Intelligent inspection system of electric power material storage warehouse

Abstract

The invention relates to the technical field of electric power material storage warehouse inspection, in particular to an intelligent inspection system of an electric power material storage warehouse, which comprises a database module used for storing comprehensive information of the electric power material storage warehouse, and a model construction module used for constructing a three-dimensional space model of the electric power material storage warehouse according to the comprehensive information; according to the invention, the feature vectors of the plurality of sub-tasks in each inspection area are constructed and the plurality of sub-tasks are clustered, so that a plurality of relatively balanced task sets can be obtained, the plurality of sub-tasks can be reasonably distributed, all unmanned aerial vehicles in the inspection area can be utilized to the maximum extent, the operation efficiency of the unmanned aerial vehicles is improved, and the working efficiency of the unmanned aerial vehicles is improved. According to the method, the unmanned aerial vehicle can inspect more sub-tasks within the average flight duration, path planning can be carried out on the task set, the shooting point position path and the inspection path of the sub-tasks are balanced, and integrity and continuity of inspection data corresponding to each sub-task are facilitated while the inspection efficiency is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of inspection of electric power material storage depots, and in particular to an intelligent inspection system of electric power material storage depots. Background Art

[0002] Since the existing power material storage depots occupy a large area and store a large variety of power materials, some power material storage depots use relevant personnel to conduct inspections, which wastes human resources and causes the inspection time to be too long. This method has been gradually replaced by drone inspections. In order to avoid the above problems, some power material storage depots also use drones for inspections. However, when allocating multiple inspection tasks of the power material storage depots, if the inspection tasks are assigned to drones only according to the release order of multiple inspection tasks, when the power material storage depots are large, it is difficult for each drone to complete a large-scale inspection task. Therefore, how to split and reorganize multiple inspection tasks so that drones can quickly complete all inspection tasks becomes a difficult problem. In addition, when performing path planning, how to ensure the inspection efficiency while ensuring the integrity and continuity of the inspection data corresponding to each subtask also becomes a difficult problem. Summary of the invention

[0003] In view of the above-mentioned shortcomings of the prior art, the present invention provides an intelligent inspection system for electric power material storage depots to solve the above-mentioned problems.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0005] The present invention provides an intelligent inspection system for a power material reserve depot, comprising a database module for storing comprehensive information of the power material reserve depot, and further comprising:

[0006] A model building module, which is used to build a three-dimensional spatial model of the power material storage depot based on comprehensive information;

[0007] A channel establishment module, which is used to establish a three-dimensional safe flight channel map based on a three-dimensional space model;

[0008] An area division module is used to divide the three-dimensional space model according to the comprehensive information and the three-dimensional safe flight channel map, thereby obtaining multiple inspection areas;

[0009] A task allocation module is used to receive, split and reorganize multiple inspection tasks in each inspection area, thereby obtaining multiple task sets, wherein the multiple task sets are obtained by clustering the split inspection tasks in each inspection area through an improved K-means algorithm;

[0010] The abnormality judgment module is used to obtain the planned paths of multiple task sets and execute the task sets according to the planned paths, thereby obtaining inspection data, and comparing the inspection data with the preset inspection standards to determine whether there are any abnormalities in the inspection data.

[0011] Furthermore, the comprehensive information includes power material information, transportation and storage equipment information, machine nest information, and drone information in the machine nest;

[0012] The power material information includes the power material name, power material specification, power material quantity, power material location and power material status;

[0013] The information of the transport and storage equipment includes the name of the transport and storage equipment, the specifications of the transport and storage equipment, and the quantity of the transport and storage equipment;

[0014] The nest information includes the nest name, nest location, and the number of drones in the nest;

[0015] Drone information includes drone name, drone specifications, and drone status.

[0016] Furthermore, the method of constructing a three-dimensional space model of the power material storage depot based on the comprehensive information includes:

[0017] Obtain architectural drawings of the power material storage depot, and construct a basic three-dimensional framework of the power material storage depot based on the architectural drawings;

[0018] Use a depth camera to photograph the power materials, transportation and storage equipment, and machine nests in the power material storage depot, and then obtain a power material model, a transportation and storage equipment model, and a machine nest model;

[0019] The power material model, transportation and storage equipment model and machine nest model are filled into the basic three-dimensional framework to obtain a three-dimensional space model.

[0020] Furthermore, the method of establishing a three-dimensional safe flight channel diagram based on the three-dimensional space model includes:

[0021] Obtain the minimum safe passage space required for the UAV to fly, and thereby obtain multiple spatial regions in the three-dimensional space model that are greater than or equal to the minimum safe passage space;

[0022] Preset channel construction conditions, and form a safe flight channel with multiple continuous space areas that meet the channel construction conditions. The construction condition is that the specification of the drop space corresponding to the connection between two adjacent minimum safe passage spaces is greater than or equal to the specification of the minimum safe passage space;

[0023] All safe flight channels in the three-dimensional space model are extracted to obtain a three-dimensional safe flight channel map, in which each safe flight channel in the three-dimensional safe flight channel map can be connected to the drone nest directly or through other safe flight channels.

[0024] Furthermore, the method of dividing the three-dimensional space model according to the comprehensive information and the three-dimensional safe flight channel diagram to obtain multiple inspection areas includes:

[0025] Obtain the average flight distance of the drone and the length of each safe flight channel in the three-dimensional safe flight channel map, and associate the length with the corresponding safe flight channel;

[0026] Use the graph traversal algorithm to traverse the three-dimensional safe flight channel graph, and take the position with a distance less than or equal to the average flight distance and the largest straight-line distance to the drone nest as the boundary point;

[0027] Obtain boundary points in multiple directions around the drone nest and connect multiple boundary points to form the initial inspection area of ​​the drone nest;

[0028] Obtain overlapping areas of the initial inspection areas corresponding to multiple drone nests, preset attribution rules, and attribute the overlapping areas to the corresponding initial inspection areas according to the attribution rules, thereby obtaining multiple inspection areas;

[0029] Among them, the attribution rule is: first, the power materials and their corresponding areas in the overlapping areas of multiple initial inspection areas are classified into the initial inspection area with the same power materials, and then the remaining power materials and their corresponding areas in the overlapping areas are classified into the adjacent and smallest initial inspection area.

[0030] Furthermore, the method of receiving, splitting and reorganizing multiple inspection tasks in each inspection area to obtain multiple task sets includes:

[0031] Receive multiple inspection tasks in each inspection area, and split each inspection task in each inspection area into multiple subtasks, wherein each subtask corresponds to an electric power material and an inspection item of the electric power material;

[0032] For each subtask, create a feature vector Vi = [oi, di, pi], where oi is the normalized position of the ith subtask, di is the normalized remaining time of the ith subtask, and pi is the normalized complexity level of the ith subtask;

[0033] All subtasks in each inspection area are clustered using the improved K-means algorithm, and the objective function is defined. Similar subtasks are then grouped together to obtain multiple task sets.

[0034] Among them, the expression of the objective function is:

[0035]

[0036] Where Jminimize is the objective function, K represents the total number of task sets after clustering, Cj is the jth task set, Vi is the feature vector of the i-th subtask, μj is the center point of the jth task set, ||Vi-μi|| 2 Represents the Euclidean distance between the feature vector Vi of the i-th subtask and the center point μj of the j-th task set.

[0037] Furthermore, the clustering process includes:

[0038] Calculate the initial K value and randomly select K initial center points, calculate the distance D(Vi, μj) between each subtask Vi and all center points, and assign Vi to the cluster Cj with the smallest distance;

[0039] Recalculate the center point of each cluster, and then the center point of each cluster is:

[0040]

[0041] Where μj is the center point of the jth cluster, nj is the number of subtasks in the jth cluster;

[0042] Obtain the estimated total inspection time Tz for each cluster, the average flight time Tavg of the drone, and the maximum round-trip time Tmr for the drone to reach the center point of each cluster and complete the return trip;

[0043] For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz, where Tc∈[Tcmin, Tcmax], Tcmin is the minimum elastic time, and Tcmax is the maximum elastic time;

[0044] For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz. If Tc>Tcmax, add the subtasks in the other closest clusters to the current cluster. If Tc<Tcmin, add the subtasks in the current cluster to the other closest clusters. If it cannot be added to other clusters, add a new cluster.

[0045] Repeat the above process until all clusters satisfy Tc∈[Tcmin, Tcmax], that is, clustering is completed. Each cluster that completes clustering is a task set, where Tcmin is the minimum elastic time and Tcmax is the maximum elastic time.

[0046] Furthermore, the method of calculating the distance D(Vi, μj) between each subtask Vi and all center points includes:

[0047] Preset distance functions:

[0048] D(Vi,μj)=ω1×|Voi-μjoi|+ω2×|Vdi-μjdi|+ω3×|Vpi-μjpi|,

[0049] where D(Vi,μj) represents the distance between the feature vector of the ith subtask and the center point of the jth task set, Voi and μjoi represent the position of the ith subtask and the center point of the jth task set, Vdi and μjdi represent the remaining time of the ith subtask and the center point of the jth task set, Vpi and μjpi represent the complexity level of the ith subtask and the center point of the jth task set, ω1, ω2 and ω3 are the position weight factor, remaining time weight factor and complexity level weight factor of the ith subtask and the center point of the jth task set, respectively.

[0050] Furthermore, the method for obtaining the complexity level of the subtask includes:

[0051] Obtain multiple shooting points corresponding to the inspection items of each power material, and use the minimum number of shots required to complete all inspection items within the three-dimensional safe flight channel map as the number of shooting points;

[0052] Obtain the minimum number of shooting points corresponding to all power inspection items, sort the minimum number of shooting points in ascending order, and then obtain a sequence list of the minimum number of shooting points;

[0053] The number of complexity levels is preset, and the minimum shooting point number sequence list is evenly divided according to the number of complexity levels, thereby obtaining multiple intervals, each of which corresponds to a complexity level, wherein the more the number of shooting points is, the higher the complexity level is.

[0054] Furthermore, the method of obtaining the planned paths of multiple task sets and executing the task sets according to the planned paths, thereby obtaining inspection data, and comparing the inspection data with preset inspection standards to determine whether the inspection data is abnormal includes:

[0055] Obtain the shooting points of electric power materials on the safe flight channel corresponding to each subtask in the task set, and the minimum shooting points required to complete the inspection project to build a point set, and then obtain multiple point sets;

[0056] Use the TSP algorithm to determine the shooting point sequence of each point set, and then obtain the shooting point path of the subtask;

[0057] Use the nearest neighbor algorithm to determine the inspection order of each subtask in the task set and obtain the inspection path;

[0058] Fill the shooting point path of each subtask into the inspection path to obtain the initial complete path;

[0059] Exchange the inspection order of the subtasks in the initial complete path, and use the initial complete path with the shortest total path length as the planned path for the task set;

[0060] Multiple task sets are assigned to drones in the inspection area one by one. The drones inspect the subtasks in the task set by planning paths, and then obtain inspection data. Among them, each inspection item of each power material has a corresponding inspection standard.

[0061] Compare the inspection data of the subtask with the corresponding inspection standard. If the two are the same, there is no abnormality. Otherwise, there is an abnormality.

[0062] The process of exchanging the inspection order of subtasks in the initial complete path is as follows:

[0063] Randomly swap the inspection order of U subtasks, record the total path length of the initial complete path after each swap, and associate the swapped subtask inspection order with the shortened distance of the total path length;

[0064] The inspection order of the subtasks exchanged each time is sorted in descending order according to the shortened distance of the total path length to obtain an optimization list, and the initial complete path corresponding to the first ranked in the optimization list is used as the basic exchange path; based on the basic exchange path, the inspection order of U subtasks is exchanged again until the inspection order of the exchanged subtasks is exchanged S times in a row or its shortened distance is less than the preset distance threshold, that is, the exchange is stopped.

[0065] Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects:

[0066] By constructing a three-dimensional spatial model and a three-dimensional safe flight channel map of the power material storage depot, it is possible to screen out spatial areas that meet the minimum safe passage space specifications required for drone flight, so as to improve the flight safety of drones during inspections. By dividing the inspection area within the power material storage depot, it is possible to avoid the situation where drones cannot return to the machine nest after completing the inspection mission, which is conducive to the subsequent path planning of drones.

[0067] By constructing feature vectors of multiple subtasks in each inspection area and clustering the multiple subtasks, a relatively balanced set of tasks can be obtained. Therefore, the present invention can not only reasonably allocate multiple subtasks, but also maximize the use of all drones in the inspection area, improve the operating efficiency of drones, and enable drones to inspect more subtasks within the average flight time, so as to quickly complete all inspection tasks.

[0068] The TSP algorithm and the nearest neighbor algorithm can determine the optimal shooting point sequence and subtask inspection sequence, which can improve the overall inspection efficiency; by calculating the values ​​of U and S according to the average number of subtasks in the task set and the average number of shooting points corresponding to each subtask, U and S can be adapted to different task sets and balance the computational complexity and optimization effect; thus, a balance can be achieved between the shooting point path and the inspection path of the subtask in the planned path, which ensures the inspection efficiency while also facilitating the integrity and continuity of the inspection data corresponding to each subtask. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0070] Figure 1 It is the overall module block diagram of the present invention;

[0071] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] The present invention will be further described below in conjunction with the embodiments.

[0074] Scenario Overview:

[0075] The existing power material storage depots have a large area and store a large number of power materials. Some power material storage depots use relevant personnel for inspection, which wastes human resources and causes the inspection time to be too long. This method has been gradually replaced by drone inspections. In order to avoid the above problems, some power material storage depots also use drones for inspections. However, when allocating multiple inspection tasks of the power material storage depot, if the inspection tasks are assigned to drones only according to the release order of multiple inspection tasks, when the power material storage depot is large, it is difficult for each drone to complete a large-scale inspection task. Therefore, how to split and reorganize multiple inspection tasks so that drones can quickly complete all inspection tasks becomes a problem. In addition, when performing path planning, how to ensure the inspection efficiency while ensuring the integrity and continuity of the inspection data corresponding to each subtask also becomes a problem;

[0076] By constructing a three-dimensional spatial model and a three-dimensional safe flight channel map of an electric power material storage depot, the present application can screen out spatial areas that meet the minimum safe passage space specifications required for drone flight, so as to improve the flight safety of drones during inspections. By dividing the inspection areas within the electric power material storage depot, it can avoid the situation where the drone cannot return to the machine nest after completing the inspection task, which is beneficial to the path planning of subsequent drones. By constructing feature vectors of multiple subtasks in each inspection area and clustering the multiple subtasks, a relatively balanced multiple task sets can be obtained, which can not only reasonably allocate multiple subtasks, but also maximize the use of all drones in the inspection area, improve the operating efficiency of drones, and enable drones to inspect more subtasks within the average flight time, so as to quickly complete all inspection tasks; and the task set can also be path planned to achieve a balance between the shooting point path of the subtask and the inspection path, while ensuring the inspection efficiency, it is also beneficial to the integrity and continuity of the inspection data corresponding to each subtask.

[0077] Reference Figure 1 and Figure 2 :An intelligent inspection system for power material storage depots, comprising at least:

[0078] A database module that stores comprehensive information of power material reserves;

[0079] Comprehensive information includes power material information, transportation and storage equipment information, machine nest information, and drone information in the machine nest;

[0080] The power material information includes the power material name, power material specification, power material quantity, power material location and power material status (power material is in normal, damaged, under repair or any other condition);

[0081] The information of transportation and storage equipment includes the name of transportation and storage equipment, the specifications of transportation and storage equipment and the quantity of transportation and storage equipment (transportation and storage equipment includes material lifting equipment, material storage equipment (shelves), material transportation equipment, etc.);

[0082] The nest information includes the nest name, nest location, and the number of drones in the nest;

[0083] Drone information includes drone name, drone specifications, and drone status (drone is normal, charging, under maintenance, or any other).

[0084] The model building module constructs a three-dimensional spatial model of the power material storage depot based on the comprehensive information, including:

[0085] Obtain architectural drawings of the power material storage depot, and construct a basic three-dimensional framework of the power material storage depot based on the architectural drawings;

[0086] Use a depth camera to photograph the power materials, transportation and storage equipment, and machine nests in the power material storage depot, and then obtain a power material model, a transportation and storage equipment model, and a machine nest model;

[0087] Fill the power material model, the transportation and storage equipment model and the machine nest model into the basic three-dimensional framework (establish a unified coordinate system, and fill the power material model, the transportation and storage equipment model and the machine nest model into the basic three-dimensional framework according to the coordinate system), and then obtain a three-dimensional space model, wherein the three-dimensional space model is updated in real time with the changes in the quantity and location of power materials;

[0088] Specifically, by constructing a three-dimensional space model of the power material reserve warehouse based on the comprehensive information of the power material reserve warehouse, the structure of the power material reserve warehouse can be obtained, which is beneficial for the system to update the three-dimensional space model according to changes in the quantity and location of power materials in the power material reserve warehouse, and then facilitate the real-time update of the three-dimensional safe flight channel map following the root three-dimensional space model.

[0089] The channel establishment module establishes a three-dimensional safe flight channel map based on the three-dimensional space model, including:

[0090] Obtain the minimum safe passage space required for the UAV to fly, and thereby obtain multiple spatial regions in the three-dimensional space model that are greater than or equal to the minimum safe passage space;

[0091] Preset channel construction conditions, and form a safe flight channel with multiple continuous space areas that meet the channel construction conditions. The construction condition is that the specification of the drop space corresponding to the connection between two adjacent minimum safe passage spaces is greater than or equal to the specification of the minimum safe passage space;

[0092] Extract all the safe flight channels in the three-dimensional space model, and then obtain a three-dimensional safe flight channel map, wherein each safe flight channel in the three-dimensional safe flight channel map can be connected to the drone nest directly or through other safe flight channels;

[0093] It should be noted that the minimum safe passage space is based on the passage space with less than the average number of collisions of the drone in the historical flight process, and its passage space is a three-dimensional space composed of a cuboid;

[0094] Specifically, by setting up a three-dimensional safe flight channel map, it is possible to obtain the paths that drones can use during inspections in the power material storage depot; by screening space areas that meet the minimum safe passage space specifications for drone flight, it can ensure that drones will not collide due to lack of space during flight, thereby improving the flight safety of drones during inspections; by having each safe flight channel directly or indirectly connected to the drone nest, it can ensure that each safe flight channel in the three-dimensional safe flight channel map can fly; this is in turn conducive to the subsequent division of inspection areas and the path planning of drones.

[0095] The area division module divides the three-dimensional space model according to the comprehensive information and the three-dimensional safe flight channel map to obtain multiple inspection areas, including:

[0096] Get the average flight distance of the drone and the length of each safe flight channel in the three-dimensional safe flight channel diagram, and associate the length with the corresponding safe flight channel, where the average flight distance of the drone = (average flight time of the drone * average flight speed of the drone) / 2, divided by 2 includes the return trip of the drone;

[0097] Use the graph traversal algorithm to traverse the three-dimensional safe flight channel graph, and take the position with a distance less than or equal to the average flight distance and the largest straight-line distance to the drone nest as the boundary point;

[0098] Obtain boundary points in multiple directions around the drone nest and connect multiple boundary points to form the initial inspection area of ​​the drone nest;

[0099] Obtain overlapping areas of the initial inspection areas corresponding to multiple drone nests, preset attribution rules, and attribute the overlapping areas to the corresponding initial inspection areas according to the attribution rules, thereby obtaining multiple inspection areas;

[0100] The attribution rule is as follows: first, the power materials and their corresponding areas in the overlapping areas of multiple initial inspection areas are classified into the initial inspection area with the same power materials, and then the remaining power materials and their corresponding areas in the overlapping areas are classified into the adjacent and smallest initial inspection area;

[0101] Specifically, by obtaining the average flight distance of the UAV and a three-dimensional safe flight channel map, and accurately determining the boundary points through a graph traversal algorithm (such as BF or DFS), it is possible to ensure that the area division is reasonable and efficient, and avoid the situation where the power material storage depot has uncovered areas due to the limitation of the average flight distance of the UAV; by dividing the average flight distance of the UAV by 2, it is possible to avoid the situation where the UAV cannot return to the nest after completing the inspection mission; by setting the attribution rules, it is possible to ensure that the inspection area of ​​the UAV is not missed and has no redundancy, so as to improve the inspection efficiency and resource utilization of the UAV.

[0102] The task allocation module receives, splits and reorganizes multiple inspection tasks in each inspection area to obtain multiple task sets, including:

[0103] Receive multiple inspection tasks in each inspection area, and split each inspection task in each inspection area into multiple subtasks, wherein each subtask corresponds to an electric power material and an inspection item of the electric power material;

[0104] For each subtask, create a feature vector Vi = [oi, di, pi], where oi is the normalized position of the ith subtask, di is the normalized remaining time of the ith subtask, and pi is the normalized complexity level of the ith subtask;

[0105] Among them, Min-Max is used for normalization. For example, the normalized remaining time di is: Where d`i is the normalized complexity level of the ith subtask, dmin is the minimum complexity level of all subtasks, and dmax is the maximum complexity level of all subtasks;

[0106] The normalized complexity level pi is: Among them, p`i is the normalized complexity level of the ith subtask, pmin is the minimum complexity level of all subtasks, and pmax is the maximum complexity level of all subtasks;

[0107] All subtasks in each inspection area are clustered using the improved K-means algorithm, and the objective function is defined. Similar subtasks are then grouped together to obtain multiple task sets.

[0108] Among them, the expression of the objective function is:

[0109]

[0110] Where Jminimize is the objective function, K represents the total number of task sets after clustering, Cj is the jth task set, Vi is the feature vector of the i-th subtask, μj is the center point of the jth task set, ||Vi-μi|| 2Represents the Euclidean distance between the feature vector Vi of the i-th subtask and the center point μj of the j-th task set;

[0111] The clustering process includes:

[0112] Calculate the initial K value and randomly select K initial center points, calculate the distance D(Vi, μj) between each subtask Vi and all center points, and assign Vi to the cluster Cj with the smallest distance;

[0113] Recalculate the center point of each cluster, and then the center point of each cluster is:

[0114]

[0115] Where μj is the center point of the jth cluster, nj is the number of subtasks in the jth cluster;

[0116] Obtain the estimated total inspection time Tz for each cluster, the average flight time Tavg of the drone, and the maximum round-trip time Tmr for the drone to reach the center point of each cluster and complete the return trip;

[0117] For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz, where Tc∈[Tcmin, Tcmax], Tcmin is the minimum elastic time, and Tcmax is the maximum elastic time;

[0118] For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz. If Tc>Tcmax, add the subtasks in the other closest clusters to the current cluster. If Tc<Tcmin, add the subtasks in the current cluster to the other closest clusters. If it cannot be added to other clusters, add a new cluster.

[0119] Repeat the above process until all clusters satisfy Tc∈[Tcmin, Tcmax], that is, clustering is completed. Each cluster that completes clustering is a task set, where Tcmin is the minimum elastic time and Tcmax is the maximum elastic time.

[0120] It should be noted that the initial K value is obtained by dividing the sum of the number of all subtasks in the inspection area by the average number of historical subtasks of the drone;

[0121] The minimum and maximum flexible time are obtained by obtaining multiple groups of historical inspection tasks of drones, obtaining the difference between the estimated total inspection time and the actual total inspection time corresponding to each group of historical inspection tasks, calculating the average and standard deviation of the differences corresponding to the multiple groups of historical inspection tasks, and taking the sum of the average and the standard deviation as the maximum flexible time, and taking the value of the average minus the standard deviation as the minimum flexible time;

[0122] The method of calculating the distance D(Vi, μj) between each subtask Vi and all center points includes:

[0123] Preset distance functions:

[0124] D(Vi,μj)=ω1×|Voi-μjoi|+ω2×|Vdi-μjdi|+ω3×|Vpi-μjpi|,

[0125] Where D(Vi,μj) represents the distance between the feature vector of the ith subtask and the center point of the jth task set, Voi and μjoi represent the position of the ith subtask and the center point of the jth task set respectively (the position difference between Voi and μjoi can be calculated by the Euclidean distance algorithm), Vdi and μjdi represent the remaining time of the ith subtask and the remaining time of the center point of the jth task set respectively, Vpi and μjpi represent the complexity level of the ith subtask and the complexity level of the center point of the jth task set respectively, ω1, ω2 and ω3 are the position weight factor, remaining time weight factor and complexity level weight factor of the ith subtask and the center point of the jth task set respectively, and the sum of ω1, ω2 and ω3 is 1;

[0126] Specifically, by forming a feature vector with the location, remaining time and complexity level of the subtask, it helps to combine similar tasks into a task set. By optimizing the task set based on location, remaining time and complexity level, the flight path and time of the UAV can be effectively reduced, and the inspection efficiency can be improved. By considering the elastic time between the total inspection time and the average flight time of the UAV in the clustering process, the system can balance the number of subtasks in each task set, maximize the use of the average flight time of each UAV, and avoid idle or overloaded UAVs.

[0127] The method of obtaining the complexity level of the subtask includes:

[0128] Obtain multiple shooting points corresponding to the inspection items of each power material, and take the minimum number of shots required to complete all inspection items within the three-dimensional safe flight channel map as the number of shooting points, where the shooting points are the shooting positions required to complete the inspection items;

[0129] Obtain the minimum number of shooting points corresponding to all power inspection items, sort the minimum number of shooting points in ascending order, and then obtain a sequence list of the minimum number of shooting points;

[0130] Preset the number of complexity levels, and evenly divide the minimum number of shooting points according to the number of complexity levels to obtain multiple intervals, each of which corresponds to a complexity level, where the more shooting points there are, the higher the complexity level (for example, the number of shooting points corresponding to the four subtasks is 2, 5, 9, and 14, and the intervals of complexity levels are [1,3], [4,6], [7,9], [10,12], and [13,15] respectively; then, the complexity levels of the four subtasks are level 1 complexity level, level 2 complexity level, level 4 complexity level, and level 5 complexity level respectively);

[0131] Specifically, by determining the complexity level through shooting points, the complexity level assessment of each power material can be more standardized. By considering the shooting points within the three-dimensional safe flight channel map, the safety of drone inspections can be effectively guaranteed. By determining the minimum number of shots required to complete the inspection project, unnecessary duplication of work can be avoided and the inspection efficiency can be improved.

[0132] The abnormality judgment module obtains the planned paths of multiple task sets and executes the task sets according to the planned paths, thereby obtaining inspection data, and compares the inspection data with the preset inspection standards to determine whether there are abnormalities in the inspection data, including:

[0133] Obtain the shooting points of electric power materials on the safe flight channel corresponding to each subtask in the task set, and the minimum shooting points required to complete the inspection project to build a point set, and then obtain multiple point sets;

[0134] Use the TSP algorithm to determine the shooting point sequence of each point set, and then obtain the shooting point path of the subtask;

[0135] Use the nearest neighbor algorithm to determine the inspection order of each subtask in the task set and obtain the inspection path;

[0136] Fill the shooting point path of each subtask into the inspection path to obtain the initial complete path;

[0137] Exchange the inspection order of the subtasks in the initial complete path, and use the initial complete path with the shortest total path length as the planned path for the task set;

[0138] The process of exchanging the inspection order of subtasks in the initial complete path is as follows:

[0139] Randomly swap the inspection order of U subtasks, record the total path length of the initial complete path after each swap, and associate the swapped subtask inspection order with the shortened distance of the total path length;

[0140] The inspection order of each exchanged subtask is sorted in descending order according to the shortened distance of the total path length to obtain an optimized list, and the initial complete path corresponding to the first ranked in the optimized list is used as the basic exchange path; the inspection order of U subtasks is exchanged again based on the basic exchange path until the inspection order of the exchanged subtask is exchanged S times in a row or the shortened distance is less than the preset distance threshold, that is, the exchange is stopped;

[0141] The way to obtain U and S is:

[0142] In the formula, is the average number of subtasks in the task set, is the average number of shooting points corresponding to each subtask in the task set, Indicates that the lower limit of U is 2, and U follows the average number of subtasks in the task set increases with the increase of Indicates that the lower limit of S is 5, and S follows the average number of shooting points corresponding to each subtask in the task set 2 and 5 can also be modified by relevant personnel according to actual conditions.

[0143] Specifically, the TSP algorithm and the nearest neighbor algorithm can determine the optimal shooting point sequence and subtask inspection sequence, which can improve the overall inspection efficiency. By exchanging and optimizing the subtask sequence multiple times, the total path length can be continuously shortened, and finally a flight path close to the global optimal is obtained to reduce the flight distance of the drone. By ensuring that each shooting point is on a safe flight channel, the accident risk of the drone is low when performing tasks. By calculating the values ​​of U and S according to the average number of subtasks in the task set and the average number of shooting points corresponding to each subtask, U and S can adapt to different task sets and balance the computational complexity and optimization effect. By recording and sorting each The shortened distance after the first exchange can identify the inspection order of the exchanged subtasks that has the greatest impact on the shortening of the total path length. By taking the initial complete path corresponding to the first position in the sequence as the basic exchange path, the inspection order of the subsequent exchanged subtasks can be exchanged on this basis, so as to more quickly obtain the initial complete path with the shortest total path length as the planning path of the task set; making the optimization process of the entire initial complete path more intelligent; and furthermore, a balance can be achieved between the shooting point path and the inspection path of the subtask, ensuring the inspection efficiency while also ensuring the integrity of the point set corresponding to each subtask, which is conducive to ensuring the integrity and continuity of the inspection data corresponding to each subtask;

[0144] Multiple task sets are assigned to drones in the inspection area one by one. The drones inspect the subtasks in the task set by planning paths, and then obtain inspection data. Among them, each inspection item of each power material has a corresponding inspection standard.

[0145] Compare the inspection data of the subtask with the corresponding inspection standard. If the two are the same, there is no abnormality. Otherwise, there is an abnormality.

[0146] Specifically, by having corresponding inspection standards for each type of power material, the consistency and reliability of the inspection of each type of power material is ensured. By comparing the inspection data with the inspection standards, it is possible to quickly and intuitively judge whether there are any abnormalities in the power materials.

[0147] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent inspection system for a power material storage depot, comprising: A database module, which is used to store comprehensive information of power material reserves; A model building module, which is used to build a three-dimensional spatial model of the power material storage depot based on comprehensive information; A channel establishment module, which is used to establish a three-dimensional safe flight channel map based on a three-dimensional space model; The area division module is used to divide the three-dimensional space model according to the comprehensive information and the three-dimensional safe flight channel map, thereby obtaining multiple inspection areas, and is characterized in that it also includes: A task allocation module is used to receive, split and reorganize multiple inspection tasks in each inspection area, thereby obtaining multiple task sets, wherein the multiple task sets are obtained by clustering the split inspection tasks in each inspection area through the K-means algorithm; The abnormality determination module is used to obtain the planned paths of multiple task sets and execute the task sets according to the planned paths, thereby obtaining inspection data and determining whether there is an abnormality based on the inspection data, wherein: By acquiring multiple point sets, determining the shooting point order of each point set, obtaining the shooting point path of the subtask, and thereby determining the inspection order to obtain the inspection path, the initial complete path is obtained based on the inspection path to define the planned path of the task set, and the inspection data of the subtask is obtained through the planned path.

2. The intelligent inspection system for electric power material storage depot according to claim 1 is characterized in that: The comprehensive information includes power material information, transportation and storage equipment information, machine nest information, and drone information in the machine nest; The power material information includes the power material name, power material specification, power material quantity, power material location and power material status; The information of the transport and storage equipment includes the name of the transport and storage equipment, the specifications of the transport and storage equipment, and the quantity of the transport and storage equipment; The nest information includes the nest name, nest location, and the number of drones in the nest; Drone information includes drone name, drone specifications, and drone status.

3. The intelligent inspection system for electric power material storage depot according to claim 2 is characterized in that: The method of establishing a three-dimensional safe flight channel diagram based on the three-dimensional space model includes: Obtain the minimum safe passage space required for the UAV to fly, and thereby obtain multiple spatial regions in the three-dimensional space model that are greater than or equal to the minimum safe passage space; Preset channel construction conditions, and form a safe flight channel with multiple continuous space areas that meet the channel construction conditions. The construction condition is that the specification of the drop space corresponding to the connection between two adjacent minimum safe passage spaces is greater than or equal to the specification of the minimum safe passage space; All safe flight channels in the three-dimensional space model are extracted to obtain a three-dimensional safe flight channel map, in which each safe flight channel in the three-dimensional safe flight channel map can be connected to the drone nest directly or through other safe flight channels.

4. The intelligent inspection system for electric power material storage depot according to claim 2 is characterized in that: The method of dividing the three-dimensional space model according to the comprehensive information and the three-dimensional safe flight channel diagram to obtain multiple inspection areas includes: Obtain the average flight distance of the drone and the length of each safe flight channel in the three-dimensional safe flight channel map, and associate the length with the corresponding safe flight channel; Use the graph traversal algorithm to traverse the three-dimensional safe flight channel graph, and take the position with a distance less than or equal to the average flight distance and the largest straight-line distance to the drone nest as the boundary point; Obtain boundary points in multiple directions around the drone nest and connect multiple boundary points to form the initial inspection area of ​​the drone nest; Obtain overlapping areas of the initial inspection areas corresponding to multiple drone nests, preset attribution rules, and attribute the overlapping areas to the corresponding initial inspection areas according to the attribution rules, thereby obtaining multiple inspection areas; Among them, the attribution rule is: first, the power materials and their corresponding areas in the overlapping areas of multiple initial inspection areas are classified into the initial inspection area with the same power materials, and then the remaining power materials and their corresponding areas in the overlapping areas are classified into the adjacent and smallest initial inspection area.

5. The intelligent inspection system for electric power material storage depot according to claim 2 is characterized in that: The method of receiving, splitting and reorganizing multiple inspection tasks in each inspection area to obtain multiple task sets includes: Receive multiple inspection tasks in each inspection area, and split each inspection task in each inspection area into multiple subtasks, wherein each subtask corresponds to an electric power material and an inspection item of the electric power material; For each subtask, create a feature vector Vi = [oi, di, pi], where oi is the normalized position of the ith subtask, di is the normalized remaining time of the ith subtask, and pi is the normalized complexity level of the ith subtask; All subtasks in each inspection area are clustered using the improved K-means algorithm, and the objective function is defined. Similar subtasks are then grouped together to obtain multiple task sets. Among them, the expression of the objective function is: Where Jminimize is the objective function, K represents the total number of task sets after clustering, Cj is the jth task set, Vi is the feature vector of the i-th subtask, μj is the center point of the jth task set, ||Vi-μi|| 2 Represents the Euclidean distance between the feature vector Vi of the i-th subtask and the center point μj of the j-th task set.

6. The intelligent inspection system for electric power material storage depot according to claim 5 is characterized in that: The clustering process includes: Calculate the initial K value and randomly select K initial center points, calculate the distance D(Vi, μj) between each subtask Vi and all center points, and assign Vi to the cluster Cj with the smallest distance; Recalculate the center point of each cluster, and then the center point of each cluster is: Where μj is the center point of the jth cluster, nj is the number of subtasks in the jth cluster; Obtain the estimated total inspection time Tz for each cluster, the average flight time Tavg of the drone, and the maximum round-trip time Tmr for the drone to reach the center point of each cluster and complete the return trip; For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz, where Tc∈[Tcmin, Tcmax], Tcmin is the minimum elastic time, and Tcmax is the maximum elastic time; For each cluster, calculate its elastic time Tc, Tc = Tavg-Tmr-Tz. If Tc>Tcmax, add the subtasks in the other closest clusters to the current cluster. If Tc<Tcmin, add the subtasks in the current cluster to the other closest clusters. If it cannot be added to other clusters, add a new cluster. Repeat the above process until all clusters satisfy Tc∈[Tcmin, Tcmax], that is, clustering is completed. Each cluster that completes clustering is a task set, where Tcmin is the minimum elastic time and Tcmax is the maximum elastic time.

7. The intelligent inspection system for electric power material storage depot according to claim 6 is characterized in that: The method of calculating the distance D(Vi, μj) between each subtask Vi and all center points includes: Preset distance functions: D(Vi,μj)=ω1×|Voi-μjoi|+ω2×|Vdi-μjdi|+ω3×|Vpi-μjpi|, where D(Vi,μj) represents the distance between the feature vector of the ith subtask and the center point of the jth task set, Voi and μjoi represent the position of the ith subtask and the center point of the jth task set, Vdi and μjdi represent the remaining time of the ith subtask and the center point of the jth task set, Vpi and μjpi represent the complexity level of the ith subtask and the center point of the jth task set, ω1, ω2 and ω3 are the position weight factor, remaining time weight factor and complexity level weight factor of the ith subtask and the center point of the jth task set, respectively.

8. The intelligent inspection system for electric power material storage depot according to claim 5 is characterized in that: The method for obtaining the complexity level of the subtask includes: Obtain multiple shooting points corresponding to the inspection items of each power material, and use the minimum number of shots required to complete all inspection items within the three-dimensional safe flight channel map as the number of shooting points; Obtain the minimum number of shooting points corresponding to all power inspection items, sort the minimum number of shooting points in ascending order, and then obtain a sequence table of the minimum number of shooting points; The number of complexity levels is preset, and the minimum shooting point number sequence list is evenly divided according to the number of complexity levels, thereby obtaining multiple intervals, each of which corresponds to a complexity level.

9. The intelligent inspection system for electric power material storage depot according to claim 8, characterized in that: The method of determining whether there is an abnormality is specifically as follows: Acquire the shooting points of electric power materials on the safe flight channel corresponding to each subtask in the task set, and construct a point set with the minimum shooting points required to complete the inspection project, and obtain multiple point sets; Use the TSP algorithm to determine the shooting point sequence of each point set, and then obtain the shooting point path of the subtask; Use the nearest neighbor algorithm to determine the inspection order of each subtask in the task set and obtain the inspection path; Fill the shooting point path of each subtask into the inspection path to obtain the initial complete path; Exchange the inspection order of the subtasks in the initial complete path, and use the initial complete path with the shortest total path length as the planned path for the task set; Multiple task sets are assigned to drones in the inspection area one by one. The drones inspect the subtasks in the task set by planning paths to obtain inspection data. Compare the inspection data of the subtask with the corresponding inspection standard to determine whether there is any abnormality; The process of exchanging the inspection order of subtasks in the initial complete path is as follows: Randomly swap the inspection order of U subtasks, record the total path length of the initial complete path after each swap, and associate the swapped subtask inspection order with the shortened distance of the total path length; The inspection order of the subtasks exchanged each time is sorted in descending order according to the shortened distance of the total path length to obtain an optimization list, and the initial complete path corresponding to the first ranked in the optimization list is used as the basic exchange path; based on the basic exchange path, the inspection order of U subtasks is exchanged again until the inspection order of the exchanged subtasks is exchanged S times in a row or its shortened distance is less than the preset distance threshold, that is, the exchange is stopped.

10. The intelligent inspection system for electric power material storage depot according to claim 9, characterized in that: The acquisition method of U and S is: In the formula, is the average number of subtasks in the task set, is the average number of shooting points corresponding to each subtask in the task set, Indicates that the lower limit of U is 2, and U follows the average number of subtasks in the task set increases with the increase of Indicates that the lower limit of S is 5, and S follows the average number of shooting points corresponding to each subtask in the task set increases with the increase of .

Images