Unmanned aerial vehicle photovoltaic inspection route planning method based on spatial clustering and path optimization

Abstract

The invention provides an unmanned aerial vehicle photovoltaic inspection route planning method based on spatial clustering and path optimization, and belongs to the technical field of unmanned aerial vehicle aerial survey and intelligent planning. The method comprises the following steps: acquiring a photovoltaic panel layout file of a photovoltaic power station, and analyzing polygon coordinates of each photovoltaic panel; automatically identifying a UTM projection band according to the coordinates and carrying out coordinate conversion; combining adjacent photovoltaic panels into a plurality of areas based on a spatial clustering algorithm; calculating a minimum area bounding rectangle for each region and determining an optimal rotation angle; generating a coverage grid according to the camera parameters, and screening a grid center point in the area as a waypoint; and finally, sequencing the waypoints through a path optimization algorithm, and outputting a complete unmanned aerial vehicle route file. The method can minimize the number of images, improve the aerial photography efficiency and reduce redundant flight and image repetition on the premise of ensuring complete coverage of a photovoltaic area, and has relatively high automation and practical values.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) application technology, and in particular relates to a UAV route planning method for automated inspection of photovoltaic power plants. Background Technology

[0002] In new energy technologies, the operation and maintenance of photovoltaic power plants is a crucial link in ensuring power generation efficiency, and drone inspection is an important means to replace manual labor and improve operation and maintenance efficiency. The strategy of drone inspection is to take pictures of each photovoltaic panel by carrying high-definition or thermal imaging cameras to detect faults such as hot spots and microcracks. In order to achieve fully automated operation, a flight path that can fly over all photovoltaic panels needs to be planned in advance.

[0003] However, large-scale photovoltaic power plants typically occupy a large area, and the distribution of photovoltaic panels (or photovoltaic arrays) has two main characteristics: first, macroscopic dispersion, meaning that photovoltaic panels are divided into multiple discontinuous areas according to the terrain and planning; second, microscopic regularity, meaning that within each area, photovoltaic panels are usually deployed in rows and columns.

[0004] Existing flight path planning methods mostly employ a simple "bow"-shaped full-area coverage pattern. This method treats the entire plant area or several manually selected large areas as a single rectangle for indiscriminate coverage. Taking a power plant covering 1 square kilometer but with only 40% photovoltaic panel coverage as an example, this means that up to 60% of the flight path involves flying over open spaces, roads, or unused areas between photovoltaic panels. This ineffective flight significantly wastes the valuable flight endurance of drones, resulting in a limited area that can be covered in a single operation, increasing the total number of sorties and operating costs. This problem of wasted endurance is particularly prominent for ultra-large power plants.

[0005] Meanwhile, when the factory area is divided into multiple zones, existing methods typically require manual specification of the flight sequence or a simple fixed sequence from zone 1 to zone 2. This connection method does not consider the spatial topological relationships between zones, resulting in excessively long transfer paths for drones between zones, further reducing overall operational efficiency.

[0006] Many existing drone inspection solutions lack precise and intelligent route optimization tailored to the actual distribution of photovoltaic panels, resulting in wasted flight time and low operational efficiency. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention proposes a drone route planning method for automated inspection of photovoltaic power plants, specifically including the following steps: Step 1: Obtain an electronic drawing file containing the geographical locations of multiple photovoltaic panels, parse the file, automatically determine the UTM projection zone where the photovoltaic panel is located based on the longitude range of the photovoltaic panel coordinates, and use the UTM projection method to convert the latitude and longitude coordinates of the photovoltaic panel into plane rectangular coordinates. Step 2: Based on the transformed planar coordinate information, the centroid of each photovoltaic panel polygon is used as its spatial location representative point. The principle of spatial proximity is used to automatically merge photovoltaic panels with similar locations to form a photovoltaic cluster. Step 3: Calculate the average principal orientation angle for each photovoltaic cluster, including calculating the principal orientation angle for each photovoltaic panel (N in total) based on its geometric shape or coordinate data. Find all of them separately. and The sum of these values ​​is then substituted into the atan2 function to obtain the average angle. The calculation formula is as follows: in, It is the average principal orientation angle of the cluster. It is the orientation angle of the long side of each photovoltaic panel. It represents the number of photovoltaic panels in the cluster.

[0008] Step 4: Based on the average main direction angle of the photovoltaic cluster, generate an internal coverage sub-flight route for each photovoltaic cluster; Step 5: Based on the enhanced nearest neighbor algorithm, determine the access order of all sub-routes, splice the sub-routes into a single continuous total route, and control the UAV to return to the takeoff point after all sub-routes have been visited.

[0009] Furthermore, the spatial clustering described in step 2 specifically includes: Establish a two-dimensional spatial grid index covering the entire photovoltaic field area; A breadth-first search algorithm is used to calculate the Euclidean distance between the centroids of all photovoltaic panels, and those distances less than a preset threshold are selected. The photovoltaic panels are considered connected and merged into the same photovoltaic cluster; The threshold The calculation formula is: in, The length of the photovoltaic panel, is the clustering coefficient, with a value ranging from 1.2 to 1.5.

[0010] Furthermore, the process of generating internal coverage sub-routes in step 4 includes: The geometric outline of the photovoltaic cluster is buffered and expanded outward, with the buffer distance set at 5m; Set the coordinates of the expanded boundary points as the origin, and define the envelope points within the boundary. Rotate around the origin The angle (i.e., rotating back against the photovoltaic average direction) is used to obtain the coordinates of the envelope point in the rotating coordinate system. The calculation formula is as follows: After rotation, the photovoltaic array is aligned with the x′ axis in the average principal direction in the rotating coordinate system; Calculate the axis-aligned bounding box range in the rotated coordinate system. Within this range, several lines parallel to the y′ axis are generated along the x′ direction with step sizes, and candidate waypoints are then generated on these lines based on the point spacing. .

[0011] Candidate waypoints Rotate back to the original coordinate system and filter out the points located inside the outline of the photovoltaic cluster as the final waypoint set by the inclusion relationship between the points and the polygon.

[0012] Among them, candidate waypoints The formula for calculating the rotation back to the original coordinate system is: Furthermore, the grid row spacing and column spacing of the candidate waypoints are jointly determined by the focal length of the UAV camera, flight altitude, forward overlap rate and lateral overlap rate, and the forward overlap rate and lateral overlap rate are set to 10%, and the flight altitude is set to 35 meters.

[0013] Furthermore, the enhanced nearest neighbor algorithm described in step 5 includes: Take the drone's takeoff point as the current location ; For each unvisited sub-route Calculate the distance from the current position to its entry point. and exit point The Euclidean distance is calculated using the following formula: Select the sub-path that minimizes the distance and its endpoint as the next destination; If the selected point is the exit point, then connect to the sub-route in the reverse direction; otherwise, connect in the forward direction and update the current position to the corresponding other endpoint. Repeat the above process until all sub-paths have been visited, and finally control the drone to return to the takeoff point.

[0014] Furthermore, the electronic drawing file is in JSON format and contains the latitude and longitude coordinates of the polygon vertices of each photovoltaic panel.

[0015] Furthermore, the route file output by the method is compatible with the DJI Sikong platform import format.

[0016] Furthermore, the drone is equipped with a thermal imaging camera with a resolution of 640×512 and a focal length of 9.1 mm.

[0017] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the unmanned aerial vehicle photovoltaic inspection route planning method described in any one of the claims.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention significantly reduces the ineffective flight path of UAVs in photovoltaic inspection operations by spatially clustering discrete photovoltaic panels, generating sub-flight routes for overlapping regions, and intelligently optimizing the connection sequence between regions. This greatly improves the operational efficiency and endurance utilization of a single flight. This method not only achieves fully automated planning from blueprints to flight routes, avoiding the inefficiency and oversights of manual planning, but also further reduces the total task time by optimizing multi-regional transfer paths. This effectively reduces the overall operating cost of inspections, significantly improving the practicality and economy of UAV automated inspection technology. The effectiveness of this invention has been verified through implementation. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method. Figure 2 Example of a photovoltaic panel location distribution diagram; Figure 3 Example diagram for planned flight routes; Figure 4 A diagram illustrating the import of DJI Sikong (a flight path control technology) into flight routes. Detailed Implementation

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0021] The idea behind this invention is to spatially cluster macroscopically discrete photovoltaic panels into several compact operational clusters, generate an internal coverage route for each cluster that fits its shape and orientation, and finally intelligently connect the routes of all clusters through a global path optimization algorithm to form a single continuous mission route with the shortest total distance.

[0022] Explanation of terminology in this invention UTM is an abbreviation for Universal Transverse Mercator projection. It is a globally used planar projection coordinate system used to divide the Earth's surface into 60 longitudinal projection zones according to longitude, converting latitude and longitude coordinates (spherical coordinates) into plane rectangular coordinates (X, Y). 50N refers to the Northern Hemisphere, 114°E to 120°E.

[0023] Theoretical explanation of the invention: In the field of UAV flight path planning, traditional grid coverage algorithms, while simple, are highly efficient due to their direct correlation with the density of targets within the planned area. For scenarios like photovoltaic power plants, where targets (solar panels) are concentrated in a localized area with large open spaces between regions, traditional algorithms generate numerous invalid paths over these open areas, significantly wasting the UAV's endurance.

[0024] The core of this invention lies in decomposing a large-scale, low-density "sparse" covering problem into multiple small-scale, high-density "dense" covering subproblems, and solving the optimal connection between the subproblems.

[0025] Spatial clustering: Its effectiveness stems from the "island-like" layout characteristics of photovoltaic panels. Through clustering algorithms, the system can automatically identify these "islands" (i.e., photovoltaic clusters), allowing subsequent route planning to focus entirely on these effective areas, fundamentally avoiding ineffective coverage of the "ocean" (open land).

[0026] Rotating Bounding Box and Fitting Flight Path: For a rectangular photovoltaic cluster tilted relative to due north, using a "bow"-shaped flight path aligned with due north results in a suboptimal path length and number of turns. This invention calculates the cluster's principal orientation angle and aligns the flight path with the direction the rectangular photovoltaic cluster is tilted relative to due north. This is equivalent to rotating the "coordinate system" to the direction that best fits the target, thus achieving coverage with the fewest flight paths and the shortest total length—a manifestation of local optimization.

[0027] Enhanced Nearest Neighbor Algorithm: When multiple independent clusters exist, their access order is a typical Traveling Salesman Problem (TSP). The enhanced nearest neighbor algorithm used in this invention is an efficient heuristic solution. It not only finds the next cluster closest to the current location, but also innovatively considers both the "entry" and "exit" points into the cluster, and allows for the "reverse" execution of the cluster's sub-routes based on the optimal decision. This flexibility greatly shortens the transition paths between regions, achieving near-optimal global paths.

[0028] Example 1 The following is the processing flow of a UAV photovoltaic inspection route planning method based on spatial clustering and path optimization provided by this invention: Step 1: In a factory area in Suzhou, photovoltaic (PV) panels are distributed across an area of ​​approximately 4 hectares, totaling 4400 panels. An electronic drawing file containing the geographical locations of multiple PV panels is obtained. This file (JSON format) is parsed to obtain the geographical coordinates (latitude and longitude) of each PV panel. Subsequently, based on the longitude range of the PV panel coordinates, its UTM projection zone is automatically determined (e.g., eastern China is typically located in the UTM 50N zone), and this parameter is used to convert all latitude and longitude coordinates to UTM Cartesian coordinates. The distribution of the PV panels is shown below. Figure 2 As shown.

[0029] The drone used in this invention is a DJI M3T, with a thermal imaging camera resolution of 640*512 and a focal length of 9.1 mm. The flight altitude is set to 35 meters. In addition, to reduce the number of images captured during the operation and maximize operational efficiency, the forward and lateral overlap rates are set to 10%. The drone's flight speed is set to 5 meters per second, and the takeoff point is manually specified by the user according to the site conditions.

[0030] Step 2: Based on the transformed planar coordinate information, the centroid of each photovoltaic panel polygon is used as its spatial location representative point, and a two-dimensional spatial grid index covering the entire factory area is established (grid size is 5m×5m) to accelerate proximity queries. Subsequently, photovoltaic panels with similar locations are automatically merged to form photovoltaic clusters using the principle of spatial proximity.

[0031] Specifically, clustering is performed using a breadth-first search (BFS) algorithm, which calculates the Euclidean distance between the centroids of all photovoltaic panels and groups those with a distance less than a preset threshold. (In this embodiment) Photovoltaic panels with a diameter of 6m are considered connected and merged into the same photovoltaic cluster. Ultimately, 4400 photovoltaic panels are clustered into two photovoltaic clusters. The threshold is calculated using the following formula: in, The length of the photovoltaic panel, is the clustering coefficient, with a value ranging from 1.2 to 1.5.

[0032] Step 3, calculate the average principal orientation angle for each photovoltaic cluster, including: For each photovoltaic panel (N in total), calculate the principal orientation angle based on its geometric shape or coordinate data. Find all of them separately. and The sum of these values ​​is then substituted into the atan2 function to obtain the average angle. ; In this step, since the photovoltaic panel is rectangular and the coordinates of the four points of the rectangle are known, the angle of the direction of the long side of the rectangle is used as the orientation angle. .

[0033] The calculation formula is as follows: In the formula, The average principal orientation angle of the photovoltaic cluster. It is the orientation angle of the long side of each photovoltaic panel. It represents the number of photovoltaic panels in the cluster.

[0034] Finally, in this embodiment of the invention, the average principal orientation angle of the photovoltaic cluster is calculated to be 84.7 degrees after summing the orientation angle vectors of each photovoltaic panel in each photovoltaic cluster and then inputting the result into the atan2 function. Step 4: Based on the average principal orientation angle of the photovoltaic cluster, generate an internal coverage sub-flight route for each photovoltaic cluster. First, the geometric contour of each photovoltaic cluster is buffered and expanded outwards, with a buffer distance of 5m (this value is not directly related to camera parameters; it primarily serves to establish a safe operating boundary. This empirical value aims to offset shooting deviations caused by UAV positioning errors and flight attitude swaying. By expanding the flight area, it ensures that the camera's field of view completely covers the edge of the photovoltaic panel, thus fundamentally avoiding missed shots in the boundary area). Then, the coordinates of the expanded boundary points are set as the origin, and all envelope points within the boundary are... Rotate around the origin The angle (i.e., rotating back against the photovoltaic average direction) is used to obtain the coordinates of the envelope point in the rotating coordinate system. The calculation formula is as follows: After rotation, the photovoltaic array is aligned with the x′ axis in the average principal direction in the rotating coordinate system; The axis-aligned bounding box range is then calculated in the rotating coordinate system. Within this range, several lines parallel to the y′ axis are generated along the x′ direction with step size, and candidate waypoints (x′′, y′′) are generated on the lines according to the point spacing.

[0035] And the generated candidate waypoints Rotate back to the original coordinate system using the following formula: Then, by using the inclusion relationship function between points and polygons, points located inside the outline of the photovoltaic cluster are selected as the final set of waypoints.

[0036] Step 5: Based on the enhanced nearest neighbor algorithm, determine the visiting order of all sub-routes and concatenate them into a single continuous overall route. The total length of the route generated by this method is approximately 2.9 km, with a total of 114 waypoints. Figure 3 As shown, it can cover all photovoltaic panels within the factory area, enabling efficient drone inspection. The results of importing the flight path into DJI's Sikong platform are as follows: Figure 4 As shown.

[0037] The specific process of the enhanced nearest neighbor algorithm is as follows: Take the drone's takeoff point as the current location For each unvisited sub-route Calculate the distance from the current position to its entry point. and exit point Euclidean distance: Choose the route that minimizes the distance as the next destination: In the formula, The optimal next sub-route is selected based on the enhanced nearest neighbor algorithm. Represents the selected optimal sub-route The endpoint that is determined to be the optimal connection point.

[0038] If the selected point is the exit point, the flight path is connected in the reverse direction; otherwise, it is connected in the forward direction, and the current position is updated to the corresponding other endpoint. Repeat the above steps until all sub-flight paths are marked as "visited." At this point, a continuous overall flight path has been generated. After the drone completes all filming, it returns to the starting point and ends the inspection mission.

[0039] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the invention.

Claims

1. A method for planning unmanned aerial vehicle (UAV) photovoltaic inspection routes based on spatial clustering and path optimization, characterized in that, Includes the following steps: Step 1: Obtain an electronic drawing file containing the geographical locations of multiple photovoltaic panels, parse the file, automatically determine the UTM projection zone where the photovoltaic panel is located based on the longitude range of the photovoltaic panel coordinates, and use the UTM projection method to convert the latitude and longitude coordinates of the photovoltaic panel into plane rectangular coordinates. Step 2: Based on the transformed planar coordinate information, the centroid of each photovoltaic panel polygon is used as its spatial location representative point. The principle of spatial proximity is used to automatically merge photovoltaic panels with similar locations to form a photovoltaic cluster. Step 3: Calculate the average principal orientation angle for each photovoltaic cluster, including calculating the principal orientation angle for each of the N photovoltaic panels in each cluster based on their geometry or coordinate data. ; Calculate the values ​​of all photovoltaic panels respectively and The sum of these values ​​is then substituted into the atan2 function to obtain the average angle of all photovoltaic panels. The calculation formula is as follows: ； in, It is the average principal orientation angle of the photovoltaic cluster. It is the orientation angle of the long side of each photovoltaic panel. It refers to the number of photovoltaic panels in the cluster; Step 4: Based on the average main direction angle of the photovoltaic cluster, generate an internal coverage sub-flight route for each photovoltaic cluster; Step 5: Based on the enhanced nearest neighbor algorithm, determine the access order of all sub-routes, splice the sub-routes into a single continuous total route, and control the UAV to return to the takeoff point after all sub-routes have been visited.

2. The method according to claim 1, characterized in that, The spatial clustering described in step 2 specifically includes: Establish a two-dimensional spatial grid index covering the entire photovoltaic field area; A breadth-first search algorithm is used to calculate the Euclidean distance between the centroids of all photovoltaic panels, and those distances less than a preset threshold are selected. The photovoltaic panels are considered connected and merged into the same photovoltaic cluster; The threshold The calculation formula is: ； in, The length of the photovoltaic panel, is the clustering coefficient, with a value ranging from 1.2 to 1.

5.

3. The method according to claim 1, characterized in that, The process of generating internal coverage sub-routes in step 4 includes: The geometric outline of the photovoltaic cluster is buffered and expanded outward, with the buffer distance set at 5m; Set the coordinates of the expanded boundary points as the origin, and define the envelope points within the boundary. Rotate around the origin Angle, to obtain the coordinates of the envelope point in the rotating coordinate system. The calculation formula is as follows: ； After rotation, the photovoltaic array is aligned with the x′ axis in the average principal direction in the rotating coordinate system; Calculate the axis-aligned bounding box range in the rotated coordinate system. Within this range, several lines parallel to the y′ axis are generated along the x′ direction with step sizes, and candidate waypoints are then generated on these lines based on the point spacing. ; Candidate waypoints Rotate back to the original coordinate system and filter out points located inside the outline of the photovoltaic cluster as the final waypoint set based on the inclusion relationship between the points and the polygon; Among them, candidate waypoints The formula for calculating the rotation back to the original coordinate system is: 。 4. The method according to claim 3, characterized in that, The grid row spacing and column spacing of the candidate waypoints are jointly determined by the focal length of the UAV camera, flight altitude, forward overlap rate and lateral overlap rate, and the forward overlap rate and lateral overlap rate are set to 10%, and the flight altitude is set to 35 meters.

5. The method according to claim 1, characterized in that, The enhanced nearest neighbor algorithm described in step 5 includes: Take the drone's takeoff point as the current location ; For each unvisited sub-route Calculate the distance from the current position to its entry point. and exit point The Euclidean distance is calculated using the following formula: Select the sub-path that minimizes the distance and its endpoint as the next destination; If the selected point is the exit point, then connect to the sub-route in the reverse direction; otherwise, connect in the forward direction and update the current position to the corresponding other endpoint. Repeat the above process until all sub-paths have been visited, and finally control the drone to return to the takeoff point.

6. The method according to claim 1, characterized in that, The electronic drawing file is in JSON format and contains the latitude and longitude coordinates of the polygon vertices of each photovoltaic panel.

7. The method according to claim 1, characterized in that, The flight path files output by the method are compatible with the DJI Sikong platform import format.

8. The method according to claim 1, characterized in that, The drone is equipped with a thermal imaging camera with a resolution of 640×512 and a focal length of 9.1 mm.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the UAV photovoltaic inspection route planning method as described in any one of claims 1 to 8.

Images