Power distribution network active operation and maintenance method and system based on point cloud model

Abstract

The invention relates to the technical field of three-dimensional point cloud, and discloses a power distribution network active operation and maintenance method and system based on a point cloud model, and the method comprises the steps: carrying out the flight inspection of power distribution network equipment through employing a point cloud sensor carried by an unmanned plane, and collecting the three-dimensional point cloud data of the power distribution network equipment in real time; performing multi-scale processing on the three-dimensional point cloud data, calculating a comprehensive health index of the power distribution network equipment, identifying a fault mode of the equipment and performing positioning; and generating a targeted operation and maintenance strategy according to the health index and the fault mode of the power distribution network equipment. The inspection efficiency is remarkably improved, the labor cost is reduced, the safety risk is reduced, the equipment state can be monitored in real time, and potential faults can be found in time. The optimized flight path and operation and maintenance decision support system improves the resource utilization rate, ensures efficient and safe operation of the power distribution network, and provides an innovative solution for intelligent operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional point cloud technology, and in particular to a distribution network active operation and maintenance method and system based on a point cloud model. Background Art

[0002] With the development of the power industry, especially the increasing scale and complexity of distribution networks, traditional manual inspection and maintenance methods can no longer meet the needs of efficient operation of modern distribution networks. The operating status of distribution network equipment directly affects the stability and reliability of power supply, so timely monitoring and maintenance of the health status of equipment has become an important task of modern power systems. Traditional distribution network operation and maintenance mainly rely on manual inspections, regular inspections and conventional maintenance methods, which are not only inefficient, but also have great safety risks and human errors.

[0003] In recent years, with the rapid development of UAV technology, LiDAR technology and point cloud data processing technology, intelligent distribution network inspection systems based on these advanced technologies have gradually become a research hotspot. As a flexible and low-cost flying platform, UAVs can carry LiDAR sensors to efficiently collect three-dimensional data of distribution networks, quickly obtain spatial information of distribution network equipment, and provide accurate data for subsequent equipment health assessment and fault diagnosis. However, how to organically combine UAV technology with point cloud data processing technology and provide efficient operation and maintenance decision support based on these data is still a major challenge in the current field of intelligent operation and maintenance of distribution networks.

[0004] Traditional point cloud data processing methods mainly focus on the modeling and analysis of static objects. However, in the process of distribution network inspection, the collection, processing and analysis of point cloud data involve a large number of equipment and complex environmental factors. How to achieve real-time and accurate equipment health monitoring and fault diagnosis has become a key technical problem. In addition, traditional path planning methods often rely too much on manual experience in distribution network inspections and lack comprehensive optimization of multiple factors such as flight distance and energy consumption. How to design an efficient flight path, reduce flight time and energy consumption, and ensure coverage of all key equipment is also an urgent problem to be solved.

[0005] At present, the fault pattern recognition technology in distribution network operation and maintenance mainly relies on manual experience or diagnostic systems based on simple algorithms. Although these methods can identify some common faults, they are still limited in fault diagnosis and precise positioning of complex equipment. Therefore, how to combine point cloud data for more intelligent and automated fault pattern recognition has become an important issue to improve the efficiency and accuracy of distribution network operation and maintenance. Summary of the invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by the present invention is: the efficiency and accuracy of path planning, health monitoring, fault diagnosis and operation and maintenance decision optimization in distribution network equipment inspection.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: a distribution network active operation and maintenance method based on a point cloud model, comprising:

[0009] Using drones equipped with point cloud sensors to perform flight inspections on distribution network equipment, and collecting three-dimensional point cloud data of the distribution network equipment in real time;

[0010] Perform multi-scale processing on the three-dimensional point cloud data, calculate the comprehensive health index of the distribution network equipment, identify the failure mode of the equipment and locate it;

[0011] Generate targeted operation and maintenance strategies based on the health index and failure mode of the distribution network equipment.

[0012] As a preferred solution of the active operation and maintenance method of the distribution network based on the point cloud model described in the present invention, wherein: the real-time acquisition of the three-dimensional point cloud data of the distribution network equipment includes accurately calibrating the basic information of key equipment, buildings and tall trees in the distribution network through GPS, laser radar and geographic information system technology;

[0013] The basic information includes location, size and relative height to the set plane;

[0014] Using the laser radar on the drone to collect the three-dimensional point cloud data of the distribution network equipment to generate a three-dimensional model;

[0015] Determine the take-off location of the drone and the end point of the flight mission;

[0016] Identify critical nodes in the flight path;

[0017] The key nodes include the locations of key equipment, buildings and tall trees in the distribution network;

[0018] Use path optimization algorithms to automatically plan flight paths based on the location layout of key equipment, buildings and tall trees, flight targets and environmental factors;

[0019] The flight goal is to minimize the flight path of the drone and minimize the battery energy consumption of the drone;

[0020] Use LiDAR to avoid obstacles in real time and automatically adjust flight altitude based on device type and surrounding environment;

[0021] Combined with the feedback information from GPS and IMU, the flight status and flight path of the drone can be monitored in real time.

[0022] As a preferred solution of the active operation and maintenance method of the distribution network based on the point cloud model described in the present invention, wherein: automatically planning the flight path includes listing the position coordinates of all key nodes;

[0023] Calculate and record the distance and energy consumption between nodes;

[0024] Generate the optimal flight path by combining genetic algorithm and ant colony algorithm;

[0025] The genetic algorithm includes randomly generating 30 initial flight paths, each gene of the path represents a key node;

[0026] Set the crossover rate to 80% and the mutation rate to 20%.

[0027] The fitness of each path is calculated, taking into account the distance and energy consumption of the path. The formula of the fitness function is expressed as:

[0028] Fitness=α×Distance+β×Energy Consumption

[0029] Among them, α and β represent weight coefficients, Distance represents path distance, and Energy Consumption represents energy consumption;

[0030] Based on the initial flight path generated by the genetic algorithm, the path pheromone concentration between nodes is initialized to 0.5;

[0031] According to the fitness function, the roulette wheel selection method is used to select the path with higher fitness, and 15 initial flight paths are selected and crossed using the partial mapping crossover method;

[0032] For some chromosomes, the exchange mutation operation is used to mutate and randomly exchange the positions of two nodes in the path;

[0033] Evaluate the fitness of the newly generated path and update the population;

[0034] The new population is selected according to the fitness value, the paths with high fitness are retained, and the paths with low fitness are eliminated.

[0035] As a preferred solution of the active operation and maintenance method of the distribution network based on the point cloud model described in the present invention, wherein: the ant colony algorithm includes, according to 30 initial flight paths generated by the genetic algorithm, using the initial pheromone of the ant colony algorithm as the initial pheromone, and distributing the pheromone concentration to each path;

[0036] Set up 50 ants, each ant represents a potential flight path;

[0037] Each ant starts from the starting point and selects the next node based on the pheromone concentration and heuristic information between the current nodes until the entire path is completed;

[0038] The heuristic information includes the shortest distance and the minimum energy consumption, which can be expressed as:

[0039]

[0040] Among them, d ij represents the distance between node i and node j, e ij represents the energy consumption between node i and node j, α and β are weight factors that control the contribution of distance and energy consumption to heuristic information;

[0041] Each ant calculates the probability of selecting the next node based on the pheromone concentration and heuristic information. The formula is expressed as:

[0042]

[0043] Among them, τ ij represents the pheromone concentration, η ij represents heuristic information, α and β are weight coefficients;

[0044] Volatile pheromones reduce the impact of outdated paths. The formula is:

[0045]

[0046] Where ρ represents the pheromone volatilization rate.

[0047] According to the fitness of the path, pheromone is added, and the formula is expressed as:

[0048] τ ij =τ ij +Δτ ij

[0049] Among them, Δτ ij It represents the amount of pheromone added by ants according to the path length and energy consumption.

[0050] When the ant colony algorithm meets the termination condition, the iteration stops;

[0051] In each generation of the genetic algorithm, paths with higher fitness are selected and the pheromone concentrations of these paths are increased to enhance the probability of the ant colony algorithm selecting these paths;

[0052] In each round of iteration of the ant colony algorithm, paths with higher pheromone concentrations are selected as elite individuals of the genetic algorithm and added to the population of the genetic algorithm.

[0053] The genetic algorithm and the ant colony algorithm work synchronously in each iteration:

[0054] Perform crossover, mutation and fitness evaluation in genetic algorithms and update the population

[0055] After each round of ants complete the path construction, they update the pheromone and generate a new path;

[0056] The termination condition includes terminating the optimization process when both the genetic algorithm and the ant colony algorithm reach a maximum number of iterations of 1000;

[0057] The optimal flight path is selected from the genetic algorithm and the ant colony algorithm as the optimal flight path of the UAV.

[0058] As a preferred solution of the active operation and maintenance method of the distribution network based on the point cloud model described in the present invention, wherein: multi-scale processing of the three-dimensional point cloud data includes pre-processing the collected three-dimensional point cloud data;

[0059] Categorize key equipment in the distribution network into large equipment and small equipment, and extract multi-scale point cloud data;

[0060] The multi-scale point cloud data includes large-scale data and small-scale data;

[0061] For large devices, a 15 cm voxel grid is used for fast segmentation to obtain large-scale data;

[0062] The large-scale data includes the macroscopic outline, key information and main structure of the equipment;

[0063] For small devices, a 1 cm voxel grid is used to obtain small-scale data;

[0064] The small-scale data include the morphology and damage area of ​​the device surface;

[0065] For large-scale data, the Euclidean clustering method is used to extract the overall outline of the equipment, and the curvature edge detection method is used to extract the appearance and structural features of large equipment for macroscopic modeling;

[0066] For small-scale data, curvature analysis is used to identify surface damage, and morphological methods are used to enhance surface details, highlight the surface features of tiny extracted small devices, and perform fine modeling.

[0067] As a preferred solution of the active operation and maintenance method of the distribution network based on the point cloud model described in the present invention, the comprehensive health index includes a structural integrity index, a surface damage index, an environmental impact index, an equipment aging index and a fault history index;

[0068] The structural integrity index includes measuring the condition of the overall structure, supporting frame, surface damage, etc. of the equipment;

[0069] The surface damage index includes measuring the degree of cracks, corrosion, decay and wear on the surface of the equipment;

[0070] The environmental impact index includes considering the impact of the external environment on the health of key equipment in the distribution network;

[0071] The equipment aging index includes the impact of the equipment's age and maintenance history on the health of key equipment in the distribution network;

[0072] The fault history index includes evaluating the current health status of key equipment in the distribution network based on the equipment's fault records and maintenance history;

[0073] The formula for calculating the comprehensive health index is expressed as:

[0074] H=w1·S structural +w2·S surface +w3·S environment +w4·S age +w5·S history

[0075] w1+w2+w3+w4+w5=1

[0076] Among them, H represents the comprehensive health index, S structural represents the structural integrity index, w1 represents the weight of the structural integrity index; S surface represents the surface damage index, w2 represents the weight of the surface damage index; S environment represents the environmental impact index, w3 represents the weight of the environmental impact index; S age represents the equipment aging index, w4 represents the weight of the equipment aging index; S history represents the fault history index, w5 represents the weight of the fault history index;

[0077] Identifying equipment failures involves building a recognition model using random forest combined with support vector machine algorithms, training the extracted appearance structural features of large equipment and surface features of small equipment, and identifying failure modes. The formula is:

[0078]

[0079] Among them, y represents the classification result of the failure mode, X i represents the i-th feature of the distribution network equipment, α i Represents the weight coefficient of the i-th random forest tree; Represents the random forest model for feature X i The classification result; β represents the weight coefficient of the support vector machine, w represents the weight vector of the support vector machine; b represents the bias term of the support vector machine; Let \(\varPhi\) denote the normalization function, \(K(X,X')\) denote the Gaussian kernel function, \(\langle w,X\rangle\) denote the inner product calculation of the support vector machine; \(\text{sign}(z)\) denotes the category that returns the decision result;

[0080] When \(y = 0\), it indicates that no fault occurs;

[0081] When \(y = 1\), it indicates that cracks occur in the key equipment of the distribution network, and the fault score is 5 points;

[0082] When \(y = 2\), it indicates that the key equipment of the distribution network is corroded, and the fault score is 4 points;

[0083] When \(y = 3\), it indicates that the key equipment of the distribution network is deformed, and the fault score is 3 points;

[0084] When \(y = 4\), it indicates that the key equipment of the distribution network is worn, and the fault score is 2 points;

[0085] Locating the key equipment with faults includes, for each identified fault, using the spatial coordinates of each point in the point cloud data and combining with the fault type to label the key equipment of the distribution network where the fault occurs.

[0086] As a preferred solution of the active operation and maintenance method for the distribution network based on the point cloud model described in the present invention, among them: generating targeted operation and maintenance strategies includes, according to the health index and fault mode of the distribution network equipment, designing a priority ranking model, and the formula is expressed as:

[0087] \(P = w_1\cdot H+w_2\cdot F\) mode

[0088] \(w_1 + w_2=1\)

[0089] Among them, \(H\) represents the comprehensive health index, \(w_1\) represents the weight of the comprehensive health index; \(F\) mode represents the fault mode score, and \(w_2\) represents the weight of the fault mode score;

[0090] When \(0\leq P\leq10\), it indicates that there is no fault in the distribution network, regular inspections are carried out, continuous monitoring is carried out, and mild maintenance is carried out appropriately;

[0091] When \(10\lt P\leq20\), it indicates that medium faults occur in the distribution network, monitoring is strengthened, medium-scale repairs and preventive maintenance are carried out;

[0092] When \(20\lt P\leq30\), it indicates that high-level faults occur in the distribution network, emergency repairs are carried out, shutdown processing is carried out if necessary, and resource support is increased;

[0093] When \(30\lt P\), it indicates that serious faults occur in the distribution network, the machine is immediately shut down, the equipment is replaced, and the catastrophic event handling program is started.

[0094] A distribution network active operation and maintenance system based on a point cloud model, wherein:

[0095] The data acquisition module uses a drone equipped with a point cloud sensor to perform flight inspections on distribution network equipment and collect three-dimensional point cloud data of the distribution network equipment in real time;

[0096] A data processing module performs multi-scale processing on the three-dimensional point cloud data, calculates the comprehensive health index of the distribution network equipment, identifies the failure mode of the equipment and locates it;

[0097] The strategy generation module generates a targeted operation and maintenance strategy according to the health index and failure mode of the distribution network equipment.

[0098] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0099] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of any one of the methods of the present invention.

[0100] Beneficial effects of the invention: The active distribution network operation and maintenance method based on the point cloud model provided by the invention realizes the automated inspection, intelligent path planning, accurate health monitoring and fault diagnosis of distribution network equipment by combining drones, point cloud technology, genetic algorithm and ant colony algorithm. It significantly improves the inspection efficiency, reduces labor costs, reduces safety risks, and can monitor the equipment status in real time and detect potential faults in time. The optimized flight path and operation and maintenance decision support system improve resource utilization, ensure the efficient and safe operation of the distribution network, and provide innovative solutions for intelligent operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0102] Figure 1 An overall flow chart of a distribution network active operation and maintenance method based on a point cloud model provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0103] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0104] Example 1, reference Figure 1 , is an embodiment of the present invention, and provides a distribution network active operation and maintenance method based on a point cloud model, comprising:

[0105] S1: Use a drone equipped with a point cloud sensor to perform a flight inspection of the distribution network equipment and collect three-dimensional point cloud data of the distribution network equipment in real time.

[0106] The real-time collection of three-dimensional point cloud data of the distribution network equipment includes accurately calibrating basic information of key equipment, buildings and tall trees in the distribution network through GPS, lidar and geographic information system technology.

[0107] The basic information includes position, size and relative height to a set plane.

[0108] The laser radar on the drone is used to collect the three-dimensional point cloud data of the distribution network equipment to generate a three-dimensional model.

[0109] Determine the drone takeoff location and the end point of the flight mission.

[0110] Identify critical nodes in the flight path.

[0111] The key nodes include the locations of key equipment, buildings and tall trees in the distribution network.

[0112] Use path optimization algorithms to automatically plan flight paths based on the location layout of key equipment, buildings and tall trees, flight targets and environmental factors.

[0113] The flight objectives are to minimize the flight path of the drone and minimize the battery energy consumption of the drone.

[0114] Using LiDAR, obstacle avoidance is performed in real time, combined with automatic adjustment of flight altitude based on device type and surrounding environment.

[0115] Combined with the feedback information from GPS and IMU, the flight status and flight path of the drone can be monitored in real time.

[0116] Automatically planning the flight path includes listing the location coordinates of all key nodes.

[0117] Calculate and record the distance and energy consumption between nodes.

[0118] Genetic algorithm and ant colony algorithm are combined to generate the optimal flight path.

[0119] The genetic algorithm includes randomly generating 30 initial flight paths, and each gene of the path represents a key node.

[0120] Set the crossover rate to 80% and the mutation rate to 20%.

[0121] The fitness of each path is calculated, taking into account the distance and energy consumption of the path. The formula of the fitness function is expressed as:

[0122] Fitness=α×Distance+β×Energy Consumption

[0123] Among them, α and β represent weight coefficients, Distance represents path distance, and Energy Consumption represents energy consumption.

[0124] Based on the initial flight path generated by the genetic algorithm, the path pheromone concentration between nodes is initialized to 0.5.

[0125] According to the fitness function, the roulette wheel selection method is used to select the path with higher fitness, and 15 initial flight paths are selected and crossed using the partial mapping crossover method.

[0126] For some chromosomes, the exchange mutation operation is used to mutate and randomly exchange the positions of two nodes in the path.

[0127] Evaluate the fitness of the newly generated paths and update the population.

[0128] The new population is selected according to the fitness value, the paths with high fitness are retained, and the paths with low fitness are eliminated.

[0129] The ant colony algorithm includes,30 initial flight paths generated by the genetic algorithm, which will be used as the initial pheromone of the ant colony algorithm, and the pheromone concentration will be distributed to each path.

[0130] Set up 50 ants, each ant represents a potential flight path.

[0131] Each ant starts from the starting point and selects the next node based on the pheromone concentration and heuristic information between the current nodes until the entire path is completed.

[0132] The heuristic information includes the shortest distance and the minimum energy consumption, which can be expressed as:

[0133]

[0134] Among them, dij represents the distance between node i and node j, e ij represents the energy consumption between node i and node j, and α and β are weight factors that control the contribution of distance and energy consumption to heuristic information.

[0135] Each ant calculates the probability of selecting the next node based on the pheromone concentration and heuristic information. The formula is expressed as:

[0136]

[0137] Among them, τ ij represents the pheromone concentration, η ij represents the heuristic information, α and β are weight coefficients.

[0138] Volatile pheromones reduce the impact of outdated paths. The formula is:

[0139]

[0140] Where ρ is the pheromone volatilization rate.

[0141] According to the fitness of the path, pheromone is added, and the formula is expressed as:

[0142] τ ij =τ ij +Δτ ij

[0143] Among them, Δτ ij It represents the amount of pheromone added by ants according to the path length and energy consumption.

[0144] When the ant colony algorithm meets the termination condition, the iteration stops.

[0145] In each generation of the genetic algorithm, paths with higher fitness are selected and the pheromone concentrations of these paths are increased to enhance the probability of the ant colony algorithm selecting these paths.

[0146] In each round of iteration of the ant colony algorithm, paths with higher pheromone concentrations are selected as elite individuals of the genetic algorithm and added to the population of the genetic algorithm.

[0147] The genetic algorithm and the ant colony algorithm work synchronously in each round of iteration.

[0148] Perform crossover, mutation, and fitness evaluation in the genetic algorithm to update the population.

[0149] After each round of ants complete the path construction, they update the pheromone and generate a new path.

[0150] The termination condition includes terminating the optimization process when both the genetic algorithm and the ant colony algorithm reach a maximum number of iterations of 1000.

[0151] The optimal flight path is selected from the genetic algorithm and the ant colony algorithm as the optimal flight path of the UAV.

[0152] Furthermore, by combining drones, lidar, genetic algorithms and ant colony algorithms, the flight path optimization of distribution network inspections is achieved, maximizing inspection efficiency and saving energy. By accurately collecting the three-dimensional point cloud data of distribution network equipment and combining environmental factors, the shortest path and the flight route with the lowest energy consumption are automatically planned to ensure that the inspection task is completed efficiently under limited battery energy. The combination of genetic algorithms and ant colony algorithms effectively solves the multi-objective problem in path optimization, selects the optimal path through multiple iterative optimizations, and improves the intelligence level and practicality of drone inspections.

[0153] S2: Perform multi-scale processing on the three-dimensional point cloud data, calculate the comprehensive health index of the distribution network equipment, identify the failure mode of the equipment and locate it.

[0154] The multi-scale processing of the three-dimensional point cloud data includes preprocessing the collected three-dimensional point cloud data.

[0155] The key equipment in the distribution network is divided into large equipment and small equipment, and multi-scale point cloud data is extracted.

[0156] The multi-scale point cloud data includes large-scale data and small-scale data.

[0157] For large devices, a 15 cm voxel grid is used for fast segmentation to obtain large-scale data.

[0158] The large-scale data includes the macroscopic outline, key information and main structure of the equipment.

[0159] For small devices, a 1 cm voxel grid is used to obtain small-scale data.

[0160] The small-scale data includes the morphology and damage area of ​​the device surface.

[0161] For large-scale data, the Euclidean clustering method is used to extract the overall outline of the equipment, and the curvature edge detection method is used to extract the appearance and structural features of large equipment for macroscopic modeling.

[0162] For small-scale data, curvature analysis is used to identify surface damage, and morphological methods are used to enhance surface details, highlight the surface features of tiny extracted small devices, and perform fine modeling.

[0163] The comprehensive health index includes structural integrity index, surface damage index, environmental impact index, equipment aging index and failure history index.

[0164] The structural integrity index includes measuring the overall structure of the equipment, the supporting frame, surface damage and other aspects.

[0165] The surface damage index includes measuring the degree of cracks, corrosion, decay and wear on the surface of the equipment.

[0166] The environmental impact index includes consideration of the impact of the external environment on the health of key equipment in the distribution network.

[0167] The equipment aging index includes the impact of the equipment's service life and maintenance history on the health of key equipment in the distribution network.

[0168] The fault history index includes evaluating the current health status of key equipment in the distribution network based on the fault records and maintenance history of the equipment.

[0169] The formula for calculating the comprehensive health index is expressed as:

[0170] H=w1·S structural +w2·S surface +w3·S environment +w4·S age +w5·S history

[0171] w1+w2+w3+w4+w5=1

[0172] Among them, H represents the comprehensive health index, S structural represents the structural integrity index, w1 represents the weight of the structural integrity index; S surface represents the surface damage index, w2 represents the weight of the surface damage index; S environment represents the environmental impact index, w3 represents the weight of the environmental impact index; S age represents the equipment aging index, w4 represents the weight of the equipment aging index; S history represents the fault history index, and w5 represents the weight of the fault history index.

[0173] Identifying equipment failures involves building a recognition model using random forest combined with support vector machine algorithms, training the extracted appearance structural features of large equipment and surface features of small equipment, and identifying failure modes. The formula is:

[0174]

[0175] Among them, y represents the classification result of the failure mode, X i represents the i-th feature of the distribution network equipment, α i Represents the weight coefficient of the i-th random forest tree; Represents the random forest model for feature X iThe classification result; β represents the weight coefficient of the support vector machine, w represents the weight vector of the support vector machine; b represents the bias term of the support vector machine; represents the normalization function, K(X,X′) represents the Gaussian kernel function,<w,X> Represents the inner product calculation of the support vector machine; sign(z) represents the category of the returned decision result.

[0176] When y=0, it indicates that no fault occurs.

[0177] When y=1, it means that cracks occur in key equipment of the distribution network, and the fault score is 5 points.

[0178] When y=2, it means that the key equipment of the distribution network is corroded and the fault score is 4 points.

[0179] When y=3, it means that the key equipment of the distribution network is deformed and the fault score is 3 points.

[0180] When y=4, it means that the key equipment of the distribution network is worn out and the fault score is 2 points.

[0181] The positioning of the key equipment causing the fault includes, for each identified fault, using the spatial coordinates of each point in the point cloud data, combined with the fault type, to mark the key equipment of the distribution network where the fault occurs.

[0182] Furthermore, accurate evaluation and positioning of distribution network equipment can be achieved through multi-scale point cloud data processing, comprehensive health index calculation and fault pattern recognition. By distinguishing large and small equipment and processing point cloud data with different resolutions, the macro and micro characteristics of the equipment can be captured in detail. Combining multi-dimensional indicators such as structural integrity, surface damage, environmental impact, equipment aging and fault history, a comprehensive health assessment is provided, and machine learning algorithms are used to identify and classify fault patterns, and finally the fault location is accurately located. This comprehensive approach improves the accuracy and timeliness of fault diagnosis and provides a scientific basis for intelligent operation and maintenance and resource optimization of distribution networks.

[0183] S3: Generate a targeted operation and maintenance strategy based on the health index and failure mode of the distribution network equipment. Generating a targeted operation and maintenance strategy includes designing a priority ranking model based on the health index and failure mode of the distribution network equipment, and the formula is expressed as:

[0184] P=w1·H+w2·F mode

[0185] w1+w2=1

[0186] Among them, H represents the comprehensive health index, w1 represents the weight of the comprehensive health index; F mode represents the failure mode score, and w2 represents the weight of the failure mode score.

[0187] When 0 ≤ P ≤ 10, it indicates that the distribution network has no faults, and regular inspections, continuous monitoring, and appropriate minor maintenance are carried out.

[0188] When 10 < P ≤ 20, it indicates that the distribution network has medium faults, and enhanced monitoring, medium-scale repairs, and preventive maintenance are carried out.

[0189] When 20 < P ≤ 30, it indicates that the distribution network has high-level faults, and emergency repairs are carried out. Shut down the machine if necessary, and increase resource support.

[0190] When 30 < P, it indicates that the distribution network has severe faults, and the machine is shut down immediately, the equipment is replaced, and the catastrophic event handling procedure is started.

[0191] Furthermore, by combining the health index and fault mode of the distribution network equipment, a priority ranking model is developed to achieve targeted operation and maintenance strategies. According to different fault levels, the operation and maintenance measures are automatically adjusted to ensure a rapid response and appropriate handling methods when faults occur. This method helps to optimize the allocation of operation and maintenance resources, improve the fault response efficiency, reduce the downtime, and ensure the stability and security of the distribution network through real-time monitoring and precise decision-making.

[0192] Embodiment 2, an embodiment of the present invention, provides an active operation and maintenance system for a distribution network based on a point cloud model, including:

[0193] A data acquisition module uses an unmanned aerial vehicle equipped with a point cloud sensor to conduct flight inspections on the distribution network equipment and collect three-dimensional point cloud data of the distribution network equipment in real time.

[0194] A data processing module performs multi-scale processing on the three-dimensional point cloud data, calculates the comprehensive health index of the distribution network equipment, and identifies and locates the fault modes of the equipment.

[0195] A strategy generation module generates targeted operation and maintenance strategies based on the health index and fault mode of the distribution network equipment.

[0196] Embodiment 3, an embodiment of the present invention, is different from the previous two embodiments in that:

[0197] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0198] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0199] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0200] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiment, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (P genetic algorithm), a field programmable gate array (FP genetic algorithm), etc.

[0201] Example 4: Select a test area containing multiple power distribution equipment (such as substations, distribution boxes, switch cabinets, etc.). Use drones equipped with lidar sensors to conduct flight inspections of distribution network equipment. Use GPS, lidar, and geographic information system (GIS) technology to accurately calibrate the basic information of key equipment, buildings, and tall trees in the area, including the location information, size, and relative height of the equipment to the set plane. Provide the necessary geographic and equipment basic data for subsequent three-dimensional modeling.

[0202] The drone conducts inspections according to a predetermined flight path. The path planning takes into account environmental factors, such as the impact of buildings and trees, and uses genetic algorithms and ant colony algorithms to optimize the flight path. During the flight, the lidar sensor collects 3D point cloud data of the distribution network equipment in real time and generates a complete 3D point cloud model. These point cloud data will provide basic data for equipment health assessment, fault identification and positioning.

[0203] The collected 3D point cloud data is analyzed at multiple scales through the data processing module. For large equipment (such as substations, etc.), a 15cm voxel grid is used for coarse-grained segmentation to extract the macroscopic outline and main structure of the equipment; for small equipment (such as distribution boxes, etc.), a 1cm voxel grid is used for fine segmentation to identify surface morphology and minor damage areas. The Euclidean clustering method is used to extract contours of large-scale data, and the curvature analysis method is used to identify surface damage of small-scale data.

[0204] Based on the multi-scale processed point cloud data, the comprehensive health index of the distribution network equipment is calculated. The health index takes into account the structural integrity, surface damage, environmental impact, equipment aging, and failure history of the equipment. The extracted equipment features are trained using the random forest combined with the support vector machine (SVM) algorithm to identify the failure modes of the equipment, including cracks, corrosion, deformation, and wear.

[0205] According to the health index and failure mode of the equipment, the operation and maintenance strategy is generated through the priority sorting model. The priority sorting model is designed by weighted calculation based on the comprehensive health index and failure mode score to give priority to high-risk failures. The experimental results are shown in Table 1.

[0206] Table 1 Experimental data table

[0207]

[0208] As can be seen from the table, the failure mode score of substation A is 5, indicating that there are serious cracks in the equipment, which requires priority treatment, so its operation and maintenance priority (77.5) is higher.

[0209] The failure mode score of distribution box B is 4, indicating that the equipment has been corroded. Although it is not as serious as cracks, it also needs to be repaired in time. Its operation and maintenance priority is 62.0, which is relatively high.

[0210] The failure mode score of distribution cabinet C is 3, indicating that the equipment is deformed. This is a medium failure with a low priority. The operation and maintenance priority is 45.5.

[0211] The failure mode score of substation D is 0, which means that the equipment has not failed. Therefore, its operation and maintenance priority is 87.5. Although the health index is high, the priority is low because there is no failure.

[0212] By adding the fault mode score, the present invention can refine the fault degree of the distribution network equipment and accurately adjust the operation and maintenance priority based on the score. Compared with the manual evaluation or simple health index scoring in the prior art, this fault mode scoring method can more accurately reflect the actual fault condition of the equipment and can flexibly adjust the operation and maintenance strategy. Through the refined fault mode scoring (cracks, corrosion, deformation, etc.), the health status of the equipment can be more comprehensively evaluated to avoid missed diagnosis or misdiagnosis.

[0213] The combination of failure mode scoring and comprehensive health index makes the operation and maintenance priority more scientific and reasonable. Through weighted calculation, it can ensure that serious faulty equipment is handled first, reducing the risk of equipment failure.

[0214] Through intelligent calculation of operation and maintenance priorities, operation and maintenance resources can be reasonably allocated to ensure that high-priority equipment is repaired in a timely manner and avoid unnecessary waste of resources.

[0215] Traditional methods often rely on manual inspection and subjective judgment, and the accuracy and efficiency of fault diagnosis are low. However, by introducing fault mode scoring, the present invention can not only diagnose equipment failures more accurately, but also adjust the operation and maintenance strategy according to the actual fault conditions of the equipment, thereby improving the accuracy and efficiency of operation and maintenance. Compared with traditional static operation and maintenance strategies, the present invention combines point cloud data and fault mode scoring to dynamically adjust the operation and maintenance strategy according to the actual conditions of the equipment, ensuring that the equipment is inspected and maintained at the most appropriate time, thereby reducing the incidence of equipment failures.

[0216] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution network active operation and maintenance method based on point cloud model, characterized in that: include: Using drones equipped with point cloud sensors to conduct flight inspections of distribution network equipment, and collect three-dimensional point cloud data of the distribution network equipment in real time; Perform multi-scale processing on the three-dimensional point cloud data, calculate the comprehensive health index of the distribution network equipment, identify the failure mode of the equipment and locate it; Generate targeted operation and maintenance strategies based on the health index and failure mode of the distribution network equipment.

2. The active operation and maintenance method for distribution network based on point cloud model according to claim 1, characterized in that: The real-time acquisition of the three-dimensional point cloud data of the distribution network equipment includes accurately calibrating the basic information of key equipment, buildings and tall trees in the distribution network through GPS, laser radar and geographic information system technology; The basic information includes location, size and relative height to the set plane; Using the laser radar on the drone to collect the three-dimensional point cloud data of the distribution network equipment to generate a three-dimensional model; Determine the take-off location of the drone and the end point of the flight mission; Identify critical nodes in the flight path; The key nodes include the locations of key equipment, buildings and tall trees in the distribution network; Use path optimization algorithms to automatically plan flight paths based on the location layout of key equipment, buildings and tall trees, flight targets and environmental factors; The flight goal is to minimize the flight path of the drone and minimize the battery energy consumption of the drone; Use LiDAR to avoid obstacles in real time and automatically adjust flight altitude based on device type and surrounding environment; Combined with the feedback information from GPS and IMU, the flight status and flight path of the drone can be monitored in real time.

3. The active operation and maintenance method for distribution network based on point cloud model according to claim 2, characterized in that: Automatic flight path planning includes listing the location coordinates of all key nodes; Calculate and record the distance and energy consumption between nodes; Generate the optimal flight path by combining genetic algorithm and ant colony algorithm; The genetic algorithm includes randomly generating 30 initial flight paths, each gene of the path represents a key node; Set the crossover rate to 80% and the mutation rate to 20%. The fitness of each path is calculated, taking into account the distance and energy consumption of the path. The formula of the fitness function is expressed as: Fitness=α×Distance+β×Energy Consumption Among them, α and β represent weight coefficients, Distance represents path distance, and Energy Consumption represents energy consumption; Based on the initial flight path generated by the genetic algorithm, the path pheromone concentration between nodes is initialized to 0.5; According to the fitness function, the roulette wheel selection method is used to select the path with higher fitness, and 15 initial flight paths are selected and crossed using the partial mapping crossover method; For some chromosomes, the exchange mutation operation is used to mutate and randomly exchange the positions of two nodes in the path; Evaluate the fitness of the newly generated path and update the population; The new population is selected according to the fitness value, the paths with high fitness are retained, and the paths with low fitness are eliminated.

4. The active operation and maintenance method for distribution network based on point cloud model according to claim 3, characterized in that: The ant colony algorithm includes, according to 30 initial flight paths generated by the genetic algorithm, using the initial pheromone of the ant colony algorithm as the initial pheromone, and distributing the pheromone concentration to each path; Set up 50 ants, each ant represents a potential flight path; Each ant starts from the starting point and selects the next node based on the pheromone concentration and heuristic information between the current nodes until the entire path is completed; The heuristic information includes the shortest distance and the minimum energy consumption, which can be expressed as: Among them, d ij represents the distance between node i and node j, e ij represents the energy consumption between node i and node j, α and β are weight factors that control the contribution of distance and energy consumption to heuristic information; Each ant calculates the probability of selecting the next node based on the pheromone concentration and heuristic information. The formula is expressed as: Among them, τ ij represents the pheromone concentration, η ij represents heuristic information, α and β are weight coefficients; Volatile pheromones reduce the impact of outdated paths. The formula is: in, Indicates the pheromone volatilization rate. According to the fitness of the path, pheromone is added, and the formula is expressed as: t ij =t ij +Δt ij Among them, Δτ ij It represents the amount of pheromone added by ants according to the path length and energy consumption. When the ant colony algorithm meets the termination condition, the iteration stops; In each generation of the genetic algorithm, paths with higher fitness are selected and the pheromone concentrations of these paths are increased to enhance the probability of the ant colony algorithm selecting these paths; In each round of iteration of the ant colony algorithm, paths with higher pheromone concentrations are selected as elite individuals of the genetic algorithm and added to the population of the genetic algorithm. The genetic algorithm and the ant colony algorithm work synchronously in each iteration: Perform crossover, mutation and fitness evaluation in genetic algorithms and update the population After each round of ants complete the path construction, they update the pheromone and generate a new path; The termination condition includes terminating the optimization process when both the genetic algorithm and the ant colony algorithm reach a maximum number of iterations of 1000; The optimal flight path is selected from the genetic algorithm and the ant colony algorithm as the optimal flight path of the UAV.

5. The active operation and maintenance method for distribution network based on point cloud model according to claim 4, characterized in that: The multi-scale processing of the three-dimensional point cloud data includes preprocessing the collected three-dimensional point cloud data; Categorize key equipment in the distribution network into large equipment and small equipment, and extract multi-scale point cloud data; The multi-scale point cloud data includes large-scale data and small-scale data; For large devices, a 15 cm voxel grid is used for fast segmentation to obtain large-scale data; The large-scale data includes the macroscopic outline, key information and main structure of the equipment; For small devices, a 1 cm voxel grid is used to obtain small-scale data; The small-scale data include the morphology and damage area of ​​the device surface; For large-scale data, the Euclidean clustering method is used to extract the overall outline of the equipment, and the curvature edge detection method is used to extract the appearance and structural features of large equipment for macroscopic modeling; For small-scale data, curvature analysis is used to identify surface damage, and morphological methods are used to enhance surface details, highlight the surface features of tiny extracted small devices, and perform fine modeling.

6. The active operation and maintenance method for distribution network based on point cloud model according to claim 5, characterized in that: The comprehensive health index includes structural integrity index, surface damage index, environmental impact index, equipment aging index and failure history index; The structural integrity index includes measuring the condition of the overall structure, supporting frame, surface damage, etc. of the equipment; The surface damage index includes measuring the degree of cracks, corrosion, decay and wear on the surface of the equipment; The environmental impact index includes considering the impact of the external environment on the health of key equipment in the distribution network; The equipment aging index includes the impact of the equipment's age and maintenance history on the health of key equipment in the distribution network; The fault history index includes evaluating the current health status of key equipment in the distribution network based on the equipment's fault records and maintenance history; The formula for calculating the comprehensive health index is expressed as: H=w1·S structural +w2·S surface +w3·S environment +w4·S age +w5·S history w1+w2+w3+w4+w5=1 Among them, H represents the comprehensive health index, S structural represents the structural integrity index, w1 represents the weight of the structural integrity index; S surface represents the surface damage index, w2 represents the weight of the surface damage index; S environment represents the environmental impact index, w3 represents the weight of the environmental impact index; S age represents the equipment aging index, w4 represents the weight of the equipment aging index; S history represents the fault history index, w5 represents the weight of the fault history index; The faults of the identification device include constructing an identification model by combining the random forest and support vector machine algorithms, training the extracted appearance structure features of large-scale devices and surface features of small-scale devices, and identifying fault modes, which is expressed by the formula: Among them, y represents the classification result of the failure mode, X i represents the i-th feature of the distribution network equipment, α i Represents the weight coefficient of the i-th random forest tree; Represents the random forest model for feature X i The classification result; β represents the weight coefficient of the support vector machine, w represents the weight vector of the support vector machine; b represents the bias term of the support vector machine; represents the normalization function, K(X,X′) represents the Gaussian kernel function,<w,X> Indicates the inner product calculation of the support vector machine; sign(z) indicates the category of the returned decision result; When y = 0, it indicates that no fault occurs; When y = 1, it indicates that cracks occur in the key devices of the distribution network, and the fault score is 5 points; When y = 2, it indicates that the key devices of the distribution network are corroded, and the fault score is 4 points; When y = 3, it indicates that the key devices of the distribution network are deformed, and the fault score is 3 points; When y = 4, it indicates that the key devices of the distribution network are worn, and the fault score is 2 points; Locating the key devices with faults includes, for each identified fault, using the spatial coordinates of each point in the point cloud data and combining with the fault type to label the key devices of the distribution network where the fault occurs.

7. The active operation and maintenance method for distribution network based on point cloud model according to claim 6, characterized in that: Generating targeted operation and maintenance strategies includes designing a priority ranking model according to the health index and fault mode of the distribution network devices, which is expressed by the formula: P=w1·H+w2·F mode w1 + w2 = 1 Among them, H represents the comprehensive health index, w1 represents the weight of the comprehensive health index; F mode represents the failure mode score, w2 represents the weight of the failure mode score; When 0 ≤ P ≤ 10, it indicates that the distribution network has no faults, and regular inspections, continuous monitoring, and timely minor maintenance are carried out; When 10 < P ≤ 20, it indicates that medium faults occur in the distribution network, and enhanced monitoring, medium-scale repairs, and preventive maintenance are carried out; When 20 < P ≤ 30, it indicates that high-level faults occur in the distribution network, and emergency repairs are carried out, shutdown processing is carried out if necessary, and resource support is increased; When 30 < P, it indicates that serious faults occur in the distribution network, and the equipment is immediately shut down, replaced, and a catastrophic event handling procedure is started.

8. A distribution network active operation and maintenance system based on a point cloud model, characterized in that: A data acquisition module uses a drone equipped with a point cloud sensor to conduct flight inspections on distribution network devices and real-time collect three-dimensional point cloud data of the distribution network devices; A data processing module performs multi-scale processing on the three-dimensional point cloud data, calculates the comprehensive health index of the distribution network devices, identifies the fault modes of the devices, and locates them; A strategy generation module generates targeted operation and maintenance strategies according to the health index and fault mode of the distribution network devices.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, it implements the steps of the distribution network active operation and maintenance method according to any one of claims 1 to 7 based on the point cloud model.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the distribution network active operation and maintenance method according to any one of claims 1 to 7 based on the point cloud model.

Images