Power transmission line short circuit fault prediction method and system

Abstract

The invention provides a power transmission line short-circuit fault prediction method and system, and relates to the technical field of data processing, and the method comprises the steps: carrying out the smooth iteration through a gradient optimization strategy based on the initial space risk correction factors of adjacent sub-region units, so as to obtain the space risk correction factors; based on the short-circuit probability predicted value and the spatial risk correction factor, calculating a linkage risk value of the target segmented region in combination with the power grid topological structure data; generating a risk level label and a disposal strategy instruction of the target segmented region based on the linkage risk value; when the output risk level label is a high risk, triggering integrity verification of the multi-dimensional monitoring data; and when the monitoring data is found to be missing through verification, dynamically complementing the multi-dimensional monitoring data of the target segmented region by using the monitoring data and historical data of the adjacent segmented regions, and re-collecting the complemented data. According to the invention, the accuracy and reliability of fault prediction are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for predicting short-circuit faults in power transmission lines. Background Technology

[0002] With the continuous expansion and intelligent development of power systems, short-circuit fault prediction in transmission lines has become a crucial link in ensuring the safe and stable operation of the power grid. However, transmission lines are exposed to complex and ever-changing natural and social environments, resulting in frequent short-circuit faults. Once a short-circuit fault occurs, it can not only cause local power supply interruptions but also trigger a chain reaction in the power grid, leading to large-scale blackouts and causing huge economic losses and social impacts. Therefore, achieving accurate prediction of short-circuit faults in transmission lines and taking preventive and responsive measures in advance is of great practical significance for ensuring the safe and stable operation of the power system.

[0003] Existing fault prediction methods suffer from numerous problems and are insufficient to meet the requirements of safe and stable operation of modern power systems. Therefore, there is an urgent need to invent a method for predicting short-circuit faults in transmission lines that can comprehensively consider multi-dimensional data, including electrical parameters, terrain, and vegetation factors, while possessing the ability to perform refined spatial regional analysis and effectively handle data gaps or anomalies. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for predicting short-circuit faults in power transmission lines, which improves the accuracy and reliability of fault prediction.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: A method for predicting short-circuit faults in transmission lines, the method comprising: Step 1: Collect multi-dimensional monitoring data of the target segmented area in real time; Step 2: Based on multi-dimensional monitoring data, generate predicted short-circuit probability values ​​for the target segmented regions; Step 3: Set three preset spatial reference points within the target segmentation area to construct the spatial analysis area; Step 4: Divide the spatial analysis area into multiple sub-regions, extract the elevation data sequence of each sub-region, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-region based on the vegetation cover image data, input the terrain slope feature value and vegetation density level into the risk correction mapping table to generate the initial spatial risk correction factor; based on the initial spatial risk correction factor of adjacent sub-regions, perform smooth iteration through the gradient optimization strategy to obtain the spatial risk correction factor. Step 5: Based on the predicted short-circuit probability and spatial risk correction factor, and combined with power grid topology data, calculate the cascading risk value of the target segment region; Step 6: Based on the chain risk value, generate risk level labels and handling strategy instructions for the target segmented areas; Step 7: When the output risk level label is high risk, trigger the integrity verification of the multidimensional monitoring data; Step 8: When the verification finds that the monitoring data is missing, use the monitoring data of the adjacent segment area and historical data to dynamically supplement the multidimensional monitoring data of the target segment area, and re-collect the supplemented data.

[0006] Secondly, a short-circuit fault prediction system for transmission lines includes: The acquisition module is used to collect multi-dimensional monitoring data of the target segmented area in real time; The detection module is used to generate short-circuit probability prediction values ​​for the target segmented area based on multi-dimensional monitoring data; The analysis module is used to set three preset spatial reference points within the target segmented area to construct a spatial analysis region; The optimization module is used to divide the spatial analysis area into multiple sub-regions, extract the elevation data sequence of each sub-region, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-region based on the vegetation cover image data, input the terrain slope feature value and vegetation density level into the risk correction mapping table to generate the initial spatial risk correction factor; based on the initial spatial risk correction factor of adjacent sub-regions, the module performs smooth iteration through the gradient optimization strategy to obtain the final spatial risk correction factor. The calculation module is used to calculate the cascading risk value of the target segment area based on the predicted short-circuit probability value and the spatial risk correction factor, combined with the power grid topology data. The processing module is used to generate risk level labels and handling strategy instructions for target segmented areas based on the chain risk value; The verification module is used to trigger integrity verification of the multidimensional monitoring data when the output risk level label is high risk. The data processing module is used to dynamically supplement the multidimensional monitoring data of the target segment area by using monitoring data from adjacent segment areas and historical data when the verification finds that the monitoring data is missing, and then re-collect the supplemented data.

[0007] The above-described solution of the present invention has at least the following beneficial effects: Real-time acquisition of multi-dimensional monitoring data of target segmented areas and generation of short-circuit probability prediction values ​​overcomes the limitations of traditional methods that rely on monitoring a single electrical parameter. By combining electrical parameters with terrain and vegetation environmental data, the operating status of transmission lines is reflected, effectively avoiding misjudgments caused by load fluctuations and improving the accuracy of short-circuit fault prediction.

[0008] By setting preset spatial reference points within the target segmented area to construct a spatial analysis area and dividing it into zones, the elevation data of each sub-region unit is extracted to calculate the terrain slope characteristic value. Combined with vegetation cover image data, the vegetation density level is identified, and finally, a spatial risk correction factor is generated. The spatial analysis method can accurately identify the differences in short-circuit risk caused by differences in terrain and vegetation in different areas, thereby improving operation and maintenance efficiency and resource utilization efficiency. Attached Figure Description

[0009] Figure 1 This is a schematic flowchart of a short-circuit fault prediction method for power transmission lines provided by an embodiment of the present invention.

[0010] Figure 2 This is a schematic diagram of a short-circuit fault prediction system for power transmission lines provided in an embodiment of the present invention. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a method for predicting short-circuit faults in transmission lines, the method comprising the following steps: Step 1: Collect multi-dimensional monitoring data of the target segmented area in real time; Step 2: Based on multi-dimensional monitoring data, generate predicted short-circuit probability values ​​for the target segmented regions; Step 3: Set three preset spatial reference points within the target segmentation area to construct the spatial analysis area; Step 4: Divide the spatial analysis area into multiple sub-regions, extract the elevation data sequence of each sub-region, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-region based on the vegetation cover image data, input the terrain slope feature value and vegetation density level into the risk correction mapping table to generate the initial spatial risk correction factor; based on the initial spatial risk correction factor of adjacent sub-regions, perform smooth iteration through the gradient optimization strategy to obtain the spatial risk correction factor. Step 5: Based on the predicted short-circuit probability and spatial risk correction factor, and combined with power grid topology data, calculate the cascading risk value of the target segment region; Step 6: Based on the chain risk value, generate risk level labels and handling strategy instructions for the target segmented areas; Step 7: When the output risk level label is high risk, trigger the integrity verification of the multidimensional monitoring data; Step 8: When the verification finds that the monitoring data is missing, use the monitoring data of the adjacent segment area and historical data to dynamically supplement the multidimensional monitoring data of the target segment area, and re-collect the supplemented data.

[0013] In this embodiment of the invention, multi-dimensional monitoring data of the target segment area is collected in real time, and short-circuit probability prediction values ​​are generated, which changes the limitation of traditional methods that rely on monitoring a single electrical parameter. By combining electrical parameters with terrain and vegetation environment data, the operating status of the transmission line is reflected, effectively avoiding misjudgments caused by load fluctuation factors and improving the accuracy of short-circuit fault prediction.

[0014] A spatial analysis area is constructed by setting preset spatial reference points within the target segmented area and then dividing it into zones. Elevation data of each sub-region is extracted to calculate topographic slope characteristic values, and vegetation density levels are identified by combining this with vegetation cover image data, ultimately generating a spatial risk correction factor. This spatial analysis method can accurately identify differences in short-circuit risk caused by topographic and vegetation variations in different areas, improving operation and maintenance efficiency and resource utilization efficiency.

[0015] In a preferred embodiment of the present invention, step 1, real-time acquisition of multi-dimensional monitoring data of the target segmented region, includes: Step 11: Calculate the spatial positioning coordinates of the three fixed monitoring points based on the tower coordinates and conductor direction of the target segment area; Step 12: Based on the positioning coordinates, drive the sensor groups at each monitoring point to synchronously collect raw meteorological data, raw electrical data, and raw image data; Step 13: Perform timestamp alignment on the raw data to generate time-synchronized monitoring data frames; Step 14: Perform vegetation encroachment analysis on the raw image data in the monitoring data frame and output the real-time distance value between the conductor and the vegetation. Step 15: Perform structured fusion processing on the time-synchronized monitoring data frames and real-time distance values ​​to construct a multi-dimensional monitoring data matrix. Extract the meteorological raw data channels from the monitoring data frames to generate time series vectors for wind speed, rainfall intensity, and temperature; extract the electrical raw data channels from the monitoring data frames to generate time series vectors for current RMS value and harmonic distortion rate; construct time series vectors using the real-time distance values ​​between the conductor and vegetation as independent environmental monitoring channels; and align the meteorological vectors, electrical vectors, and distance vectors spatially based on a unified timestamp benchmark to generate a multi-dimensional monitoring data matrix. Step 16: Verify the data integrity of the matrix. If the missing rate exceeds the set threshold, trigger the data re-acquisition command for the specified monitoring points.

[0016] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 11 above, the location of the poles is measured on-site using a high-precision GNSS measuring instrument to obtain three-dimensional coordinate information such as latitude, longitude, and altitude. If there are already power line design drawings, the pole coordinate data can be directly extracted from them, and the number and relative position relationship of each pole can be recorded to form a list of pole coordinates. By reviewing the line completion data, the actual direction of the conductor in space is obtained; adjacent towers are connected to form a spatial line segment representing the direction of the conductor, and the horizontal angle and vertical tilt angle are determined. The spacing is set according to the line operating environment and monitoring needs. In areas with complex terrain, dense trees, or those susceptible to meteorological disasters, the spacing is set at 30-50 meters to allow for denser monitoring. In areas with stable environments, the spacing can be widened to 80-100 meters. For example, a 50-meter spacing is set in mountainous areas with trees, and a 100-meter spacing is set in open plains.

[0017] Using the coordinates of the starting tower as a reference point, the preliminary plane coordinates of the monitoring points are calculated along the direction of the traverse and according to the set spacing. Treating the horizontal projection of the traverse as a straight line, and based on the horizontal angle of the traverse direction, the offset distance of each monitoring point relative to the reference tower in the east-west and north-south directions is determined using the side length relationships of right triangles, thus obtaining the preliminary plane coordinates. Collect data on conductor type, specifications (such as diameter, weight per unit length), design tension, span length, etc.; consider factors such as ambient temperature and wind force, and estimate conductor sag by referring to historical operating data or using empirical formulas; for example, conductor sag will increase significantly in hot and windy weather, and the impact of changes on the location of monitoring points should be carefully considered. Based on the sag analysis results of the conductor, the preliminary plane coordinates are corrected vertically (elevation). A perpendicular line is drawn from each monitoring point to the sag curve of the conductor to determine the actual vertical distance between the monitoring point and the conductor. Combined with the altitude of the benchmark tower, the precise altitude of the monitoring point is calculated, and a three-dimensional coordinate containing both plane and elevation information is obtained.

[0018] In step 12 above, multiple types of sensor groups (meteorological sensors, electrical parameter sensors, and image acquisition equipment) are installed at each monitoring point, and the sensor coordinates are matched with the calculated positioning coordinates; a unified clock source (such as a GPS clock) is used to calibrate the time of all sensors to ensure that the acquisition time is consistent; a synchronous trigger signal (such as a timing pulse) is sent to drive each sensor to start data acquisition at the same time point to avoid timing deviation.

[0019] In step 13 above, a precise timestamp (accurate to the millisecond level) is added to the raw data collected by each sensor to record the actual time when the data was generated; for the differences in sampling frequency of different sensors (e.g., meteorological data once per second, image data 30 times per second), the low-frequency data is aligned to a unified time series by interpolation or resampling, based on the highest frequency data; the aligned data is aggregated according to a fixed time window (e.g., 1 minute) to generate a monitoring data frame containing multiple types of data (each frame corresponds to a unified timestamp).

[0020] In step 14 above, Gaussian filtering or median filtering is used to remove random noise (such as sensor thermal noise and transmission interference) from the image while preserving image edge details; histogram equalization is used to enhance the overall contrast of the image, and local adaptive histogram equalization (CLAHE) is used to enhance dark details in scenes with uneven lighting.

[0021] Image distortion correction is performed based on camera calibration parameters (intrinsic parameter matrix, distortion coefficients) to eliminate the effects of radial and tangential lens distortion; RGB images are converted to HSV or Lab color spaces to enhance the contrast between conductors and background (e.g., steel-cored aluminum stranded wires are more easily distinguished in specific color channels); Canny operators are used to extract image edges, and Hough transform is used to detect straight line segments. Line segment groups that meet conductor characteristics (e.g., continuous, parallel, specific length) are selected, and cluster analysis is performed on the detected line segments. The conductor curve model (catenary or parabola) is fitted using the least squares method to eliminate falsely detected non-conductor edges.

[0022] Extract spectral features (such as NDVI vegetation index), texture features (gray-level co-occurrence matrix), and shape features (area, perimeter) from the image; set a threshold based on the NDVI index (usually NDVI>0.3 indicates a vegetation area) to initially segment suspected vegetation areas; extract and match feature points (such as SIFT, ORB) from multiple frames or binocular camera images to generate a disparity map; combine camera intrinsic and extrinsic parameters and disparity information to calculate the 3D coordinates of each point in the image using triangulation principles to construct a sparse point cloud; fuse multi-view point cloud data and use Poisson surface reconstruction or moving least squares method to generate a dense 3D model to complete occluded areas; map the detected wires and vegetation areas in the 2D image to the 3D point cloud model to obtain spatial coordinates.

[0023] Using KD-tree or octree data structures, the algorithm quickly searches for the nearest point pair between the conductor point cloud and the vegetation point cloud, calculates the Euclidean distance between the nearest point pair, considers safety margins (such as the requirement in power line design specifications that the minimum vertical distance between the conductor and the tree is 6 meters), and outputs real-time safety distance assessment results. The algorithm compares the distance calculation results with those of multiple consecutive frames of images and filters out outliers. The algorithm also periodically uses UAV LiDAR scanning to acquire real-world environmental data, compares it with the algorithm's calculation results, and updates the model parameters to improve accuracy.

[0024] In step 15 above, the meteorological raw data channel is separated from the monitoring data frame, and the real-time values ​​of wind speed, rainfall intensity, and temperature are extracted to generate corresponding time series vectors in chronological order (e.g., [wind speed 1, wind speed 2, ...]). The electrical raw data channel is separated, and the real-time values ​​of current RMS value and harmonic distortion rate are extracted to generate electrical parameter time series vectors. The vegetation distance value is used as an independent channel, and a distance time series vector is generated according to the timestamp. Based on a unified timestamp, the meteorological vector, electrical vector, and distance vector are aligned on the time axis (ensuring that multi-dimensional data at the same time point correspond) and combined to form a multi-dimensional monitoring data matrix (rows represent time, and columns represent different types of parameters).

[0025] Step 16 above, based on the constructed multidimensional monitoring data matrix, checks the integrity of each data channel at each time point hour by hour in chronological order. This includes checking whether there are valid data records for wind speed, rainfall intensity, and temperature in the meteorological data channel; effective current values ​​and harmonic distortion rates in the electrical data channel; and for the vegetation distance channel. If a data channel lacks valid data at a given time point, that time point is marked as a missing data point. The total number of time points in the multidimensional monitoring data matrix is ​​then counted; this number represents the theoretical number of times the monitoring system should collect data within a certain time period. The number of marked missing points is then... Divide the quantity by the total number of points to obtain the data missing rate. For example, if there are 100 time points in total and 8 are missing points, the missing rate is 8%. Set a reasonable missing rate threshold based on the importance of the line operation and the usage scenario of the monitoring data. For critical transmission lines, to ensure the accuracy and reliability of data analysis, the missing rate threshold may be set at 3%-5%. For general monitoring areas, the threshold can be appropriately relaxed to 8%-10%. Compare the calculated missing rate with the set threshold. If the missing rate exceeds the threshold, it means that the integrity of the collected data cannot meet the requirements, and data re-collection is required.

[0026] By examining the monitoring data corresponding to the missing points, and determining which specific monitoring points are experiencing data loss based on the data source, for example, if meteorological data for a certain period is missing, the monitoring point that collected the meteorological data can be located. A data re-acquisition command is sent to the identified monitoring point with missing data via a wireless network communication system. The command includes the re-acquisition time range, typically from the start time to the end time of the missing data, instructing the monitoring point's sensor array to re-acquire data within the specified time period to fill in the missing portion. After the monitoring point completes the data re-acquisition, the newly acquired data is transmitted back to the system and integrated with the original data. The integrity of the integrated data matrix is ​​then checked again following the above process, and the missing rate is calculated to ensure that the data loss problem is resolved and meets the needs of subsequent data analysis and line status assessment.

[0027] In a preferred embodiment of the present invention, step 2, generating a predicted short-circuit probability value for the target segmented region based on multi-dimensional monitoring data, includes: Step 21: Extract meteorological monitoring channels from the multidimensional monitoring data matrix and generate meteorological feature vectors; Step 22: Extract the power grid operation monitoring channel from the multidimensional monitoring data matrix, calculate the fluctuation variance of the effective current value and the frequency of exceeding the harmonic distortion rate, and generate an electrical anomaly feature vector. Step 23: Extract the environmental monitoring channel of the multidimensional monitoring data matrix, and mark the status of the real-time distance value between the conductor and the vegetation. When the real-time distance value is less than the safety threshold, it is marked as a high vegetation risk state; when the real-time distance value is greater than or equal to the safety threshold, it is marked as a low vegetation risk state. Step 24: Input the meteorological feature vector into the meteorological short-circuit probability prediction model and output the short-circuit probability value of meteorological factors; input the electrical anomaly feature vector into the electrical fault probability mapping table and output the short-circuit probability value of electrical factors. Step 25: Correct the short-circuit probability value of meteorological factors based on the vegetation risk status label. Under high vegetation risk status, the short-circuit probability value of meteorological factors is increased by one level; under low vegetation risk status, the original probability value is maintained. Step 26: Combine the corrected short-circuit probability values ​​of meteorological factors and electrical factors to generate the final short-circuit probability prediction value.

[0028] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 21 above, the meteorological monitoring channel is found from the multidimensional monitoring data matrix, and monitoring data including wind speed, rainfall intensity, and temperature are extracted. The meteorological data are arranged in chronological order, and the wind speed, rainfall intensity, and temperature data corresponding to each time point are combined into a data group. The data groups of all time points are integrated to form a complete meteorological feature vector, which records the meteorological changes of the target segment area over a period of time.

[0029] Step 22 above, determining the statistical time period: Based on the line operation characteristics and monitoring needs, select a suitable statistical time period, such as 1 hour, 1 day, or 1 week. The time period should reflect the normal fluctuation pattern of the current, taking into account possible special operating conditions; within the selected time period, add up all the effective current values, and then divide by the total number of data points to obtain the average value of the effective current values ​​within the time period. The average value represents the average level of the current during the time period; calculate the difference between each effective current value and the average value, square the difference, and then calculate the average of the squared values. The result is the fluctuation variance of the effective current value; the larger the fluctuation variance, the more severe the fluctuation of the current within the time period, and the more likely there is an electrical anomaly; based on the harmonic control standards formulated by the power company, determine the normal range of the harmonic distortion rate. Different voltage levels and different types of power grids have different requirements for the harmonic distortion rate, and specific standards need to be determined according to the actual situation; check the harmonic distortion rate data at each time point one by one to determine whether it exceeds the normal reference standard range. For each data point that exceeds the standard, increment the frequency count by 1. After the statistics are completed, the total number of times the harmonic distortion rate exceeds the standard within the time period is obtained. The number of times reflects the severity of the harmonic problem in the power grid. The calculated current effective value fluctuation variance and the statistically obtained harmonic distortion rate exceeding the standard frequency are combined together in a specific order. For example, the fluctuation variance data is placed first, followed by the exceeding frequency data. The combined data is used as a whole to form an electrical anomaly feature vector, which reflects the current fluctuation and harmonic anomaly situation during the operation of the power grid in the target segment area, providing key electrical feature basis for subsequent short circuit probability calculation.

[0030] In step 23 above, based on power line design specifications and operational experience, the safe distance threshold between the conductor and vegetation is determined; the real-time distance values ​​between the conductor and vegetation in each environmental monitoring channel are checked one by one, and the real-time distance values ​​are compared with the safe threshold; if the real-time distance value is less than the safe threshold, the vegetation status is marked as a high vegetation risk status; if the real-time distance value is greater than or equal to the safe threshold, it is marked as a low vegetation risk status.

[0031] Step 24 above involves collecting meteorological data from the target segment area and surrounding areas over the past few years, including wind speed, rainfall intensity, temperature, humidity, and lightning information, compiling records of short-circuit events that occurred within the corresponding time period, and determining the meteorological conditions at the time of each short-circuit event.

[0032] Choose a suitable machine learning algorithm (such as random forest, support vector machine) or deep learning model (such as recurrent neural network, RNN), use historical meteorological data as input features and whether a short circuit event has occurred as the output label, and train the model. During training, continuously adjust the model parameters and optimize the model performance so that it can accurately learn the correlation between meteorological conditions and short circuit events. Input the generated meteorological feature vector into the trained meteorological short circuit probability prediction model. The model will analyze and process the data in the meteorological feature vector, and based on the learned patterns, output a value between 0 and 1 through internal calculation logic. The value is the probability value of meteorological factors causing short circuits. For example, an output value of 0.3 indicates that the probability of a short circuit in the area under meteorological conditions is 30%. Summarize the electrical anomaly characteristic data of the target segment area power grid under different operating conditions (such as different combinations of current effective value fluctuation variance and harmonic distortion rate exceedance frequency), and the number of short circuit faults occurring under the corresponding conditions. Perform statistical analysis on the data, and calculate... Calculate the frequency of short-circuit faults under each combination of electrical anomalies. Based on the statistical analysis results, establish an electrical fault probability mapping table. In the mapping table, each row corresponds to a specific combination of electrical anomalies (i.e., the range of current effective value fluctuation variance and the range of harmonic distortion rate exceeding the standard), and each column corresponds to different short-circuit probability levels (e.g., low probability, medium probability, high probability). Map the frequency of short-circuit faults under each combination of electrical anomalies to the corresponding short-circuit probability level according to certain rules (e.g., frequency threshold division), thus completing the construction of the mapping table. Match the generated electrical anomaly feature vectors with the electrical fault probability mapping table to determine the range of current effective value fluctuation variance and harmonic distortion rate exceeding the standard in the electrical anomaly feature vectors that fall into the mapping table. Find the corresponding row in the mapping table, and the short-circuit probability level corresponding to the row is the electrical factor short-circuit probability value. For example, if the combination corresponding to the electrical anomaly feature vector is displayed as "medium probability" in the mapping table, then the output electrical factor short-circuit probability value is the probability value corresponding to the medium probability level.

[0033] In step 25 above, check the vegetation risk status marking results to determine whether it is in a high vegetation risk state or a low vegetation risk state; if it is in a high vegetation risk state, increase the short-circuit probability value of meteorological factors by one level (for example, from low probability to medium probability); if it is in a low vegetation risk state, keep the short-circuit probability value of meteorological factors unchanged.

[0034] Step 26 above involves collecting short-circuit fault records for the target segment area over the past few years, including the time, specific location, and fault type (caused by meteorological factors or electrical factors) for each fault. Complete meteorological monitoring data (wind speed, rainfall intensity, temperature), power grid operation monitoring data (current RMS value, harmonic distortion rate), and environmental monitoring data (distance between conductors and vegetation) for the corresponding time period are then compiled. Short-circuit faults are categorized and statistically analyzed according to their inducing factors, such as the number of short circuits caused by meteorological factors and the number of short circuits caused by electrical factors, determining the proportion of different factors in historical faults. Combining historical fault data with actual line operation, the roles of meteorological and electrical factors in inducing short-circuit faults are assessed. The impact of extreme weather conditions (such as heavy rain and strong winds) on line insulation and conductor sway, as well as the potential risks of abnormal electrical parameters (excessive current fluctuations, harmonic exceedances) to equipment aging and insulation breakdown, are analyzed. From a data perspective, statistical analysis methods are used to study the probability of short circuits caused by meteorological factors. The correlation between meteorological and electrical factors' short-circuit probability values ​​and actual short-circuit events is determined. The weights of meteorological and electrical factors in short-circuit probability prediction are established. If historical data shows that meteorological factors account for 60% of short circuits and electrical factors for 40%, and the risk of meteorological factors remains high under the line environment, the initial weights for meteorological factors can be set at 0.6 and electrical factors at 0.4. The initial weights are verified by applying them to some historical data for short-circuit probability prediction simulation. The prediction results are compared with actual short-circuit situations, and the weights are fine-tuned based on error feedback until the prediction effect reaches a better level. The corrected meteorological and electrical short-circuit probability values ​​are then weighted according to the determined weights. Different risk levels are assigned based on the final predicted short-circuit probability value. For example, a predicted value of 0-0.2 indicates low risk, 0.2-0.5 indicates medium risk, and 0.5-1 indicates high risk. Corresponding risk level prompts are output to intuitively reflect the possibility of short circuits occurring in the target segment area.

[0035] In a preferred embodiment of the present invention, step 3 involves setting three preset spatial reference points within the target segmentation region to construct a spatial analysis region, including: Step 31: Based on the tower coordinates and conductor routing data of the target segmented area, select the starting tower position, the lowest point of the intermediate sag, and the ending tower position of the segmented path as the spatial reference point generation benchmark. Step 32: Set the coordinates of the starting tower position as the first spatial reference point, the coordinates of the lowest point of the intermediate sag as the second spatial reference point, and the coordinates of the ending tower position as the third spatial reference point. Step 33: Connect the first spatial reference point, the second spatial reference point, and the third spatial reference point to form a spatial triangular topology network; Step 34: Receive spatial triangular topology network data, calculate the minimum circumcircle of the spatial triangular topology network, determine the minimum circular boundary containing three spatial reference points, and record the center coordinates and radius of the minimum circumcircle; add the circumcircle radius to the preset buffer distance parameter to obtain the expanded buffer radius; generate a circular spatial analysis region with the center position as the center and the buffer radius as the boundary; verify whether the circular spatial analysis region completely covers the conductor path of the target segmented line; if there are uncovered segments, increase the buffer radius according to the set ratio and regenerate; if it completely covers the target segmented line, output the spatial analysis region and the set of boundary coordinates of the region.

[0036] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 31 above, using a high-precision GNSS measuring instrument (such as RTK equipment), during periods of good weather and no electromagnetic interference, measure each tower in the target segment area. During measurement, fix the measuring instrument antenna to the top of the tower or a specific marked point, continuously observe and record for at least 5 minutes, and take the average value as the tower's three-dimensional coordinates (latitude, longitude, and altitude), ensuring the measurement error is controlled within centimeters. If as-built drawings of the power line are available, compare the actual measured coordinates with the coordinates marked on the drawings. For towers with significant deviations, re-check the measurement process or contact the design unit to confirm the original data and make corrections. To correct errors in drawings or record actual coordinate changes, an accurate list of tower coordinates should be compiled. Based on tower numbering, connect the coordinates of adjacent towers sequentially in a Geographic Information System (GIS) or professional drawing software to draw the conductor's path on a two-dimensional plane, marking the span length between towers and the conductor suspension point height. Combined with tower elevation data, the two-dimensional path should be expanded into a three-dimensional spatial curve. Considering the conductor's thermal expansion and contraction characteristics at different temperatures, and referencing historical operating temperature data, the three-dimensional curve should be fine-tuned to simulate the conductor's spatial shape during actual operation. Conductor type information (e.g., LGJ-400 / 3) should be collected. 5) Obtain environmental temperature and wind speed meteorological data based on parameters such as material (steel-cored aluminum stranded wire), weight per unit length, design tension, and span length; using the catenary equation or parabolic approximation model (selecting the appropriate model based on the span length; generally, a parabolic model can be used for spans less than 200 meters, and a catenary model for spans greater than 200 meters), substitute the collected parameters into the model, and through iterative calculation, find the position with the maximum conductor sag. The corresponding three-dimensional coordinates are the coordinates of the lowest point of sag. For example, in the catenary model, the sag point position needs to be continuously adjusted, and the sag value at different positions needs to be calculated until the maximum value is found; utilize a drone equipped with a laser... The LiDAR scanner scans the traverse to acquire actual 3D point cloud data. The calculated coordinates of the lowest sag point are compared with the point cloud data. If the deviation exceeds the allowable range (e.g., 0.5 meters), the parameter input is rechecked or the model calculation process is adjusted until the calculation result matches the actual situation. On the drawn 3D traverse path diagram, the starting tower position, the verified intermediate lowest sag point position, and the ending tower position are marked. These three points constitute the spatial reference points. The 3D coordinates of the three reference points, their relative positions on the path diagram, and the corresponding span information are compiled into a table for easy reference in subsequent steps. In step 32 above, the acquired coordinate data of the starting tower, intermediate lowest sag point, and ending tower are checked to confirm the consistency of the coordinate system (e.g., WGS84, CGCS2000) and units (latitude and longitude in degrees, altitude in meters). If discrepancies exist, all coordinates are unified to the standard coordinate system specified by the project using a coordinate transformation tool. The coordinate data is then screened again to remove outliers caused by measurement errors or data transmission errors (e.g., latitude, longitude, or altitude data that significantly deviates from the normal range).For missing or inaccurate data, if it is tower coordinates, re-measure in the field; if it is the coordinates of the lowest point of sag, review the calculation process or use data from adjacent towers for interpolation estimation; find the record of the starting tower for the corresponding target segment area from the compiled tower coordinate list and determine the three-dimensional coordinate information; mark the three-dimensional coordinates (latitude, longitude, altitude) of the starting tower as the "first spatial reference point" and add a note in the data file or geographic information system, indicating that it is the reference benchmark at the beginning of the line, to facilitate subsequent data association and retrieval; find the coordinate data of the ending tower for the target segment area in the tower coordinate list and compare it with the field measurement records or the line Design drawings to ensure coordinates match actual locations; mark the three-dimensional coordinates of the terminating tower as the "third spatial reference point" and establish a correlation with the coordinates of the starting tower, recording the span length between the two towers and the conductor routing angle; calculate the straight-line distance and azimuth between the three reference points and compare them with the actual line design parameters; if the distance or angle deviation exceeds the allowable range (e.g., distance deviation exceeds 1%, angle deviation exceeds 1°), re-examine the coordinate data or measurement process and make corrections; use the three-dimensional point cloud data obtained by LiDAR scanning to calibrate the coordinates of the three reference points and project the reference point coordinates into the point cloud data.

[0037] In step 33 above, based on the geographical range of the target segmented area, select a suitable projection surface (such as a Gaussian projection plane or a UTM projection zone) to ensure that the projected planar coordinates accurately reflect the actual spatial relationships. For small areas (such as within 10km), the horizontal plane can be used directly as the projection surface, ignoring the influence of the Earth's curvature. Extract the elevation component from the three-dimensional coordinates (latitude, longitude, and altitude) of the three reference points separately and temporarily retain it for subsequent three-dimensional analysis. Project only the latitude and longitude coordinates onto a two-dimensional plane to generate planar coordinates (X, Y). Confirm that the three reference points use the same coordinate system (such as UTM50N) after projection. If there are differences in coordinate systems, perform a unified transformation using coordinate transformation parameters (seven or three parameters) to ensure the accuracy of the relative positions between points. Calculate the planar distance between any two points after projection and compare it with the actual measured ground distance (measured using GPS track or total station). The error should be controlled within 1%. If the deviation is large, recheck the projection parameters or the original coordinate data. Connect the first reference point, the second reference point, and the third reference point sequentially using straight line segments to form a closed triangle. When connecting the wires, follow the order of "start-middle-end" to ensure that the sides of the triangle correspond to the physical paths of the wire segments (starting segment, sag segment, and ending segment).

[0038] In CAD drawing software or a GIS platform, import the planar coordinates of three reference points, use the "three points form a plane" function to generate a triangular geometry, and automatically calculate the length of each side and the interior angles. Calculate the lengths of the three sides of the triangle, ensuring that the sum of any two sides is greater than the third side (e.g., if the length from the first to the second side is 150 meters, from the second to the third side is 200 meters, and from the first to the third side is 300 meters, verify that 150 + 200 > 300). Measure the three interior angles; the sum of the interior angles should be 180° ± 0.5°. If the deviation exceeds the threshold, check for input errors in the reference point coordinates. Overlay the triangular topology with the 3D path of the traverse line to confirm that the three sides of the triangle are... Whether the main segment area of ​​the conductor corresponds to the lowest point of the sag is located inside or on the side of the triangle, to avoid the topology being disconnected from the actual line; re-associate the elevation data of the three vertices of the triangle to form a spatial triangular surface with elevation, for example, the first reference point (starting tower) is at an elevation of 120 meters, the second reference point (lowest point of sag) is at an elevation of 115 meters, and the third reference point (ending tower) is at an elevation of 122 meters, forming a spatial triangular topology with undulations; in three-dimensional space, check whether the triangular plane intersects with the actual sag curve of the conductor to ensure that the triangular topology can truly reflect the spatial shape of the conductor segments, providing an accurate benchmark for subsequent minimum circumcircle calculation.

[0039] In step 34 above, observe the constructed spatial triangular topological network, confirm the relative positional relationship of the three reference points on the plane, and determine the type of triangle (acute, right, or obtuse triangle). This is because the circumcenters of different types of triangles are located differently: the circumcenter of an acute triangle is inside the triangle, the circumcenter of a right triangle is at the midpoint of the hypotenuse, and the circumcenter of an obtuse triangle is outside the triangle. Circumcenter location: The circumcenter of the triangle is determined through ruler and compass construction or coordinate calculation. If ruler and compass construction is used, the perpendicular bisectors of the three sides of the triangle are drawn; the intersection of these perpendicular bisectors is the circumcenter. If coordinate calculation is used, the plane coordinates (X0, Y0) of the circumcenter are calculated using the formula for calculating the coordinates of the circumcenter (involving the relationship between vertex coordinates and side lengths). The distance from the circumcenter to any reference point is measured and marked as the radius R0 of the minimum circumcircle. This can be done directly using measuring tools (such as the distance measurement function in GIS software) or by calculating the distance between two points based on coordinates. A preset buffer distance parameter D is obtained from the system configuration or project specifications. This parameter is usually set based on factors such as the line voltage level and the complexity of the surrounding environment. For example, the buffer distance for a 110kV line in a mountainous environment might be set to 20 meters, while in a plain area it might be set to 15 meters. The minimum circumcircle radius R0 is added to the preset buffer distance parameter D to obtain the expanded buffer radius R, i.e., R = R0 + D. Using the calculated circumcenter coordinates (X0, Y0) as the center of the circular region, and the buffer radius R as the boundary, a circular region is drawn on the plane. The circle drawing function of drawing software (such as CAD or ArcGIS) can be used to generate a circular spatial analysis region by inputting the center coordinates and radius value.

[0040] Project the 3D traverse path of the target segmented route onto the same plane as the circular area to obtain the line segment representation of the traverse path on the plane; check each part of the projected traverse path line segment one by one to determine whether it is completely inside the circular area; the relationship between the line segment and the circular boundary can be viewed using the spatial analysis function of software (such as ArcGIS spatial query tool); if it is found that there are traverse path projected line segments not covered by the circular area, increase the buffer radius R by a set ratio (such as 10%) to obtain a new radius R' = R × 1.1. With the original center as the center, redraw the circular area with the new radius, and verify the coverage again. Repeat this process until the traverse path is completely covered by the circular area; record the center coordinates (X0, Y0), radius R, and boundary coordinate set of the finally generated circular spatial analysis area; the boundary coordinate set can be obtained by taking points at certain angular intervals (such as every 5 degrees) on the circular boundary and recording the coordinates (X0, Y0, R). i Y i The relevant data (center coordinates, radius, and boundary coordinates) of the spatial analysis area are obtained. These data are then exported as a file in a specified format (such as CSV or SHP) for subsequent use in line monitoring, hazard analysis, and other tasks.

[0041] In this embodiment of the invention, the triangular topology and minimum circumcircle algorithm ensure that the spatial analysis area closely fits the conductor path, avoiding monitoring blind spots and improving fault location accuracy. The buffer radius expansion mechanism can cover the spatial displacement of the conductor caused by changes in sag and wind sway, ensuring the effectiveness of long-term monitoring. The regular geometric shape of the circular area facilitates sensor deployment and data acquisition, reduces redundant monitoring points, and lowers operation and maintenance costs. Combined with altitude data, the planar circular area can be expanded into a three-dimensional spatial cylinder, providing a benchmark for the spatial distance analysis between the conductor and surrounding objects (such as trees and buildings), and assisting in early warning of potential hazards.

[0042] In a preferred embodiment of the present invention, step 4 involves dividing the spatial analysis area into multiple sub-regional units, extracting elevation data sequences from each sub-regional unit, calculating terrain slope feature values ​​based on a gradient optimization strategy, identifying the vegetation density level of each sub-regional unit based on vegetation cover image data, inputting the terrain slope feature values ​​and vegetation density levels into a risk correction mapping table to generate an initial spatial risk correction factor; and obtaining a spatial risk correction factor by performing smooth iterations based on the initial spatial risk correction factors of adjacent sub-regional units, including: Step 41: Receive the set of boundary coordinates of the spatial analysis region, and divide the spatial analysis region into equal-area grids based on the preset grid size to generate multiple sub-region units; Step 42: Extract elevation monitoring point data within the sub-region unit and construct an elevation data sequence sorted by spatial location; calculate the elevation difference between adjacent points in the elevation data sequence based on the gradient optimization strategy, and take the maximum value of the absolute value of the elevation difference as the terrain slope feature value. Step 43: Retrieve vegetation cover images from environmental monitoring data; identify the percentage of vegetation pixels within a unit and classify vegetation density levels according to a preset ratio range; Step 44: Input the terrain slope feature value and vegetation density level into the preset risk correction mapping table, and query to obtain the initial spatial risk correction factor; Step 45: Receive the initial spatial risk correction factor set, which contains the initial factor values ​​of all sub-region units; select the sub-region unit to be processed as the central unit, and obtain the initial factor values ​​of the eight adjacent sub-region units; calculate the arithmetic mean of the initial factor values ​​of the central unit and the eight adjacent units, and record it as the neighborhood mean; compare the initial factor value of the central unit with the neighborhood mean; if the central value is greater than the neighborhood mean, decrease the central unit factor value by a set step size; if the central value is less than the neighborhood mean, increase the central unit factor value by the same step size; if the central value is equal to the neighborhood mean, keep the factor value unchanged; use the adjusted central unit factor value as the new factor value for the next iteration. Step 46: Iterate until the change in all unit factor values ​​is less than the convergence threshold, and output the final spatial risk correction factor.

[0043] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 41 above, based on the actual area of ​​the spatial analysis area and the monitoring accuracy requirements, a preset grid size (such as 10 meters × 10 meters or 20 meters × 20 meters) is set. For example, if the area is 1 square kilometer, a 10-meter grid can be divided into 100 × 100 units.

[0044] Based on the set of regional boundary coordinates, starting from the southwest corner vertex, a regular rectangular grid is generated by dividing the grid at equal intervals along the east-west and north-south directions according to the grid size. Each sub-region unit defines its boundary with the coordinates of its lower left and upper right corners (e.g., X1, Y1 to X2, Y2) and assigns a unique number. In step 42 above, the elevation coordinates (X, Y, Z) of the sub-region unit are obtained through ground elevation monitoring stations, UAV lidar scanning, or satellite remote sensing DEM data deployed within the region. The elevation data is checked to identify and remove outliers that deviate significantly from the surrounding terrain (e.g., sudden rises / falls due to sensor malfunctions). The mean elevation of each point and its five nearest neighbors is calculated. If the difference exceeds three times the standard deviation, it is considered an outlier and replaced with the mean of the nearest points. If the elevation points within the unit are sparse (e.g., less than five), the inverse distance is used. Distance-weighted interpolation (IDW) or Kriging interpolation is used to generate dense elevation data with a grid resolution of 1 meter based on known points, ensuring that each sub-region cell contains at least 20 elevation points. All elevation points within the cell are sorted by planar coordinates, first by X-coordinate (east-west direction) from smallest to largest, and if the X-coordinates are the same, then by Y-coordinate (north-south direction) from smallest to largest, forming an ordered list arranged from southwest to northeast. It is ensured that adjacent points after sorting are spatially close (i.e., the difference between the horizontal and vertical coordinates does not exceed 1 / 2 of the grid side length). If there are points with excessively large intervals, virtual interpolation points are inserted to ensure the continuity of the sequence. Starting from the first sorted point, the elevation difference between the point and the next point is calculated sequentially, i.e., |Z(i+1)-Z(i)|, where i is the sequence index (i=1, 2, ..., n-1). Traverse all adjacent points for elevation differences and find the difference with the largest absolute value as the terrain slope feature value of the sub-region unit; for example, the maximum elevation difference is 20 meters, so the terrain slope feature value is recorded as 20 meters (if the grid side length is 5 meters, the actual slope is 20 meters / 5 meters = 4, but here we directly record the absolute value of the elevation difference, and the specific slope conversion will be handled in the subsequent risk model). If a sub-region unit is located at the boundary of the spatial analysis area, and the edge point has only one adjacent point, then only the elevation difference on one side is calculated, which does not affect the selection of the maximum value (because the slope feature value focuses on the maximum undulation within the unit, not the integrity of the boundary); in addition to calculating by sorting in a single direction, the elevation difference is calculated repeatedly by different sorting methods such as prioritizing the north-south direction and the diagonal direction to ensure that the maximum elevation difference is not missed. For example, sort by Y coordinate first and then calculate. If a larger elevation difference is found, the feature value is updated; according to the overall terrain characteristics of the region (such as a preset maximum elevation difference threshold of 50 meters in mountainous areas and 20 meters in plains), the feature values ​​of abnormally high values ​​are verified a second time to avoid the distortion of feature values ​​caused by local terrain changes (such as a single steep slope). If necessary, the second largest value is taken or neighborhood mean smoothing is performed.

[0045] In step 43 above, vegetation cover images (such as RGB images or NDVI index images with a resolution of 1 meter) are retrieved from drone aerial photography or satellite remote sensing, and the images are denoised, radiometrically corrected, and geometrically registered to ensure alignment with the coordinates of the sub-region units.

[0046] Within the image range of each unit, the ratio of vegetation pixels to total pixels is calculated by segmenting the image using color thresholds (such as the green component threshold in RGB) or NDVI thresholds (such as NDVI>0.3 indicating vegetation). The image is then divided into levels according to preset ratio ranges: 0-30% is low density, 30-70% is medium density, and 70%-100% is high density. For example, if the vegetation pixel ratio of a certain unit is 65%, it is classified as medium density.

[0047] In step 44 above, a risk correction mapping table is preset with the terrain slope feature value range as rows (e.g., 0-5 meters, 5-10 meters, >10 meters) and vegetation density level as columns (low / medium / high). The corresponding initial correction factor (e.g., 1.0, 1.5, 2.0) is stored in the cross cells. The initial factor is obtained by querying the mapping table according to the slope feature value range and vegetation density level of the cell.

[0048] In step 45 above, with the current processing unit as the center, select eight neighboring units (top, bottom, left, right, and diagonal) to form a 3×3 neighborhood window. If the unit is located at the boundary of the region, take the existing neighboring units. Collect the initial factor values ​​of all units in the neighborhood and calculate the arithmetic mean. For example, if the factor of the central unit is 1.5, the factors of the eight neighboring units are [1.2, 1.3, 1.4, 1.6, 1.7, 1.4, 1.3, 1.2], and the mean is 1.4. If the central value 1.5 > the mean 1.4, decrease it by 5% (1.5 × 0.95 = 1.425). If the central value is less than the mean, increase it by 5%. If they are equal, keep it unchanged. Record the adjusted factor value as the result of this round of iteration for the unit.

[0049] In step 46 above, after traversing all units, calculate the maximum change in all unit factor values ​​in each iteration (e.g., the maximum value of |new value - old value|); when the maximum change is less than the convergence threshold (e.g., 0.05), terminate the iteration.

[0050] Generate a thematic layer or table containing coordinates of all sub-region units, terrain slope, vegetation density level, and final spatial risk correction factor. The higher the factor value, the greater the risk correction weight.

[0051] In this embodiment of the invention, by using gridded partitioning and topography-vegetation dual-factor coupling analysis, the accuracy of spatial risk assessment is improved from the regional level to the sub-regional level (e.g., 10m×10m), accurately locating high-risk areas; by using neighborhood mean iterative adjustment, abrupt changes in the spatial distribution of risk factors are avoided, making the correction factors more consistent with the gradual characteristics of topography and vegetation (e.g., smooth changes in factors in transitional areas between hillsides and flatlands); by integrating elevation data and vegetation images, the impact of topographic slope on vegetation growth rate is quantified (e.g., vegetation roots are more likely to damage slope stability on steep slopes), achieving multi-dimensional cross-validation of risk assessment; risk correction factors can be directly superimposed into the short-circuit probability prediction model, prioritizing inspections of areas with high factor values ​​(e.g., steep slopes with high vegetation density), optimizing the efficiency of operation and maintenance resource allocation.

[0052] In a preferred embodiment of the present invention, step 5, based on the predicted short-circuit probability value and the spatial risk correction factor, combined with power grid topology data, calculates the cascading risk value of the target segment region, including: Step 51: Receive the predicted short-circuit probability value of the target segment region and the corresponding spatial risk correction factor; multiply the predicted short-circuit probability value of the target segment region by the corresponding spatial risk correction factor to generate the basic risk value of the segment region. Step 52: Call the power grid topology data to identify the direct connection relationships of the target segment region in the power grid; determine the set of downstream adjacent segment regions with direct electrical connections; Step 53: Obtain the basic risk value of the downstream region, multiply the basic risk value of the downstream region by the preset impact propagation coefficient to obtain the risk contribution value of the downstream region to the target segment region; sum up the risk contribution values ​​of all downstream adjacent segment regions to obtain the total risk impact of the neighborhood. Step 54: Add the base risk value to the sum of the neighboring risk impacts to obtain the final cascading risk value of the target segmented region; Step 55: Output the chain risk values ​​for all segmented regions, which are used for quantitative assessment of the chain propagation risk of power grid short-circuit faults.

[0053] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: Step 51 above analyzes historical short-circuit fault data, equipment aging, real-time operating status (such as load current and voltage fluctuations), and external environmental factors (such as weather conditions and lightning activity frequency) of the target segment area to obtain a prediction value. For example, a machine learning model is trained on data such as the number of short circuits per year in the region over the past ten years, equipment maintenance records, and seasonal high-incidence periods of faults to predict the probability of short circuit occurrence during a period (such as the peak summer high-temperature load period), with a value range of 0 to 1 (0 indicates no possibility of fault, and 1 indicates that a fault will definitely occur).

[0054] The correction factor is used to quantify the impact of the geographical environment and spatial layout of the target segment area on short-circuit risk. In mountainous areas and areas with many trees, the correction factor is higher than that in plain areas because the lines are easily affected by vegetation or have a high probability of lightning strikes. In areas around substations and areas with dense cable trenches, the correction factor is even higher because the equipment spacing is small and the possibility of fault spread is high. In coastal areas with high humidity and salt spray all year round, or in plateau areas that are prone to icing in winter, the insulation performance of equipment deteriorates faster, and the correction factor needs to be adjusted upward. In areas where landslides, floods and other disasters have occurred and caused short-circuit faults, the correction factor will be dynamically adjusted according to the frequency and degree of impact of the disaster. The correction factor is usually preset by power grid operation and maintenance experience combined with geographic information system (GIS) data, and the value range is generally 1 to 2 (1 indicates no additional environmental risk, and 2 indicates a significantly increased risk).

[0055] For each target segment area, short-circuit probability prediction values ​​are retrieved from the power grid monitoring system (e.g., real-time acquisition via SCADA system or output from offline prediction models), and corresponding spatial risk correction factors are matched from the GIS database (e.g., automatically associated with preset environmental risk levels based on regional coordinates). The short-circuit probability prediction values ​​and spatial risk correction factors are multiplied to obtain the basic risk value of the area.

[0056] In step 52 above, the power grid topology data (such as node connection relationships and line routes) is read to determine the location of the target segment area in the power grid and its direct electrical connection relationships.

[0057] Based on the power flow direction or topology logic of the power grid, identify the downstream segment regions that are directly connected to the target region and form a "downstream adjacent segment region set". For example, if the target region is connected to two regions A and B through a line, and the current flows from the target region to A and B, then A and B are downstream adjacent regions.

[0058] In step 53 above, for each determined downstream adjacent region, the calculated basic risk value is obtained; the basic risk value of each downstream region is multiplied by a preset impact propagation coefficient (the coefficient reflects the intensity of risk propagation from downstream to the target region, usually a constant between 0 and 1, such as 0.3) to obtain the risk contribution value of a single downstream region to the target region; the risk contribution values ​​of all downstream adjacent regions are added together to obtain the total risk impact of the neighborhood.

[0059] In step 54 above, the basic risk value of the target segmented area is added to the sum of the risk impacts of neighboring areas to obtain the final cascading risk value.

[0060] Step 55 above summarizes and outputs the chain risk values ​​of all segmented areas to form the risk quantification results of each area of ​​the power grid, providing data for risk assessment of the chain propagation of short-circuit faults.

[0061] In this embodiment of the invention, short-circuit probability prediction and spatial correction factors are combined to consider both the probability of fault occurrence and the influence of objective factors such as geographical environment, making risk assessment more realistic. By identifying upstream and downstream connections through the power grid topology, the propagation impact of neighboring risks is quantified, avoiding isolated assessments of individual area risks and better reflecting the "chain propagation" characteristics of power grid faults. Through the calculation of chain risk values, key areas with high risk and significant influence from neighboring areas can be identified, providing accurate basis for power grid operation and maintenance and fault prevention strategies (such as strengthening inspections and optimizing protection configurations). By converting risk into quantifiable values, power grid managers can compare the risk levels of different areas, formulate priority-based risk management plans, and improve the reliability and safety of power grid operation.

[0062] In a preferred embodiment of the present invention, step 6, based on the cascading risk value, generates a risk level label and a handling strategy instruction for the target segmented area, including: Step 61: Receive the chain risk values ​​of all target segment regions; preset low-risk threshold range, medium-risk threshold range and high-risk threshold range to form risk level classification rules; Step 62: Compare the chain risk value of each target segment area with the preset risk threshold range; if the chain risk value falls into the low risk threshold range, generate a low risk level label; if the chain risk value falls into the medium risk threshold range, generate a medium risk level label; if the chain risk value falls into the high risk threshold range, generate a high risk level label. Step 63: Invoke the preset risk-strategy mapping table to map low-risk level labels to basic inspection strategy instructions; map medium-risk level labels to enhanced monitoring strategy instructions; and map high-risk level labels to emergency response strategy instructions. Step 64: Based on the power grid topology data, perform coordination verification on the handling strategy instructions for adjacent target segment areas. If adjacent segment areas are all marked with high-risk level labels, upgrade the handling strategy instructions to regional linkage handling instructions and add cross-segment fault isolation schemes; output the final risk level labels of each target segment area and the corresponding handling strategy instructions to the power grid dispatch terminal.

[0063] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: Step 61 above, based on historical short-circuit fault data and risk assessment experience of the power grid, establishes three risk threshold ranges: Low-risk threshold range: such as [0, 0.4], corresponding to areas with extremely low probability of fault chain propagation; Medium-risk threshold range: such as (0.4, 0.7], corresponding to areas with certain risk of cascading faults and requiring attention; High-risk threshold range: such as (0.7, 1], corresponding to areas where faults are highly likely to spread and may cause large-scale power outages. The threshold division needs to balance safety and operation and maintenance costs. For example, by statistically analyzing the distribution of cascading risk values ​​in various regions over the past three years, the top 20% of high-risk values ​​are set as high-risk ranges, the middle 50% as medium-risk ranges, and the bottom 30% as low-risk ranges. The three threshold ranges are then mapped one-to-one with the "low / medium / high" risk levels to form a definite level division rule (such as a closed interval that includes the left side but not the right side, or the boundary is adjusted according to actual needs).

[0064] In step 62 above, for each target segmented area (e.g., ordered by area number or geographical distribution), the chain risk value is extracted and compared with the three-level threshold interval; it is then determined whether it falls into the low-risk interval: if the risk value ≤ 0.4 (e.g., 0.35, 0.4), a "low-risk level label" is directly generated; if it exceeds the low-risk upper limit (>0.4), it is further determined whether it falls into the medium-risk interval: if the risk value ≤ 0.7 (e.g., 0.5, 0.7), a "medium-risk level label" is generated; if it exceeds the medium-risk upper limit (>0.7) and the risk value ≤ 1 (because the maximum probability value is 1), a "high-risk level label" is generated; Example: The chain risk value for area A is 0.2. 8. Since 0.28 ≤ 0.4, it is marked as "low risk"; the chain risk value of region B is 0.63, since 0.4 < 0.63 ≤ 0.7, it is marked as "medium risk"; the chain risk value of region C is 0.79, since 0.7 < 0.79 ≤ 1, it is marked as "high risk"; if the risk value of a certain region is exactly equal to the threshold boundary (such as 0.4 or 0.7), it is uniformly classified according to the preset rules (such as "including the left boundary but excluding the right boundary") to avoid ambiguity; if a risk value exceeds the preset range (such as a calculation error causing a risk value of 1.2, theoretically the probability value should not exceed 1), the system automatically triggers the data verification mechanism, marks it as "data abnormal" and temporarily classifies it as the highest risk level.

[0065] In step 63 above, the power grid operation and maintenance expert team, based on historical fault handling experience, equipment importance, and scheduling rules, pre-determines a mapping table containing four columns: "risk level label", "disposal strategy instruction", "strategy execution priority", and "applicable scenario description".

[0066] Quarterly manual inspection (recording equipment temperature, insulation resistance, and contact tightness), automatic generation of inspection reports, and marking of abnormal data (such as triggering a yellow warning when the load exceeds 80% of the rated value); daily remote monitoring (real-time collection of current, voltage, and partial discharge signals through IoT sensors), activation of AI fault prediction model (updating risk prediction values ​​hourly), and real-time push notifications of abnormalities to maintenance personnel's mobile phones (such as immediate alarm when the temperature suddenly rises by 10°C).

[0067] Emergency response strategy: covering arriving on-site within 4 hours and shutting down the power for repair (prioritizing dispatching the nearest maintenance team), carrying spare equipment components (such as fuses and surge arresters), and simultaneously developing a fault isolation plan (determining the tripping sequence of sectional switches and the switching logic of backup power).

[0068] For each area risk level label (e.g., "high risk"), retrieve the corresponding "handling strategy instruction" column from the mapping table; if the label is "low risk", directly retrieve the "basic inspection strategy" and include it in the next quarter's inspection plan according to "three-level priority", without immediate intervention; if the label is "medium risk", trigger the dual strategy of "enhanced monitoring + special investigation", increase the area data collection frequency through the dispatch system (e.g., from once every 15 minutes to once every 5 minutes); dispatch the maintenance team to complete the special equipment inspection within 72 hours (focusing on cable joints, insulators and other easily aging components); if the label is "high risk", activate "emergency shutdown maintenance + contingency plan activation", send a shutdown application to the dispatch terminal (with equipment number, risk value and suggested power outage period), and immediately shut down the area equipment after confirmation, while retrieving the pre-prepared "High-risk Area Fault Emergency Response Plan" (e.g., activate the interconnection switch of the adjacent area and switch to backup power supply); a medium-risk area (label "high risk") When a medium-risk area is matched with "enhanced monitoring + special investigation", the system automatically allocates a dedicated monitoring channel, tracks the load curve in real time, and generates a special investigation work order (the work order includes: checking the lubrication of the circuit breaker operating mechanism, testing the operation time of the relay protection device, and recording the manufacturing year of the equipment); when a high-risk area (labeled "high-risk") triggers "emergency shutdown for maintenance", the dispatch system immediately locks the power supply line of the area, sends an emergency work order containing location information, risk level, and maintenance tool list to the operation and maintenance department, and simultaneously notifies the dispatcher to prepare to start the backup power supply plan; if there are exclusive strategies for special equipment types (such as transformers, cable lines) in the mapping table, the exclusive strategy is matched first (for example, high-risk areas with oil-containing equipment need to add oil chromatography analysis items); the strategy instructions include "mandatory actions" and "optional actions", such as "emergency shutdown" in high-risk areas is a mandatory action, and "start backup power supply" is an optional action (whether to execute depends on the real-time load of the power grid).

[0069] Step 64 above, based on power grid topology data (such as a node connection table), obtains all directly adjacent areas of the target segment (including upstream, downstream, and same-level interconnection areas). If the target area is high-risk, and any adjacent area is also high-risk (e.g., area A and area B are directly connected via a busbar, and both are marked as high-risk), it is determined to be an "adjacent high-risk cluster." The original "emergency handling strategy" for a single area is upgraded to a "regional coordinated handling instruction," and an additional cross-segment fault isolation scheme is added (such as predefined segment switch tripping logic and automatic switching procedures for backup lines). The final risk level label (e.g., "high risk") and handling strategy instructions (e.g., "regional coordinated handling + emergency maintenance + cross-segment isolation") are sent to the power grid dispatch terminal, along with geographic location and strategy execution priority.

[0070] In this embodiment of the invention, continuous risk values ​​are transformed into intuitive "low / medium / high" level labels by dividing threshold intervals, making it easier for maintenance personnel to quickly understand the severity of risks. Different risk levels correspond to differentiated handling strategies (from routine inspections to emergency response), avoiding "one-size-fits-all" maintenance and improving resource utilization efficiency. For example, high-risk areas directly trigger emergency maintenance to prevent faults from evolving into large-scale power outages. Through topological correlation collaborative verification, high-risk clusters are identified and upgraded to regional linkage strategies, and cross-segment fault isolation schemes are planned in advance to effectively block the chain propagation path of faults (such as when adjacent high-risk areas fail simultaneously, the fault propagation channel is quickly cut off through a preset isolation scheme). Finally, the output risk labels and strategy instructions are directly connected to the power grid dispatching system to form a closed-loop management of "risk assessment - strategy generation - execution implementation," providing automated support for real-time dispatching and maintenance plan formulation, and improving the speed and safety of power grid emergency response.

[0071] In a preferred embodiment of the present invention, step 7, when the output risk level label is high risk, triggers an integrity check of the multidimensional monitoring data, including: Step 71: Based on the high-risk level label, extract multi-dimensional monitoring data associated with the target segmented areas; Step 72: Perform time window verification on the multidimensional monitoring data to verify whether the timestamps of the data are within the preset valid monitoring time range; Step 73: Perform missing item verification on the data that passed the time window verification, check the field integrity of each data source item by item, and identify missing data items; if there are missing data items, generate a data missing alarm command and start the supplementary collection process, calling the backup monitoring node or the data proxy node of the adjacent segment area to fill in the missing items. Step 74: Repeat the verification process for the supplemented data until the multidimensional monitoring data completely passes the integrity verification or reaches the preset maximum retry threshold. Step 75: If data is still missing after reaching the maximum retry threshold, mark the data in the high-risk segment as "incomplete" and append the list of missing data items to the emergency response strategy instruction and output it to the power grid dispatch terminal.

[0072] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 71 above, for the target segment area marked as "high risk", retrieve directly related multi-dimensional monitoring data, including transformer oil temperature, line load current, circuit breaker opening and closing status, insulator leakage current, etc. (from SCADA system or IoT sensors); regional humidity, temperature, and wind speed (used to determine equipment insulation performance and external causes, such as icing and lightning strike risk); short circuit fault records and protection device operation logs of the area and adjacent areas within the past year; equipment installation location coordinates, line direction, and surrounding terrain (obtained through GIS system, used to analyze the rationality of spatial risk correction factors).

[0073] Using a unique regional code (such as equipment ID or segment line number), relevant data is retrieved in batches from the real-time database, historical database, and spatial database of the power grid data center. If a high-risk area is "110kV Line A, Section 03", the load current (850A), bus voltage (108kV), and cable joint temperature (65℃) are automatically extracted; three overload alarm records in the past 30 days; logs of brief grounding faults caused by icing last winter; humidity of 85% (which can easily lead to insulation degradation); and altitude of 1200 meters (risk of corona discharge in high-altitude areas).

[0074] Step 72 above, "Effective monitoring time window" is the data within 7 days before the evaluation time (which can be adjusted according to equipment characteristics, such as 24 hours for high-frequency fault equipment and 30 days for old equipment), is used to ensure the timeliness of the data.

[0075] In step 73 above, for the data that passes through the time window, the integrity of each field is checked according to the preset data template. For example, the equipment operation data template requires that it must include four fields: "monitoring time", "current", "voltage" and "temperature". If a data item is missing the "temperature" field, it is marked as "missing data item".

[0076] In step 74 above, based on the stability of the power grid monitoring system and the actual efficiency of data replenishment, the preset threshold range is 3-5 times (typically 3 times); the average time for a single replenishment process (e.g., 1 minute for calling a backup node, 2 minutes for a neighboring agent) ensures that the total time does not exceed 15 minutes (meeting the time margin of "4-hour emergency response" in high-risk areas); to avoid excessive system resource consumption due to infinite retries (e.g., communication bandwidth being occupied by replenishment requests for a long time, affecting real-time monitoring services); the single retry interval is set to 3-10 minutes (typically 5 minutes), and is dynamically adjusted according to the following conditions: when the backup node responds quickly (e.g., local redundant sensors), the interval can be shortened to 3 minutes; when the neighboring agent node needs to communicate across regions, the interval is extended to 8 minutes; if the power grid data center load rate exceeds 70%, the interval is automatically extended to 10 minutes to avoid increasing the system load.

[0077] The first retry (counter=1, first attempt within the threshold range) will prioritize calling the local backup monitoring node (such as the cable joint temperature field) again after waiting for a 5-minute interval if the system finds missing data (e.g., missing cable joint temperature field) and the backup monitoring node has been started but failed to collect data. For example, if the main temperature sensor in target area A fails and the backup sensor does not respond due to loose wiring, the system will automatically send a hardware reset command during the first retry to restart the backup sensor and request data. If the backup sensor recovers and returns a temperature value of 25℃ (temperature difference ≤10% with adjacent areas), the missing item will be determined to be filled, and the system will proceed to the final data integrity verification.

[0078] The second retry (counter=2, intermediate attempts within the threshold range) occurs if the first retry fails (e.g., backup sensor hardware is damaged and unable to return data). An 8-minute interval is waited (extended due to cross-regional communication requirements). Data proxy nodes in adjacent segmented regions are then invoked. For example, if region A and region B are directly connected via a bus, and temperature sensors of the same type in region B are operating normally, based on the power grid topology (distance between the two regions ≤ 500 meters, ambient temperature difference ≤ 5℃), temperature data from region B is extracted as a proxy value (with a "data proxy source" label, such as "taken from sensor number 2 in adjacent region B"). In addition to field integrity, data validity checks are added (e.g., proxy temperature value). Check if the deviation from the equipment's rated operating temperature is within ±20% to avoid data distortion due to environmental differences; if there is no similar monitoring equipment in area B or the agent data deviation exceeds the threshold, record "neighbor agent failure", increment the counter by 1, and proceed to the third retry; for the third retry (counter=3, the last attempt within the threshold range), if the first two retry attempts fail (e.g., backup node failure, no valid data from the neighbor agent), wait for a 10-minute interval (maximum interval, allowing sufficient time for cross-system data retrieval), and then initiate cross-system data supplementation; retrieve historical maintenance records from the equipment ledger system to obtain the "rated insulation temperature" of the equipment in the segmented area as a theoretical reference value (e.g., the maximum rated temperature of the cable is 7). 0℃, used to fill in missing real-time temperature data); extract real-time humidity and wind speed data of the area from the meteorological system, and combine with the equipment thermal balance model to estimate the possible temperature range (e.g., when humidity is 80% and wind speed is 2m / s, the equipment temperature is approximately ambient temperature +15℃); mark the supplementary data as "estimated value" or "historical reference value" and indicate it in the data missing list (e.g., "temperature data is a theoretical estimate and needs to be verified by on-site measurement"); if there is still missing core data (e.g., the "protection action time" in the fault log cannot be completed by any system), or the maximum number of retries (3 times) is reached, terminate the retry process; generate a detailed log after each retry, including the number of retry (1st / 3rd time), the initiation time, etc. Time (e.g., 10:05, 10:13, 10:23), duration (45 seconds, 90 seconds, 180 seconds); data source for re-collection (backup node / neighborhood agent / cross-system), success status (yes / no), unresolved missing items (e.g., "temperature data ×, protection action time ×"); if a retry takes more than 1.5 times the interval (e.g., more than 7.5 minutes at a 5-minute interval), an automatic "data re-collection timeout" alarm is triggered, checking for hardware failures on the backup node (e.g., loose sensor power adapter); whether the neighborhood agent node restricts data access due to network firewall policies; and whether there are permission abnormalities on the cross-system data interface (e.g., expired access token for the ledger system).

[0079] In step 75 above, for high-risk areas where data is still missing even after reaching the retry limit, the monitoring data is marked as "incomplete" and the missing fields (such as "cable joint temperature, fault duration") are recorded.

[0080] The list of missing data items is embedded in the emergency response strategy instructions, for example: "Inspect 110kV Line A, Section 03, on-site within 4 hours to investigate potential equipment hazards"; followed by: "On-site inspection within 4 hours (missing data items: cable joint temperature, fault duration record), with a focus on verifying the temperature sensor and log entry module"; when output to the power grid dispatch terminal, missing items are highlighted in red, and missing data is prioritized for collection during inspection (e.g., manually recording temperature using a portable thermometer); in this embodiment of the invention, multi-dimensional data integrity is forcibly verified for high-risk areas to avoid misjudgment of the fault cause due to missing data (e.g., no temperature data available). This approach helps to identify potential equipment overheating risks, improving the reliability of risk assessments. By supplementing missing data (e.g., by calling proxy data from adjacent areas), it ensures that emergency response plans are based on comprehensive information (e.g., by combining accurate load current and historical fault records to reasonably schedule power outage maintenance time). For data that cannot be supplemented, it identifies the key information that maintenance personnel need to collect on-site, forming a closed loop of "data verification - missing data supplementation - on-site verification" to reduce handling loopholes caused by data issues. Prioritizing the use of backup nodes and neighboring proxy data to supplement missing items avoids blindly increasing hardware investment, reducing maintenance costs while ensuring data quality.

[0081] In a preferred embodiment of the present invention, step 8, when the verification finds that monitoring data is missing, dynamically supplements the multidimensional monitoring data of the target segment area using monitoring data from adjacent segment areas and historical data, and re-collects the supplemented data, including: Step 81: When a missing data item is identified, initiate the dynamic data completion process; Step 82: Based on the type of missing data item, extract the corresponding data from the real-time monitoring data of adjacent segment areas or the historical monitoring data of the target segment area; Step 83: Perform spatiotemporal alignment processing on the extracted data and map it to the monitoring timestamp and spatial location of the target segment region; Step 84: Inject the spatiotemporally aligned data into the multidimensional monitoring dataset of the target segmented region to generate the completed dataset; Step 85: Trigger the supplementary data collection process, and call the data proxy node to perform a second integrity check on the completed dataset; Step 86: If the second verification passes, mark the completed data as "verified"; if the verification fails, return to execute a new round of completion operation until the verification passes or the maximum number of retries threshold is reached.

[0082] In this embodiment of the invention, the above steps can be implemented through the following steps, specifically as follows: In step 81 above, when missing data items are detected in the target segment area (such as an empty "cable joint temperature" field in the equipment operation data or a missing "protection action time" record in the historical fault data), the system automatically marks the missing item type (real-time data missing / historical data missing) and activates the dynamic completion process to generate a "data completion task sheet", which includes: target area identifier (such as "110kV B line 05 segment"), missing data item name (such as "insulator leakage current"), missing time range (such as no valid data in the last hour); priority marking (missing items in high-risk areas are marked as "urgent", and those in medium and low-risk areas are marked as "normal".

[0083] In step 82 above, real-time monitoring data is extracted from adjacent segment areas (directly electrically connected areas determined according to the power grid topology, such as similar equipment within 500 meters upstream and downstream). For example, if the current sensor in target area A fails, the real-time current value of the adjacent area B (belonging to the same bus and 300 meters away) is extracted (it is necessary to ensure that the load characteristics of the two are similar, such as the proportion of industrial load exceeding 60%).

[0084] In step 83 above, the timestamps of the extracted real-time data from the adjacent areas are adjusted to the missing timestamps of the target area. For example, when the current data (1200A) collected in the adjacent area B at 10:05 is mapped to the missing time point of 10:00 in the target area A, it is necessary to combine the load change trend (such as the load increasing linearly from 10:00 to 10:10, and the current at 10:00 is estimated to be about 1150A).

[0085] In step 84 above, the spatiotemporally aligned data is inserted into the multidimensional monitoring dataset of the target area. Missing fields are marked as "complete data" and a source description is attached (such as "[Complete data for adjacent area B]" or "[Mapping of historical data from 2024]"). Example: Original missing dataset: [Time: 10:00, Current: Missing, Voltage: 110kV, Temperature: Missing], Completed dataset: [Time: 10:00, Current: [Complete data for area B] 1150A, Voltage: 110kV, Temperature: [Mapping of historical data] 60℃]. Ensure that the units and precision of the completed data are consistent with the original data (e.g., the current unit is uniformly "A", and one decimal place is retained).

[0086] In step 85 above, check whether the timestamp of the completed data is within the valid monitoring window of the target area (e.g., the last 7 days); whether the completed data and fields conform to physical laws (e.g., when the current increases, the temperature should rise synchronously, with a deviation of no more than ±15%); whether the data from adjacent areas comes from trusted nodes (e.g., sensors within the equipment calibration cycle); and whether the historical data has been denoised (removing outliers during fault periods).

[0087] If the secondary verification in step 86 above is normal (e.g., the power calculation results of the completed current and voltage values ​​match the rated capacity), the data is marked as "verified" and allowed for risk assessment and response strategy formulation. If the completed data is found to be unreasonable (e.g., the current value deviation exceeds 30% due to significant differences in load types between adjacent areas), the process is returned to a different data source (e.g., extracted from data from a more distant similar area or a more recent historical period) and the completion process is re-executed. A maximum retry threshold (3 times) is set. If all 3 attempts fail, the completed data is retained but marked as "awaiting manual confirmation," and the response strategy notes that the data may contain errors (e.g., "Temperature data is a neighborhood estimate and needs to be verified by actual measurement during on-site maintenance").

[0088] In this embodiment of the invention, the problem of missing monitoring data in high-risk areas is solved by dynamically fusing real-time and historical data from adjacent areas, avoiding misjudgments of risks due to missing key data (such as missing potential equipment overheating hazards); the optimal data source is automatically selected based on data type (real-time / historical) and spatiotemporal characteristics (distance, operating conditions) to improve the accuracy of data completion (e.g., prioritizing real-time data from adjacent areas on the same busbar rather than inefficient data from distant areas); secondary verification ensures that the completed data conforms to physical laws and business logic, avoiding the introduction of erroneous data that interferes with decision-making (e.g., preventing incorrect current values ​​due to differences in load type); it reduces the cost of manual intervention, allowing high-risk areas to quickly obtain usable data to support the timely formulation of emergency response strategies (e.g., accurately determining whether immediate power outage maintenance is necessary based on completed temperature data); and verification is performed through data proxy nodes to avoid duplicate data collection and waste of system resources, while utilizing historical and neighboring data to reduce hardware redundancy investment (e.g., reducing the over-configuration of backup sensors).

[0089] like Figure 2As shown, embodiments of the present invention also provide a short-circuit fault prediction system for transmission lines, comprising: a data acquisition module for real-time acquisition of multi-dimensional monitoring data of a target segmented area; a detection module for generating a predicted short-circuit probability value for the target segmented area based on the multi-dimensional monitoring data; an analysis module for setting three preset spatial reference points within the target segmented area to construct a spatial analysis area; and an optimization module for partitioning the spatial analysis area to generate multiple sub-regional units, extracting elevation data sequences from each sub-regional unit, calculating terrain slope feature values ​​based on a gradient optimization strategy, identifying the vegetation density level of each sub-regional unit based on vegetation cover image data, and inputting the terrain slope feature values ​​and vegetation density levels into a risk correction mapping table to generate an initial spatial risk correction mapping table. The system comprises the following modules: a positive factor, a calculation module, and a processing module. The calculation module calculates the cascading risk value of the target segment region based on the short-circuit probability prediction, the spatial risk correction factor, and the power grid topology data. The processing module generates risk level labels and handling strategy instructions for the target segment region based on the cascading risk value. The verification module triggers a completeness verification of the multidimensional monitoring data when the output risk level label is high-risk. The organization module dynamically completes the multidimensional monitoring data of the target segment region using monitoring data from adjacent segment regions and historical data when the verification reveals missing monitoring data, and then re-collects the completed data.

[0090] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles 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 predicting short-circuit faults in transmission lines, characterized in that, The method includes: Step 1: Collect multi-dimensional monitoring data of the target segmented area in real time; Step 2: Based on multi-dimensional monitoring data, generate predicted short-circuit probability values ​​for the target segmented regions; Step 3: Set three preset spatial reference points within the target segmentation area to construct the spatial analysis area; Step 4: Divide the spatial analysis area into multiple sub-regions, extract the elevation data sequence of each sub-region, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-region based on the vegetation cover image data, input the terrain slope feature value and vegetation density level into the risk correction mapping table to generate the initial spatial risk correction factor; based on the initial spatial risk correction factor of adjacent sub-regions, perform smooth iteration through the gradient optimization strategy to obtain the spatial risk correction factor. Step 5: Based on the predicted short-circuit probability and spatial risk correction factor, and combined with power grid topology data, calculate the cascading risk value of the target segment region; Step 6: Based on the chain risk value, generate risk level labels and handling strategy instructions for the target segmented areas; Step 7: When the output risk level label is high risk, trigger the integrity verification of the multidimensional monitoring data; Step 8: When the verification finds that the monitoring data is missing, use the monitoring data of the adjacent segment area and historical data to dynamically supplement the multidimensional monitoring data of the target segment area, and re-collect the supplemented data.

2. The method for predicting short-circuit faults in transmission lines according to claim 1, characterized in that, Step 1: Real-time acquisition of multi-dimensional monitoring data of the target segmented region, including: Step 11: Calculate the spatial positioning coordinates of the three fixed monitoring points based on the tower coordinates and conductor direction of the target segment area; Step 12: Based on the positioning coordinates, drive the sensor groups at each monitoring point to synchronously collect raw meteorological data, raw electrical data, and raw image data; Step 13: Perform timestamp alignment on the raw data to generate time-synchronized monitoring data frames; Step 14: Perform vegetation encroachment analysis on the raw image data in the monitoring data frame and output the real-time distance value between the conductor and the vegetation. Step 15: Perform structured fusion processing on the time-synchronized monitoring data frames and real-time distance values ​​to construct a multi-dimensional monitoring data matrix. Extract the meteorological raw data channels from the monitoring data frames to generate time series vectors for wind speed, rainfall intensity, and temperature; extract the electrical raw data channels from the monitoring data frames to generate time series vectors for current RMS value and harmonic distortion rate; construct time series vectors using the real-time distance values ​​between the conductor and vegetation as independent environmental monitoring channels; and align the meteorological vectors, electrical vectors, and distance vectors spatially based on a unified timestamp benchmark to generate a multi-dimensional monitoring data matrix. Step 16: Verify the data integrity of the matrix. If the missing rate exceeds the set threshold, trigger the data re-acquisition command for the specified monitoring points.

3. The method for predicting short-circuit faults in transmission lines according to claim 2, characterized in that, Step 2: Based on multi-dimensional monitoring data, generate predicted short-circuit probability values ​​for the target segmented region, including: Step 21: Extract meteorological monitoring channels from the multidimensional monitoring data matrix and generate meteorological feature vectors; Step 22: Extract the power grid operation monitoring channel from the multidimensional monitoring data matrix, calculate the fluctuation variance of the effective current value and the frequency of exceeding the harmonic distortion rate, and generate an electrical anomaly feature vector. Step 23: Extract the environmental monitoring channel of the multidimensional monitoring data matrix, and mark the status of the real-time distance value between the conductor and the vegetation. When the real-time distance value is less than the safety threshold, it is marked as a high vegetation risk state; when the real-time distance value is greater than or equal to the safety threshold, it is marked as a low vegetation risk state. Step 24: Input the meteorological feature vector into the meteorological short-circuit probability prediction model and output the short-circuit probability value of meteorological factors; input the electrical anomaly feature vector into the electrical fault probability mapping table and output the short-circuit probability value of electrical factors. Step 25: Correct the short-circuit probability value of meteorological factors based on the vegetation risk status label. Under high vegetation risk status, the short-circuit probability value of meteorological factors is increased by one level; under low vegetation risk status, the original probability value is maintained. Step 26: Combine the corrected short-circuit probability values ​​of meteorological factors and electrical factors to generate the final short-circuit probability prediction value.

4. The method for predicting short-circuit faults in transmission lines according to claim 3, characterized in that, Step 3: Set three preset spatial reference points within the target segmentation area to construct the spatial analysis region, including: Step 31: Based on the tower coordinates and conductor routing data of the target segmented area, select the starting tower position, the lowest point of the intermediate sag, and the ending tower position of the segmented path as the spatial reference point generation benchmark. Step 32: Set the coordinates of the starting tower position as the first spatial reference point, the coordinates of the lowest point of the intermediate sag as the second spatial reference point, and the coordinates of the ending tower position as the third spatial reference point. Step 33: Connect the first spatial reference point, the second spatial reference point, and the third spatial reference point to form a spatial triangular topology network; Step 34: Receive spatial triangular topology network data, calculate the minimum circumcircle of the spatial triangular topology network, determine the minimum circular boundary containing three spatial reference points, and record the center coordinates and radius of the minimum circumcircle; add the circumcircle radius to the preset buffer distance parameter to obtain the expanded buffer radius; generate a circular spatial analysis region with the center position as the center and the buffer radius as the boundary; verify whether the circular spatial analysis region completely covers the conductor path of the target segmented line; if there are uncovered segments, increase the buffer radius according to the set ratio and regenerate; if it completely covers the target segmented line, output the spatial analysis region and the set of boundary coordinates of the region.

5. The method for predicting short-circuit faults in transmission lines according to claim 4, characterized in that, Step 4: Divide the spatial analysis area into multiple sub-regional units, extract the elevation data sequence of each sub-regional unit, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-regional unit based on the vegetation cover image data, input the terrain slope feature value and vegetation density level into the risk correction mapping table, and generate the initial spatial risk correction factor. Based on the initial spatial risk correction factor of adjacent sub-region units, a gradient optimization strategy is used for smooth iteration to obtain the spatial risk correction factor, including: Step 41: Receive the set of boundary coordinates of the spatial analysis region, and divide the spatial analysis region into equal-area grids based on the preset grid size to generate multiple sub-region units; Step 42: Extract elevation monitoring point data within the sub-region unit and construct an elevation data sequence sorted by spatial location; calculate the elevation difference between adjacent points in the elevation data sequence based on the gradient optimization strategy, and take the maximum value of the absolute value of the elevation difference as the terrain slope feature value. Step 43: Retrieve vegetation cover images from environmental monitoring data; identify the percentage of vegetation pixels within a unit and classify vegetation density levels according to a preset ratio range; Step 44: Input the terrain slope feature value and vegetation density level into the preset risk correction mapping table, and query to obtain the initial spatial risk correction factor; Step 45: Receive the initial spatial risk correction factor set, which contains the initial factor values ​​of all sub-region units; select the sub-region unit to be processed as the central unit, and obtain the initial factor values ​​of the eight adjacent sub-region units; calculate the arithmetic mean of the initial factor values ​​of the central unit and the eight adjacent units, and record it as the neighborhood mean; compare the initial factor value of the central unit with the neighborhood mean; if the central value is greater than the neighborhood mean, decrease the central unit factor value by a set step size; if the central value is less than the neighborhood mean, increase the central unit factor value by the same step size; if the central value is equal to the neighborhood mean, keep the factor value unchanged; use the adjusted central unit factor value as the new factor value for the next iteration. Step 46: Iterate until the change in all unit factor values ​​is less than the convergence threshold, and output the final spatial risk correction factor.

6. The method for predicting short-circuit faults in transmission lines according to claim 5, characterized in that, Step 5: Based on the predicted short-circuit probability and spatial risk correction factor, combined with power grid topology data, calculate the cascading risk value of the target segment region, including: Step 51: Receive the predicted short-circuit probability value of the target segment region and the corresponding spatial risk correction factor; multiply the predicted short-circuit probability value of the target segment region by the corresponding spatial risk correction factor to generate the basic risk value of the segment region. Step 52: Call the power grid topology data to identify the direct connection relationships of the target segment region in the power grid; determine the set of downstream adjacent segment regions with direct electrical connections; Step 53: Obtain the basic risk value of the downstream region, multiply the basic risk value of the downstream region by the preset impact propagation coefficient to obtain the risk contribution value of the downstream region to the target segment region; sum up the risk contribution values ​​of all downstream adjacent segment regions to obtain the total risk impact of the neighborhood. Step 54: Add the base risk value to the sum of the neighboring risk impacts to obtain the final cascading risk value of the target segmented region; Step 55: Output the chain risk values ​​for all segmented regions, which are used for quantitative assessment of the chain propagation risk of power grid short-circuit faults.

7. The method for predicting short-circuit faults in transmission lines according to claim 6, characterized in that, Step 6: Based on the chain risk value, generate risk level labels and handling strategy instructions for the target segmented areas, including: Step 61: Receive the chain risk values ​​of all target segment regions; preset low-risk threshold range, medium-risk threshold range and high-risk threshold range to form risk level classification rules; Step 62: Compare the chain risk value of each target segment area with the preset risk threshold range; if the chain risk value falls into the low risk threshold range, generate a low risk level label; if the chain risk value falls into the medium risk threshold range, generate a medium risk level label; if the chain risk value falls into the high risk threshold range, generate a high risk level label. Step 63: Invoke the preset risk-strategy mapping table to map low-risk level labels to basic inspection strategy instructions; map medium-risk level labels to enhanced monitoring strategy instructions; and map high-risk level labels to emergency response strategy instructions. Step 64: Based on the power grid topology data, perform coordination verification on the handling strategy instructions for adjacent target segment areas. If adjacent segment areas are all marked with high-risk level labels, upgrade the handling strategy instructions to regional linkage handling instructions and add cross-segment fault isolation schemes; output the final risk level labels of each target segment area and the corresponding handling strategy instructions to the power grid dispatch terminal.

8. The method for predicting short-circuit faults in transmission lines according to claim 7, characterized in that, Step 7: When the output risk level label is high risk, trigger the integrity verification of the multidimensional monitoring data, including: Step 71: Based on the high-risk level label, extract multi-dimensional monitoring data associated with the target segmented areas; Step 72: Perform time window verification on the multidimensional monitoring data to verify whether the timestamps of the data are within the preset valid monitoring time range; Step 73: Perform missing item verification on the data that passed the time window verification, check the field integrity of each data source item by item, and identify missing data items; if there are missing data items, generate a data missing alarm command and start the supplementary collection process, calling the backup monitoring node or the data proxy node of the adjacent segment area to fill in the missing items. Step 74: Repeat the verification process for the supplemented data until the multidimensional monitoring data completely passes the integrity verification or reaches the preset maximum retry threshold. Step 75: If data is still missing after reaching the maximum retry threshold, mark the data in the high-risk segment as "incomplete" and append the list of missing data items to the emergency response strategy instruction and output it to the power grid dispatch terminal.

9. The method for predicting short-circuit faults in transmission lines according to claim 8, characterized in that, Step 8: When the verification reveals missing monitoring data, dynamically supplement the multidimensional monitoring data of the target segment area using monitoring data from adjacent segment areas and historical data, and then re-collect the supplemented data, including: Step 81: When a missing data item is identified, initiate the dynamic data completion process; Step 82: Based on the type of missing data item, extract the corresponding data from the real-time monitoring data of adjacent segment areas or the historical monitoring data of the target segment area; Step 83: Perform spatiotemporal alignment processing on the extracted data and map it to the monitoring timestamp and spatial location of the target segment region; Step 84: Inject the spatiotemporally aligned data into the multidimensional monitoring dataset of the target segmented region to generate the completed dataset; Step 85: Trigger the supplementary data collection process, and call the data proxy node to perform a second integrity check on the completed dataset; Step 86: If the second verification passes, mark the completed data as "verified"; if the verification fails, return to execute a new round of completion operation until the verification passes or the maximum number of retries threshold is reached.

10. A short-circuit fault prediction system for transmission lines, wherein the system implements the method as described in any one of claims 1 to 9, characterized in that, include: The acquisition module is used to collect multi-dimensional monitoring data of the target segmented area in real time; The detection module is used to generate short-circuit probability prediction values ​​for the target segmented area based on multi-dimensional monitoring data; The analysis module is used to set three preset spatial reference points within the target segmented area to construct a spatial analysis region; The optimization module is used to divide the spatial analysis area into multiple sub-regional units, extract the elevation data sequence of each sub-regional unit, calculate the terrain slope feature value based on the gradient optimization strategy, identify the vegetation density level of each sub-regional unit based on the vegetation cover image data, and input the terrain slope feature value and vegetation density level into the risk correction mapping table to generate the initial spatial risk correction factor. Based on the initial spatial risk correction factor of adjacent sub-region units, a gradient optimization strategy is used to perform smooth iteration to obtain the spatial risk correction factor. The calculation module is used to calculate the cascading risk value of the target segment area based on the predicted short-circuit probability value and the spatial risk correction factor, combined with the power grid topology data. The processing module is used to generate risk level labels and handling strategy instructions for target segmented areas based on the chain risk value; The verification module is used to trigger integrity verification of the multidimensional monitoring data when the output risk level label is high risk. The data processing module is used to dynamically supplement the multidimensional monitoring data of the target segment area by using monitoring data from adjacent segment areas and historical data when the verification finds that the monitoring data is missing, and then re-collect the supplemented data.

Images