Photovoltaic module fault detection system, method and device based on infrared image hot spot detection, and storage medium

Abstract

The invention provides a photovoltaic module fault detection system, method and device based on infrared image hot spot detection, and a storage medium, and belongs to the field of photovoltaic module fault detection. The system comprises a temperature difference boundary extraction module, an image sequence registration module, a hot spot track identification module, an abnormal dynamic screening module and a fault hot spot confirmation module. Pixel difference screening is carried out by setting a temperature difference threshold value, a boundary communication structure is established, frame-level displacement calculation between time sequence images is introduced to realize coordinate alignment, hot spot center points in continuous frames are extracted to form a motion path, abnormal point screening and time node labeling are carried out in combination with path displacement characteristics, and the time sequence image is obtained. According to the method, the boundary area growth rate and temperature change double factors are fused to screen fault areas, dynamic tracking and accurate detection of abnormal hot spots are achieved, the logic relevance between abnormal behavior recognition and fault judgment is enhanced, and the integrity of photovoltaic module fault information extraction and the accuracy of hot spot recognition are guaranteed.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic module fault detection, and in particular to a photovoltaic module fault detection system and method, device, and storage medium based on infrared image hot spot detection. Background Art

[0002] The field of photovoltaic module fault detection involves technologies for monitoring and diagnosing the operating status of modules in photovoltaic power generation systems. Various detection methods are used to identify abnormalities within or on the module surface, such as hot spots, damage, obstructions, and burned-out junction boxes, to ensure the system's operational efficiency and equipment safety. Photovoltaic module fault detection encompasses multiple aspects, including electrical performance testing, image recognition and analysis, temperature distribution monitoring, infrared imaging, and data acquisition and processing. It often combines automated equipment and image processing methods to quickly and accurately identify the location and type of faults.

[0003] Among them, the photovoltaic module fault detection system based on infrared image hot spot detection mainly targets hot spots caused by local overheating problems caused by factors such as micro-cracks in the cell, internal short circuits, and obstructions during the operation of the photovoltaic module. The principle of this system is: use an infrared thermal imager to collect thermal images of the surface of the photovoltaic module, and use the image grayscale value to determine the existence of local overheating areas, namely hot spots, and then perform local area statistics and temperature gradient judgments to determine the distribution of hot spots, and then identify faulty components.

[0004] The aforementioned methods rely primarily on image grayscale threshold settings and local area statistics to identify hot spot areas, providing static fault diagnosis. They struggle to accurately depict the spatial extension of hot spot boundaries. The lack of a unified registration method between images at different time points makes it impossible to track the movement of hot spots over time. The lack of a dynamic analysis mechanism for identifying hot spot trends makes it difficult to capture hot spot behavior changes in a timely manner. For example, when a photovoltaic module develops a tiny crack or is partially obscured, the hot spot's initial manifestations are mild, and traditional methods can easily overlook or misjudge it, delaying fault confirmation and maintenance response, impacting the efficiency and stability of the photovoltaic power generation system. Summary of the Invention

[0005] In order to solve the above technical problems, the present application proposes a photovoltaic module fault detection system and method, device, and storage medium based on infrared image hot spot detection.

[0006] The technical solution adopted in this application is: a photovoltaic module fault detection system based on infrared image hot spot detection, comprising:

[0007] Temperature difference boundary extraction module: It is used to screen pixel pairs with temperature differences greater than the temperature gradient threshold based on the infrared temperature data in the infrared image of the photovoltaic module, record the positions of the pixel pairs, and continuously calculate the pixel temperature differences between each two pixel pairs, screen the connected areas of the hot spot boundaries, and generate the hot spot boundary area distribution results;

[0008] Image sequence registration module: used to obtain the positions of the four corner points of the photovoltaic module frame in different frames based on the infrared images of the photovoltaic module at multiple monitoring times, and calculate the average inter-frame corner point displacement value to complete the coordinate alignment and obtain the multi-time series component image registration results;

[0009] Hot spot trajectory identification module: This module is used to extract the hot spot center point position within the hot spot boundary area distribution results and the multi-time component image registration results, record the inter-frame coordinates of the hot spot center point, calculate the displacement path based on the coordinate change between every two center points in adjacent frames, and generate the hot spot cross-frame path distribution structure;

[0010] Abnormal dynamic screening module: This module is used to compare the displacements of all frames in the path one by one based on the distribution structure of the hot spot cross-frame path, calculate the average displacement length of the hot spot per frame in the entire path, screen abnormal nodes and mark them, and obtain the annotation set of the hot spot abnormal motion frames;

[0011] Fault hot spot confirmation module: It is used to measure the area of the hot spot boundary structure and calculate the hot spot boundary growth rate based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results. It screens the hot spot areas where the hot spot boundary growth rate in two consecutive frames is higher than the hot spot expansion rate threshold and the temperature difference is higher than the temperature gradient threshold, confirms them as faulty hot spots, and outputs the photovoltaic module fault identification results.

[0012] Furthermore, the temperature difference boundary extraction module includes:

[0013] Temperature difference calculation submodule: The infrared image is divided into N 3×3 pixel blocks. Based on the infrared temperature data of each pixel in the infrared image, the infrared temperature values of the central pixel of each pixel block and its eight neighboring pixels are read. The temperature difference of each pair of central and neighboring pixels is calculated. The obtained temperature difference is compared with the temperature gradient threshold. The pixel pairs with temperature differences greater than the temperature gradient threshold are screened to obtain the number of high-gradient pixel pairs.

[0014] Continuous pixel segment screening submodule: Based on the number of high-gradient pixel pairs, the module traverses the image row by row, determines the spatial positional relationship between any two high-gradient pixel pairs in the horizontal, vertical, and diagonal directions, performs directional continuity judgment on adjacent high-gradient pixel pairs, and counts the lengths of pixel segments that meet the continuity conditions. It then screens for connected area segments whose length is not less than the minimum recognizable hot spot size to obtain directional pixel connected segment length data.

[0015] Hot spot boundary identification submodule: Based on the directional pixel connected segment length data, all pixel segments that meet the minimum identifiable hot spot size are regionally fused, the corresponding boundary contour information is read and boundary range statistics are performed, the distribution position of the continuous boundary area in the image is determined, and the hot spot boundary area distribution result is obtained.

[0016] Furthermore, the image sequence registration module includes:

[0017] Image acquisition submodule: Based on the hot spot boundary area distribution results, it obtains infrared images of photovoltaic modules at multiple monitoring times, numbers each frame of the image and records the timestamp information, sorts the images by time sequence number, counts the boundary coordinates of the photovoltaic module area in each frame of the image, establishes an image sequence index table, and obtains image sequence index data;

[0018] Corner detection submodule: Extracts the component boundary area in each frame of the image based on the image sequence index data, combines the grayscale gradient change values of the four vertex areas in the boundary contour, and calculates the coordinate position of the local gradient maximum point in the area in each frame of the image in turn, obtains the four corner point candidate coordinate set, and generates the image frame corner point coordinate data;

[0019] Coordinate alignment submodule: Based on the image frame corner coordinate data, the average inter-frame corner point displacement value is calculated according to the coordinate difference of the corresponding four corner points in two adjacent frames of images, and inverse pixel offset processing is performed on each frame of image. That is, the coordinate values of all pixels in the current image are combined with the corresponding displacement values for addition and subtraction compensation, and the image is subjected to sub-pixel coordinate resampling and interpolation correction to obtain the multi-time series component image registration result.

[0020] Furthermore, the calculation formula of the average inter-frame corner point displacement value is as follows:

[0021] ;

[0022] in, Indicates the Image at time The average inter-frame corner displacement value, 、 Respectively represent the image Corner point at time The horizontal and vertical coordinates of 、 Indicates the coordinate position of the corresponding corner point at the previous moment, and 4 is the number of corner points in each image.

[0023] Furthermore, the hot spot trajectory recognition module includes:

[0024] Hot spot extraction submodule: This module is used to traverse each frame of the image sequence based on the multi-time component image registration results, extract all pixel sets within the hot spot boundary area distribution results, determine the pixel coordinates corresponding to the maximum value and use it as the hot spot center point position of the current image frame, record the row and column index coordinates of the hot spot center point, and obtain the hot spot center point coordinate sequence;

[0025] Coordinate recording submodule: Based on the hot spot center point coordinate sequence, the hot spot center point coordinate values between two consecutive frames are extracted in image time order. The difference between the horizontal and vertical coordinates of the hot spot center in adjacent frames is calculated, and the translation component of the hot spot coordinates between frames is recorded. The infrared temperature value of the pixel corresponding to the hot spot center point in each frame is extracted. A joint record set of horizontal and vertical coordinate changes, hot spot area changes, and temperature changes is established to obtain the hot spot path change across frames.

[0026] Path generation submodule: According to the cross-frame path change of the hot spot, the horizontal coordinate difference, vertical coordinate difference, hot spot area change value and temperature value difference between all frames are integrated, and the frames are arranged in sequence according to the frame number. The coordinate points of the hot spot center point are connected in sequence to form a cross-frame continuous path line of the hot spot center point, and the cross-frame path distribution structure of the hot spot is obtained.

[0027] Furthermore, the abnormal dynamic screening module includes:

[0028] Displacement calculation submodule: Calculate the spatial displacement of hot spots between corresponding frames based on the center point coordinate data of all consecutive frames in the hot spot cross-frame path distribution structure, organize all the hot spot spatial displacement values between frames into an array in the order of frame numbers, and calculate the average displacement length per unit frame of the path to obtain the average displacement length per unit frame data;

[0029] Abnormal identification submodule: Based on the average displacement length of the unit frame, the inter-frame displacement value of each segment in the hot spot path is read, and the difference between each segment value and the average value is calculated to determine whether it exceeds twice the average threshold. If it exceeds, the corresponding frame number is marked as an abnormal node. All marked frames are searched frame by frame, and all abnormal marking results are accumulated to obtain a list of hot spot abnormal motion frame numbers;

[0030] Information annotation submodule: According to the hot spot abnormal motion frame number list, the hot spot boundary information and center point coordinate data in the corresponding frame image are extracted, a square frame coverage area is generated for the abnormal frame image, the frame index information is synchronously read and the frame number and abnormal identification label text are drawn in the upper left corner of the image, a hot map annotation output image is generated for each abnormal frame, and a hot spot abnormal motion frame annotation set is obtained.

[0031] Furthermore, the fault hot spot confirmation module includes:

[0032] Boundary measurement submodule: Based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, the closed pixel set formed by the hot spot boundary of each image frame is obtained. The number of all pixels in the hot spot boundary area of each frame is accumulated, and the hot spot boundary area value of the current frame is counted. The time series is constructed in the order of frame numbers to obtain the hot spot boundary area sequence;

[0033] Growth screening submodule: Based on the hot spot boundary area sequence, the boundary area values of all consecutive frame images are read. The growth rate of the hot spot boundary between adjacent frames is calculated using the standard hot spot boundary growth rate calculation formula. Each frame's growth rate value is judged one by one to see if it exceeds the hot spot expansion rate threshold specified by the standard. The hot spot maximum temperature difference between the current frame and the previous frame is also obtained to determine whether it exceeds the temperature gradient threshold. If both conditions are met, the current frame number is recorded as a candidate frame. All frame numbers that meet the screening conditions constitute a candidate set, and a sequence of fault candidate frame numbers is obtained.

[0034] Fault identification submodule: According to the fault candidate frame number sequence, read the hot spot center point coordinate change trend and hot spot boundary structure of the corresponding frame image, detect the continuous change characteristics of the hot spot area in the spatial structure of the candidate frame, and judge whether the boundary expansion direction and the hot spot temperature rise direction are consistent. If the hot spot area contour deformation direction and the hot spot temperature rise center point drift direction in two consecutive frames are spatially consistent, the candidate frame is confirmed as a fault hot spot area, and the photovoltaic module fault identification result is obtained.

[0035] A photovoltaic module fault detection method based on infrared image hot spot detection, based on the photovoltaic module fault detection system based on infrared image hot spot detection, the method comprises the following steps:

[0036] S1: Based on the infrared temperature data in the infrared image of the photovoltaic module, the pixel pairs with temperature differences greater than the temperature gradient threshold are screened, the positions of the pixel pairs are recorded, and the pixel temperature differences between each two pixel pairs are continuously calculated to screen the connected areas of the hot spot boundaries and generate the hot spot boundary area distribution results;

[0037] S2: Based on the infrared images of the photovoltaic modules at multiple monitoring moments, the positions of the four corner points of the photovoltaic module frame in different frames are obtained, and the average inter-frame corner point displacement value is calculated to complete the coordinate alignment and obtain the multi-time series module image registration result;

[0038] S3: Based on the hot spot boundary area distribution results and the multi-time component image registration results, the hot spot center point position within the hot spot boundary area distribution results is extracted, the inter-frame coordinates of the hot spot center point are recorded, and the displacement path is calculated based on the coordinate change between every two center points in adjacent frames to generate the hot spot cross-frame path distribution structure;

[0039] S4: Based on the distribution structure of the hot spot path across frames, the displacements between all frames in the path are compared one by one, the average displacement length of the hot spot per frame in the entire path is calculated, abnormal nodes are screened and marked, and a set of annotations of hot spot abnormal motion frames is obtained;

[0040] S5: Based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, the area of the hot spot boundary structure is measured and the hot spot boundary growth rate is calculated. The hot spot area with a hot spot boundary growth rate higher than the hot spot expansion rate threshold in two consecutive frames and a temperature difference higher than the temperature gradient threshold is screened to confirm it as a faulty hot spot, and the photovoltaic module fault identification result is output.

[0041] A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0042] A computer-readable storage medium stores a computer program / instruction thereon, which implements the steps of the method when executed by a processor.

[0043] Compared with the prior art, the present application has the following beneficial effects: by setting a temperature difference threshold to perform pixel difference screening and establish a boundary connectivity structure, frame-level displacement calculation between time-series images is introduced to achieve coordinate alignment, the center points of hot spots in continuous frames are extracted to form a motion path during path tracing, and the path displacement characteristics are combined to perform abnormal point screening and time node marking in dynamic identification. In the judgment process, the boundary area growth rate and temperature change dual factors are integrated to screen the fault area, and dynamic tracking and accurate confirmation of abnormal hot spots are achieved, which improves the recognition ability of the hot spot evolution process in different operation stages, enhances the logical correlation between abnormal behavior recognition and fault judgment, ensures the integrity of photovoltaic module fault information extraction and the accuracy of hot spot recognition, and promotes the expansion of fault location from single-frame recognition to multi-frame evolution trend judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present application will be further described below with reference to the accompanying drawings:

[0045] Figure 1 A flowchart of the operation of this application system;

[0046] Figure 2 Flowchart for the temperature difference boundary extraction module operation;

[0047] Figure 3 Flowchart for the image sequence registration module operation;

[0048] Figure 4 Flowchart for the operation of the hot spot track identification module;

[0049] Figure 5Flowchart for the operation of abnormal dynamic screening module;

[0050] Figure 6 A flowchart for the operation of the fault hot spot confirmation module;

[0051] Figure 7 This is the system block diagram of this application. DETAILED DESCRIPTION

[0052] like Figures 1 to 7 As shown, the present application provides a photovoltaic module fault detection system based on infrared image hot spot detection, comprising:

[0053] Temperature difference boundary extraction module: The infrared image is divided into N 3×3 pixel blocks, and the infrared temperature data of all pixels in the infrared image of the photovoltaic module are obtained. The temperature difference between the central pixel of each pixel block and the average temperature of its eight neighboring pixels is calculated. The pixel pairs with temperature difference greater than the temperature gradient threshold (≥3°C / pixel set according to the IEC62446-3:2017 standard) are screened. The infrared image is traversed and the positions of pixel pairs with temperature difference meeting the conditions are recorded. The pixel temperature difference is continuously calculated in the horizontal, vertical and diagonal directions for the pixel pairs. The length of the pixel segment with adjacent temperature difference continuously meeting the temperature gradient threshold is counted. The hot spot boundary connected area that is not less than the minimum recognizable hot spot size (≥5×5 pixel matrix set according to the UL1703 standard) is screened to generate the hot spot boundary area distribution result.

[0054] Image sequence registration module: Acquire infrared images of photovoltaic modules at multiple monitoring times, detect the positions of the four corner points of the photovoltaic module frame in the image, compare the coordinate differences of the four corner points in different frame images, calculate the average inter-frame corner point displacement value, and align the image coordinates in the time series dimension (sub-pixel alignment: error ≤ 0.5 pixel) to obtain the multi-time series module image registration results.

[0055] Hot spot trajectory identification module: It is used to extract the coordinates of the highest temperature pixel within the hot spot boundary area distribution results as the position of the hot spot center point based on the hot spot boundary area distribution results and the multi-time component image registration results, record the inter-frame coordinates of the hot spot center point in chronological order, calculate the displacement path based on the coordinate changes between each two center points, and generate the hot spot cross-frame path distribution structure.

[0056] Abnormal dynamic screening module: It is used to compare the displacement between all frames in the path one by one according to the distribution structure of the hot spot cross-frame path, calculate the average displacement length of the hot spot per frame in the entire path, screen out abnormal nodes that exceed twice the average displacement, associate the boundary position and time index in the image where the abnormal node is located, and mark them (in accordance with the infrared thermal image annotation specification of ISO18436-6:2014), and obtain the annotation set of the hot spot abnormal motion frame.

[0057] Fault hot spot confirmation module: It is used to measure the area of the hot spot boundary area and compare the degree of change with the area of the previous frame based on the annotation set of the hot spot abnormal motion frame and the hot spot boundary area distribution results, calculate the hot spot boundary growth rate, and screen the hot spot areas with a hot spot boundary growth rate higher than the hot spot expansion rate threshold (IECTS62910:2017 stipulates ≥15% / min) in two consecutive frames and a temperature difference higher than the temperature gradient threshold. The hot spot area is confirmed as a faulty hot spot, thereby obtaining the photovoltaic module fault identification result.

[0058] The hot spot boundary area distribution results include the coordinates of the temperature difference abnormal area, the length of the boundary connected segment, the minimum hot spot size identifier, and the edge connection structure; the multi-time component image registration results include the photovoltaic component frame feature point alignment information, the inter-frame coordinate transformation vector, the registration error range, and the image sequence index; the hot spot cross-frame path distribution structure includes the hot spot center point coordinate sequence, the inter-frame path displacement vector, the path continuity index, and the hot spot movement direction trend; the hot spot abnormal motion frame-time annotation set includes the abnormal displacement frame index, the corresponding hot spot boundary box position, the annotation timestamp, and the annotation compliance identifier; the photovoltaic component fault identification results include the fault hot spot position identifier, the boundary growth rate record, the temperature difference compliance analysis result, and the fault confirmation result.

[0059] See also Figure 2 ,The temperature difference boundary extraction module includes : a temperature difference calculation submodule, a continuous pixel segment screening submodule, and a hot spot boundary recognition submodule.

[0060] The temperature difference calculation submodule is based on the infrared temperature data of each pixel in the infrared image. It reads the infrared temperature values of the pixel and its eight neighboring pixels, calculates the temperature difference for each pair of center-neighborhood pixels, compares the obtained temperature difference with the temperature gradient threshold, and screens out pixel pairs with temperature differences greater than the temperature gradient threshold to obtain the number of high-gradient pixel pairs.

[0061] Specifically, the image's temperature matrix data is first collected to identify the infrared temperature value of each pixel within the image. For example, if the temperature of pixel P1 in the image region is 41.2°C, the corresponding average temperature of its eight neighboring pixels (hereinafter referred to as the neighborhood average temperature value) is 38.5°C. The pixel and its neighborhood average temperature value are then combined to construct a pairwise center-to-neighborhood temperature difference combination. For P1, for example, its temperature difference is 41.2°C minus the neighborhood average temperature value of 38.5°C, resulting in a temperature difference of 2.7°C. A similar operation is then performed for other pixels in the image, recording all center-to-neighborhood temperature differences and constructing a unified temperature difference matrix. The temperature difference value of each pair of center-to-neighborhood pixels is then compared against a set temperature gradient threshold (set at 3°C / pixel according to the IEC62446-3:2017 standard) to determine whether the temperature difference between each pair of center-to-neighborhood pixels satisfies the temperature gradient threshold. As shown in Table 1 below, the difference between pixel P2 and the average temperature of its neighborhood is 4.7°C, meeting the condition. The difference between pixel P3 and the average temperature of its neighborhood is 5.3°C, also meeting the condition. However, pixel P1, P4, and P5 do not meet the condition. Therefore, pixel P2 and P3 are selected as target pixels that meet the temperature difference judgment. Based on this screening operation, the number of center-neighborhood pixels in the current image that meet the threshold can be counted to obtain the number of high-gradient pixel pairs. For example, if the image matrix contains a total of 50,000 pixels, and 3,156 pixel pairs have a temperature difference exceeding 3°C after counting, the positions of these pixel pairs are recorded and stored in an index table for subsequent module calls. The number of high-gradient pixel pairs indicates the number of pixels in the image with significant thermal gradient mutations at local locations, which serves as an important basis for subsequent judgment of local hot spot formation.

[0062] Table 1 Temperature difference judgment pixel data table

[0063] ;

[0064] As shown in Table 1, the temperature difference between pixels P2 and P3 exceeds 3° C., and they are selected as a high-gradient pixel pair.

[0065] The continuous pixel segment screening submodule traverses the image row by row based on the number of high-gradient pixel pairs, determines the spatial position relationship between each pair of high-gradient pixel pairs in the horizontal, vertical and diagonal directions, makes directional continuity judgments on adjacent high-gradient pixel pairs, and counts the lengths of pixel segments that meet the continuity conditions. It screens connected area segments whose length is not less than the minimum identifiable hot spot size to obtain directional pixel connected segment length data.

[0066] Specifically, according to the number of high-gradient pixel pairs obtained, its coordinate position in the two-dimensional image matrix needs to be identified during the image traversal process, and all high-gradient pixel pairs are arranged in order by rows, columns, and diagonals according to the image scanning direction and their continuity is checked segment by segment. The sliding window method is used to scan each row of the image in turn. For example, starting from the 10th row, check whether there are adjacent high-gradient pixel pairs in the row. If the position numbers between two or more adjacent high-gradient pixel pairs differ by 1, it is considered that there is a continuous pixel segment. If the length exceeds the set threshold, it is considered to meet the hot spot size requirements. This operation is carried out in the horizontal direction, vertical direction and two diagonal directions in turn, and the length value of each high-gradient pixel pair in each direction is compared to determine whether it meets the minimum identification hot spot size conditions. The size is set to 5×5 pixels according to the UL1703 standard, that is, the length of the continuous pixel segment shall not be less than 5 pixels. The following test results can be referred to in this implementation, as shown in Table 2 below:

[0067] Table 2 Pixel connected segment determination table

[0068] ;

[0069] As shown in Table 2, the pixel lengths of segments S2, S3, and S5 are 6, 7, and 5, respectively, all reaching or exceeding 5 pixels. Therefore, they are determined to be connected segments that meet the hot spot size. In this way, all directional continuous pixel segments that meet the hot spot size in the current image can be screened out, and the directional pixel connected segment length data can be obtained.

[0070] The hot spot boundary recognition submodule performs regional fusion on all pixel segments that meet the minimum identifiable hot spot size based on the directional pixel connected segment length data, reads the corresponding boundary contour information and performs boundary range statistics, determines the distribution position of the continuous boundary area in the image, and obtains the hot spot boundary area distribution result.

[0071] Based on the directional pixel connected segment length data obtained above, we further extract the boundary information of all pixel segments that meet the hot spot size requirements from the image. The position coordinates of the starting and ending pixels of each connected region in the image matrix are recorded, and the bounding box or circumscribed contour of each connected region is generated. The boundary span is calculated by taking the pixel difference between the starting and ending pixel coordinates of the boundary, and each region is labeled and classified. For example, the starting pixel coordinates of hot spot number H1 are (45, 60) and the ending coordinates are (50, 65), with a boundary span of 5 pixels. This boundary coordinate information allows for further spatial aggregation analysis to identify the distribution of hot spot boundaries throughout the image. This is shown in Table 3 below.

[0072] Table 3 Hot spot boundary area parameters

[0073] ;

[0074] As shown in Table 3, the start and end pixels covered by different hot spot boundaries are clearly marked. Based on this data, the hot spot boundary area distribution results can be obtained.

[0075] See also Figure 3 ,The image sequence registration module includes : image acquisition submodule, corner point detection submodule, and coordinate alignment submodule.

[0076] The image acquisition submodule obtains infrared images of photovoltaic modules at multiple monitoring moments based on the distribution results of hot spot boundary areas, numbers each frame of image and records timestamp information, sorts the images by time sequence number, counts the boundary coordinates of the photovoltaic module area in each frame of image, establishes an image sequence index table, and obtains image sequence index data.

[0077] Specifically, infrared images of PV panels at multiple monitoring times are acquired. Each frame is numbered and its corresponding acquisition time is recorded. The acquisition time unit is set to milliseconds, and the numbering format is F1 to Fn. After recording the number, the basic parameter information of each frame is retrieved, including resolution specifications, image size, and pixel unit conversion factor. The image coordinate system is then mapped to the actual physical area of the panel in a proportional manner to obtain the regional boundary value of the PV panel frame on the image. The panel area boundary is determined by the pixel coordinates of the upper left corner and the lower right corner. For example, the component region boundaries of image number F1 are the upper left coordinates (120, 90) and the lower right coordinates (520, 380). Therefore, it can be determined that the width of the component region is 400 pixels and the height is 290 pixels. Under the condition of an image resolution of 640×480 pixels, it occupies approximately 60.4% of the entire image area. The coordinate values and distribution ranges of all pixels in the region are called, and a mapping structure is established for parameters such as image acquisition time, number, coordinate range, and resolution. The index relationship between time series and images is constructed, and all images are sorted in ascending order based on time values to ensure that the time axis structure is continuous and monotonic. An image sequence index table is generated based on the sorted image sequence. The index table uses the image number as the primary key and contains field information such as image acquisition time, resolution, and component region boundary value. It is used for coordinate call in downstream image registration tasks. Finally, the image sequence index data shown in Table 4 is obtained.

[0078] Table 4 Infrared image sequence parameters

[0079] .

[0080] The corner detection submodule extracts the component boundary area in each frame image based on the image sequence index data, combines the grayscale gradient change values of the four vertex areas in the boundary contour, and calculates the coordinate position of the local gradient maximum in the area in each frame image in turn, obtains the four corner point candidate coordinate set, and generates the image frame corner point coordinate data.

[0081] Specifically, based on the image sequence index data, the PV module border region corresponding to each frame of the image is extracted. After demarcating the PV module border region, regional pixel blocks are obtained from its four corners: the upper left corner, upper right corner, lower left corner, and lower right corner. The grayscale values of a 5×5 pixel window in each regional pixel block are sampled. The grayscale gradients of the pixels within each regional pixel block in the horizontal and vertical directions are calculated. The coordinates of the pixel corresponding to the maximum grayscale change between all pixels in the regional pixel block and its neighborhood are extracted as candidate corner points. This corner point extraction operation is performed sequentially for each frame of the image, obtaining a total of four candidate corner points. The local contrast mean and gradient direction consistency of the corner point candidate set are analyzed. If the gradient direction angle deviation of a candidate point is within 5° and the contrast difference is greater than 20 grayscale levels, it is considered a valid corner point. Finally, the coordinates of the four corner points are extracted from each frame of the image, and an image corner point coordinate record table is constructed to obtain the image frame corner point coordinate data, as shown in Table 5 below.

[0082] Table 5 Image frame corner point coordinate data table

[0083] .

[0084] The coordinate alignment submodule is based on the image frame corner coordinate data. According to the coordinate difference of the corresponding four corner points in two adjacent frames, the average inter-frame corner displacement value is calculated using formula (1). Inverse pixel offset processing is performed on each frame of the image, that is, the coordinate values of all pixels in the current image are combined with the corresponding displacement values for addition and subtraction compensation. The image is subjected to sub-pixel coordinate resampling and interpolation correction to obtain the multi-time series component image registration result.

[0085] (1);

[0086] in, Indicates the Image at time The average inter-frame corner displacement value, 、 Respectively represent the image Corner point at time The horizontal and vertical coordinates of 、 Indicates the coordinate position of the corresponding corner point at the previous moment, and 4 is the number of corner points in each image.

[0087] According to the image frame corner point coordinate data, the difference operation is performed on the four corner point coordinates of each frame image and its previous frame image, and the Euclidean distance is used as the pixel displacement index of the corresponding corner points between the two frames of image. With the previous frame image The corresponding corner point coordinates are expressed as and ,in Represents the four corner points of the image, corresponding to the upper left corner, upper right corner, lower left corner, and lower right corner respectively. Formula (1) is used to calculate the average inter-frame corner point displacement value.

[0088] The average inter-frame corner point displacement value refers to the average spatial displacement of the corresponding four photovoltaic module corner points in the image coordinate system in two adjacent frames of infrared images, with the unit being pixels. This value is obtained by measuring the coordinate change of each corner point in two consecutive frames and taking its Euclidean distance, and then averaging the displacement values of the four corner points. It reflects the degree of positional offset of the photovoltaic module image in the time series due to external environmental factors (such as camera shake, installation angle change or image acquisition error). It is an important indicator for measuring whether spatial offset occurs in the image sequence. The larger the value, the more obvious the positional change between the images, which needs to be corrected through coordinate compensation and alignment operations. This ensures that the multi-time series images have a unified spatial reference, which helps to ensure the continuity and accuracy of subsequent image overlay, hot spot tracking and other processing.

[0089] Formula (1) is based on the spatial geometric changes between the corresponding corner points in two frames of images. First, the horizontal and vertical offsets of each corner point between two adjacent frames are obtained by calculating the coordinate difference, which are and , square the offsets in these two directions and add them together, that is , used to represent the square of the Euclidean distance between two points in two-dimensional space, and then the actual straight-line distance between pixels is obtained by square root operation , which reflects the change in the position of each corner point in the image sequence. In order to avoid the offset in the positive and negative directions, an absolute value operation is added to ensure that the displacement of all corner points is non-negative. Then the displacement of the four corner points is summed, that is, , and finally multiply by the coefficient The average value is obtained. This operation is used to comprehensively describe the average displacement trend of the entire frame image at the four corner points, which is the key reference for achieving spatial consistency registration between image sequence frames.

[0090] Taking images F1 and F2 as an example, the coordinates of the four corner points are as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] Substituting into formula (1) we get:

[0096] ;

[0097] The same logic is applied to subsequent frames to obtain the average frame corner displacement values shown in Table 6.

[0098] Table 6 Average inter-frame corner point displacement value table

[0099] ;

[0100] The results show that the motion trend of the image on the time axis can be used for subsequent pixel-level coordinate alignment (sub-pixel alignment error ≤ 0.5 pixel) operations to complete the acquisition of multi-temporal component image registration results.

[0101] See also Figure 4 ,The hot spot trajectory recognition module includes : a hot spot extraction submodule, a coordinate recording submodule, and a path generation submodule.

[0102] The hot spot extraction submodule is used to traverse each frame of the image sequence based on the multi-time component image registration results, extract all pixel sets within the hot spot boundary area distribution results, determine the pixel coordinates corresponding to the maximum value and use it as the hot spot center point position of the current image frame, record the row and column index coordinates of the hot spot center point, and obtain the hot spot center point coordinate sequence.

[0103] Specifically, obtain the multi-time component image registration result, traverse all frame images in the image sequence, and extract the distribution range of the hot spot boundary area based on the registered pixel coordinates, call the closed polygon area of ​​the hot spot boundary in each frame image, and crop and extract its pixel set from the entire image, sort the infrared temperature values of all pixels in the extracted area in ascending order, count the maximum value position in the temperature array, and record the two-dimensional coordinate point position corresponding to the maximum temperature and use it as the center point of the hot spot. The corresponding coordinate point is recorded with the image pixel index; if there are multiple highest values with the same temperature value, then press The frontmost pixel is selected as the center point in the order of row first and column second best. For example, in the image numbered F2, there are a total of 1335 regional pixels, of which the maximum temperature is 75.1℃, and its corresponding coordinate point is (312, 239). The hot spot center point is calibrated in the above manner, and the images F1 to F5 are processed in sequence. The number of boundary pixels, the highest temperature value and the corresponding center coordinate position in each frame image are summarized to establish a hot spot center point record table, as shown in Table 7. Through this table, the hot spot spatial position and the maximum value of the thermal parameter corresponding to each frame image can be clearly determined, and the hot spot center point coordinate sequence can be obtained.

[0104] Table 7 Hot spot center point extraction data table

[0105] ;

[0106] As shown in Table 7, the hot spot area of each frame image has a slight position drift in the time series, the temperature change range is ±2.3℃, and the spatial position is mainly concentrated in the central area.

[0107] The coordinate recording submodule is based on the hot spot center point coordinate sequence, extracts the hot spot center point coordinate values between two consecutive frames of images in image time sequence, calculates the difference between the horizontal and vertical coordinates of the hot spot center of adjacent frames (i.e., two consecutive frames of images), records the translation component of the hot spot coordinates between frames, extracts the infrared temperature values of the pixels corresponding to the hot spot center points of each frame, establishes a joint record set of horizontal and vertical coordinate changes, hot spot area changes and temperature changes, and obtains the change in the hot spot path across frames.

[0108] Specifically, based on the coordinate sequence of the hot spot center point, the horizontal and vertical coordinate values of the center point of each frame image are paired in chronological order, and the inter-frame horizontal coordinate difference Δx and vertical coordinate difference Δy between adjacent frames are extracted and recorded as the spatial translation change value of each frame respectively. At the same time, the pixel number change value Δs in the hot spot area is counted, that is, the pixels in the hot spot boundary area in each frame image are counted, and the difference in the number of boundary pixels between adjacent frames is calculated to reflect the shrinkage or expansion trend of the hot spot area. At the same time, the infrared temperature value T corresponding to the pixel of the hot spot center point of each frame is extracted. , calculate the temperature change value ΔT between frames, and express the degree of temperature increase or decrease in numerical form. As shown in Table 7, the temperature change value from F2 to F3 is 72.8-75.1=-2.3℃, indicating a downward trend in the hot spot temperature. The spatial drift is the horizontal difference of -3 pixels and the vertical difference of +2 pixels. The area change value is 1342-1335=+7 pixels. A hot spot cross-frame path change set containing the quaternion structure of Δx, Δy, Δs, and ΔT is constructed to form a basic data set of the inter-frame change path, and the hot spot cross-frame path change amount is obtained.

[0109] The path generation submodule integrates the horizontal coordinate difference, vertical coordinate difference, hot spot area change value and temperature value difference between all frames according to the cross-frame path change of the hot spot, arranges them in sequence according to the frame number, and connects the coordinate points of the hot spot center point in sequence to form a continuous cross-frame path line of the hot spot center point, thereby obtaining the cross-frame path distribution structure of the hot spot.

[0110] Specifically, according to the path change of the hot spot across frames, a hot spot trajectory sequence is constructed with the image frame number as the index, and the coordinate displacement data Δx and Δy of the center point between each frame are accumulated and plotted on a two-dimensional coordinate system to form the horizontal and vertical trajectory distribution of the hot spot in the image sequence. The center points of each frame are connected with a straight line in chronological order to construct a continuous trajectory segment, and the temperature value and hot spot area are marked at each trajectory node to judge the trend of trajectory change. If the direction of the horizontal displacement difference of the center point of three consecutive frames is consistent and the difference is greater than 2 pixels, the trajectory segment is marked as a continuous offset interval. If the absolute value of the temperature change between frames exceeds 2°C, it is marked as a temperature jump interval. The path structure is divided according to the above classification rules to form a hot spot cross-frame trajectory line generated by a combination of spatial position, time sequence and thermal characteristics. A complete two-dimensional trajectory structure is established to obtain the hot spot cross-frame path distribution structure.

[0111] See also Figure 5 ,The abnormal dynamic screening module includes : a displacement calculation submodule, an anomaly identification submodule and an information labeling submodule.

[0112] The displacement calculation submodule calculates the spatial displacement value of the hot spot between corresponding frames based on the center point coordinate data of all consecutive frames in the hot spot cross-frame path distribution structure, organizes all the hot spot spatial displacement values between frames into an array in the order of frame numbers, and calculates the average displacement length per unit frame of the path to obtain the average displacement length per unit frame data.

[0113] Specifically, call the center point coordinate sequence in the hot spot cross-frame path distribution structure, extract the center point coordinates in each pair of consecutive frame numbers, obtain the horizontal difference and vertical difference respectively, and use the coordinate difference to calculate the displacement distance of the hot spot between the corresponding frames, store all inter-frame displacement results in the structure list, and then sum all the displacement values in the list, and then divide them by the number of valid frame segments, that is, the total number of frames minus one, to obtain the average displacement length per unit frame, in pixels. For example, assuming the frame sequence is F1 to F5, with a total of 4 displacement segments, namely 2.236 pixels from F1 to F2, 3.606 pixels from F2 to F3, 2.236 pixels from F3 to F4, and 2.236 pixels from F4 to F5, the total displacement is 10.314 pixels. There are 4 valid frame segments, so the average displacement length per frame is calculated to be 10.314 ÷ 4 = 2.578 pixels. The Euclidean distance between the coordinates of two points is used as the inter-frame displacement indicator in the calculation. All differences must be rounded to three decimal places. Abnormal error points should be checked and excluded or reviewed in the final average result. The average displacement per frame in Table 8 is calculated based on Table 7. As shown in Table 8, the horizontal difference and vertical difference between each frame and the calculated inter-frame displacement value are clearly listed in the table to obtain the average displacement length data per frame.

[0114] Table 8 Unit frame average displacement calculation table

[0115] ;

[0116] As shown in Table 8, the displacement values are within the normal displacement range, with a maximum value of 3.606 pixels and an average value of 2.578 pixels.

[0117] The anomaly identification submodule reads the displacement values between each frame in the hot spot path based on the average displacement length data of the unit frame, performs a difference operation between each value and the average value, and determines whether it exceeds twice the average threshold. If it exceeds, the corresponding frame number is marked as an abnormal node, and all marked frames are searched frame by frame. All abnormal marking results are accumulated to obtain a list of hot spot abnormal motion frame numbers.

[0118] Specifically, based on the average displacement length data of the unit frame, all inter-frame displacement values are obtained and compared one by one with the average value. A threshold screening operation is performed on each displacement value, and the judgment condition is set to whether it exceeds twice the average value, that is, exceeds 5.156 pixels. If it exceeds twice the average value, the subsequent frame number corresponding to the segment is marked as an abnormal frame, and the abnormal frame is marked as an abnormal frame, and an abnormal frame number array is generated. At the same time, a binding relationship is established between the frame number and its corresponding inter-frame displacement value. During the screening process, if the displacement value of a frame is exactly equal to 5.156 pixels, it is classified as a normal frame to avoid misjudgment. For example, in frame F2-F3, the displacement is 3.606 pixels, which is below the threshold and does not belong to an abnormal frame. However, if a segment with an inter-frame displacement value of 6.5 pixels appears, the frame number of the subsequent image is determined to be an abnormal frame. All abnormal frame results are sorted and formed into a hot spot abnormal node sequence, resulting in a list of hot spot abnormal motion frame numbers.

[0119] The information annotation submodule extracts the hot spot boundary information and center point coordinate data in the corresponding frame image based on the hot spot abnormal motion frame number list, generates a square frame coverage area for the abnormal frame image, synchronously reads the frame index information and draws the frame number and abnormal identification label text in the upper left corner of the image, generates a hot map annotation output image for each abnormal frame, and obtains the annotation set of the hot spot abnormal motion frame.

[0120] Specifically, according to the list of hot spot abnormal motion frame numbers, the image coordinate reading operation is performed on each abnormal frame image, and the coordinate position of the calibrated hot spot center point and the hot spot area boundary coordinates in the frame image are called. According to the ISO18436-6:2014 standard mid-infrared image annotation specification, the abnormal point is marked in the form of a square annotation box. The size of the annotation box extends 20 pixels outward from the center point, that is, the side length is 40 pixels, and the frame number and "abnormal" identification text are added to the upper left corner of the image. All annotation content is overlaid on the original image data in a layer manner, and an annotated image set is output. An index structure is established, and the image number, annotation center point coordinates, annotation range size and corresponding inter-frame displacement value are recorded as structure fields, which are combined into an abnormal frame annotation information set to obtain a hot spot abnormal motion frame annotation set.

[0121] See also Figure 6 ,The fault hot spot confirmation module includes : a boundary measurement submodule, a growth screening submodule, and a fault identification submodule.

[0122] The boundary measurement submodule obtains the closed pixel set composed of the hot spot boundary of each image frame based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, accumulates the number of all pixels in the hot spot boundary area of each frame image, counts the hot spot boundary area value of the current frame, constructs a time series in the order of frame numbers, and obtains the hot spot boundary area sequence.

[0123] Specifically, based on the annotation set of hot spot abnormal motion frames and the distribution results of hot spot boundary areas, the pixel set of the hot spot boundary area is extracted frame by frame, the number of pixels in the boundary area of each frame is accumulated and the boundary area value is generated in pixels. A time series structure is constructed, and the image frame numbers must be strictly matched to avoid frame breaks or crossovers between frames. The boundary area array structure is established in numbered order, and the maximum temperature value and center coordinate point of the hot spot area corresponding to each frame are synchronously extracted to construct a comprehensive measurement data structure. In the actual implementation process, combined with the image size of the infrared thermal image being 640×480, the conventional size of the boundary area is generally between 1000 and 2000 pixels. The following parameters are extracted from image frames F1 to F5 respectively: the boundary area of F1 is 1200 pixels, the maximum temperature is 72.5°C, and the center coordinates are (310, 240); the boundary area of F2 is 1220 pixels, the maximum temperature is 73.8°C, and the center coordinates are (312, 241); the boundary area of F3 is 1450 pixels, the maximum temperature is 76.9°C, and the center coordinates are (313, 243); the boundary area of F4 is 1630 pixels, the maximum temperature is 79.5°C, and the center coordinates are (315, 245); the boundary area of F5 is 1695 pixels, the maximum temperature is 80.7°C, and the center coordinates are (317, 247). The data results are shown in Table 9 below.

[0124] Table 9 Hot spot boundary area and temperature data

[0125] ;

[0126] As shown in Table 9, both the area and temperature have an increasing trend, which provides a data basis for the subsequent joint screening of growth rate and temperature difference to obtain the hot spot boundary area sequence.

[0127] The growth screening submodule reads the boundary area values of all consecutive frame images based on the hot spot boundary area sequence, calculates the growth rate of the hot spot boundary between adjacent frames using the standard hot spot boundary growth rate calculation formula (2), and determines the growth rate value of each frame one by one. Whether it exceeds the hot spot expansion rate threshold of 15% specified in the standard, and at the same time obtain the hot spot maximum temperature difference ΔT between the current frame and the previous frame to determine whether it exceeds the temperature gradient threshold. If both conditions are met, that is If ΔT ≥ 3°C, the current frame number is recorded as a candidate frame, and all frame numbers that meet the screening conditions constitute a candidate set, and a fault candidate frame number sequence is obtained.

[0128] (2);

[0129] in, Indicates the current frame boundary growth rate, and They represent the boundary area values of the hot spots in the current frame and the previous frame respectively, and ΔT represents the difference between the maximum temperature values of the hot spots in adjacent frames.

[0130] Specifically, based on the constructed hot spot boundary area sequence, the area value of each pair of consecutive frames is called, and the boundary growth rate is calculated frame by frame according to formula (2). The following calculations are performed respectively: the frame F2 growth rate is , the frame F3 growth rate is , the frame F4 growth rate is , the frame F5 growth rate is At the same time, the temperature difference is extracted: the temperature difference between F2 and F1 is 1.3℃, the temperature difference between F3 and F2 is 3.1℃, the temperature difference between F4 and F3 is 2.6℃, and the temperature difference between F5 and F4 is 1.2℃. Double threshold screening is performed, and the area growth rate threshold is set to 15%, and the temperature difference threshold is 2.5℃. In the judgment, only frame F3 meets the conditions that the growth rate is greater than 15% and the temperature difference is greater than 2.5℃. Other frames do not meet one or both of these conditions. Therefore, only frame F3 is marked as a candidate frame, and a unit set is constructed to obtain the fault candidate frame number sequence.

[0131] The fault identification submodule reads the hot spot center coordinate change trend and hot spot boundary structure of the corresponding frame image according to the fault candidate frame number sequence, detects the continuous change characteristics of the hot spot area in the spatial structure of the candidate frame, and determines whether the conditions of the boundary expansion direction and the hot spot temperature rise direction are simultaneously met. If the hot spot area contour deformation direction and the hot spot temperature rise center point drift direction in two consecutive frames are spatially consistent, the candidate frame is confirmed as a faulty hot spot area, and the photovoltaic module fault identification result is obtained.

[0132] Specifically, based on the obtained fault candidate frame number sequence, the hot spot center coordinate value and boundary geometric center coordinate data of the candidate frame image are further extracted, and a two-dimensional vector model of the center point drift direction and boundary change direction between adjacent frames is constructed. The direction consistency judgment is performed by calculating the vector cosine angle shown in formula (3).

[0133] (3).

[0134] in, is the current frame coordinate value, are the coordinate values of adjacent frames.

[0135] Perform a sample calculation: the center coordinates of frame F3 are (313, 243), the center coordinates of frame F2 are (312, 241), and the drift vector is (1, 2). If the drift vector of the boundary geometric center of frame F3 compared to frame F2 is (2, 3), then substitute into the formula:

[0136] ;

[0137] Calculate the angle , the judgment condition θ≤30° is met, and the judgment direction is consistent. Frame F3 is confirmed as the faulty hot spot frame. Its frame number, temperature, area, and coordinate information are recorded, and the structural data is output to obtain the PV module fault identification result. This result shows that fault determination requires the three combined conditions of high boundary expansion rate, sharp temperature increase, and clear expansion direction.

[0138] This embodiment also proposes a photovoltaic module fault detection method based on infrared image hot spot detection. Based on the above-mentioned photovoltaic module fault detection system based on infrared image hot spot detection, the method includes the following steps:

[0139] S1: Based on the infrared temperature data in the infrared image of the photovoltaic module, the pixel pairs with temperature differences greater than the temperature gradient threshold are screened, the positions of the pixel pairs are recorded, and the pixel temperature differences between each two pixel pairs are continuously calculated to screen the connected areas of the hot spot boundaries and generate the hot spot boundary area distribution results;

[0140] S2: Based on the infrared images of the photovoltaic modules at multiple monitoring moments, the positions of the four corner points of the photovoltaic module frame in different frames are obtained, and the average inter-frame corner point displacement value is calculated to complete the coordinate alignment and obtain the multi-time series module image registration result;

[0141] S3: Based on the hot spot boundary area distribution results and the multi-time component image registration results, the hot spot center point position within the hot spot boundary area distribution results is extracted, the inter-frame coordinates of the hot spot center point are recorded, and the displacement path is calculated based on the coordinate change between every two center points in adjacent frames to generate the hot spot cross-frame path distribution structure;

[0142] S4: Based on the distribution structure of the hot spot path across frames, the displacements between all frames in the path are compared one by one, the average displacement length of the hot spot per frame in the entire path is calculated, abnormal nodes are screened and marked, and a set of annotations of hot spot abnormal motion frames is obtained;

[0143] S5: Based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, the area of the hot spot boundary structure is measured and the hot spot boundary growth rate is calculated. The hot spot area with a hot spot boundary growth rate higher than the hot spot expansion rate threshold in two consecutive frames and a temperature difference higher than the temperature gradient threshold is screened to confirm it as a faulty hot spot, and the photovoltaic module fault identification result is output.

[0144] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0145] This embodiment further provides a computer-readable storage medium having a computer program / instruction stored thereon, which implements the steps of the method when the computer program / instruction is executed by a processor.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A photovoltaic module fault detection system based on infrared image hot spot detection, characterized by: include: Temperature difference boundary extraction module: It is used to screen pixel pairs with temperature differences greater than the temperature gradient threshold based on the infrared temperature data in the infrared image of the photovoltaic module, record the positions of the pixel pairs, and continuously calculate the pixel temperature differences between each two pixel pairs, screen the connected areas of the hot spot boundaries, and generate the hot spot boundary area distribution results; Image sequence registration module: used to obtain the positions of the four corner points of the photovoltaic module frame in different frames based on the infrared images of the photovoltaic module at multiple monitoring times, and calculate the average inter-frame corner point displacement value to complete the coordinate alignment and obtain the multi-time series component image registration results; Hot spot trajectory identification module: This module is used to extract the hot spot center point position within the hot spot boundary area distribution results and the multi-time component image registration results, record the inter-frame coordinates of the hot spot center point, calculate the displacement path based on the coordinate change between every two center points in adjacent frames, and generate the hot spot cross-frame path distribution structure; Abnormal dynamic screening module: This module is used to compare the displacements of all frames in the path one by one based on the distribution structure of the hot spot cross-frame path, calculate the average displacement length of the hot spot per frame in the entire path, screen abnormal nodes and mark them, and obtain the annotation set of the hot spot abnormal motion frames; Fault hot spot confirmation module: It is used to measure the area of the hot spot boundary structure and calculate the hot spot boundary growth rate based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results. It screens the hot spot areas where the hot spot boundary growth rate in two consecutive frames is higher than the hot spot expansion rate threshold and the temperature difference is higher than the temperature gradient threshold, confirms them as faulty hot spots, and outputs the photovoltaic module fault identification results.

2. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 1, characterized in that: The temperature difference boundary extraction module includes: Temperature difference calculation submodule: The infrared image is divided into N 3×3 pixel blocks. Based on the infrared temperature data of each pixel in the infrared image, the infrared temperature values of the central pixel of each pixel block and its eight neighboring pixels are read. The temperature difference of each pair of central and neighboring pixels is calculated. The obtained temperature difference is compared with the temperature gradient threshold. The pixel pairs with temperature differences greater than the temperature gradient threshold are screened to obtain the number of high-gradient pixel pairs. Continuous pixel segment screening submodule: Based on the number of high-gradient pixel pairs, the module traverses the image row by row, determines the spatial positional relationship between any two high-gradient pixel pairs in the horizontal, vertical, and diagonal directions, performs directional continuity judgment on adjacent high-gradient pixel pairs, and counts the lengths of pixel segments that meet the continuity conditions. It then screens for connected area segments whose length is not less than the minimum recognizable hot spot size to obtain directional pixel connected segment length data. Hot spot boundary identification submodule: Based on the directional pixel connected segment length data, all pixel segments that meet the minimum identifiable hot spot size are regionally fused, the corresponding boundary contour information is read and boundary range statistics are performed, the distribution position of the continuous boundary area in the image is determined, and the hot spot boundary area distribution result is obtained.

3. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 1, characterized in that: The image sequence registration module includes: Image acquisition submodule: Based on the hot spot boundary area distribution results, it obtains infrared images of photovoltaic modules at multiple monitoring times, numbers each frame of the image and records the timestamp information, sorts the images by time sequence number, counts the boundary coordinates of the photovoltaic module area in each frame of the image, establishes an image sequence index table, and obtains image sequence index data; Corner detection submodule: Extracts the component boundary area in each frame of the image based on the image sequence index data, combines the grayscale gradient change values of the four vertex areas in the boundary contour, and calculates the coordinate position of the local gradient maximum point in the area in each frame of the image in turn, obtains the four corner point candidate coordinate set, and generates the image frame corner point coordinate data; Coordinate alignment submodule: Based on the image frame corner coordinate data, the average inter-frame corner point displacement value is calculated according to the coordinate difference of the corresponding four corner points in two adjacent frames of images, and inverse pixel offset processing is performed on each frame of image. That is, the coordinate values of all pixels in the current image are combined with the corresponding displacement values for addition and subtraction compensation, and the image is subjected to sub-pixel coordinate resampling and interpolation correction to obtain the multi-time series component image registration result.

4. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 3 is characterized by: The calculation formula of the average inter-frame corner displacement value is as follows: ； in, Indicates the Image at time The average inter-frame corner displacement value, 、 Respectively represent the image Corner point at time The horizontal and vertical coordinates of 、 Indicates the coordinate position of the corresponding corner point at the previous moment, and 4 is the number of corner points in each image.

5. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 1, characterized in that: The hot spot trajectory recognition module includes: Hot spot extraction submodule: This module is used to traverse each frame of the image sequence based on the multi-time component image registration results, extract all pixel sets within the hot spot boundary area distribution results, determine the pixel coordinates corresponding to the maximum value and use it as the hot spot center point position of the current image frame, record the row and column index coordinates of the hot spot center point, and obtain the hot spot center point coordinate sequence; Coordinate recording submodule: Based on the hot spot center point coordinate sequence, the hot spot center point coordinate values between two consecutive frames are extracted in image time order. The difference between the horizontal and vertical coordinates of the hot spot center in adjacent frames is calculated, and the translation component of the hot spot coordinates between frames is recorded. The infrared temperature value of the pixel corresponding to the hot spot center point in each frame is extracted. A joint record set of horizontal and vertical coordinate changes, hot spot area changes, and temperature changes is established to obtain the hot spot path change across frames. Path generation submodule: According to the cross-frame path change of the hot spot, the horizontal coordinate difference, vertical coordinate difference, hot spot area change value and temperature value difference between all frames are integrated, and the frames are arranged in sequence according to the frame number. The coordinate points of the hot spot center point are connected in sequence to form a cross-frame continuous path line of the hot spot center point, and the cross-frame path distribution structure of the hot spot is obtained.

6. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 1, characterized in that: The abnormal dynamic screening module includes: Displacement calculation submodule: Calculate the spatial displacement of hot spots between corresponding frames based on the center point coordinate data of all consecutive frames in the hot spot cross-frame path distribution structure, organize all the hot spot spatial displacement values between frames into an array in the order of frame numbers, and calculate the average displacement length per unit frame of the path to obtain the average displacement length per unit frame data; Abnormal identification submodule: Based on the average displacement length of the unit frame, the inter-frame displacement value of each segment in the hot spot path is read, and the difference between each segment value and the average value is calculated to determine whether it exceeds twice the average threshold. If it exceeds, the corresponding frame number is marked as an abnormal node. All marked frames are searched frame by frame, and all abnormal marking results are accumulated to obtain a list of hot spot abnormal motion frame numbers; Information annotation submodule: According to the hot spot abnormal motion frame number list, the hot spot boundary information and center point coordinate data in the corresponding frame image are extracted, a square frame coverage area is generated for the abnormal frame image, the frame index information is synchronously read and the frame number and abnormal identification label text are drawn in the upper left corner of the image, a hot map annotation output image is generated for each abnormal frame, and a hot spot abnormal motion frame annotation set is obtained.

7. The photovoltaic module fault detection system based on infrared image hot spot detection according to claim 1, characterized in that: The fault hot spot confirmation module includes: Boundary measurement submodule: Based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, the closed pixel set formed by the hot spot boundary of each image frame is obtained. The number of all pixels in the hot spot boundary area of each frame is accumulated, and the hot spot boundary area value of the current frame is counted. The time series is constructed in the order of frame numbers to obtain the hot spot boundary area sequence; Growth screening submodule: Based on the hot spot boundary area sequence, the boundary area values of all consecutive frame images are read. The growth rate of the hot spot boundary between adjacent frames is calculated using the standard hot spot boundary growth rate calculation formula. Each frame's growth rate value is judged one by one to see if it exceeds the hot spot expansion rate threshold specified by the standard. The hot spot maximum temperature difference between the current frame and the previous frame is also obtained to determine whether it exceeds the temperature gradient threshold. If both conditions are met, the current frame number is recorded as a candidate frame. All frame numbers that meet the screening conditions constitute a candidate set, and a sequence of fault candidate frame numbers is obtained. Fault identification submodule: According to the fault candidate frame number sequence, read the hot spot center point coordinate change trend and hot spot boundary structure of the corresponding frame image, detect the continuous change characteristics of the hot spot area in the spatial structure of the candidate frame, and judge whether the boundary expansion direction and the hot spot temperature rise direction are consistent. If the hot spot area contour deformation direction and the hot spot temperature rise center point drift direction in two consecutive frames are spatially consistent, the candidate frame is confirmed as a fault hot spot area, and the photovoltaic module fault identification result is obtained.

8. A photovoltaic module fault detection method based on infrared image hot spot detection, characterized by: Based on the photovoltaic module fault detection system based on infrared image hot spot detection according to any one of claims 1 to 7, the method comprises the following steps: S1: Based on the infrared temperature data in the infrared image of the photovoltaic module, the pixel pairs with temperature differences greater than the temperature gradient threshold are screened, the positions of the pixel pairs are recorded, and the pixel temperature differences between each two pixel pairs are continuously calculated to screen the connected areas of the hot spot boundaries and generate the hot spot boundary area distribution results; S2: Based on the infrared images of the photovoltaic modules at multiple monitoring moments, the positions of the four corner points of the photovoltaic module frame in different frames are obtained, and the average inter-frame corner point displacement value is calculated to complete the coordinate alignment and obtain the multi-time series module image registration result; S3: Based on the hot spot boundary area distribution results and the multi-time component image registration results, the hot spot center point position within the hot spot boundary area distribution results is extracted, the inter-frame coordinates of the hot spot center point are recorded, and the displacement path is calculated based on the coordinate change between every two center points in adjacent frames to generate the hot spot cross-frame path distribution structure; S4: Based on the distribution structure of the hot spot path across frames, the displacements between all frames in the path are compared one by one, the average displacement length of the hot spot per frame in the entire path is calculated, abnormal nodes are screened and marked, and a set of annotations of hot spot abnormal motion frames is obtained; S5: Based on the hot spot abnormal motion frame annotation set and the hot spot boundary area distribution results, the area of the hot spot boundary structure is measured and the hot spot boundary growth rate is calculated. The hot spot area with a hot spot boundary growth rate higher than the hot spot expansion rate threshold in two consecutive frames and a temperature difference higher than the temperature gradient threshold is screened to confirm it as a faulty hot spot, and the photovoltaic module fault identification result is output.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to claim 8.

10. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to claim 8 are implemented.

Images