Automatic inspection methods, devices, storage media and electronic equipment for photovoltaic power plants

Abstract

This disclosure provides an automatic inspection method, device, storage medium, and electronic device for photovoltaic (PV) power plants. The method includes: acquiring a two-dimensional orthophoto image and a CAD design image of the PV power plant; associating the PV strings in the two-dimensional orthophoto image with the PV strings in the CAD design image to obtain a first association relationship; triggering a drone inspection when a defective string with abnormal power data is present; acquiring inspection images taken by the drone after the inspection; mapping the defective string onto the two-dimensional orthophoto image according to the first association relationship; mapping the defective string onto the inspection image according to a second association relationship between the two-dimensional orthophoto image and the inspection image, and marking and displaying the location of the defective string in the inspection image. This disclosure associates power data monitoring with drone inspection, improving inspection efficiency and accuracy, while mapping and marking / displaying defective strings onto the inspection image makes the display of defective strings more intuitive.

Description

Technical Field

[0001] This disclosure relates to the field of photovoltaic technology, and in particular to an automatic inspection method, device, storage medium and electronic equipment for photovoltaic power plants. Background Technology

[0002] Photovoltaic (PV) inspection involves controlling drones to fly over PV power plants along an inspection route. By collecting visible light images of the PV plants (i.e., inspection images), drones identify defects, pinpointing defective strings and generating a defect detection report. However, drone inspections are inefficient and can only detect and locate defects from a single image dimension, compromising accuracy. Furthermore, defective strings are only visible in two-dimensional orthophotos, making it difficult for maintenance personnel to visually verify and address defects on-site. Summary of the Invention

[0003] The purpose of this disclosure is to provide an automatic inspection method, device, storage medium, and electronic device for photovoltaic power plants, in order to solve the problems of low efficiency, low accuracy, and unchanged presentation mode in the prior art of photovoltaic inspection.

[0004] The embodiments of this disclosure adopt the following technical solution: an automatic inspection method for a photovoltaic power station, comprising: acquiring a two-dimensional orthophoto image and a CAD design image of the photovoltaic power station, and associating the photovoltaic strings in the two-dimensional orthophoto image and the photovoltaic strings in the CAD design image to obtain a first association relationship; monitoring the power data of the photovoltaic strings in the photovoltaic power station, and triggering a drone inspection when there is a defective string with abnormal power data; acquiring an inspection image taken by the drone after the inspection; mapping the position of the defective string to the two-dimensional orthophoto image according to the first association relationship; mapping the defective string to the inspection image according to a second association relationship between the two-dimensional orthophoto image and the inspection image, and marking and displaying the position of the defective string in the inspection image.

[0005] This disclosure also provides an automatic inspection device for a photovoltaic power station, comprising: a first image acquisition module, configured to acquire a two-dimensional orthophoto image and a CAD design image of the photovoltaic power station, and associate the photovoltaic strings in the two-dimensional orthophoto image with the photovoltaic strings in the CAD design image to obtain a first association relationship; an inspection module, configured to monitor the power data of the photovoltaic strings in the photovoltaic power station, and trigger a drone inspection when there is a defective string with abnormal power data; a second image acquisition module, configured to acquire an inspection image taken by the drone after inspection; a first mapping module, configured to map the position of the defective string to the two-dimensional orthophoto image according to the first association relationship; and a second mapping module, configured to map the defective string to the inspection image according to a second association relationship between the two-dimensional orthophoto image and the inspection image, and mark and display the position of the defective string in the inspection image.

[0006] This disclosure also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described automatic inspection method for photovoltaic power stations.

[0007] This disclosure also provides an electronic device, including at least a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program in the memory to implement the steps of the above-described automatic inspection method for photovoltaic power stations.

[0008] The beneficial effects of this disclosure are as follows: By linking power data-based monitoring with UAV-based inspection, real-time monitoring of defective strings within the photovoltaic power station can be achieved, triggering UAV inspections and improving the efficiency and accuracy of photovoltaic power station inspections. Simultaneously, by combining the correlation between photovoltaic strings in CAD design images and two-dimensional orthophoto images, defective strings are mapped to two-dimensional orthophoto images. Then, based on the correlation between the two-dimensional orthophoto images and UAV inspection images, defective strings are mapped to inspection images for marking and display, making the display of defective strings more intuitive and easier for maintenance personnel to view. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of the automatic inspection method for photovoltaic power stations in the first embodiment of this disclosure;

[0011] Figure 2 This is a schematic diagram of the structure of the automatic inspection device for a photovoltaic power station in the second embodiment of this disclosure. Detailed Implementation

[0012] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0013] Photovoltaic (PV) inspection involves controlling drones to fly over PV power plants along an inspection route. By collecting visible light images of the PV plants (i.e., inspection images), drones identify defects, pinpointing defective strings and generating a defect detection report. However, drone inspections are inefficient and can only detect and locate defects from a single image dimension, compromising accuracy. Furthermore, defective strings are only visible in two-dimensional orthophotos, making it difficult for maintenance personnel to visually verify and address defects on-site.

[0014] To address the aforementioned problems, the first embodiment of this disclosure provides an automatic inspection method for photovoltaic power stations, the flowchart of which is shown below. Figure 1 As shown, it mainly includes steps S10 to S50:

[0015] S10: Obtain the two-dimensional orthophoto image and CAD design image of the photovoltaic power station, and associate the photovoltaic strings in the two-dimensional orthophoto image and the photovoltaic strings in the CAD design image to obtain the first association relationship.

[0016] A 2D orthophoto (DOM) of a photovoltaic (PV) power plant is an orthophoto obtained by photographing the entire PV power plant area using a drone equipped with a camera, and then processing the images using intelligent mapping software. During the 2D orthophoto construction process, the software can also collect 3D point clouds of the PV power plant to obtain a digital display of the plant. After acquiring the DOM, it can be divided into subarrays based on the actual arrangement of the PV arrays within the power plant, resulting in subarray DOMs. Subsequent automated inspections will then use these subarrays as the inspection scope.

[0017] Subsequently, the photovoltaic (PV) strings in the DOM are identified. Specifically, any image segmentation model can be used to segment the PV strings from the DOM, and after segmentation, the first pixel coordinates of each PV string in the DOM are obtained based on the pixel coordinate system in the DOM image. Specifically, the image segmentation model used in this embodiment can be the deep learning semantic segmentation model DeeplapV3+ as the algorithm model to accurately extract the strings in the DOM. After selecting the model, a training set must first be constructed. Here, DOM images with extracted strings are selected as the training set. Since the size of a single DOM is relatively large, in this embodiment, the DOM needs to be cropped into small images for training. Here, a 512*512 pixel sliding window is used, and the movement step is set to 400. After obtaining the cropped sub-images, data augmentation is needed to better train the model. First, the Canny operator is used to extract the edges of the strings in the image and save them as single-channel boundary images. These images are then stitched together with the original image to create a 4-channel sub-image. This process increases the edge features of the strings in the image, making it easier to extract complete strings and avoiding overlapping segmentation results. Second, the sub-images augmented by the Canny operator are further augmented using data augmentation methods such as random inversion, brightness adjustment, color gamut changes, and noise addition. This results in the construction of a string segmentation dataset for better string extraction.

[0018] Once the algorithm model is trained, it can be used to identify and extract strings from the domain image (DOM) of a photovoltaic power station. For larger DOM images, the same cropping method as the training set is used: a 512*512 pixel sliding window with a step size of 400 pixels is used to crop the DOM into several smaller images. These smaller images are then input into the algorithm model, which outputs the corresponding string recognition results. Specifically, strings in the smaller images are identified as 1, and the background is identified as 0. Using the coordinates of the top-left corner of each smaller image as a reference, it is then restored to the DOM image. After overlaying, the background images still result in 0, but any string element will always result in 1 regardless of overlay. This process yields the string recognition and extraction for the entire DOM image. The first pixel coordinate of each string is used as the bounding box coordinates corresponding to its edge in the DOM pixel coordinate system.

[0019] Furthermore, after obtaining the extracted string results, the strings can be sorted according to pixel coordinates, and encoded based on the sorting results in an ascending order from left to right and top to bottom. The encoding rule is: subarray-row-column, to obtain the positional encoding of the string. Since the DOM contains built-in geographic location information, this positional encoding has corresponding first pixel coordinates and first world coordinates.

[0020] The CAD design image is created during the design of the photovoltaic power station. It records the design information of the station, the design code of each string, and the world coordinates (i.e., second world coordinates) of that string during the design process. After acquiring the CAD design image, it can be segmented according to subarrays, and the strings in the CAD design image can be extracted and saved according to the subarrays. Subsequently, the second world coordinates of the photovoltaic strings in the CAD design image are mapped to the DOM image using the DOM's world coordinate system, and the second pixel coordinates corresponding to the second world coordinates of the photovoltaic strings in the DOM image are determined. However, due to differences in acquisition tools and position caused by actual construction, the first pixel coordinates directly determined in the DOM and the second pixel coordinates obtained by mapping through the CAD design image for the same photovoltaic string usually do not coincide. To achieve matching of the same photovoltaic strings, this embodiment adjusts the overall position of the second pixel coordinates of all photovoltaic strings to achieve a one-to-one matching with the strings in the DOM, thereby completing the association between the position code and the design code of the photovoltaic string. It should be noted that in actual matching, it can be performed according to subarrays, which helps to improve matching efficiency and accuracy. Specifically, the above adjustment operation can be performed using the Hungarian algorithm, or by other adjustment algorithms; this embodiment does not impose any limitations on this.

[0021] S20 monitors the power data of photovoltaic strings in the photovoltaic power station and triggers drone inspection when there are defective strings with abnormal power data.

[0022] When photovoltaic (PV) strings are operating normally, their corresponding power data, such as current and voltage values, are within the normal range. However, if a PV string malfunctions, such as being shaded by pollutants or experiencing a photovoltaic panel failure, the power data of that string will become abnormal. Therefore, this embodiment monitors the power data of the PV strings in the PV power plant in real time. When a defective string with abnormal power data is detected, a drone can be triggered for on-site inspection to help maintenance personnel determine whether the corresponding photovoltaic panel is shaded or malfunctioning.

[0023] Specifically, during real-time monitoring of power data, when the power data of a certain photovoltaic string is found to be outside the normal range, the string is identified as defective. The design code of the defective string is then obtained based on the design code recorded in the CAD design image. Subsequently, the corresponding location code and its first-world coordinates in the domain (DOM) are determined based on the design code of the defective string; this represents the actual geographical coordinates of the defective string. After determining the first-world coordinates of all defective strings, these coordinates can be used as inspection points to plan the drone inspection route. The drone is then triggered to perform inspection operations according to the route, allowing it to traverse and inspect the locations of all defective strings in a single inspection. Simultaneously, the drone is instructed to capture inspection images at each inspection point to obtain visible light images of the defective strings.

[0024] In some embodiments, when planning the inspection route of a UAV, the first world coordinates of all defect clusters can be integrated to form the minimum number of cruise points, thereby achieving the optimal planned path. Specifically, firstly, the first world coordinates of the center point of all defect clusters are used as inspection points to establish an inspection point set; then, any inspection point is selected from the inspection point set as a recording point, and the coverage area of ​​the cruise image captured by the UAV when it inspects the recording point is calculated using the camera imaging principle. It should be noted that due to the existence of positioning errors, the actual coverage area can be calculated as 90% of the theoretical coverage area to avoid an excessively large coverage area leading to increased errors in subsequent waypoint calculations; then, it is checked whether other inspection points are included in the coverage area. If no other inspection points are included in the coverage area, the current recording point is added to the waypoint set as a waypoint. If the coverage area contains other checkpoints, merge the other checkpoints with the current record point, and add the current record point as a waypoint to the waypoint set. At the same time, delete the other checkpoints already included in the coverage area from the checkpoint set. Repeat the above steps to continuously select a checkpoint from the checkpoint set as a new record point, and determine the waypoint according to its corresponding coverage area and integrate it with other checkpoints, until all checkpoints in the checkpoint set have been traversed and the checkpoint set is empty. At this point, some of the checkpoints are added to the waypoint set as waypoints, and the other part, which is located in the coverage area of ​​a certain waypoint, is directly deleted to reduce the number of waypoints and facilitate subsequent route planning.

[0025] Once the waypoint set is determined, the UAV inspection route can be planned based on the first-world coordinates of all waypoints in the set. In this embodiment, the UAV inspection route can be planned using any of the following algorithms: greedy algorithm, ant colony algorithm, backtracking algorithm, divide-and-conquer algorithm, etc. This embodiment does not limit the specific algorithm used for route planning. In actual operation, a suitable algorithm can be selected based on the number of waypoints, inspection efficiency, time requirements, etc. It should also be noted that if the number of defective strings exceeds a preset percentage of the total number of photovoltaic strings in the current photovoltaic power station, such as 60%, it indicates that most of the strings in the current photovoltaic power station are defective. At this time, the UAV can be directly started to inspect the entire power station to determine whether there is a large-scale string failure or photovoltaic panel shading caused by sand and dust.

[0026] S30 acquires inspection images taken by the drone after the inspection.

[0027] When the drone inspects a faulty cluster during the inspection process, it will take pictures of it. The pictures will reflect the actual situation of the site area containing the faulty cluster.

[0028] S40, based on the first association relationship, map the position of the defect cluster onto the two-dimensional orthophoto image.

[0029] The first association in this embodiment is the association between the two-dimensional orthophoto image of the photovoltaic power station and the CAD design image. Specifically, it's the association between the design code and location code of the photovoltaic string obtained in step S10, and the relationship between the first pixel coordinates and first world coordinates in the two-dimensional orthophoto image. After the UAV inspection, the first world coordinates and first pixel coordinates of the defective string in the DOM can be determined based on the first association according to the design code of the defective string. This maps the location of the defective string to the DOM, allowing for marking and display, enabling maintenance personnel to intuitively understand the location of the defective string through the DOM image.

[0030] S50, based on the second correlation between the two-dimensional orthophoto image and the inspection image, maps the defect string to the inspection image, and marks and displays the position of the defect string in the inspection image.

[0031] In this embodiment, in addition to displaying the defective clusters in the domain, the defective clusters can also be marked and displayed in the inspection images. This allows maintenance personnel to quickly and directly locate the faulty clusters and understand their current status based on the actual inspection images. Specifically, the UAV's flight altitude Z during the inspection process is usually fixed, and the intrinsic parameters K and extrinsic parameters T of its camera are also known parameters. Therefore, the mapping relationship between the pixel coordinate system of the inspection image actually captured by the UAV and the world coordinate system of the actual geographical location corresponding to the inspection image can be determined. The world coordinate system of the actual geographical location corresponding to the inspection image is the world coordinate system of the two-dimensional orthophoto image. Based on the camera imaging principle, the mapping relationship can be obtained as follows:

[0032]

[0033] Among them, P uv Z represents the pixel coordinates of the inspection image, Z represents the drone's flight altitude in real-world coordinates, and P represents the pixel coordinates of the inspection image. w Let P be the world coordinate system for a two-dimensional orthophoto image, K be the camera intrinsic parameter, and T be the camera extrinsic parameter. Therefore, given P... w Given K, T, and Z, P can be obtained. uv By combining the coordinates of the defect string with its first-world coordinates in the DOM, the drone's flight altitude Z, and the camera's intrinsic and extrinsic parameters K and T, the third pixel coordinates of the defect string in the inspection image can be obtained. At this point, the defect string can be marked and displayed in the inspection image based on its third pixel coordinates.

[0034] In some embodiments, photovoltaic (PV) inspection may further include a timed inspection operation that triggers a drone to inspect all PV strings in a PV power station according to a preset cycle. Defective strings are extracted through image recognition of the drone's inspection images. Then, based on the mapping relationship between the inspection images and the DOM (Distributed Domain), the corresponding defective strings are marked and displayed in the DOM. Simultaneously, when the drone's periodic inspection is triggered, power data monitoring can also be triggered. After merging and unifying the results from the two monitoring methods, all defective strings can be marked and displayed simultaneously on the DOM and the inspection images, thereby achieving a multi-dimensional data-driven PV power station inspection effect and improving inspection efficiency and identification accuracy.

[0035] This embodiment links power data monitoring with drone inspection to achieve real-time monitoring of defective strings within photovoltaic power plants and trigger drone inspections, thereby improving the efficiency and accuracy of photovoltaic power plant inspections. Simultaneously, by combining the correlation between photovoltaic strings in CAD design images and 2D orthophotos, defective strings are mapped onto 2D orthophotos. Then, based on the correlation between the 2D orthophotos and drone inspection images, defective strings are mapped onto the inspection images for marking and display, making the display of defective strings more intuitive and easier for maintenance personnel to view.

[0036] Based on the same inventive concept, the second embodiment of this disclosure provides an automatic inspection device for photovoltaic power stations, the structural schematic diagram of which is shown below. Figure 2 As shown, the system mainly includes the following sequentially coupled modules: a first image acquisition module 10, used to acquire a two-dimensional orthophoto image and a CAD design image of the photovoltaic power station, and associate the photovoltaic strings in the two-dimensional orthophoto image and the photovoltaic strings in the CAD design image to obtain a first association relationship; an inspection module 20, used to monitor the power data of the photovoltaic strings in the photovoltaic power station, and trigger a drone inspection when there are defective strings with abnormal power data; a second image acquisition module 30, used to acquire inspection images taken by the drone after inspection; a first mapping module 40, used to map the position of the defective string to the two-dimensional orthophoto image according to the first association relationship; and a second mapping module 50, used to map the defective string to the inspection image according to the second association relationship between the two-dimensional orthophoto image and the inspection image, and mark and display the position of the defective string in the inspection image.

[0037] In some embodiments, the first image acquisition module 10 is specifically used to take pictures of the photovoltaic power station using a drone, and construct a two-dimensional orthophoto of the photovoltaic power station based on the shooting results; segment the two-dimensional orthophoto image using an image segmentation model to obtain the first pixel coordinates of the photovoltaic strings in the two-dimensional orthophoto image; assign a position code to the photovoltaic strings based on the first pixel coordinates of the photovoltaic strings, and associate the position code, the first pixel coordinates, and the first world coordinates of the photovoltaic strings; acquire the CAD design image of the photovoltaic power station, and determine the second world coordinates and design code of each photovoltaic string in the CAD design image; determine the second pixel coordinates corresponding to the second world coordinates of the photovoltaic strings in the two-dimensional orthophoto image; adjust the overall position of the second pixel coordinates so that the second pixel coordinates of all photovoltaic strings match the unique first pixel coordinates, thereby completing the association between the position code and the design code of the photovoltaic strings.

[0038] In some embodiments, the inspection module 20 is specifically used to determine that the photovoltaic string is a defective string when the power data of the photovoltaic string is not within the normal range, and to obtain the design code of the defective string; to determine the corresponding location code and first world coordinates according to the design code of the defective string; to plan the inspection route of the UAV using the first world coordinates of all defective strings as inspection points; to trigger the inspection operation of the UAV according to the inspection route, and to instruct the UAV to take inspection images when it reaches the inspection point.

[0039] In some embodiments, the inspection module 20 is further specifically configured to: establish an inspection point set using the first world coordinates of the center point of all defect clusters as inspection points; calculate the coverage area of ​​the cruise image captured by the UAV when it inspects the recording point, using any inspection point in the inspection point set as a recording point; detect whether other inspection points are included in the coverage area; if no other inspection points are included in the coverage area, add the recording point as a waypoint to the waypoint set; if other inspection points are included in the coverage area, add the recording point as a waypoint to the waypoint set and delete other inspection points in the coverage area from the inspection point set; traverse all inspection points in the inspection point set until the inspection point set is empty; and plan the inspection route of the UAV based on the first world coordinates of all waypoints in the waypoint set.

[0040] In some embodiments, the inspection module 20 plans the inspection route of the UAV using any of the following algorithms: greedy algorithm, ant colony algorithm, backtracking algorithm, and divide-and-conquer algorithm.

[0041] In some embodiments, the second mapping module 50 is specifically used to establish a mapping relationship between the pixel coordinate system of the inspection image and the world coordinate system of the two-dimensional orthophoto image based on the flight altitude of the UAV when capturing the inspection image, the camera intrinsic parameters and camera extrinsic parameters of the UAV; determine the corresponding first world coordinates based on the design code of the defect string; determine the third pixel coordinates of the defect string in the inspection image based on the first world coordinates of the defect string, the flight altitude, the camera intrinsic parameters and camera extrinsic parameters of the UAV, and the mapping relationship; and mark and display the position of the defect string in the inspection image based on the third pixel coordinates.

[0042] In some embodiments, the inspection module 20 is further configured to trigger the UAV to inspect all photovoltaic strings in the photovoltaic power station according to a preset cycle; the second image acquisition module 30 is further configured to perform image recognition on the inspection images of the UAV to identify defective strings; and the second mapping module 50 is further configured to map the defective strings onto a two-dimensional orthophoto image for marking and display.

[0043] This embodiment links power data monitoring with drone inspection to achieve real-time monitoring of defective strings within photovoltaic power plants and trigger drone inspections, thereby improving the efficiency and accuracy of photovoltaic power plant inspections. Simultaneously, by combining the correlation between photovoltaic strings in CAD design images and 2D orthophotos, defective strings are mapped onto 2D orthophotos. Then, based on the correlation between the 2D orthophotos and drone inspection images, defective strings are mapped onto the inspection images for marking and display, making the display of defective strings more intuitive and easier for maintenance personnel to view.

[0044] Based on the same inventive concept, the third embodiment of this disclosure provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the automatic inspection method for photovoltaic power stations according to the first embodiment of this disclosure.

[0045] Based on the same inventive concept, the fourth embodiment of this disclosure provides an electronic device, including at least a memory and a processor. The memory stores a computer program, and the processor executes the computer program in the memory to implement the steps of the automatic inspection method for photovoltaic power stations according to the first embodiment of this disclosure.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A method for automatically inspecting a photovoltaic plant, characterized in that, The method comprises: acquiring a two-dimensional orthographic image and a CAD design image of a photovoltaic power station, and associating photovoltaic strings in the two-dimensional orthographic image with photovoltaic strings in the CAD design image to obtain a first association relationship; monitoring power data of the photovoltaic strings in the photovoltaic power station, and triggering a UAV inspection when there is a defective string with abnormal power data; acquiring an inspection image taken after the UAV inspection; mapping a position of the defective string to the two-dimensional orthographic image according to the first association relationship; mapping the defective string to the inspection image according to a second association relationship between the two-dimensional orthographic image and the inspection image, and marking and displaying the position of the defective string in the inspection image; the acquiring a two-dimensional orthographic image and a CAD design image of a photovoltaic power station, and associating photovoltaic strings in the two-dimensional orthographic image with photovoltaic strings in the CAD design image to obtain a first association relationship comprises: taking a photograph of the photovoltaic power station by a UAV, and constructing a two-dimensional orthographic image of the photovoltaic power station according to the photographing result; segmenting the two-dimensional orthographic image by an image segmentation model to obtain first pixel coordinates of the photovoltaic strings in the two-dimensional orthographic image; assigning a position code to the photovoltaic strings according to the first pixel coordinates of the photovoltaic strings, and associating the position code, the first pixel coordinates and first world coordinates of the photovoltaic strings; acquiring a CAD design image of the photovoltaic power station, and determining second world coordinates and a design code of each photovoltaic string in the CAD design image; determining second pixel coordinates corresponding to the second world coordinates of the photovoltaic strings in the two-dimensional orthographic image; adjusting the overall position of the second pixel coordinates so that the second pixel coordinates of all photovoltaic strings match the unique first pixel coordinates, to complete the association between the position code and the design code of the photovoltaic strings.

2. The method of claim 1, wherein, the monitoring power data of the photovoltaic strings in the photovoltaic power station, and triggering a UAV inspection when there is a defective string with abnormal power data comprises: determining the photovoltaic string as the defective string when the power data of the photovoltaic string is not in a normal range, and acquiring a design code of the defective string; determining corresponding position codes and first world coordinates according to the design code of the defective string; taking the first world coordinates of all the defective strings as inspection points, and planning an inspection route of the UAV; triggering an inspection operation of the UAV according to the inspection route, and instructing the UAV to take a photograph of an inspection image when the UAV inspects the inspection points.

3. The method of claim 2, wherein, the taking the first world coordinates of all the defective strings as inspection points, and planning an inspection route of the UAV comprises: taking the first world coordinates of the center points of all the defective strings as inspection points to establish an inspection point set; taking any one of the inspection points in the inspection point set as a recording point, calculating a coverage area of a cruising image taken when the UAV inspects the recording point, and detecting whether the coverage area contains other inspection points; ​ In the case that the coverage area does not contain other inspection points, the record point is placed into the waypoint set as a waypoint; In the case that the coverage area contains other inspection points, the record point is placed into the waypoint set as a waypoint, and other inspection points in the coverage area are deleted from the inspection point set; All inspection points in the inspection point set are traversed until the inspection point set is empty; The inspection route of the UAV is planned according to the first world coordinates of all waypoints in the waypoint set.

4. The method of claim 3, wherein, The inspection route of the UAV is planned according to the first world coordinates of all waypoints in the waypoint set, including: the inspection route of the UAV is planned by any one of the following algorithms: greedy algorithm, ant colony algorithm, backtracking algorithm, divide-and-conquer algorithm.

5. The method of claim 1, wherein, The second correlation between the two-dimensional orthographic image and the inspection image is used to map the defect group string into the inspection image, and the position of the defect group string in the inspection image is marked and displayed, including: A mapping relationship between the pixel coordinate system of the inspection image and the world coordinate system of the two-dimensional orthographic image is established according to the flight height of the UAV when the inspection image is taken, the camera intrinsic parameters and the camera extrinsic parameters of the UAV; The corresponding first world coordinates are determined according to the design code of the defect group string; The third pixel coordinates of the defect group string in the inspection image are determined according to the first world coordinates of the defect group string, the flight height, the camera intrinsic parameters and the camera extrinsic parameters of the UAV, and the mapping relationship; The position of the defect group string in the inspection image is marked and displayed according to the third pixel coordinates.

6. The method of claim 1 to 5, wherein, Further comprising: Triggering the UAV to inspect all photovoltaic strings in the photovoltaic power station according to a preset period; Performing image recognition on the inspection image of the UAV to identify a defective string; Mapping the defective string into the two-dimensional orthographic image for marking and display.

7. An automatic inspection device for a photovoltaic plant, characterized in that it comprises: Comprise: A first image acquisition module is configured to acquire a two-dimensional orthographic image and a CAD design image of a photovoltaic power station, and correlate photovoltaic strings in the two-dimensional orthographic image with photovoltaic strings in the CAD design image to obtain a first correlation relationship. The first image acquisition module is specifically configured to: take a photograph of the photovoltaic power station by a UAV, and construct a two-dimensional orthographic image of the photovoltaic power station according to the photographing result; segment the two-dimensional orthographic image by an image segmentation model to obtain first pixel coordinates of photovoltaic strings in the two-dimensional orthographic image; assign a position code to each photovoltaic string according to the first pixel coordinates of the photovoltaic string, and correlate the position code, the first pixel coordinates, and the first world coordinates of the photovoltaic string; acquire a CAD design image of the photovoltaic power station, and determine second world coordinates and design codes of each photovoltaic string in the CAD design image; Determine second pixel coordinates corresponding to the second world coordinates of the photovoltaic strings in the two-dimensional orthographic image; adjust the overall position of the second pixel coordinates so that the second pixel coordinates of all photovoltaic strings match the unique first pixel coordinates, to complete the correlation between the position codes and the design codes of the photovoltaic strings. The inspection module is configured to monitor power data of photovoltaic strings in the photovoltaic power station, and trigger the UAV to perform inspection when there is a defective string with abnormal power data; The second image acquisition module is configured to acquire an inspection image captured after the UAV performs inspection; The first mapping module is configured to map a position of the defective string to the two-dimensional orthographic image according to the first correlation relationship; The second mapping module is configured to map the defective string to the inspection image according to a second correlation relationship between the two-dimensional orthographic image and the inspection image, and mark and display the position of the defective string in the inspection image.

8. A storage medium storing a computer program, characterized by The computer program, when executed by a processor, implements the steps of the automatic inspection method of the photovoltaic power station according to any one of claims 1 to 6.

9. An electronic device comprising at least a memory and a processor, said memory having stored thereon a computer program, characterized in that, The processor, when executing the computer program on the memory, implements the steps of the automatic inspection method of the photovoltaic power station according to any one of claims 1 to 6.

Images