information in the correlation matching data to obtain a training data set, and the labeling content comprises the corresponding relationship of the fault type, the severe convective cause type and the characteristic parameters thereof.

[0084] Specifically, the system collects relevant data of the 500kV and above voltage level power transmission line in the province, including radar site data, satellite cloud image data, ground automatic station observation data and lightning positioning system data provided by the meteorological department, power transmission line operation data and fault records provided by the power grid company, and digital elevation model data and line corridor environment data provided by the geographic information platform.

[0085] The data preprocessing module unifies the time reference of all data to UTC+8 standard time, converts the spatial coordinates to the WGS-84 coordinate system, and performs grid division according to a 5-minute time granularity and a 1km*1km spatial granularity. The spatial correlation matching module performs fine splitting processing on the power transmission line, performs secondary splitting by identifying the terrain conversion features, environmental mutation features and strong convective sensitive terrain features in the line corridor, identifies the strong convective event and divides the core influence area, the significant influence area and the peripheral influence area, and assigns the intensity level weights 1.0, 0.7 and 0.4 respectively, and calculates the spatial correlation strength of each line section by combining the grid exposure, the terrain correction weight and the line corridor environment weight.

[0086] The time correlation screening module sets a differential time window according to the dominant type of strong convection, wherein the lightning dominant type sets a ±5-minute window, the strong wind dominant type sets a 20-minute-before-to-10-minute-after window, and the short-time heavy precipitation dominant type sets a 30-minute-before-to-20-minute-after window. By calculating the time factor and the correlation reliability, the reliability threshold is set to 0.6, and the high-reliability correlation matching data is successfully screened out, effectively eliminating the pseudo-correlation between the faults and the strong convective events caused by the aging of the line equipment and other non-strong convective factors, and the weakly correlated data with a large time difference.

[0087] The attribution labeling module systematically labels the correlation matching data, clearly distinguishes the lightning cause (the main fault types are instantaneous trip and insulator flashover), the strong wind cause (the main fault types are conductor breakage and tower tilt), the short-time heavy precipitation cause (the main fault type is tower foundation settlement) and the fault caused by the composite cause, and labels the corresponding feature parameters for each sample, such as the radar reflectivity factor peak value and the lightning intensity peak value of the lightning cause sample, the maximum wind speed peak value and the wind speed duration of the strong wind cause sample.

[0088] The method of the present application realizes fine spatial matching of the line section and the strong convective event by multi-dimensional weight comprehensive calculation, realizes accurate space-time correlation through differential time window setting and correlation reliability screening mechanism, and establishes a clear corresponding relationship between the fault type and the strong convective cause through systematic attribution labeling. Compared with the traditional method which only matches based on the time window and does not consider the spatial correlation strength and reliability screening, the training data set obtained by the method of the present application has significantly improved quality, can accurately reflect the real correlation between the strong convective weather conditions, geographical environmental factors and line faults, and provides reliable data support for constructing a high-precision fault prediction model.

[0089] Embodiment three: the third embodiment of the present application, based on the same inventive concept, the present application provides a computer readable storage medium, the computer readable storage medium stores a computer program, the computer program is executed by a processor to realize the steps of the power transmission line fault prediction model training data acquisition method of the above-mentioned embodiment.

[0090] Embodiment four: the fourth embodiment of the present application, based on the same inventive concept, the present application provides a computer device, comprising: a processor, a memory; the processor, the memory communicate with each other; the memory is used for storing instructions; the processor is used for executing the instructions in the memory, executing the power transmission line fault prediction model training data acquisition method of the above-mentioned embodiment.

[0091] It should be understood that parts of the present application can be realized by hardware, software, firmware or their combination. In the above-mentioned embodiments, a plurality of steps or methods can be realized by software or firmware stored in the memory and executed by a suitable instruction execution system. For example, if realized by hardware, and in another embodiment, any one or their combination of the following technologies known in the art can be used: discrete logic circuit with logic gate circuit for implementing logic function on data signal, special integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA) and the like.

[0092] Finally, it should be noted that: the above only for the preferred embodiments of the present application, and not for the limitation of the present application, although the above-mentioned embodiments of the present application are described in detail, for those skilled in the art, it still can be modified, or equivalent replacement of part of the technical features of the technical solutions recorded in the above-mentioned embodiments. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for acquiring training data for a transmission line fault prediction model, characterized in that, include: S100: Construct a multi-source heterogeneous raw data pool, which includes severe convective weather data, power transmission line operation and fault data, and geospatial data. S200: Perform spatiotemporal benchmark unification processing on the data in the multi-source heterogeneous raw data pool, and perform gridding based on the preset spatiotemporal granularity to obtain spatiotemporal gridded data; S300: Divide the transmission line into line segments according to the latitude and longitude of the towers and map them to the corresponding grid cells of the spatiotemporal gridded data; identify strong convective events and divide the affected areas, and assign intensity level weights to each affected area; Calculate the grid exposure of the line segment, and combine the intensity level weights and the terrain correction weights extracted based on geospatial data to calculate the spatial association strength; S400: Calculate the time factor based on the time relationship between the time of the fault and the time of the strong convection; The association credibility is calculated by weighted summation based on the spatial association strength and time factor, and association results with credibility higher than a preset threshold are selected as association matching data. S500: Attribution labeling is performed on the transmission line fault information in the associated matching data to obtain a training dataset. The labeling content includes the correspondence between fault type, strong convection cause type and its characteristic parameters.

2. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The identification of severe convective events and the delineation of affected areas include: Multi-parameter collaborative thresholds were set based on radar reflectivity factor, satellite cloud top brightness temperature, and ground automatic station observation data. The density clustering algorithm is used to aggregate adjacent grids that meet the multi-parameter collaborative threshold to identify strong convection events; Based on the intensity distribution of the severe convective event decreasing outward from the core area, the core influence area, significant influence area, and peripheral influence area are divided from the inside out, and each influence area is assigned a corresponding intensity level weight.

3. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The calculation of the grid exposure of the line segment includes: The transmission line is divided into basic segments according to the tower coordinates, and then further subdivided according to the key geographical features within the line corridor to generate several fine segments. The fine line segment is spatially cut with the spatiotemporal gridded data to obtain a sub-line segment that falls completely into a single grid cell; The weighted length is calculated based on the length of each sub-segment and the attribute parameters of the line to which it belongs; The grid exposure is obtained by summing the weighted lengths of all sub-segments falling within the same grid and dividing by the grid edge length.

4. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The terrain correction weights extracted from geospatial data are obtained in the following way: Extract the slope, aspect, and relative elevation parameters of the grid from the digital elevation model data; The weighting coefficients for the slope, aspect, and relative elevation parameters are fitted based on historical fault data. The slope, aspect, and relative altitude parameters are weighted and summed according to their corresponding weight coefficients, and then normalized to obtain the terrain correction weights.

5. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The computational spatial correlation strength includes: The environmental weight of the route corridor is obtained by weighted summation of the threat level of trees, the threat level of buildings and the terrain passage index within the route corridor. Multiply the intensity level weight, the terrain correction weight, and the route corridor environment weight to obtain the spatial comprehensive weight of the grid; The spatial correlation strength of a line segment is obtained by multiplying the grid exposure of each grid along the route with the spatial comprehensive weight of the corresponding grid and summing the results.

6. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The calculation of the time factor based on the temporal relationship between the time of the fault and the time of the strong convection includes: Differentiated time-related windows are set based on the dominant type of severe convective events; Calculate the time difference between the time of the line fault and the peak time of the strong convection; The time factor is determined based on the inverse ratio of the time difference to the total duration of the time-related window.

7. The method for acquiring training data for a transmission line fault prediction model according to claim 1, characterized in that, The attribution annotations include: Based on the operation and maintenance records, fault waveforms, or on-site investigation results in the transmission line fault data, the fault type is marked. Based on the dominant characteristics of strong convection events in the aforementioned correlation matching data, the types of strong convection triggers are labeled; The peak values ​​of meteorological parameters corresponding to the type of severe convection are extracted as the characteristic parameters.

8. A training data acquisition system for a transmission line fault prediction model, characterized in that, include: The data pool construction module is used to construct a multi-source heterogeneous raw data pool, which includes severe convective meteorological data, power transmission line operation and fault data, and geospatial data. The data preprocessing module is used to perform spatiotemporal benchmark unification processing on the data in the multi-source heterogeneous raw data pool, and to perform gridding based on a preset spatiotemporal granularity to obtain spatiotemporal gridded data. The spatial association matching module is used to divide the transmission line into line segments according to the tower latitude and longitude and map them to the corresponding grid cells of the spatiotemporal gridded data; identify strong convective events and divide the affected areas, and assign intensity level weights to each affected area; Calculate the grid exposure of the line segment, and combine the intensity level weights and the terrain correction weights extracted based on geospatial data to calculate the spatial association strength; The time-related filtering module is used to calculate the time factor based on the time relationship between the time of the fault and the time of the strong convection. The association credibility is calculated by weighted summation based on the spatial association strength and time factor, and association results with credibility higher than a preset threshold are selected as association matching data. The attribution annotation module is used to perform attribution annotation on the transmission line fault information in the associated matching data to obtain a training dataset. The annotation content includes the correspondence between fault type, strong convection cause type and its characteristic parameters.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a method for acquiring training data for a transmission line fault prediction model as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements a method for acquiring training data for a transmission line fault prediction model as described in any one of claims 1 to 7.

Images