Photovoltaic panel detection method based on infrared thermal imaging

Abstract

The invention discloses a photovoltaic panel detection method based on infrared thermal imaging, and particularly relates to the field of photovoltaic panel detection, and the method comprises the steps: photovoltaic panel region division: determining a region needing to be detected as a target data region; photovoltaic panel data acquisition: scanning a target detection area by adopting infrared optical monitoring equipment under different time period strategies, and then acquiring grating data to obtain a hot spot parameter, a cold spot parameter, a temperature parameter and an appearance parameter of the photovoltaic panel; photovoltaic panel data analysis: analyzing the collected parameters to obtain a photovoltaic panel hot spot danger coefficient, a photovoltaic panel cold spot danger coefficient, a photovoltaic panel temperature danger coefficient and an appearance damage degree danger coefficient; and data comprehensive analysis: importing the hot spot danger coefficient value of each sub-region panel, the cold spot danger coefficient value of the photovoltaic panel, the temperature danger coefficient value of the photovoltaic panel and the appearance damage degree danger coefficient value into a comprehensive analysis model to obtain a photovoltaic panel integrity index.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic panel detection, and more specifically, to a photovoltaic panel detection method based on infrared thermal imaging. Background Art

[0002] As the global demand for renewable energy increases, solar photovoltaic power generation, as a sustainable renewable energy source, has received widespread attention. However, when using photovoltaic panels, they are inevitably affected by various factors, resulting in failures. This makes fault detection of photovoltaic panels very important, and infrared imaging detection is a good method.

[0003] At present, the existing infrared photovoltaic panel detection methods include photovoltaic panel data acquisition, photovoltaic panel data analysis, and photovoltaic panel data feedback; among them, panel data acquisition is mainly used to scan the photovoltaic panel area, quickly scan the area with potential problems, so as to collect photovoltaic panel data with potential problems to facilitate the next step of analysis. Photovoltaic panel data analysis conducts a detailed analysis of the area of ​​potential problems of the photovoltaic panel collected by infrared imaging, so as to determine the specific problem situation of the potential problem area, and finally feedback the analysis results to relevant technical personnel for subsequent maintenance work.

[0004] However, there are still some shortcomings in its actual use. For example, when collecting photovoltaic panel data, the collected data is single and ambiguous, and cannot provide extensive data support and precise data requirements for data analysis, which further affects the analysis results of potential problem areas. When analyzing the data, there is a problem of low efficiency, and the problem cannot be quickly identified, wasting a certain amount of time.

[0005] Therefore, there is an urgent need to provide a photovoltaic panel detection method based on infrared thermal imaging to solve the problems of the existing infrared photovoltaic panel detection method, such as the singleness of the collected data, the ambiguity of the analysis process and the low efficiency of the analysis results. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a photovoltaic panel detection method based on infrared thermal imaging, and solves the problems raised in the above-mentioned background technology through the following scheme.

[0007] To achieve the above object, the present invention provides the following technical solution: a photovoltaic panel detection method based on infrared thermal imaging, comprising:

[0008] P1. Photovoltaic panel area division: It is used to determine the area to be detected as the target data area, and use the OpenCV image processing library based on Python language to extract frames second by second on the photovoltaic panel video stream collected in the observation area of ​​the monitoring probe; obtain image sequences with equal time intervals at different timestamps, and perform segmentation operations on the photovoltaic panel based on the semantic segmentation algorithm to divide the target detection area into target areas, and number each sub-area in the target area as 1, 2, ..., n, i∈(1, 2, ..., n) in sequence;

[0009] P2. Photovoltaic panel data collection: Use infrared optical monitoring equipment to scan the target detection area, and then collect grating data to obtain the hot spot parameters, cold spot parameters, temperature parameters and appearance parameters of the photovoltaic panel. Extensive data collection is crucial to the accuracy of the analysis;

[0010] P3. Photovoltaic panel data analysis: Analyze the collected parameters to obtain the photovoltaic panel hot spot risk factor, photovoltaic panel cold spot risk factor, photovoltaic panel temperature risk factor and appearance damage risk factor; the photovoltaic panel data analysis includes photovoltaic panel risk factor calculation model, photovoltaic panel cold spot risk factor calculation model, photovoltaic panel temperature risk factor calculation model and appearance damage risk factor calculation model. These calculation models analyze photovoltaic panels from various angles, improve the accuracy of the analysis results, and provide help for the subsequent comprehensive analysis;

[0011] P4. Comprehensive data analysis: The hot spot risk factor value of each sub-area panel, the cold spot risk factor value of the photovoltaic panel, the temperature risk factor value of the photovoltaic panel, and the appearance damage risk factor value are imported into the comprehensive analysis model to obtain the photovoltaic panel integrity index;

[0012] P5. Final evaluation: used to establish the preset value of the integrity index of the photovoltaic panel, make a comparative judgment based on the preset value of the integrity, and send out relevant signals;

[0013] P6. Real-time remote monitoring: The data from comprehensive analysis will be tracked in real time. Once a problematic photovoltaic panel is detected, the data will be immediately fed back to the relevant technical staff for processing.

[0014] Preferably, the equal area division method is to divide the area into multiple small grids according to a certain size, and then perform data collection in each small grid, which can clearly define the area for data collection and facilitate management and subsequent data processing.

[0015] Preferably, the photovoltaic panel data collection includes: collecting panel part data and non-panel part data, the panel part data includes hot spot parameters and cold spot parameters; the non-panel part data includes temperature parameters and appearance parameters; the hot spot parameters include crack length coefficient Bubble coefficient And the impurity area factor The cold spot parameters include the debris area factor The fracture depth coefficient α and the black piece area coefficient β; the temperature parameters include the inverter temperature c, the photovoltaic combiner box temperature d and the distribution cabinet temperature e; the appearance parameters include the line corrosion area η and the bracket wear volume ρ.

[0016] Preferably, the photovoltaic panel hazard factor calculation model is used to import the collected hot spot parameters into the photovoltaic panel hot spot hazard factor calculation model to obtain the photovoltaic panel hot spot hazard factor value; the photovoltaic panel cold spot hazard factor calculation model is used to import the collected cold spot parameters into the photovoltaic panel cold spot hazard factor calculation model to obtain the photovoltaic panel cold spot hazard factor value; the photovoltaic panel temperature hazard factor calculation model is used to import the collected temperature parameters into the photovoltaic panel temperature hazard factor calculation model to obtain the photovoltaic panel temperature hazard factor value; the photovoltaic panel appearance damage degree hazard factor calculation model is used to import the collected appearance parameters into the photovoltaic panel appearance damage degree hazard factor calculation model to obtain the photovoltaic panel appearance damage degree hazard factor value.

[0017] Preferably, the photovoltaic panel hot spot risk coefficient calculation model is specifically: Where u represents the hot spot hazard coefficient value of the photovoltaic panel, represents the crack length coefficient of the ith sub-region of the photovoltaic panel, represents the bubble coefficient of the ith sub-region of the photovoltaic panel, Represents the impurity area coefficient of the i-th sub-region of the photovoltaic panel.

[0018] Preferably, the photovoltaic panel cold spot risk coefficient calculation model is specifically: Where f represents the cold spot risk factor value of the photovoltaic panel, represents the fragmentation area coefficient of the ith sub-region of the photovoltaic panel, α i represents the fracture depth coefficient of the ith sub-region of the photovoltaic panel, β i Represents the black sheet area coefficient of the i-th sub-region of the photovoltaic panel.

[0019] Preferably, the photovoltaic panel temperature risk coefficient calculation model is specifically: Where x represents the temperature hazard coefficient of the photovoltaic panel, c i represents the inverter temperature of the ith sub-area of ​​the photovoltaic panel, d irepresents the temperature of the photovoltaic combiner in the i-th sub-area of ​​the photovoltaic panel, e i Indicates the temperature of the power distribution cabinet in the i-th sub-area of ​​the photovoltaic panel.

[0020] Preferably, the photovoltaic panel appearance wear risk coefficient model is specifically: Where z represents the risk factor value of the photovoltaic panel's appearance wear, η i represents the line corrosion area of ​​the ith sub-region of the photovoltaic panel, ρ i Represents the bracket wear volume of the i-th sub-area of ​​the photovoltaic panel.

[0021] Preferably, the comprehensive analysis model is specifically: Where Q is the integrity index value of the photovoltaic panel.

[0022] Preferably, the photovoltaic panel integrity preset value is marked as Q def , when Q>=Q def When Q def When , it means that the PV panel integrity index value is less than the preset value of the PV panel integrity, indicating that the PV panel is in poor condition and a warning signal is issued.

[0023] Technical effects and advantages of the present invention:

[0024] 1. The present invention divides the photovoltaic panels to be inspected into equal-area areas for inspection by dividing the photovoltaic panels into areas of equal area, which not only reduces the difficulty of inspection, but also improves the accuracy of inspection, providing convenient help for the post-processing of photovoltaic panels; through photovoltaic panel data collection, starting from various angles, including hot spot parameters, cold spot parameters, temperature parameters and appearance parameters of photovoltaic panels, multi-faceted data collection avoids the singleness of data, and provides strong data support for photovoltaic panel data analysis of photovoltaic panels;

[0025] 2. Through the analysis of photovoltaic panel data, starting from the part, the data of each sub-area is analyzed, and relevant calculation models are established, including the photovoltaic panel risk factor calculation model, the photovoltaic panel cold spot risk factor calculation model, the photovoltaic panel temperature risk factor calculation model and the appearance damage risk factor calculation model, to accurately calculate the correlation coefficient of the data and avoid the ambiguity of photovoltaic panel data analysis;

[0026] ​3. Through comprehensive analysis of the data, the data is integrated and analyzed as a whole, and a photovoltaic panel integrity calculation model is established, which further improves the accuracy of photovoltaic panel data analysis and provides great help for the comprehensive detection of photovoltaic panels; through final evaluation and real-time remote monitoring, problematic photovoltaic panels are quickly detected, and relevant warnings are issued, so that relevant personnel can promptly discover problematic photovoltaic panels, thereby improving the efficiency of detection. The present invention uses a photovoltaic panel detection method based on infrared thermal imaging. These methods well solve the problems of existing infrared photovoltaic panel detection methods, such as the singleness of collected data, the ambiguity of the analysis process, and the low efficiency of the analysis results, thereby providing help for the detection of photovoltaic panels. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0028] Figure 2 The photovoltaic panel index number P ID The unique encoding identification number F of the specific frame recognition result image at a specific time node ID Schematic diagram of the composition structure. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0030] As attached Figure 1 The photovoltaic panel detection method based on infrared thermal imaging shown includes the following steps:

[0031] P1. Photovoltaic panel area division: First, use the Python-based OpenCV image processing library to extract frames from the photovoltaic panel video stream collected in the monitoring probe observation area second by second. For example, a video stream with a frame rate of 60fps extracts one frame every 60 frames as the result output, and obtains a series of image sequences with different timestamps but equal time intervals. The photovoltaic panel is segmented and recognized based on a mature semantic segmentation algorithm. The segmentation model architecture, backbone network and compilation environment framework are not limited in this solution. The labels used for identification are positive pixels, that is, pixels in the photovoltaic panel range in the image, and negative labels are background pixels.

[0032] The image sequence containing the timestamp and pixel classification label results of semantic segmentation recognition is obtained. The background data is removed, noise is reduced, and geometric correction is performed through the OpenCV image algorithm library of the Python programming language to obtain a panel area image that meets the standard. It is generally a rectangular area with a constant aspect ratio. The specific area varies slightly due to different sizes of photovoltaic panels. The technicians need to make adjustments according to the actual situation. The photovoltaic panels in the target area are bound to the corresponding index number of the photovoltaic panel through the spatial position relationship, which is recorded as P ID , so as to establish the corresponding relationship between the detection result and the panel in the subsequent recognition and detection tasks. Then, the specific frame-by-frame recognition result image at a specific time node is assigned a unique coding identification number F ID Its specific composition is as follows Figure 2 shown.

[0033] In this embodiment, it is specifically necessary to explain that an equal area division method can be adopted, which is to divide into multiple small grids according to a certain size, and then perform data collection in each small grid. The area for data collection can be clearly defined, which is convenient for management and subsequent data processing.

[0034] P2. Photovoltaic panel data collection: Data collection mainly consists of the panel part and the non-panel part. The panel part includes hot spot parameters and cold spot parameters; the non-panel part includes temperature parameters and appearance parameters. The temperature parameters in the non-panel parameters include inverter temperature c, photovoltaic junction box temperature d, and distribution cabinet temperature e. They are collected by integrating mature sensor components and transmitted through communication means such as the Internet of Things or all-optical networks. Finally, the data set for the input model is obtained through data preprocessing; the appearance parameters include line corrosion area η and bracket wear volume ρ, which need to be summarized in combination with the observation summary evaluation table of the operation and maintenance technicians.

[0035] The panel part is mainly collected by P1 and given F ID The frame extraction recognition result image is obtained by secondary recognition and acquisition, wherein the hot spot parameter includes the crack length coefficient Bubble coefficient And the impurity area factor For crack length factor Impurity area factor It is necessary to use semantic segmentation algorithm for collection again; for the bubble number coefficient Object detection algorithms should be used for acquisition.

[0036] The cold spot parameters include the fragment area coefficient Fracture depth coefficient α and black piece area coefficient β; the fragment area coefficient and black piece area coefficient should be identified and analyzed using semantic segmentation algorithm, and the fracture depth coefficient should be calculated comprehensively in combination with the above crack length identification results. The collection strategy of the hot spot parameters is to collect at uniform time intervals during the photovoltaic power generation period, and the collection strategy of the cold spot parameters is to collect at uniform time intervals during the non-photovoltaic power generation period.

[0037] In some implementation cases, for the crack length coefficient in the above-mentioned cold spot parameter Usually, it is necessary to use a semantic segmentation algorithm to identify crack features first. It is necessary to use the Deeplab-V3+ algorithm architecture based on the ResNet-50 backbone network and use the pytorch deep learning framework based on the python language for encoding. In this embodiment, a pre-trained crack detection model is used for transfer learning to obtain better recognition results, and the model can be iterated with a lower amount of training / verification data to achieve good data acquisition results. The crack feature recognition result is usually a semantic segmentation mask image (Mask). The crack recognition result mask is converted into an output in JSON data format through the Python-based OpenCV image processing library, which records the inflection point coordinates of the boundary line of each pixel category in the image pixel coordinate system, and the boundary coordinates of the pixel semantic recognition result as crack class data are marked as P. L The above boundary coordinates are used as input and the orthogonal skeleton method provided by mature computing libraries such as scikit-image is used to calculate the average width of the cracks in the i-th sub-region of all n sub-regions detected on this panel. The longitudinal length of the crack μ i Perform calculations and calculate the crack length coefficient of the i-th sub-area of ​​the photovoltaic panel The formula is

[0038]

[0039] In this embodiment, for the impurity area coefficient in the hot spot parameter First, it is necessary to perform semantic segmentation of the image data of the impurity feature area. The technical path is the same as the aforementioned crack length coefficient, and the semantic segmentation recognition result is a mask image containing the impurity coverage area. The impurity coverage area may usually appear in multiples, so it is necessary to use a target detection algorithm to identify and count the impurity feature areas and bind them to the corresponding semantic segmentation blocks; the target detection technical solution recommends that implementers also use a pre-trained model, whose model architecture uses Faster-RCNN, and the backbone network uses mature technical routes such as VGG-19. The aforementioned semantic segmentation and target detection for the impurity feature area are also based on the acquisition and assignment of F as described in P1. IDThe frame extraction recognition result image, the recognition result of the semantic segmentation detection model is usually a mask of several closed recognition areas. Each area is converted into a labeling file in JSON data format through the OpenCV function library and assigned a first preset number j for the closed impurity feature area therein; the result of target detection and recognition is a closed range frame, which is also converted into a labeling file in JSON format, and the closed detection frame therein is assigned a second preset number i; when the closed geometric area marks of the two preset numbers satisfy that the second preset number frame area covers the first preset number frame and the covering area is greater than or equal to 80%, the value i=j is assigned. If the aforementioned recognition result that satisfies the second preset number i has a total of n impurity feature coverage areas, the total number of pixels in the semantic segmentation recognition result mask image of the i-th impurity feature coverage area is ψ i , and the currently identified collection is assigned F ID The total number of pixels of the frame extraction recognition result image is Τ i , then the impurity area coefficient of the i-th impurity coverage feature area is The formula is

[0040]

[0041] In this embodiment, for the bubble coefficient parameter in the hot spot parameter First, it is necessary to identify the bubble features of the panel, which contain two key information: quantity and area. Two deep learning image detection algorithms, semantic segmentation and target detection, are required. The implementation ideas of semantic segmentation and target detection algorithms for bubble features are similar to those of the impurity area coefficient mentioned above. The related methods have similar ideas, and the results are also a mask image containing the bubble coverage area and a closed detection frame of the bubble detection area. Similarly, the semantic segmentation and target detection for the bubble coefficient part are also based on the acquisition and assignment of F in P1. ID The frame extraction recognition result image, the recognition result of the semantic segmentation detection model is usually a mask of several closed recognition areas. Each area is converted into a label file in JSON data format through the OpenCV function library and the first preset number j is assigned to the closed bubble feature area therein; the result of target detection recognition is a closed range frame, which is also converted into a label file in JSON format, and the closed detection frame therein is assigned a second preset number i; when the closed geometric area marks of the two preset numbers satisfy that the second preset number frame area covers the first preset number frame and the covering area is greater than or equal to 90%, the value i=j is assigned. If the above recognition result that satisfies the second preset number i has a total of n bubble feature coverage areas, the total number of pixels in the semantic segmentation recognition result mask image of the i-th bubble feature coverage area is δ i, and the total number of pixels of the currently identified frame extraction recognition result image collected and assigned FID is Τ i , then the bubble coefficient of the i-th bubble feature coverage area is The formula is

[0042] In this embodiment, the fragment area coefficient in the cold spot parameter is It is necessary to first identify the fragment features of the panel. The technical path is the same as the impurity area coefficient mentioned above. The semantic segmentation recognition result is a mask image containing the impurity coverage area. Fragment areas may usually appear in multiples. Therefore, it is necessary to use a target detection algorithm to identify and count the impurity feature areas and bind them to the corresponding semantic segmentation blocks. The relevant target detection technical solution should adopt the same solution as the target detection of impurity feature areas. The semantic segmentation and target detection for the fragment feature area are also based on the acquisition and assignment of F in P1. ID The frame extraction recognition result image, the recognition result of its semantic segmentation detection model is usually a mask of several closed recognition areas. Each area is converted into a labeling file in JSON data format through the OpenCV function library and the first preset number j is assigned to the closed impurity feature area therein; the result of target detection recognition is a closed range frame, which is also converted into a labeling file in JSON format, and the closed detection frame therein is assigned a second preset number i; when the closed geometric area marks of the two preset numbers satisfy that the second preset number frame area covers the first preset number frame and the covering area is greater than or equal to 80%, the value i=j is assigned. If the above-mentioned recognition result that satisfies the second preset number i has a total of n fragment feature areas, the total number of pixels in the semantic segmentation recognition result mask image of the i-th fragment feature area is And the currently identified collection is assigned F ID The total number of pixels of the frame extraction recognition result image is Τ i , then the impurity area coefficient of the i-th impurity coverage feature area is The formula is

[0043]

[0044] In this embodiment, the black patch area coefficient β in the cold spot parameter needs to first identify the black patch features of the panel, which also contains two types of key information: quantity and area. It is necessary to use two types of deep learning image detection algorithms: semantic segmentation and target detection. The implementation ideas of semantic segmentation and target detection algorithms for black patch features are similar to those of the aforementioned bubble coefficient. The related methods have similar ideas, and the results are also a mask image containing the black patch coverage area and a closed detection frame of the black patch detection area. Similarly, the semantic segmentation and target detection for the black patch area coefficient part are also based on the acquisition and assignment of F in P1. ID The result image of the frame extraction recognition, the recognition result of the semantic segmentation detection model is usually a mask of several closed recognition areas. Each area is converted into a label file in JSON data format through the OpenCV function library and assigned a first preset number j for the closed black film feature area therein; the result of the target detection recognition is a closed range frame, which is also converted into a label file in JSON format, and the closed detection frame therein is assigned a second preset number i; when the closed geometric area marks of the two preset numbers satisfy that the second preset number frame area covers the first preset number frame and the covering area is greater than or equal to 90%, the value i=j is assigned. If the above recognition result that satisfies the second preset number i has a total of n black film feature coverage areas, the total number of pixels in the semantic segmentation recognition result mask image of the i-th black film feature coverage area is x i , and the currently identified collection is assigned F ID The total number of pixels of the frame extraction recognition result image is Τ i , then the black patch area coefficient of the i-th black patch feature coverage area is β i , the formula is

[0045] In this embodiment, the fracture depth coefficient α in the cold spot parameter needs to be further calculated in combination with the result of the crack feature identification described above. The fracture depth coefficient α needs to be combined with the average width ω of the i-th crack among all n cracks detected in the crack feature detection result. i and the mean value of pixels within the feature range of the i-th crack The calculation is performed, wherein the pixels within the crack feature range are converted through the JSON file obtained by the mask conversion of the crack feature recognition result described above and the original image for intersection calculation and image cropping before output. The skeleton line (axis) of the i-th crack is calculated based on the orthogonal skeleton line method in the scikit-image computing library, and the length of the skeleton line is l pixels. The corresponding crack width of any pixel point k of the skeleton line is calculated as (The crack width is the length from the normal line of any pixel on the skeleton line to the edges of the crack on both sides, in pixels) and the qth pixel value of the ith crack of the n cracks is the total m pixels. The fracture depth coefficient α of the i-th crack is i The formula is

[0046]

[0047] P3. Photovoltaic panel data analysis: Analyze the collected parameters to obtain the photovoltaic panel hot spot risk factor, photovoltaic panel cold spot risk factor, photovoltaic panel temperature risk factor and appearance damage risk factor; the photovoltaic panel data analysis includes photovoltaic panel risk factor calculation model, photovoltaic panel cold spot risk factor calculation model, photovoltaic panel temperature risk factor calculation model and appearance damage risk factor calculation model. These calculation models analyze photovoltaic panels from various angles, improve the accuracy of the analysis results, and provide assistance for the subsequent comprehensive analysis.

[0048] In this embodiment, it is specifically necessary to explain that the photovoltaic panel hazard coefficient calculation model is used to import the collected hot spot parameters into the photovoltaic panel hot spot hazard coefficient calculation model to obtain the photovoltaic panel hot spot hazard coefficient value; the photovoltaic panel cold spot hazard coefficient calculation model is used to import the collected cold spot parameters into the photovoltaic panel cold spot hazard coefficient calculation model to obtain the photovoltaic panel cold spot hazard coefficient value; the photovoltaic panel temperature hazard coefficient calculation model is used to import the collected temperature parameters into the photovoltaic panel temperature hazard coefficient calculation model to obtain the photovoltaic panel temperature hazard coefficient value; the photovoltaic panel appearance damage degree hazard coefficient calculation model is used to import the collected appearance parameters into the photovoltaic panel appearance damage degree hazard coefficient calculation model to obtain the photovoltaic panel appearance damage degree hazard coefficient value.

[0049] In this embodiment, it should be specifically explained that the photovoltaic panel hot spot risk coefficient calculation model is specifically: Where u represents the hot spot hazard coefficient value of the photovoltaic panel, Indicates the crack length of each sub-area of ​​the photovoltaic panel, Indicates the number of bubbles in each sub-area of ​​the photovoltaic panel, Represents the impurity area of ​​each sub-region of the photovoltaic panel.

[0050] In this embodiment, it should be specifically explained that the photovoltaic panel cold spot risk coefficient calculation model is specifically: Where f represents the cold spot risk factor value of the photovoltaic panel, represents the fragment area of ​​each sub-region of the photovoltaic panel, α i represents the fracture depth of each sub-region of the photovoltaic panel, β i Indicates the black area of ​​each sub-region of the photovoltaic panel.

[0051] In this embodiment, it should be specifically explained that the photovoltaic panel temperature risk coefficient calculation model is specifically: Where x represents the temperature hazard coefficient of the photovoltaic panel, c i Indicates the inverter temperature of each sub-area of ​​the photovoltaic panel, d i Indicates the temperature of the photovoltaic combiner in each sub-area of ​​the photovoltaic panel, ei Indicates the temperature of the power distribution cabinet in each sub-area of ​​the PV panel.

[0052] In this embodiment, it should be specifically explained that the photovoltaic panel appearance wear risk coefficient model is specifically: Where z represents the risk factor value of the photovoltaic panel's appearance wear, η i represents the line corrosion area of ​​each sub-region of the photovoltaic panel, ρ i Represents the wear volume of the bracket in each sub-area of ​​the photovoltaic panel.

[0053] P4. Comprehensive data analysis: The hot spot risk factor value of each sub-area panel, the cold spot risk factor value of the photovoltaic panel, the temperature risk factor value of the photovoltaic panel and the appearance damage risk factor value are imported into the comprehensive analysis model to obtain the photovoltaic panel integrity index.

[0054] In this embodiment, it should be specifically explained that the comprehensive analysis model is specifically: Where Q is the integrity index value of the photovoltaic panel.

[0055] P5. Final evaluation: used to establish preset values ​​for the integrity index of photovoltaic panels, make comparative judgments based on the preset values, and send out relevant signals.

[0056] In this embodiment, it should be specifically noted that the photovoltaic panel integrity preset value is marked as Q def , when Q>=Q def When Q def When , it means that the PV panel integrity index value is less than the preset value of the PV panel integrity, indicating that the PV panel is in poor condition and a warning signal is issued.

[0057] P6. Real-time remote monitoring: The data from the final evaluation feedback will be tracked in real time. Once an early warning signal is received, the data will be immediately fed back to the relevant technical personnel for processing.

[0058] ​The present invention divides the photovoltaic panels to be inspected into equal-area areas for inspection through photovoltaic panel area division, which not only reduces the difficulty of inspection, but also improves the accuracy of inspection, providing convenient help for the later processing of photovoltaic panels; through photovoltaic panel data collection, starting from various angles, including hot spot parameters, cold spot parameters, temperature parameters and appearance parameters of photovoltaic panels, multi-faceted data collection avoids the singleness of data and provides strong data support for photovoltaic panel data analysis; through photovoltaic panel data analysis, starting from the part, the data of each sub-area is analyzed, and relevant calculation models are established, including photovoltaic panel risk factor calculation model, photovoltaic panel cold spot risk factor calculation model, photovoltaic panel temperature risk factor calculation model and appearance damage risk factor calculation model, accurately The correlation coefficient of the data is calculated to avoid the ambiguity of photovoltaic panel data analysis; through comprehensive analysis of the data, the data is integrated and analyzed as a whole, and a photovoltaic panel integrity calculation model is established, which further improves the accuracy of photovoltaic panel data analysis and provides great help for the comprehensive detection of photovoltaic panels; through final evaluation and real-time remote monitoring, problematic photovoltaic panels are quickly detected, and relevant warnings are issued, so that relevant personnel can promptly discover problematic photovoltaic panels, thereby improving the efficiency of detection; the present invention uses a photovoltaic panel detection method based on infrared thermal imaging, which well solves the problems of the existing infrared photovoltaic panel detection method, such as the singleness of the collected data, the ambiguity of the analysis process and the low efficiency of the analysis results, and provides help for the detection of photovoltaic panels.

[0059] Secondly: In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0060] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic panel detection method based on infrared thermal imaging, characterized in that: include: P1. Photovoltaic panel area division: It is used to determine the area to be detected as the target data area, and use the OpenCV image processing library based on Python language to perform frame extraction operation on the photovoltaic panel video stream collected in the observation area of ​​the monitoring probe second by second; Obtain image sequences with equal time intervals at different time stamps, and perform segmentation operations on the photovoltaic panel based on the semantic segmentation algorithm to divide the target detection area into target areas, and number each sub-area in the target area as 1, 2, ..., n, i∈(1, 2, ..., n); P2. Photovoltaic panel data collection: Use infrared optical monitoring equipment to scan the target detection area, and then collect grating data to obtain the hot spot parameters, cold spot parameters, temperature parameters and appearance parameters of the photovoltaic panel; P3. Photovoltaic panel data analysis: Analyze the collected parameters to obtain the photovoltaic panel hot spot risk factor, photovoltaic panel cold spot risk factor, photovoltaic panel temperature risk factor and appearance damage risk factor; the photovoltaic panel data analysis includes photovoltaic panel risk factor calculation model, photovoltaic panel cold spot risk factor calculation model, photovoltaic panel temperature risk factor calculation model and appearance damage risk factor calculation model. These calculation models analyze photovoltaic panels from various angles, improve the accuracy of the analysis results, and provide help for the subsequent comprehensive analysis; P4. Comprehensive data analysis: The hot spot risk factor value of each sub-area panel, the cold spot risk factor value of the photovoltaic panel, the temperature risk factor value of the photovoltaic panel, and the appearance damage risk factor value are imported into the comprehensive analysis model to obtain the photovoltaic panel integrity index; P5. Final evaluation: used to establish the preset value of the integrity index of the photovoltaic panel, make a comparison and judgment based on the preset value of the integrity, and send out relevant signals; P6. Real-time remote monitoring: The data from comprehensive analysis feedback will be tracked in real time. Once a problematic photovoltaic panel is detected, the data will be immediately fed back to the relevant technical personnel for processing.

2. The equal-area division method is to divide the area into multiple small grids according to a certain size, and then collect data in each small grid. The area for data collection can be clearly defined, which is convenient for management and subsequent data processing.

3. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel data collection includes: collecting panel part data and non-panel part data, the panel part data includes hot spot parameters and cold spot parameters; the non-panel part data includes temperature parameters and appearance parameters; the hot spot parameters include crack length coefficient Bubble coefficient l and impurity area coefficient The cold spot parameters include the debris area factor The fracture depth coefficient α and the black piece area coefficient β; the temperature parameters include the inverter temperature c, the photovoltaic combiner box temperature d and the distribution cabinet temperature e; the appearance parameters include the line corrosion area η and the bracket wear volume ρ.

4. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel hazard coefficient calculation model is used to import the collected hot spot parameters into the photovoltaic panel hot spot hazard coefficient calculation model to obtain the photovoltaic panel hot spot hazard coefficient value; the photovoltaic panel cold spot hazard coefficient calculation model is used to import the collected cold spot parameters into the photovoltaic panel cold spot hazard coefficient calculation model to obtain the photovoltaic panel cold spot hazard coefficient value; the photovoltaic panel temperature hazard coefficient calculation model is used to import the collected temperature parameters into the photovoltaic panel temperature hazard coefficient calculation model to obtain the photovoltaic panel temperature hazard coefficient value; the photovoltaic panel appearance damage degree hazard coefficient calculation model is used to import the collected appearance parameters into the photovoltaic panel appearance damage degree hazard coefficient calculation model to obtain the photovoltaic panel appearance damage degree hazard coefficient value.

5. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel hot spot risk coefficient calculation model is specifically as follows: Where u represents the hot spot hazard coefficient value of the photovoltaic panel, represents the crack length coefficient of the ith sub-region of the photovoltaic panel, l i represents the bubble coefficient of the ith sub-region of the photovoltaic panel, Represents the impurity area coefficient of the i-th sub-region of the photovoltaic panel.

6. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel cold spot risk coefficient calculation model is specifically as follows: Where f represents the cold spot risk factor value of the photovoltaic panel, represents the fragmentation area coefficient of the ith sub-region of the photovoltaic panel, α i represents the fracture depth coefficient of the ith sub-region of the photovoltaic panel, β i Represents the black sheet area coefficient of the i-th sub-region of the photovoltaic panel.

7. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel temperature risk coefficient calculation model is specifically as follows: Where x represents the temperature hazard coefficient of the photovoltaic panel, c i represents the inverter temperature of the ith sub-area of ​​the photovoltaic panel, d i represents the temperature of the PV combiner in the i-th sub-area of ​​the PV panel, and ei represents the temperature of the distribution cabinet in the i-th sub-area of ​​the PV panel.

8. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel appearance wear risk coefficient model is specifically: Where z represents the risk factor value of the photovoltaic panel's appearance wear, η i represents the line corrosion area of ​​the ith sub-region of the photovoltaic panel, ρ i Represents the bracket wear volume of the i-th sub-area of ​​the photovoltaic panel.

9. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The comprehensive analysis model is specifically: Where Q is the integrity index value of the photovoltaic panel.

10. The photovoltaic panel detection method based on infrared thermal imaging according to claim 1, characterized in that: The photovoltaic panel integrity preset value is marked as Q def , when Q>=Q def When Q def When , it means that the photovoltaic panel integrity index value is less than the photovoltaic panel integrity preset value, indicating that the photovoltaic panel is in poor condition and a warning signal is issued.​

Images