An intelligent inspection system for photovoltaic panels based on drones

Abstract

The present application discloses a photovoltaic panel site intelligent inspection system based on drones, including: an inspection drone, an image acquisition module, a central control module and a personal terminal; the inspection drone is used to conduct inspections based on an inspection plan; the image acquisition module is integrated on the inspection drone, and the image acquisition module is used to collect photovoltaic panel images; the central control module is used to set the inspection plan and generate a photovoltaic panel inspection report based on the photovoltaic panel image; the personal terminal is used to visualize the photovoltaic panel inspection report and feed it back to the staff. Traditional manual inspections require a lot of manpower and time, while drone detection can quickly cover a large area of ​​photovoltaic power stations, and can complete a comprehensive inspection of photovoltaic power stations in a short time, reducing the workload of manual inspections and greatly improving detection efficiency.

Description

Technical Field

[0001] The present application belongs to the field of drone inspection technology, and specifically relates to a photovoltaic panel site intelligent inspection system based on drones. Background Art

[0002] In recent decades, great efforts in the pursuit of low-carbon energy supply and technological progress have driven the vigorous development of various renewable energy sources, such as solar energy, wind energy, ocean energy, hydropower and biomass energy, etc. At present, solar photovoltaic energy plays an increasingly important role in the global transition to clean and low-carbon energy supply.

[0003] At present, large-scale photovoltaic power station systems are usually located in plains and hilly areas without obvious shading areas. These areas are usually relatively remote, which brings challenges to photovoltaic inspection and monitoring. Traditional photovoltaic module inspections are evaluated through manual inspections, but this inspection method is inefficient and has a high error rate. It is not feasible to use manual visual evaluation in large-scale photovoltaic systems in practice. In general, there are currently the following difficulties in photovoltaic power station inspections: 1. There are few professional and technical inspection personnel; 2. The area is large and remote, and it is difficult to inspect the site; 3. The efficiency of manual inspection of hot plate inspections is relatively low; 4. The inspection covers a large area. Summary of the invention

[0004] This application aims to address the deficiencies of the prior art and proposes an intelligent inspection system for photovoltaic panel sites based on drones. The drones are equipped with high-precision camera equipment. Through the drone's infrared thermal imager, the temperature distribution of photovoltaic modules can be detected and abnormal conditions can be discovered in a timely manner. At the same time, the drone can also take high-definition photos of the surface of photovoltaic modules to capture subtle cracks, dirt and other problems.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] A photovoltaic panel site intelligent inspection system based on drones, comprising: an inspection drone, an image acquisition module, a central control module and a personal terminal;

[0007] The inspection drone is used to perform inspections based on an inspection plan;

[0008] The image acquisition module is integrated on the inspection drone, and the image acquisition module is used to acquire images of photovoltaic panels;

[0009] The central control module is used to set the inspection plan and generate a photovoltaic panel inspection report based on the photovoltaic panel image;

[0010] The personal terminal is used to visualize the photovoltaic panel inspection report and feed it back to the staff.

[0011] Preferably, the image acquisition module includes: an infrared image acquisition unit and a visible light image acquisition unit;

[0012] The infrared image acquisition unit is used to acquire infrared images of the surface of the photovoltaic panel;

[0013] The visible light image acquisition unit is used to acquire the visible light image of the surface of the photovoltaic panel.

[0014] Preferably, the central control module includes: a weather monitoring unit, a route planning unit, a plan generation unit, an image recognition unit and a report generation unit;

[0015] The meteorological monitoring unit is used to monitor the on-site weather and obtain weather information;

[0016] The route planning unit is used to plan the inspection route;

[0017] The plan generating unit is used to generate the inspection plan based on the weather information and the inspection route;

[0018] The image recognition unit is used to recognize hot spots, stains and cracks in the photovoltaic panel image to obtain an image recognition result;

[0019] The report generating unit is used to generate the photovoltaic panel inspection report based on the image recognition result.

[0020] Preferably, the image recognition unit comprises: a hot spot recognition subunit, a stain recognition subunit and a crack recognition subunit;

[0021] The hot spot recognition subunit is used to receive the infrared image, and perform hot spot recognition on the infrared image using a hot spot recognition model to obtain a hot spot recognition result;

[0022] The stain recognition subunit is used to receive the visible light image, and recognize the type of stain in the visible light image to obtain a stain recognition result;

[0023] The crack identification unit is used to receive the visible light image and perform crack identification on the visible light image using a crack detection model to obtain a crack identification result.

[0024] Preferably, the workflow of the hot spot identification subunit includes:

[0025] Preprocessing the infrared image to obtain a preprocessed infrared image;

[0026] Segmenting the hot spot area in the preprocessed infrared image, and extracting the hot spot feature vector of the hot spot area;

[0027] The hot spot feature vector is input into the trained hot spot recognition model to perform hot spot recognition to obtain the hot spot recognition result.

[0028] Preferably, the workflow of the stain identification unit includes:

[0029] Extracting a color region of a preset color range in the visible light image, and judging whether there is stain based on the color region;

[0030] If stains exist, the stain area image is extracted, and the stain area image is input into a trained stain recognition model to perform stain type recognition to obtain the stain recognition result.

[0031] Preferably, the workflow of the crack identification unit includes:

[0032] Preprocessing the visible light image to obtain a preprocessed visible light image;

[0033] Segmenting the crack region in the preprocessed visible light image, and extracting crack feature information of the crack region;

[0034] Crack identification is performed based on the crack feature information to obtain the crack identification result.

[0035] Preferably, the personal terminal comprises: a visualization unit and an alarm unit;

[0036] The visualization unit is used to visualize the photovoltaic panel inspection report to obtain a distribution map of the photovoltaic panel inspection situation;

[0037] The warning unit is used to issue a warning to the staff based on the photovoltaic panel inspection distribution map.

[0038] Compared with the prior art, the beneficial effects of this application are:

[0039] 1. Traditional manual inspections require a lot of manpower and time, while drone inspections can quickly cover a large area of ​​photovoltaic power stations and complete a comprehensive inspection of photovoltaic power stations in a short time, reducing the workload of manual inspections and greatly improving inspection efficiency;

[0040] 2. Photovoltaic power stations are usually located in remote areas, such as mountainous areas or deserts. The environment is complex and dangerous. Manual inspections have certain safety risks. Drone inspections can avoid direct contact with personnel, reduce the risk factor, and improve the safety of inspections.

[0041] 3. The drone is equipped with high-precision camera equipment. Through the drone's infrared thermal imager, the temperature distribution of photovoltaic modules can be detected and abnormal conditions can be discovered in time. At the same time, the drone can also take high-definition photos of the surface of photovoltaic modules to capture subtle cracks, dirt and other problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 Schematic diagram of the system structure of an embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0046] In this embodiment, if Figure 1 As shown, a photovoltaic panel site intelligent inspection system based on drones includes: an inspection drone, an image acquisition module, a central control module and a personal terminal.

[0047] The inspection drone is used to perform inspections based on the inspection plan. In this embodiment, the inspection drone can be a DJI drone.

[0048] The image acquisition module is integrated on the inspection drone, and the image acquisition module is used to acquire images of the photovoltaic panels. The image acquisition module includes: an infrared image acquisition unit and a visible light image acquisition unit. The infrared image acquisition unit is used to acquire infrared images of the surface of the photovoltaic panels. In this embodiment, the infrared image acquisition unit can use an infrared camera; the visible light image acquisition unit is used to acquire visible light images of the surface of the photovoltaic panels. In this embodiment, the visible light image acquisition unit can use a high-definition wireless camera.

[0049] The central control module is used to set the inspection plan and generate a photovoltaic panel inspection report based on the photovoltaic panel image. The central control module includes: a meteorological monitoring unit, a route planning unit, a plan generation unit, an image recognition unit, and a report generation unit. The meteorological monitoring unit is used to monitor the on-site weather and obtain weather information; the route planning unit is used to plan the inspection route; the plan generation unit is used to generate an inspection plan based on weather information and the inspection route; the image recognition unit is used to identify hot spots, stains, and cracks in the photovoltaic panel image and obtain image recognition results; the report generation unit is used to generate a photovoltaic panel inspection report based on the image recognition results.

[0050] The image recognition unit includes: a hot spot recognition subunit, a stain recognition subunit and a crack recognition subunit.

[0051] The hot spot recognition subunit is used to receive infrared images and use the hot spot recognition model to perform hot spot recognition on the infrared images to obtain hot spot recognition results. The workflow of the hot spot recognition subunit includes: preprocessing the infrared images to obtain preprocessed infrared images; segmenting the hot spot areas in the preprocessed infrared images and extracting the hot spot feature vectors of the hot spot areas; inputting the hot spot feature vectors into the trained hot spot recognition model to perform hot spot recognition and obtain hot spot recognition results. In this embodiment, the infrared image is first enhanced and filtered to obtain a preprocessed infrared image; the preprocessed infrared image is then grayed, and the grayed image is edge detected to segment the photovoltaic panel area; the hot spot area in the photovoltaic panel area is then segmented using the fuzzy C-means algorithm based on the least squares method to obtain the hot spot area, and the feature vector of the hot spot area is extracted. The specific feature vector includes: the ratio of the hot spot area to the photovoltaic panel area, the number of cells on the photovoltaic panel that produce hot spot failures, and the difference in temperature between the hot spot area and the normal area. The feature vector is normalized, that is, the feature vector parameters are mapped to the interval [0,1] to obtain the hot spot feature vector; the hot spot feature vector is input into the trained hot spot recognition model to perform hot spot recognition to obtain the recognition result of the hot spot fault type. The training method of the hot spot recognition model is as follows: first, infrared images of photovoltaic panels in normal working state and several hot spot fault states are collected, and these images are enhanced, infrared image filtered and grayed. The processed images are segmented to obtain training hot spot area images, and feature vectors are extracted from the hot spot area images to obtain training hot spot feature vectors. The training hot spot images and their corresponding feature vectors are divided into training set and verification set in a ratio of 7:3, respectively. The training set is input into the convolutional neural network for model training, and the verification set is input into the trained model for accuracy verification to obtain the final hot spot recognition model.

[0052] The stain recognition subunit is used to receive a visible light image, and identify the type of stains in the visible light image to obtain a stain recognition result. The working process of the stain recognition unit includes: extracting a color area of ​​a preset color range in the visible light image, and judging whether there is a stain based on the color area; if there is a stain, extracting the image of the stain area, and inputting the image of the stain area into a trained stain recognition model to identify the type of stain, and obtain a stain recognition result. In this embodiment, the color ranges of several normal photovoltaic panels are first extracted, and these color ranges are integrated to obtain a preset color range. If there is a non-preset color area in the color area, it is determined that there is a stain at the position where the visible light image is taken, otherwise, it is determined that there is no stain at the position where the visible light image is taken; after determining that there is a stain, extracting the image of the stain area, and inputting the image into the trained stain recognition model to identify the type of stain, and obtain a stain recognition result. The training method of the stain recognition model is as follows: collect several different types of photovoltaic panel stain images, mark the stains in these images, divide these images and the corresponding marks into training set and verification set in a ratio of 7:3, input the training set into the convolutional neural network for model training, and input the verification set into the trained model for accuracy verification to obtain the final stain recognition model.

[0053] The crack recognition unit is used to receive a visible light image, and use a crack detection model to perform crack recognition on the visible light image to obtain a crack recognition result. The workflow of the crack recognition unit includes: preprocessing the visible light image to obtain a preprocessed visible light image; segmenting the crack area in the preprocessed visible light image, and extracting crack feature information of the crack area; performing crack recognition based on the crack feature information to obtain a crack recognition result. In this embodiment, the visible light image is first converted from the RGB space to the HSV space, and shadows are removed using OpenCV to obtain a shadow-free image, the shadow-free image is grayed to obtain a grayed image, and the grayed image is denoised based on the median filtering method to obtain a preprocessed visible light image; the preprocessed image is segmented based on the threshold segmentation method, and crack areas are screened out to obtain a crack area image; feature extraction is performed on the crack area image, and crack recognition is performed based on the crack feature information to obtain a crack recognition result.

[0054] In this embodiment, the report generation unit can also set a positioning subunit. When the image recognition unit detects problems such as hot spots, stains and cracks on the photovoltaic panels, the positioning subunit will mark the current position of the drone on the inspection route and locate the position of the problematic photovoltaic panel to obtain the position information of the problematic photovoltaic panel.

[0055] The personal terminal is used to visualize the inspection report of photovoltaic panels and feed it back to the staff. The personal terminal includes: a visualization unit and a warning unit; the visualization unit is used to visualize the inspection report of photovoltaic panels. Specifically, based on the location information of the problematic photovoltaic panels, it is displayed in the photovoltaic power station map. Based on the hot spot recognition results, stain recognition results and crack recognition results, the hot spot location, stain location and type, and crack location are obtained respectively, and this information is displayed on the photovoltaic power station map to obtain a photovoltaic panel fault distribution map; the warning unit is used to issue warnings to the staff based on the photovoltaic panel fault distribution map.

[0056] The embodiments described above are only descriptions of the preferred methods of the present application, and are not intended to limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements made to the technical solutions of the present application by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present application.

Claims

1. An intelligent inspection system for photovoltaic panels based on drones, characterized in that: include: Inspection drones, image acquisition modules, central control modules and personal terminals; The inspection drone is used to perform inspections based on an inspection plan; The image acquisition module is integrated on the inspection drone, and the image acquisition module is used to acquire images of photovoltaic panels; The central control module is used to set the inspection plan and generate a photovoltaic panel inspection report based on the photovoltaic panel image; The personal terminal is used to visualize the photovoltaic panel inspection report and feed it back to the staff; The image acquisition module includes: an infrared image acquisition unit and a visible light image acquisition unit; The infrared image acquisition unit is used to acquire infrared images of the surface of the photovoltaic panel; The visible light image acquisition unit is used to acquire a visible light image of the surface of the photovoltaic panel; The central control module includes: a weather monitoring unit, a route planning unit, a plan generation unit, an image recognition unit and a report generation unit; The meteorological monitoring unit is used to monitor the on-site weather and obtain weather information; The route planning unit is used to plan the inspection route; The plan generating unit is used to generate the inspection plan based on the weather information and the inspection route; The image recognition unit is used to recognize hot spots, stains and cracks in the photovoltaic panel image to obtain an image recognition result; The report generating unit is used to generate the photovoltaic panel inspection report based on the image recognition result; The image recognition unit includes: a hot spot recognition subunit, a stain recognition subunit and a crack recognition subunit; The hot spot recognition subunit is used to receive the infrared image, and perform hot spot recognition on the infrared image using a hot spot recognition model to obtain a hot spot recognition result; The training method of the hot spot recognition model is as follows: first, infrared images of photovoltaic panels in normal working state and several hot spot fault states are collected, and these images are subjected to image enhancement, infrared image filtering and grayscale processing, and then the processed images are segmented to obtain training hot spot area images, and feature vectors are extracted from the hot spot area images to obtain training hot spot feature vectors, and the training hot spot images and the feature vectors corresponding thereto are divided into a training set and a verification set in a ratio of 7:3, respectively, and the training set is input into a convolutional neural network for model training, and the verification set is input into the trained model for accuracy verification, so as to obtain the final hot spot recognition model; The stain recognition subunit is used to receive the visible light image, and recognize the type of stain in the visible light image to obtain a stain recognition result; The training method of the stain recognition model is as follows: collecting a number of different types of photovoltaic panel stain images, marking the stains in these images, dividing these images and the corresponding marks into a training set and a verification set in a ratio of 7:3, inputting the training set into a convolutional neural network for model training, and inputting the verification set into the trained model for accuracy verification, thereby obtaining a final stain recognition model; The crack identification unit is used to receive the visible light image, and perform crack identification on the visible light image using a crack detection model to obtain a crack identification result; The workflow of the hot spot identification subunit includes: preprocessing the infrared image to obtain a preprocessed infrared image; segmenting the hot spot area in the preprocessed infrared image and extracting the hot spot feature vector of the hot spot area; inputting the hot spot feature vector into the trained hot spot identification model to perform hot spot identification and obtain the hot spot identification result; firstly performing image enhancement and infrared image filtering on the infrared image to obtain a preprocessed infrared image; then graying the preprocessed infrared image, performing edge detection on the grayed image, and segmenting the photovoltaic panel area; Then, the hot spot area in the photovoltaic panel area is segmented by the fuzzy C-means algorithm according to the least squares method to obtain the hot spot area and extract the feature vector of the hot spot area. The specific feature vector includes: the ratio of the hot spot area to the photovoltaic panel area, the number of cells with hot spot failures in the photovoltaic panel, and the difference between the average temperature of the hot spot area and the normal area. The feature vector is normalized, that is, the feature vector parameters are mapped to the interval [0,1] to obtain the hot spot feature vector; the hot spot feature vector is input into the trained hot spot recognition model to perform hot spot recognition and obtain the recognition result of the hot spot fault type; The working process of the stain recognition unit includes: extracting a color area of ​​a preset color range in the visible light image, and judging whether there is a stain based on the color area; if there is a stain, extracting an image of the stain area, and inputting the image of the stain area into a trained stain recognition model to identify the type of stain, and obtaining the stain recognition result; firstly extracting the color ranges of several normal photovoltaic panels, and integrating these color ranges to obtain a preset color range; if there is a non-preset color area in the color area, it is judged that there is a stain at the position where the visible light image is taken, otherwise, it is judged that there is no stain at the position where the visible light image is taken; after judging that there is a stain, extracting an image of the stain area, and inputting the image into a trained stain recognition model to identify the type of stain, and obtaining the stain recognition result; The workflow of the crack identification unit includes: preprocessing the visible light image to obtain a preprocessed visible light image; segmenting the crack area in the preprocessed visible light image, and extracting crack feature information of the crack area; identifying cracks based on the crack feature information to obtain a crack identification result; first converting the visible light image from RGB space to HSV space, and using OpenCV to remove shadows to obtain a shadow-free image, graying the shadow-free image to obtain a grayscale image, and denoising the grayscale image based on the median filtering method to obtain a preprocessed visible light image; segmenting the preprocessed image based on the threshold segmentation method, and screening out crack areas to obtain a crack area image; extracting features from the crack area image, identifying cracks based on the crack feature information, and obtaining a crack identification result.

2. According to claim 1, a photovoltaic panel site intelligent inspection system based on drones is characterized in that: The personal terminal includes: a visualization unit and an alarm unit; The visualization unit is used to visualize the photovoltaic panel inspection report to obtain a distribution map of the photovoltaic panel inspection situation; The warning unit is used to issue a warning to the staff based on the photovoltaic panel inspection distribution map.

Images