User-defined photovoltaic panel defect detection method and device based on self-cleaning equipment

Abstract

The invention belongs to the technical field of automatic detection, and discloses a user-defined photovoltaic panel defect detection method and device based on self-cleaning equipment, and the method comprises the steps: deploying an automatic defect detection device on photovoltaic panel self-cleaning equipment; acquiring a defect detection data set by utilizing an automatic defect detection device; training a defect detection model based on the defect detection data set; and carrying out photovoltaic panel defect identification by using the defect detection model. According to the invention, the automatic defect detection device is deployed on the photovoltaic panel self-cleaning equipment, the flexibility of user customization is improved to the maximum extent, and close-range image acquisition and high-precision defect detection of the photovoltaic panel are realized by using the cleaning process of the self-cleaning equipment.

Description

Technical Field

[0001] The present invention relates to the field of automatic detection technology, and in particular to a user-defined photovoltaic panel defect detection method and device based on a self-cleaning device. Background Art

[0002] Solar photovoltaic panels are the basis for the generation of electricity in photovoltaic power stations. The power generation efficiency of photovoltaic panels directly determines the power generation of a solar power station. According to practical experience, the cleanliness of the photovoltaic panel surface is proportional to the power generation efficiency. To improve the power generation efficiency, ensuring the cleanliness of photovoltaic panels is an effective way. As more and more solar power stations are built, the number of photovoltaic panels that need to be cleaned has also increased dramatically. How to quickly, effectively and intelligently clean photovoltaic panels has become a difficult problem that photovoltaic power generation companies urgently need to solve.

[0003] At the same time, due to the complex production process of solar panels, coupled with the human factors in the production, transportation and installation process, the panels are prone to various defects, which increases the breakage rate of the panels. These defects will seriously reduce the photovoltaic conversion efficiency and service life of the panels. Therefore, it is very important to detect panel defects during the production process. Currently, there are mainly the following detection methods:

[0004] (1) Manual inspection. Currently, most solar panel manufacturers use the naked eye to inspect defects on the EI image of the panel. This method is not only slow, affects production efficiency, and easily causes visual fatigue and errors, but also brings labor costs.

[0005] (2) Computer image recognition. Currently, there are many detection methods that use computer image processing technology, but these methods are only applicable to specific types of defects with obvious characteristics. For solar panels made of mainstream polycrystalline silicon, the pattern background interference is large and the characteristic defects are not obvious, so traditional methods cannot achieve effective detection.

[0006] At present, the most commonly used method is to use drones for image acquisition and defect detection. However, there are many problems in the actual application of using drones for image acquisition and defect detection, such as: (1) Some unpredictable interference will inevitably occur when the drone is flying, such as unexpected phenomena such as flocks of birds flying under the drone camera, resulting in a high error rate in image acquisition; (2) The drone needs to be maintained at a certain altitude during flight. The images collected at this altitude are lower in clarity than those collected at close range, and image distortion is common. Even with the existing high-precision image acquisition methods, the above problems cannot be avoided. Therefore, the images collected by drones are likely to be inconsistent with the actual situation, resulting in a high probability of misjudgment and poor application effect.

[0007] At the same time, defect detection is accompanied by the application of photovoltaic panels, and drones have a loss rate and their cost is relatively high.

[0008] Therefore, there is an urgent need for a photovoltaic panel real-time defect detection method that is low-cost, easy to produce and install, and has high detection accuracy. Summary of the invention

[0009] To solve the above problems, the present invention provides a customized photovoltaic panel automatic defect detection method and device combined with photovoltaic panel self-cleaning equipment, and deploys the automatic defect detection device on the photovoltaic panel self-cleaning equipment to maximize the flexibility of user customization, and utilizes the cleaning process of the self-cleaning equipment to achieve close-range image acquisition and high-precision defect detection of photovoltaic panels.

[0010] To achieve the above object, the present invention adopts the following technical solution:

[0011] In a first aspect, the present invention provides a user-defined photovoltaic panel defect detection method based on a self-cleaning device, comprising: deploying an automatic defect detection device on the photovoltaic panel self-cleaning device; acquiring a defect detection data set using the automatic defect detection device; training a defect detection model based on the defect detection data set; and identifying photovoltaic panel defects using the defect detection model.

[0012] Furthermore,

[0013] Users can customize the cleaning method, cleaning area, cleaning rounds, and image acquisition rounds of the photovoltaic panel self-cleaning equipment.

[0014] Furthermore,

[0015] Multiple rounds of image detection are carried out and a verification threshold for the number of identical results is set to verify the results of photovoltaic panel defect recognition.

[0016] Furthermore,

[0017] The result of photovoltaic panel defect identification is non-quality defect, and a secondary deep cleaning process is performed using a small area dedicated deep cleaning equipment;

[0018] The result of photovoltaic panel defect identification is a quality defect. The serial number of the work area map corresponding to the defect location is generated into a fault list and saved. Maintenance personnel confirm and repair the fault based on the fault list.

[0019] Furthermore,

[0020] The defect detection data set includes a first defect detection data set ABC1 and a second defect detection data set ABC2;

[0021] Performing sample expansion on the initial photovoltaic panel image data set to obtain a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels;

[0022] By acquiring a prior art photovoltaic panel defect image dataset, real defect feature extraction and feature classification are performed to obtain a second defect detection dataset ABC2.

[0023] Furthermore,

[0024] The CNN classification model combined with the BILSTM network is used as the first defect detection model, and the first defect detection model is trained using the first defect detection data set ABC1 to obtain a trained first defect detection model;

[0025] The multi-classification model based on Mask R-CNN is used as the second defect detection model, and the second defect detection model is trained using the second defect detection data set ABC2 to obtain a trained second defect detection model.

[0026] Furthermore,

[0027] The number of cleaning rounds is n, where n>=2.

[0028] In a third aspect, the present invention further provides a user-defined photovoltaic panel defect detection device based on a self-cleaning device, characterized in that it includes an automatic defect detection module, a defect detection data set generation module, a defect detection model training module, and a photovoltaic panel defect recognition module;

[0029] The automatic defect detection module is deployed on the photovoltaic panel self-cleaning device and is used to collect photovoltaic panel image data;

[0030] The defect detection data set generation module is used to generate a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels and a second defect detection data set ABC2 including real defect feature extraction and feature classification;

[0031] The defect detection model training module is used to train a first defect detection model and a second defect detection model according to the defect detection data set;

[0032] The photovoltaic panel defect recognition module detects photovoltaic panel image data using a defect detection model to achieve photovoltaic panel defect recognition.

[0033] In a third aspect, the present invention further provides an electronic device, comprising: a processor and a memory; the processor is coupled to the memory; wherein the processor is used to read and execute a program or instruction stored in the memory, so that the device executes the method of the first aspect.

[0034] In a fourth aspect, the present invention further provides a computer-readable storage medium storing a computer program, which implements the method of the first aspect when executed by a processor.

[0035] The technical solution provided by the present invention has at least the following technical effects or advantages:

[0036] Combining the self-cleaning equipment with the existing high-precision defect detection method, the self-cleaning equipment is used to obtain the image to be detected, which not only improves the convenience of obtaining the image, but also reduces the error of the collected image by taking the image at a close distance, thus reducing the actual application difficulty and application cost of defect detection;

[0037] By manually setting parameters such as working area, working rounds, cleaning times, image acquisition times, etc., defect recognition in a specific area and under specific settings can be achieved. By improving the flexibility of manual settings, the efficiency and accuracy of image recognition are maximized.

[0038] By screening abnormal photovoltaic panel images through the first defect detection model, the detection efficiency is improved and the training and calculation time of the model is reduced. The specific defect type is further identified through the second defect detection model. Compared with the common defect recognition model, the accuracy of defect type recognition is greatly improved.

[0039] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 This is a flow chart of a user-defined photovoltaic panel defect detection method based on a self-cleaning device in an embodiment of the present invention;

[0042] Figure 2 It is a structural schematic diagram of a user-defined photovoltaic panel defect detection device based on a self-cleaning device in an embodiment of the present invention;

[0043] Figure 3 The present invention is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The present invention provides a user-defined photovoltaic panel defect detection method based on a self-cleaning device. An automatic defect detection device is deployed on the photovoltaic panel self-cleaning device to maximize the flexibility of user customization. The cleaning process of the self-cleaning device is used to achieve close-range image acquisition and high-precision defect detection of the photovoltaic panel.

[0046] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1 As shown, the method of the present invention specifically comprises the following steps:

[0048] S1, deploy the automatic defect detection device on the photovoltaic panel self-cleaning equipment.

[0049] The above settings allow users to customize the cleaning method, cleaning area, cleaning rounds, and image acquisition rounds of the photovoltaic panel self-cleaning device according to actual needs.

[0050] (1) Customized cleaning modes mainly include: staff can define the cleaning mode of the self-cleaning equipment as cleaning mode, image acquisition mode, or cleaning plus image acquisition mode.

[0051] (2) Customized cleaning area mainly includes: establishing a self-cleaning equipment working area map based on the photovoltaic panel installation details of the current construction site. Workers can set the working area in the self-cleaning equipment to be the entire working area, multiple working areas, or a designated area in a working area map.

[0052] Wherein, a plurality of closely connected photovoltaic panels are defined as a work area map, and a full work area map including all photovoltaic panels can be generated. The partially designated area can be a work area formed based on a plurality of position points selected by a staff member in the current work area map. Furthermore, the partially designated area can also be a work area formed based on the coordinate position information by establishing a coordinate system according to the work area map and inputting the coordinates of the position by the staff member.

[0053] (3) The customized cleaning round mainly includes: the customized cleaning round refers to the staff setting the cleaning round to n times during the initialization of the self-cleaning device, where n>=2.

[0054] The self-cleaning device completes the cleaning work in the set working area map according to the cleaning logic as one round. The cleaning logic is the cleaning order of the self-cleaning device, which can be a logic of cleaning along the edge of the area first and then cleaning all the internal areas in sequence, or it can be a common cleaning logic of other self-cleaning devices.

[0055] (4) Customized image acquisition rounds mainly include: the staff sets the number of rounds of image acquisition based on the cleaning rounds in the self-cleaning equipment.

[0056] S2, obtaining a defect detection data set. The defect detection data set includes a first defect detection data set ABC1 and a second defect detection data set ABC2.

[0057] Photovoltaic panel image data is collected based on the image collection round setting.

[0058] The photovoltaic panel images are preprocessed using the image segmentation model. The collected photovoltaic panel area images are divided into multiple sub-images of the same size to obtain the initial photovoltaic panel image dataset.

[0059] The initial photovoltaic panel image data set is sample expanded to obtain a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels.

[0060] By acquiring a defect image dataset of photovoltaic panels in the prior art, including N types of data such as stains, bird droppings, and hot spot defects, feature extraction and feature classification are performed to obtain a second defect detection dataset ABC2.

[0061] As a preferred means, the image segmentation model can be any one or more combinations of instance segmentation models, and the image sample generation model is a generative adversarial network model.

[0062] S3, trains a defect detection model based on the defect detection dataset.

[0063] The CNN classification model combined with the BI LSTM network is used as the first defect detection model, and the first defect detection model is trained using the first defect detection data set ABC1 to obtain a trained first defect detection model.

[0064] The multi-classification model based on Mask R-CNN is used as the second defect detection model, and the second defect detection model is trained using the second defect detection data set ABC2 to obtain a trained second defect detection model.

[0065] S4, using the defect detection model to perform defect identification.

[0066] The images of a certain round of inspection are first passed through the first defect detection model to realize the identification of normal images and abnormal images. The abnormal images identified by the first defect detection model are input into the second defect detection model to obtain the identification results of specific defects such as stains, bird droppings, and hot spot defects.

[0067] Specifically: after obtaining the trained photovoltaic panel defect detection model, the preprocessed image is input into the first defect detection model to obtain the recognition results corresponding to the abnormal and normal photovoltaic panels; when the recognition result of the first defect detection model corresponding to the image is abnormal, the image is input into the second model again for detection to obtain the recognition result of the specific defect.

[0068] S5, defect identification result verification.

[0069] Based on the custom image acquisition round in S1 and the image recognition result in S4, multiple rounds of image detection are carried out and a verification threshold for the number of identical results is set.

[0070] The verification threshold is used to verify the defect recognition result. For example, when the number of custom image rounds is 2 or 3, it is set that when there are two identical output results in the obtained test results, the above identical output results are confirmed as the final photovoltaic panel defect recognition result. Or when the number of custom rounds is 4 or 5, it is set that when there are three identical output results in the obtained test results, the above identical output results are confirmed as the final photovoltaic panel recognition result.

[0071] S6, post-processing of results.

[0072] Stains, bird droppings, etc. that can be wiped off by deep cleaning are defined as non-quality defects, and secondary deep cleaning treatment is performed using special deep cleaning equipment for small areas.

[0073] Define real defects such as hot spot defects as quality defects, generate a fault list with the serial number of the work area map corresponding to the defect location and save it. Maintenance personnel will confirm and repair the fault based on the fault list.

[0074] Before the first defect detection model in the embodiment of the present invention is run, the CNN model is trained using the first defect detection data set ABC1. The specific training process is the prior art and is not limited here.

[0075] The preprocessed image data is input into the CNN model for image feature extraction, and then classified and identified by LSTM to distinguish normal images from abnormal images. Specifically:

[0076] The input data of the CNN model usually consists of 3 orders of magnitude tensors and 1 color channel. The image feature extraction process consists of several calculation stages. Each stage is called a layer. Connecting each stage can get the data processing flow of the CNN model:

[0077] x 1 →w 1 →x 2 →X L-1 →w L-1 →x L →w L →z.

[0078] The above formula represents the data x 1 From the first layer input, extract parameter features as tensor w 1 , the first level output is x 2 , which is also the input of the second layer. L-1 、w L-1 Represents intermediate stage data and extraction parameters.

[0079] The last layer of the model can be a loss layer to measure the difference between the target values ​​of the function, expressed as:

[0080] z→||x+y||,

[0081] Where y represents the target value.

[0082] Determine all image parameter features and combine them to form the input of the fully connected layer of the LSTM model. The predicted output of the LSTM model is:

[0083] argmax x i L .

[0084] Before the second defect detection model in the embodiment of the present invention is run, the Mask R-CNN model is trained using the second defect detection data set ABC2 to achieve recognition of normal images and abnormal images. The specific training process is not limited here.

[0085] The prediction output of the LSTM model is used as the input data of the Mask R-CNN model to achieve further segmentation and positioning of photovoltaic panel defects and accurate defect identification.

[0086] Based on the same inventive concept, see Figure 2 The present invention provides a user-defined photovoltaic panel defect detection device based on a self-cleaning device, including an automatic defect detection module, a defect detection data set generation module, a defect detection model training module, and a photovoltaic panel defect recognition module. Specifically:

[0087] The automatic defect detection module is deployed on the photovoltaic panel self-cleaning equipment to collect photovoltaic panel image data;

[0088] A defect detection data set generation module, used to generate a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels and a second defect detection data set ABC2 including real defect feature extraction and feature classification;

[0089] A defect detection model training module is used to train a first defect detection model and a second defect detection model according to a defect detection data set;

[0090] The photovoltaic panel defect recognition module uses the defect detection model to detect the photovoltaic panel image data to realize photovoltaic panel defect recognition.

[0091] A schematic diagram of the structure of an electronic device provided in an embodiment of the present invention is shown in FIG. Figure 3 As shown, the electronic device includes: a processor and a memory;

[0092] The processor is used to read and execute programs and instructions stored in the memory, so that the electronic device executes the above method embodiment.

[0093] It should be noted that, for the sake of convenience, the figure only shows the main components of the electronic device. In practical applications, the electronic device may also include components or assemblies not shown in the figure.

[0094] The present invention also provides a computer-readable storage medium storing a program or instruction. When a computer reads and executes the program or instruction, the computer executes the above method embodiment.

[0095] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A user-defined photovoltaic panel defect detection method based on a self-cleaning device, characterized in that: include: Deploy automatic defect detection devices on photovoltaic panel self-cleaning equipment; Acquiring a defect detection data set using an automatic defect detection device; Train a defect detection model based on a defect detection dataset; Photovoltaic panel defect identification using defect detection models.

2. The user-defined photovoltaic panel defect detection method based on the self-cleaning device according to claim 1, characterized in that: Users can customize the cleaning method, cleaning area, cleaning rounds, and image acquisition rounds of the photovoltaic panel self-cleaning equipment.

3. The user-defined photovoltaic panel defect detection method based on the self-cleaning device according to claim 2, characterized in that: Multiple rounds of image detection are carried out and a verification threshold for the number of identical results is set to verify the results of photovoltaic panel defect recognition.

4. The user-defined photovoltaic panel defect detection method based on the self-cleaning device according to claim 3, characterized in that: The result of photovoltaic panel defect identification is non-quality defect, and a secondary deep cleaning process is performed using a small area dedicated deep cleaning equipment; The result of photovoltaic panel defect identification is a quality defect. The serial number of the work area map corresponding to the defect location is generated into a fault list and saved. Maintenance personnel confirm and repair the fault based on the fault list.

5. The user-defined photovoltaic panel defect detection method based on the self-cleaning device according to claim 2, characterized in that: The defect detection data set includes a first defect detection data set ABC1 and a second defect detection data set ABC2; Performing sample expansion on the initial photovoltaic panel image data set to obtain a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels; By acquiring a prior art photovoltaic panel defect image dataset, real defect feature extraction and feature classification are performed to obtain a second defect detection dataset ABC2.

6. The user-defined photovoltaic panel defect detection method based on a self-cleaning device according to claim 5, characterized in that: The CNN classification model combined with the BILSTM network is used as the first defect detection model, and the first defect detection model is trained using the first defect detection data set ABC1 to obtain a trained first defect detection model; The multi-classification model based on Mask R-CNN is used as the second defect detection model, and the second defect detection model is trained using the second defect detection data set ABC2 to obtain a trained second defect detection model.

7. The user-defined photovoltaic panel defect detection method based on a self-cleaning device according to claim 2, characterized in that: The number of cleaning rounds is n, where n>=2.

8. A user-defined photovoltaic panel defect detection device based on a self-cleaning device, characterized in that: It includes automatic defect detection module, defect detection data set generation module, defect detection model training module, and photovoltaic panel defect recognition module; The automatic defect detection module is deployed on the photovoltaic panel self-cleaning device and is used to collect photovoltaic panel image data; The defect detection data set generation module is used to generate a first defect detection data set ABC1 including normal images and abnormal images of photovoltaic panels and a second defect detection data set ABC2 including real defect feature extraction and feature classification; The defect detection model training module is used to train a first defect detection model and a second defect detection model according to the defect detection data set; The photovoltaic panel defect recognition module detects photovoltaic panel image data using a defect detection model to achieve photovoltaic panel defect recognition.

9. An electronic device, characterized in that: include: Processor and memory; The processor is coupled to the memory; The processor is used to read and execute the program or instruction stored in the memory, so that the device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored, and when the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images