Intelligent power grid fault detection system and method based on multi-source heterogeneous data

Abstract

The invention discloses an intelligent power grid fault detection system and method based on multi-source heterogeneous data, and relates to the technical field of power grid fault detection, and the method comprises the steps: evaluating the fault detection accuracy of a monitoring parameter in fault range data to a marked power grid node; analyzing a node approximation degree between different power grid nodes, and evaluating an abnormal state of fault detection on the power grid nodes by the fault range data; evaluating the fault capture accuracy of the monitoring parameters in the fault range data on the target power grid node; performing anomaly identification on the power grid nodes in the power grid to obtain abnormal power grid nodes, obtaining a power grid topological graph of the power grid, and obtaining abnormal region data in the power grid according to the distribution situation of the abnormal power grid nodes in the power grid topological graph; according to the invention, fault detection is carried out on the power grid equipment to obtain abnormal power grid equipment, the abnormal power grid equipment in the power grid is maintained, rapid and accurate identification of faults in the power grid is realized, and the efficiency and accuracy of power grid fault detection are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of power grid fault detection technology, specifically to a smart power grid fault detection system and method based on multi-source heterogeneous data. Background Technology

[0002] During power grid operation, data of varying types, formats, and structures are generated from different sources and acquisition devices. Using multi-source anomaly data in power grid fault detection can effectively integrate the data, thereby solving the problems caused by a single data source. Furthermore, the fusion of multi-source data can also verify information from various parties, greatly improving the accuracy of power grid fault location and diagnosis and avoiding misjudgments. At the same time, multi-source heterogeneous data can also be used to merge different power grid data, thereby constructing a comprehensive profile of power grid faults, enabling power grid faults to be quickly and comprehensively understood.

[0003] Currently, the main method for fault detection in power grids is to monitor the equipment within the grid. When the fault-related parameters of the equipment are outside the preset range, the equipment is judged as abnormal, thus identifying the abnormal equipment in the power grid. However, in reality, the power grid structure is complex, and different equipment operates in different environments. Using preset thresholds is difficult to accurately grasp the status of the equipment. At the same time, since the power grid is connected through transmission lines, a fault in a single device may lead to abnormalities in several related devices. Traditional methods cannot identify the location of the fault in the power grid. Furthermore, because the power grid needs to transport electricity and is easily affected by natural disasters, it is necessary to check its physical integrity. However, most of the equipment in the power grid is far apart, and some are even located in inaccessible areas. If people manually check the abnormal areas of the power grid, it will undoubtedly consume a lot of manpower and resources, and it will not be possible to accurately identify the status of the equipment. Summary of the Invention

[0004] The purpose of this invention is to provide a smart grid fault detection system and method based on multi-source heterogeneous data to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a smart grid fault detection method based on multi-source heterogeneous data, the method comprising: Step S1: Obtain historical fault records and fault range data of marked power grid nodes in the power grid, evaluate the accuracy of monitoring parameters in the fault range data for fault detection of marked power grid nodes, and obtain marked fault range data; Step S2: Obtain fault range data and node data of power grid nodes in the power grid, analyze the degree of node approximation between different power grid nodes, obtain the approximate power grid node set, obtain the marked power grid node from the approximate power grid node set and record it as the approximate marked power grid node, evaluate the abnormal state of fault range data for fault detection of power grid nodes, and obtain the target power grid node. Step S3: Obtain the marked fault range data in the approximate marked power grid node of the target power grid node, evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node; Step S4: Based on the target fault range data and the marked fault range data, optimize and adjust the fault range data of the power grid nodes in the power grid, identify the abnormal power grid nodes in the power grid, obtain the abnormal power grid nodes, obtain the power grid topology map of the power grid, and obtain the abnormal area data in the power grid based on the distribution pattern of the abnormal power grid nodes in the power grid topology map. Step S5: Based on the abnormal area data, dispatch drones to photograph the power grid equipment in the power grid, acquire equipment images, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

[0006] Furthermore, step S1 includes: Step S11: Obtain the grid node in the power grid that has a fault and mark it as a marked grid node; obtain the fault range data of the marked grid node; mark the preset range of various monitoring parameters within the grid node from the fault range data. Obtain historical fault records for each marked power grid node, and extract data of various monitoring parameters within the marked power grid node from the historical fault records; Step S12: When the maximum or minimum value of a certain monitoring parameter in a certain historical fault record is outside the range of the preset fault range data, the platform issues a fault warning to the marked power grid node. If no fault is detected when the marked power grid node is subsequently inspected, then the certain historical fault record is recorded as the first historical fault record. Obtain each first historical fault record of the marked power grid node. When the maximum or minimum value of the monitoring parameter a in the first historical fault record is outside the range of the fault range data, it is determined that there is a fault detection anomaly of the monitoring parameter a in the marked power grid node. Obtain the total number of times E, the monitoring parameter a, exists in the marked power grid nodes. sum Obtain the total number γ of each first historical fault record. sum Calculate the outlier value Q of the monitoring parameter a. a =E (a，sum) / γ (a，sum) When detecting outlier Qa If the value exceeds the preset abnormal detection threshold, then the monitoring parameter a is determined to be inaccurate for detecting faults at marked power grid nodes. Obtain the range constructed by the minimum and maximum values ​​of the a-th monitoring parameter in each first historical fault record, and denot it as the marked range of the a-th monitoring parameter in the marked power grid node; Step S13: Remove each first historical fault record from the other historical fault records to obtain the second historical fault record that marks the power grid node; Calculate the range distance value S of the βth monitoring parameter in the second historical fault record. β : , Among them, H (max,β) and H (min,β) The range H of the βth monitoring parameter in the fault range data is respectively. β The maximum and minimum values ​​of H; △ For the range H β The average of the maximum and minimum values; H' (max,β) and H' (min,β) These are the maximum and minimum values ​​of the βth monitoring parameter in the second historical fault record, respectively; When the range distance value S β If the distance is less than the preset range threshold, it is determined that the βth monitoring parameter has a distance anomaly in the second historical fault record. The range anomaly value B of the β-th monitoring parameter is obtained by comparing the total number of second historical fault records showing distance anomalies in the marked power grid node with the total number of second historical fault records for the marked power grid node. β ; When the range outlier B β If the value exceeds the preset abnormal threshold, it is determined that the β-th monitoring parameter is inaccurate in detecting faults in the marked power grid node. The range constructed by the maximum and minimum values ​​of the β-th monitoring parameter within the fault range data is obtained from several second historical fault records and recorded as the marked range of the β-th monitoring parameter in the marked power grid node. Step S14: Obtain the marking range of several monitoring parameters in the marked power grid node that are determined to be inaccurate in fault detection, and replace the range of several monitoring parameters in the fault range data of the marked power grid node to obtain the marked fault range data of the marked power grid node.

[0007] Furthermore, step S2 includes: Step S21: Obtain fault range data and node data of power grid nodes, and extract data of various node indicators from the node data; Step S22: Analyze the degree of node approximation between different power grid nodes to obtain an approximate set of power grid nodes. The specific analysis process is as follows: Obtain the range L of the d-th monitoring parameter for the e-th power grid node and the e-th power grid node respectively. d =[L d,min ,L d,max ] and L e d =[L e d,min ,L e d,max ], where L d,min L d,max L represents the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the power grid node. e d,min ,L e d,max These are the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the e-th power grid node, respectively. When a certain value exists in the range L d and L e d Inside, obtain L d and L e d The intersection Y between them e d =[Y e d,min ,Y e d,max ], where Y e d,min =max{L d,min L d,max},,Y e d,max =min{L d,max L e d,max}, get range L d With L e d Intersection distance between |Y e d |=Y e d,max -Y e d,min Conversely, the intersection distance length |Y e d |=0; Calculation range L d Distance length |L d |=L d,max -L d,min Calculation range L ed Distance length |L e d |=L e d,max -L e d,min ; Calculate the approximate range C of the d-th monitoring parameter between the e-th power grid node and the d-th power grid node. e d : , Obtain the average value of the approximate range values ​​of various monitoring parameters between the power grid node and the e-th power grid node, and denot it as the approximate fault range value C between the power grid node and the e-th power grid node. e ; Step S23: Based on the node data of the power grid nodes, construct the node feature vectors of the power grid nodes, obtain the node feature vector of the e-th power grid node, calculate the cosine similarity between the power grid node and the e-th power grid node, and obtain the node approximation value F. e ; Calculate the approximate node value P between the e-th node and the power grid node. e =η C ×C e +η F ×F e , where η C η F These are the preset fault weight coefficient and node weight coefficient, η. C >0, η F >0, η C +η F =1; When the node approximation value P e If the value is greater than the preset approximation threshold, then the e-th power grid node is determined to be approximate with the power grid node, and the e-th power grid node is recorded as the approximate power grid node of the power grid node. Step S24: Obtain several approximate power grid nodes in the power grid and aggregate them to obtain approximate marked power grid nodes; Marked power grid nodes are obtained from the set of approximate power grid nodes and recorded as approximate marked power grid nodes. The total number of approximate marked power grid nodes is divided by the total number of approximate power grid nodes to obtain the abnormal value of the power grid node. When the abnormal value is greater than the preset abnormal threshold, the fault range data of the power grid node is determined. The power grid node is detected to have an abnormality and is recorded as the target power grid node.

[0008] Furthermore, step S3 includes: Step S31: Obtain the approximate marked grid nodes of the target grid node and obtain the marked fault range data of each approximate marked grid node; Step S32: Obtain the fault range data of the target power grid node, and evaluate the accuracy of the g-th monitoring parameter in the fault range data for capturing faults in the target power grid. The specific evaluation process is as follows: Obtain several monitoring parameters from the approximate marked power grid nodes that are determined to have inaccurate operational detection. Then, obtain the sum of the approximate node values ​​between each approximate marked power grid node and the target power grid node to obtain the approximate total value U. sum ; Obtain several approximate marked power grid nodes for which the g-th monitoring parameter is determined to be inaccurate in fault detection, and obtain the sum U of the node approximation values ​​between the several approximate marked power grid nodes and the target power grid node. (g,sum) Calculate the abnormal value Q of the g-th monitoring parameter in the target power grid node. g =U (g,sum) / U sum ; Step S33: When parameter Q is an outlier g If the range of the g-th monitoring parameter in the preset fault range data is less than the preset abnormal threshold, it is determined that the range of the g-th monitoring parameter in the preset fault range data accurately captures the fault of the target power grid node; otherwise, it is determined that the range of the g-th detection parameter in the preset fault range data does not accurately capture the fault of the target power grid node, and the g-th monitoring parameter is recorded as the abnormal monitoring parameter of the target power grid node. Step S34: Obtain a certain abnormal monitoring parameter of the target power grid node, obtain the average of the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record it as the range reference value ζ; Obtain the absolute value of the difference between the range baseline value ζ and the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record this as the upper limit distance τ of a certain abnormal monitoring parameter in the approximate marked power grid node. max and the lower limit distance τ min ; Calculate the range reference value ζ´ of a certain abnormal monitoring parameter in the target power grid node: , Where, ζ i Let ζ be the range reference value of a certain abnormal monitoring parameter in the i-th approximate marker of the target power grid node; P i Let be the node approximation value between the target power grid node and the i-th approximate marked power grid node; j is the total number of all approximate marked power grid nodes; Step S35: Calculate the upper limit distance τ´ of a certain abnormal monitoring parameter in the target power grid node.max and the lower limit distance τ´ min To obtain the target range W=[ζ´-τ´ of a certain abnormal monitoring parameter in the target power grid node. min ,ζ´+τ´ max ]; Obtain the target range of each abnormal monitoring parameter in the target power grid node. Use the target range of each abnormal monitoring parameter to replace the range of each abnormal monitoring parameter in the fault range data of the target power grid node. Record the replaced fault range data as the target fault range data of the target power grid node. In the above steps, the target power grid node is obtained based on the marked power grid node that has an abnormal fault detection. This is because some power grid nodes in the power grid do not have faults, and it is impossible to accurately determine the fault detection accuracy of the fault range data set by the platform based on the actual historical records. However, based on the fault range data between the power grid node and the marked power grid node and the similarity of the nodes, the power grid node that may have an abnormal fault detection can be obtained, which is the target power grid node of this application. The fault range data is obtained based on the marked fault range data of the approximate marked power grid node of the target power grid node. The fault detection range is determined based on the approximate relationship between the target power grid node and different approximate marked power grid nodes, as well as the various monitoring parameters in different approximate marked power grid nodes. This allows for obtaining the range that can capture the fault of the target power grid node to the greatest extent, which greatly ensures the accuracy of subsequent judgment of abnormalities of power grid nodes in the power grid.

[0009] Furthermore, step S4 includes: Step S41: Obtain the fault range data of each power grid node in the power grid, obtain the target fault range data of each target power grid node in the power grid, and obtain the marked fault range data of each marked power grid node in the power grid. The fault range data of power grid nodes in the power grid are optimized and adjusted, specifically as follows: The fault range data of the marked power grid nodes and the target fault range data of the target power grid nodes are used to replace the fault range data of the corresponding power grid nodes in each power grid node. Step S42: Use the fault range data of each replaced power grid node to perform fault detection on the power grid. When the maximum and minimum values ​​of a certain monitoring parameter in a certain power grid node are not within the range of the optimized fault range data, it is determined that a certain power grid node is abnormal and the certain power grid node is recorded as an abnormal power grid node. Step S43: Obtain the power grid topology map. Based on the distribution of abnormal power grid nodes in the power grid topology map, obtain the abnormal areas in the power grid. Collect and aggregate the abnormal areas in the power grid to obtain the abnormal area data of the power grid.

[0010] Furthermore, step S5 includes: Step S51: Obtain abnormal area data of the power grid, obtain various abnormal areas of the power grid from the abnormal area data, obtain the power grid equipment in the abnormal areas, and use a drone to take pictures of the power grid equipment in the abnormal areas to obtain equipment images; Step S52: Preprocess the equipment image and use a preset image recognition model to input the preprocessed equipment image into the detection model to perform fault detection on the power grid equipment, identify abnormal power grid equipment, and dispatch maintenance personnel to repair the abnormal power grid equipment in the power grid.

[0011] Based on the above method, a smart grid fault detection system based on multi-source heterogeneous data is also proposed. The system includes a range acquisition module, an anomaly identification module, and a fault detection module. The range acquisition module is used to acquire the marked fault range data of the marked power grid nodes in the power grid, evaluate the abnormal state of the fault range data for fault detection of the power grid nodes, obtain the target power grid node, and acquire the target fault range data of the target power grid node. The anomaly identification module is used to identify anomalies in the power grid nodes based on the target fault range data and the marked fault range data, thereby obtaining the abnormal power grid nodes and obtaining the abnormal area data in the power grid based on the distribution pattern of the abnormal power grid nodes in the power grid topology map. The fault detection module is used to use drones to photograph power grid equipment in the power grid, acquire equipment images, detect faults in the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

[0012] Furthermore, the range acquisition module includes a marker range acquisition unit, a fault detection anomaly assessment unit, and a target range acquisition unit; The marking range acquisition unit is used to evaluate the accuracy of fault detection of the marked power grid nodes by the monitoring parameters in the fault range data, and to mark the fault range data; The fault detection anomaly assessment unit is used to acquire approximate marked grid nodes and, based on the marked fault range data of the approximate marked grid nodes, assess the abnormal state of the grid nodes for fault detection using the fault range data, and obtain the target grid node. The target range acquisition unit is used to evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node.

[0013] Furthermore, the anomaly detection module includes a range optimization unit and an anomaly detection unit; The range optimization unit is used to acquire the marked fault range data of the marked power grid nodes, acquire the target fault range data of the target power grid node, and optimize and adjust the fault range data of the power grid nodes in the power grid. The anomaly identification unit is used to identify anomalies in the power grid nodes based on the optimized fault range data of the power grid nodes, obtain the abnormal power grid nodes, and obtain the abnormal area data in the power grid by combining the power grid topology map.

[0014] Furthermore, the fault detection module includes a fault detection unit; The fault detection unit is used to acquire images of power grid equipment in the abnormal area based on the abnormal area data, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

[0015] Compared with existing technologies, the beneficial effects of this invention are: This invention achieves accurate detection of power grid faults. Starting from the key power grid nodes that actually transport electricity in the power grid, and considering that the scope of anomaly detection performed by the platform on power grid nodes may not actually match the reality, historical fault records of marked power grid nodes that have experienced faults are acquired. Based on these historical fault records, the accuracy of fault detection of power grid nodes using the platform's preset fault range data is evaluated, thereby obtaining marked fault range data for marked power grid nodes. Furthermore, based on the degree of node similarity between power grid nodes without faults and marked power grid nodes, the risk of fault detection is obtained. The system targets dangerous power grid nodes and adjusts the platform's preset fault range data for these nodes based on monitoring parameters in the fault range data. It also optimizes the fault range data of power grid nodes within the power grid, accurately detecting faults at these nodes. Furthermore, based on the power grid topology map, it quickly and accurately identifies areas with equipment faults within the power grid and rapidly detects faults in power grid equipment within abnormal areas using images captured by drones. By leveraging electrical quantity characteristics, topological relationships, and image information within the power grid, it achieves rapid and accurate identification of faults in the power grid, significantly improving the efficiency and accuracy of power grid fault detection. Attached Figure Description

[0016] Figure 1 This is a method logic diagram of the smart grid fault detection method based on multi-source heterogeneous data of the present invention; Figure 2 This is a flowchart of the module of the smart grid fault detection system based on multi-source heterogeneous data of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: Figures 1-2 As shown, this invention provides a technical solution: a smart grid fault detection method based on multi-source heterogeneous data, the method comprising: Step S1: Obtain historical fault records and fault range data of marked power grid nodes in the power grid, evaluate the accuracy of monitoring parameters in the fault range data for fault detection of marked power grid nodes, and obtain marked fault range data; Step S1 includes: Step S11: Obtain the grid node in the power grid that has a fault and mark it as a marked grid node; obtain the fault range data of the marked grid node; mark the preset range of various monitoring parameters within the grid node from the fault range data. For example, the various monitoring parameters include voltage and current parameters at power grid nodes; Obtain historical fault records for each marked power grid node, and extract data of various monitoring parameters within the marked power grid node from the historical fault records; Step S12: When the maximum or minimum value of a certain monitoring parameter in a certain historical fault record is outside the range of the preset fault range data, the platform issues a fault warning to the marked power grid node. If no fault is detected when the marked power grid node is subsequently inspected, then the certain historical fault record is recorded as the first historical fault record. Obtain each first historical fault record of the marked power grid node. When the maximum or minimum value of the monitoring parameter a in the first historical fault record is outside the range of the fault range data, it is determined that there is a fault detection anomaly of the monitoring parameter a in the marked power grid node. Obtain the total number of times E, the monitoring parameter a, exists in the marked power grid nodes. sum Obtain the total number γ of each first historical fault record. sum Calculate the outlier value Q of the monitoring parameter a. a =E (a，sum) / γ (a，sum) When detecting outlier Q a If the value exceeds the preset abnormal detection threshold, then the monitoring parameter a is determined to be inaccurate for detecting faults at marked power grid nodes. Obtain the range constructed by the minimum and maximum values ​​of the a-th monitoring parameter in each first historical fault record, and denot it as the marked range of the a-th monitoring parameter in the marked power grid node; Step S13: Remove each first historical fault record from the other historical fault records to obtain the second historical fault record that marks the power grid node; Calculate the range distance value S of the βth monitoring parameter in the second historical fault record. β : , Among them, H (max,β) and H (min,β) The range H of the βth monitoring parameter in the fault range data is respectively. β The maximum and minimum values ​​of H; △ For the range H β The average of the maximum and minimum values; H' (max,β) and H' (min,β) These are the maximum and minimum values ​​of the βth monitoring parameter in the second historical fault record, respectively; When the range distance value S β If the distance is less than the preset range threshold, it is determined that the βth monitoring parameter has a distance anomaly in the second historical fault record. The range anomaly value B of the β-th monitoring parameter is obtained by comparing the total number of second historical fault records showing distance anomalies in the marked power grid node with the total number of second historical fault records for the marked power grid node. β ; When the range outlier B β If the value exceeds the preset abnormal threshold, it is determined that the β-th monitoring parameter is inaccurate in detecting faults in the marked power grid node. The range constructed by the maximum and minimum values ​​of the β-th monitoring parameter within the fault range data is obtained from several second historical fault records and recorded as the marked range of the β-th monitoring parameter in the marked power grid node. Step S14: Obtain the marking range of several monitoring parameters in the marked power grid node that are determined to be inaccurate in fault detection, and replace the range of several monitoring parameters in the fault range data of the marked power grid node to obtain the marked fault range data of the marked power grid node. Step S2: Obtain fault range data and node data of power grid nodes in the power grid, analyze the degree of node approximation between different power grid nodes, obtain the approximate power grid node set, obtain the marked power grid node from the approximate power grid node set and record it as the approximate marked power grid node, evaluate the abnormal state of fault range data for fault detection of power grid nodes, and obtain the target power grid node. Step S2 includes: Step S21: Obtain fault range data and node data of power grid nodes, and extract data of various node indicators from the node data; For example, various node indicators include node degree (the number of edges directly connected to the grid node) and node voltage level (the specific voltage level in the grid node, such as 10kV, 110kV, etc.). Step S22: Analyze the degree of node approximation between different power grid nodes to obtain an approximate set of power grid nodes. The specific analysis process is as follows: Obtain the range L of the d-th monitoring parameter for the e-th power grid node and the e-th power grid node respectively. d =[L d,min ,L d,max ] and L e d =[L e d,min ,L e d,max ], where L d,min L d,max L represents the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the power grid node. e d,min ,L e d,max These are the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the e-th power grid node, respectively. When a certain value exists in the range L d and L e d Inside, obtain L d and L e d The intersection Y between them e d =[Y e d,min ,Y e d,max ], where Y e d,min =max{L d,min L d,max},,Y e d,max =min{L d,max L e d,max}, get range L d With L e d Intersection distance between |Y e d |=Y e d,max -Y e d,minConversely, the intersection distance length |Y e d |=0; Calculation range L d Distance length |L d |=L d,max -L d,min Calculation range L e d Distance length |L e d |=L e d,max -L e d,min ; Calculate the approximate range C of the d-th monitoring parameter between the e-th power grid node and the d-th power grid node. e d : , Obtain the average value of the approximate range values ​​of various monitoring parameters between the power grid node and the e-th power grid node, and denot it as the approximate fault range value C between the power grid node and the e-th power grid node. e ; Step S23: Based on the node data of the power grid nodes, construct the node feature vectors of the power grid nodes, obtain the node feature vector of the e-th power grid node, calculate the cosine similarity between the power grid node and the e-th power grid node, and obtain the node approximation value F. e ; Calculate the approximate node value P between the e-th node and the power grid node. e =η C ×C e +η F ×F e , where η C η F These are the preset fault weight coefficient and node weight coefficient, η. C >0, η F >0, η C +η F =1; When the node approximation value P e If the value is greater than the preset approximation threshold, then the e-th power grid node is determined to be approximate with the power grid node, and the e-th power grid node is recorded as the approximate power grid node of the power grid node. Step S24: Obtain several approximate power grid nodes in the power grid and aggregate them to obtain approximate marked power grid nodes; Marked power grid nodes are obtained from the set of approximate power grid nodes and recorded as approximate marked power grid nodes. The total number of approximate marked power grid nodes is divided by the total number of approximate power grid nodes to obtain the abnormal value of the power grid node. When the abnormal value is greater than the preset abnormal threshold, the fault range data of the power grid node is determined. If the fault detection of the power grid node shows an anomaly, the power grid node is recorded as the target power grid node. Step S3: Obtain the marked fault range data in the approximate marked power grid node of the target power grid node, evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node; Step S3 includes: Step S31: Obtain the approximate marked grid nodes of the target grid node and obtain the marked fault range data of each approximate marked grid node; Step S32: Obtain the fault range data of the target power grid node, and evaluate the accuracy of the g-th monitoring parameter in the fault range data for capturing faults in the target power grid. The specific evaluation process is as follows: Obtain several monitoring parameters from the approximate marked power grid nodes that are determined to have inaccurate operational detection. Then, obtain the sum of the approximate node values ​​between each approximate marked power grid node and the target power grid node to obtain the approximate total value U. sum ; Obtain several approximate marked power grid nodes for which the g-th monitoring parameter is determined to be inaccurate in fault detection, and obtain the sum U of the node approximation values ​​between the several approximate marked power grid nodes and the target power grid node. (g,sum) Calculate the abnormal value Q of the g-th monitoring parameter in the target power grid node. g =U (g,sum) / U sum ; Step S33: When parameter Q is an outlier g If the range of the g-th monitoring parameter in the preset fault range data is less than the preset abnormal threshold, it is determined that the range of the g-th monitoring parameter in the preset fault range data accurately captures the fault of the target power grid node; otherwise, it is determined that the range of the g-th detection parameter in the preset fault range data does not accurately capture the fault of the target power grid node, and the g-th monitoring parameter is recorded as the abnormal monitoring parameter of the target power grid node. Step S34: Obtain a certain abnormal monitoring parameter of the target power grid node, obtain the average of the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record it as the range reference value ζ; Obtain the absolute value of the difference between the range baseline value ζ and the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record this as the upper limit distance τ of a certain abnormal monitoring parameter in the approximate marked power grid node.max and the lower limit distance τ min ; Calculate the range reference value ζ´ of a certain abnormal monitoring parameter in the target power grid node: , Where, ζ i Let ζ be the range reference value of a certain abnormal monitoring parameter in the i-th approximate marker of the target power grid node; P i Let be the node approximation value between the target power grid node and the i-th approximate marked power grid node; j is the total number of all approximate marked power grid nodes; Step S35: Calculate the upper limit distance τ´ of a certain abnormal monitoring parameter in the target power grid node. max and the lower limit distance τ´ min To obtain the target range W=[ζ´-τ´ of a certain abnormal monitoring parameter in the target power grid node. min ,ζ´+τ´ max ]; For example, the upper limit distance τ´ max and the lower limit distance τ´ min The specific calculation formula is as follows: Wherein, the upper limit distance τ´ max The calculation formula is: , Where, τ （i,max） The upper limit distance of a certain abnormal monitoring parameter in the i-th approximate marked power grid node of the target power grid node; Wherein, the lower limit distance τ´ min The calculation formula is: , Where, τ （i,mmin） The lower limit distance of a certain abnormal monitoring parameter in the i-th approximate marked power grid node of the target power grid node; Obtain the target range of each abnormal monitoring parameter in the target power grid node. Use the target range of each abnormal monitoring parameter to replace the range of each abnormal monitoring parameter in the fault range data of the target power grid node. Record the replaced fault range data as the target fault range data of the target power grid node. Step S4 includes: Step S41: Obtain the fault range data of each power grid node in the power grid, obtain the target fault range data of each target power grid node in the power grid, and obtain the marked fault range data of each marked power grid node in the power grid. The fault range data of power grid nodes in the power grid are optimized and adjusted, specifically as follows: The fault range data of the marked power grid nodes and the target fault range data of the target power grid nodes are used to replace the fault range data of the corresponding power grid nodes in each power grid node. Step S42: Use the fault range data of each replaced power grid node to perform fault detection on the power grid. When the maximum and minimum values ​​of a certain monitoring parameter in a certain power grid node are not within the range of the optimized fault range data, it is determined that a certain power grid node is abnormal and the certain power grid node is recorded as an abnormal power grid node. Step S43: Obtain the power grid topology map of the power grid. Based on the distribution of abnormal power grid nodes in the power grid topology map, obtain the abnormal areas in the power grid. Collect and aggregate the abnormal areas in the power grid to obtain the abnormal area data of the power grid. For example, a power grid topology diagram is a topology diagram constructed based on the topological relationships of the physical structure of the power grid, which clearly shows the connection relationships between the various parts of the power grid; The power grid topology mainly includes power grid nodes and edges. Power grid nodes include the outlets of power plants, substations, etc. The edges in a power grid include transmission lines, distribution lines, poles, transformers, etc. For example, to obtain a power grid topology map, and based on the distribution of abnormal power grid nodes in the topology map, to identify abnormal regions in the power grid, the specific acquisition process is as follows: To identify abnormal power grid nodes on the power grid topology map, when an abnormal power grid node is connected to another power grid node through a path in the power grid topology map, the area formed by the path between the abnormal power grid node and the other abnormal power grid node is denoted as an abnormal area. Step S5: Based on the abnormal area data, dispatch drones to photograph the power grid equipment in the power grid, acquire equipment images, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid. Step S5 includes: Step S51: Obtain abnormal area data of the power grid, obtain various abnormal areas of the power grid from the abnormal area data, obtain the power grid equipment in the abnormal areas, and use a drone to take pictures of the power grid equipment in the abnormal areas to obtain equipment images; Step S52: Preprocess the equipment image and use a preset image recognition model to input the preprocessed equipment image into the detection model to perform fault detection on the power grid equipment, identify abnormal power grid equipment, and dispatch maintenance personnel to repair the abnormal power grid equipment in the power grid. For example, the preprocessing of equipment images includes resizing the equipment images and image enhancement; For example, the detection model is built using a convolutional neural network model, trained using a large number of pre-captured images of power grid equipment, including both normal and damaged images, and the model's recognition accuracy is evaluated by accuracy (the proportion of images correctly identified) and precision (the proportion of samples that the model predicts to be "damaged" but are actually damaged). Based on the above method, a smart grid fault detection system based on multi-source heterogeneous data is also proposed. The system includes a range acquisition module, an anomaly identification module, and a fault detection module. The range acquisition module is used to acquire the marked fault range data of the marked power grid nodes in the power grid, evaluate the abnormal state of the fault range data for fault detection of the power grid nodes, obtain the target power grid node, and acquire the target fault range data of the target power grid node. The anomaly identification module is used to identify anomalies in the power grid nodes based on the target fault range data and the marked fault range data, thereby obtaining the abnormal power grid nodes and obtaining the abnormal area data in the power grid based on the distribution pattern of the abnormal power grid nodes in the power grid topology map. The fault detection module is used to use drones to photograph power grid equipment in the power grid, acquire equipment images, detect faults in the power grid equipment, identify abnormal power grid equipment, and repair abnormal power grid equipment in the power grid. The range acquisition module includes a marker range acquisition unit, a fault detection anomaly assessment unit, and a target range acquisition unit. The marking range acquisition unit is used to evaluate the accuracy of fault detection of the marked power grid nodes by the monitoring parameters in the fault range data, and to mark the fault range data; The fault detection anomaly assessment unit is used to acquire approximate marked grid nodes and, based on the marked fault range data of the approximate marked grid nodes, assess the abnormal state of the grid nodes for fault detection using the fault range data, and obtain the target grid node. The target range acquisition unit is used to evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node. The anomaly detection module includes a range optimization unit and an anomaly detection unit. The range optimization unit is used to acquire the marked fault range data of the marked power grid nodes, acquire the target fault range data of the target power grid node, and optimize and adjust the fault range data of the power grid nodes in the power grid. The anomaly identification unit is used to identify anomalies in the power grid nodes based on the optimized fault range data of the power grid nodes, obtain the abnormal power grid nodes, and combine the power grid topology map to obtain the abnormal area data in the power grid. The fault detection module includes a fault detection unit; The fault detection unit is used to acquire images of power grid equipment in the abnormal area based on the abnormal area data, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

[0019] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A smart grid fault detection method based on multi-source heterogeneous data, characterized in that, The method includes: Step S1: Obtain historical fault records and fault range data of marked power grid nodes in the power grid, evaluate the accuracy of monitoring parameters in the fault range data for fault detection of marked power grid nodes, and obtain marked fault range data; Step S2: Obtain fault range data and node data of power grid nodes in the power grid, analyze the degree of node approximation between different power grid nodes, obtain the approximate power grid node set, obtain the marked power grid node from the approximate power grid node set and record it as the approximate marked power grid node, evaluate the abnormal state of fault range data for fault detection of power grid nodes, and obtain the target power grid node. Step S3: Obtain the marked fault range data in the approximate marked power grid node of the target power grid node, evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node; Step S4: Based on the target fault range data and the marked fault range data, optimize and adjust the fault range data of the power grid nodes in the power grid, identify the abnormal power grid nodes in the power grid, obtain the abnormal power grid nodes, obtain the power grid topology map of the power grid, and obtain the abnormal area data in the power grid based on the distribution pattern of the abnormal power grid nodes in the power grid topology map. Step S5: Based on the abnormal area data, dispatch drones to photograph the power grid equipment in the power grid, acquire equipment images, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

2. The smart grid fault detection method based on multi-source heterogeneous data according to claim 1, characterized in that, Step S1 includes: Step S11: Obtain the grid node in the power grid that has a fault and mark it as a marked grid node; obtain the fault range data of the marked grid node; mark the preset range of various monitoring parameters within the grid node from the fault range data. Obtain historical fault records for each marked power grid node, and extract data of various monitoring parameters within the marked power grid node from the historical fault records; Step S12: When the maximum or minimum value of a certain monitoring parameter in a certain historical fault record is outside the range of the preset fault range data, the platform issues a fault warning to the marked power grid node. If no fault is detected when the marked power grid node is subsequently inspected, then the certain historical fault record is recorded as the first historical fault record. Obtain each first historical fault record of the marked power grid node. When the maximum or minimum value of the monitoring parameter a in the first historical fault record is outside the range of the fault range data, it is determined that there is a fault detection anomaly of the monitoring parameter a in the marked power grid node. Obtain the total number of times E, the monitoring parameter a, exists in the marked power grid nodes. sum Obtain the total number γ of each first historical fault record. sum Calculate the outlier value Q of the monitoring parameter a. a =E (a，sum) / γ (a，sum) When detecting outlier Q a If the value exceeds the preset abnormal detection threshold, then the monitoring parameter a is determined to be inaccurate for detecting faults at marked power grid nodes. Obtain the range constructed by the minimum and maximum values ​​of the a-th monitoring parameter in each first historical fault record, and denot it as the marked range of the a-th monitoring parameter in the marked power grid node; Step S13: Remove each first historical fault record from the other historical fault records to obtain the second historical fault record that marks the power grid node; Calculate the range distance value S of the βth monitoring parameter in the second historical fault record. β : ， Among them, H (max,β) and H (min,β) The range H of the βth monitoring parameter in the fault range data is respectively. β The maximum and minimum values ​​of H; △ For the range H β The average of the maximum and minimum values; H' (max,β) and H' (min,β) These are the maximum and minimum values ​​of the βth monitoring parameter in the second historical fault record, respectively; When the range distance value S β If the distance is less than the preset range threshold, it is determined that the βth monitoring parameter has a distance anomaly in the second historical fault record. The range anomaly value B of the β-th monitoring parameter is obtained by comparing the total number of second historical fault records showing distance anomalies in the marked power grid node with the total number of second historical fault records for the marked power grid node. β ; When the range outlier B β If the value exceeds the preset abnormal threshold, it is determined that the β-th monitoring parameter is inaccurate in detecting faults in the marked power grid node. The range constructed by the maximum and minimum values ​​of the β-th monitoring parameter within the fault range data is obtained from the plurality of second historical fault records and recorded as the marked range of the β-th monitoring parameter in the marked power grid node. Step S14: Obtain the marking range of several monitoring parameters in the marked power grid node that are determined to be inaccurate in fault detection, and replace the range of the several monitoring parameters in the fault range data of the marked power grid node to obtain the marked fault range data of the marked power grid node.

3. The smart grid fault detection method based on multi-source heterogeneous data according to claim 2, characterized in that, Step S2 includes: Step S21: Obtain fault range data and node data of power grid nodes, and extract data of various node indicators from the node data; Step S22: Analyze the degree of node approximation between different power grid nodes to obtain an approximate set of power grid nodes. The specific analysis process is as follows: Obtain the range L of the d-th monitoring parameter for the e-th power grid node and the e-th power grid node respectively. d =[L d,min ,L d,max ] and L e d =[L e d,min ,L e d,max ], where L d,min L d,max Let L be the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the power grid node. e d,min ,L e d,max These are the minimum and maximum threshold values ​​of the d-th monitoring parameter in the fault range data of the e-th power grid node, respectively. When a certain value exists in the range L d and L e d Inside, obtain L d and L e d The intersection Y between them e d =[Y e d,min ,Y e d,max ], where Y e d,min =max{L d,min L d,max },,Y e d,max =min{L d,max L e d,max }, get range L d With L e d Intersection distance between |Y e d |=Y e d,max -Y e d,min Conversely, the intersection distance length |Y e d |=0; Calculation range L d Distance length | L d |=L d,max -L d,min Calculation range L e d Distance length | L e d |=L e d,max -L e d,min ; Calculate the approximate range C of the d-th monitoring parameter between the e-th power grid node and the d-th power grid node. e d : ， Obtain the average value of the approximate range values ​​of various monitoring parameters between the power grid node and the e-th power grid node, and denot it as the approximate fault range value C between the power grid node and the e-th power grid node. e ; Step S23: Based on the node data of the power grid nodes, construct the node feature vectors of the power grid nodes, obtain the node feature vector of the e-th power grid node, calculate the cosine similarity between the power grid node and the e-th power grid node, and obtain the node approximation value F. e ; Calculate the approximate node value P between the e-th node and the power grid node. e =η C ×C e +η F ×F e , where η C η F These are the preset fault weight coefficient and node weight coefficient, η. C >0, η F >0, η C +η F =1; When the node approximation value P e If the value is greater than the preset approximation threshold, then the e-th power grid node is determined to be approximate with the power grid node, and the e-th power grid node is recorded as the approximate power grid node of the power grid node. Step S24: Obtain several approximate power grid nodes in the power grid and aggregate them to obtain approximate marked power grid nodes; Marked power grid nodes are obtained from the set of approximate power grid nodes and recorded as approximate marked power grid nodes. The total number of approximate marked power grid nodes is divided by the total number of approximate power grid nodes to obtain the abnormal value of the power grid node. When the abnormal value is greater than the preset abnormal threshold, the fault range data of the power grid node is determined. The power grid node is detected to have an abnormality and is recorded as the target power grid node.

4. The smart grid fault detection method based on multi-source heterogeneous data according to claim 3, characterized in that, Step S3 includes: Step S31: Obtain each approximate marked power grid node of the target power grid node, and obtain the marked fault range data of each approximate marked power grid node; Step S32: Obtain the fault range data of the target power grid node, and evaluate the accuracy of the g-th monitoring parameter in the fault range data for capturing faults in the target power grid. The specific evaluation process is as follows: Obtain several monitoring parameters from the approximate marked power grid nodes that are determined to have inaccurate operational detection. Then, obtain the sum of the approximate node values ​​between each approximate marked power grid node and the target power grid node to obtain the approximate total value U. sum ; Obtain several approximate marked power grid nodes for which the g-th monitoring parameter is determined to be inaccurate in fault detection, and obtain the sum U of the node approximation values ​​between the several approximate marked power grid nodes and the target power grid node. (g,sum) Calculate the abnormal value Q of the g-th monitoring parameter in the target power grid node. g =U (g,sum) / U sum ; Step S33: When parameter Q is an outlier g If the range of the g-th monitoring parameter in the preset fault range data is less than the preset abnormal threshold, it is determined that the range of the g-th monitoring parameter in the preset fault range data accurately captures the fault of the target power grid node; otherwise, it is determined that the range of the g-th detection parameter in the preset fault range data does not accurately capture the fault of the target power grid node, and the g-th monitoring parameter is recorded as the abnormal monitoring parameter of the target power grid node. Step S34: Obtain a certain abnormal monitoring parameter of the target power grid node, obtain the average of the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record it as the range reference value ζ; Obtain the absolute value of the difference between the range baseline value ζ and the maximum and minimum values ​​of a certain abnormal monitoring parameter within the range of the marked fault range data for a certain approximate marked power grid node, and record this as the upper limit distance τ of a certain abnormal monitoring parameter in the approximate marked power grid node. max and the lower limit distance τ min ; Calculate the range reference value ζ´ of a certain abnormal monitoring parameter in the target power grid node: ， Where, ζ i Let ζ be the range reference value of a certain abnormal monitoring parameter in the i-th approximate marker of the target power grid node; P i Let be the node approximation value between the target power grid node and the i-th approximate marked power grid node; j is the total number of the various approximate marked power grid nodes; Step S35: Calculate the upper limit distance τ´ of a certain abnormal monitoring parameter in the target power grid node. max and the lower limit distance τ´ min To obtain the target range W=[ζ´-τ´ of a certain abnormal monitoring parameter in the target power grid node. min ,ζ´+τ´ max ]; Obtain the target range of each abnormal monitoring parameter in the target power grid node. Use the target range of each abnormal monitoring parameter to replace the range of each abnormal monitoring parameter in the fault range data of the target power grid node. Record the replaced fault range data as the target fault range data of the target power grid node.

5. The smart grid fault detection method based on multi-source heterogeneous data according to claim 4, characterized in that, Step S4 includes: Step S41: Obtain the fault range data of each power grid node in the power grid, obtain the target fault range data of each target power grid node in the power grid, and obtain the marked fault range data of each marked power grid node in the power grid. The fault range data of power grid nodes in the power grid are optimized and adjusted, specifically as follows: The fault range data of the marked power grid nodes and the target fault range data of the target power grid nodes are used to replace the fault range data of the corresponding power grid nodes in each power grid node. Step S42: Use the fault range data of each replaced power grid node to perform fault detection on the power grid. When the maximum and minimum values ​​of a certain monitoring parameter in a certain power grid node are not within the range of the optimized fault range data, it is determined that the power grid node is abnormal and the power grid node is recorded as an abnormal power grid node. Step S43: Obtain the power grid topology map. Based on the distribution of abnormal power grid nodes in the power grid topology map, obtain the abnormal areas in the power grid. Collect and aggregate the abnormal areas in the power grid to obtain the abnormal area data of the power grid.

6. The smart grid fault detection method based on multi-source heterogeneous data according to claim 5, characterized in that, Step S5 includes: Step S51: Obtain abnormal area data of the power grid, obtain various abnormal areas of the power grid from the abnormal area data, obtain the power grid equipment in the abnormal areas, and use a drone to take pictures of the power grid equipment in the abnormal areas to obtain equipment images; Step S52: Preprocess the equipment image and use a preset image recognition model to input the preprocessed equipment image into the detection model to perform fault detection on the power grid equipment, identify abnormal power grid equipment, and dispatch maintenance personnel to repair the abnormal power grid equipment in the power grid.

7. A smart grid fault detection system based on multi-source heterogeneous data, used to execute the smart grid fault detection method based on multi-source heterogeneous data as described in any one of claims 1-6, characterized in that, The system includes a range acquisition module, an anomaly identification module, and a fault detection module; The range acquisition module is used to acquire the marked fault range data of the marked power grid nodes in the power grid, evaluate the abnormal state of the fault range data for fault detection of the power grid nodes, obtain the target power grid node, and acquire the target fault range data of the target power grid node. The anomaly identification module is used to identify anomalies in power grid nodes based on target fault range data and marked fault range data, obtain abnormal power grid nodes, and obtain abnormal area data in the power grid based on the distribution pattern of abnormal power grid nodes in the power grid topology map. The fault detection module is used to use a drone to photograph the power grid equipment in the power grid, acquire equipment images, detect faults in the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

8. The smart grid fault detection system based on multi-source heterogeneous data according to claim 7, characterized in that, The range acquisition module includes a marker range acquisition unit, a fault detection and anomaly assessment unit, and a target range acquisition unit; The marking range acquisition unit is used to evaluate the accuracy of fault detection of the marked power grid node by the monitoring parameters in the fault range data, and to mark the fault range data. The fault detection anomaly assessment unit is used to acquire approximate marked power grid nodes, and based on the marked fault range data of the approximate marked power grid nodes, assess the abnormal state of fault detection of the power grid nodes by the fault range data to obtain the target power grid node. The target range acquisition unit is used to evaluate the accuracy of the monitoring parameters in the fault range data for fault capture of the target power grid node, and obtain the target fault range data of the target power grid node.

9. The smart grid fault detection system based on multi-source heterogeneous data according to claim 7, characterized in that, The anomaly identification module includes a range optimization unit and an anomaly identification unit; The range optimization unit is used to acquire the marked fault range data of the marked power grid node, acquire the target fault range data of the target power grid node, and optimize and adjust the fault range data of the power grid node in the power grid. The anomaly identification unit is used to identify anomalies in the power grid nodes based on the optimized fault range data of the power grid nodes, obtain abnormal power grid nodes, and obtain abnormal area data in the power grid by combining the power grid topology map.

10. The smart grid fault detection system based on multi-source heterogeneous data according to claim 7, characterized in that, The fault detection module includes a fault detection unit; The fault detection unit is used to acquire images of power grid equipment in the abnormal area of ​​the abnormal area data based on the abnormal area data, perform fault detection on the power grid equipment, identify abnormal power grid equipment, and repair the abnormal power grid equipment in the power grid.

Images