A photovoltaic box transformer type switch cabinet temperature field three-dimensional reconstruction method

Abstract

This application provides a method for three-dimensional reconstruction of the temperature field of a photovoltaic transformer switchgear, comprising: acquiring first infrared temperature images of the switchgear from multiple angles; preprocessing the images to enhance edge information and improve contrast, resulting in a preprocessed infrared temperature image; distinguishing target and background regions through image segmentation, and extracting temperature field feature points based on the target region; constructing an initial population based on the texture information of the preprocessed image, and building a three-dimensional geometric model of the switchgear using finite element method and selecting a mesh type; conducting thermal simulation on the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part; matching the temperature field feature points with the three-dimensional geometric model and the theoretical temperature distribution field through feature matching; and completing the three-dimensional reconstruction of the switchgear temperature field based on the matching results. This method can achieve high-precision three-dimensional visualization reconstruction of the switchgear temperature field, providing reliable data support for equipment thermal status monitoring and fault early warning.

Description

Technical Field

[0001] This application relates to the field of photovoltaic equipment testing and 3D modeling technology, specifically to a 3D reconstruction method for the temperature field of a photovoltaic box-type switchgear. Background Technology

[0002] Photovoltaic transformer substations are the core power distribution equipment in photovoltaic power supply systems, responsible for the transmission and distribution of electrical energy and overload and short-circuit protection. Their operational stability directly determines the safe and efficient operation of the photovoltaic power supply system. Because photovoltaic systems operate under complex outdoor conditions for extended periods, and components such as busbars, circuit breakers, and contactors within the substations must withstand high currents for extended periods, they are prone to localized overheating due to increased contact resistance, insulation aging, and poor heat dissipation. Failure to accurately capture the temperature field distribution and its changing trends in a timely manner can lead to component burnout, substation malfunctions, and even system shutdowns, resulting in significant economic losses. Therefore, achieving accurate monitoring and three-dimensional reconstruction of the temperature field in photovoltaic transformer substations has significant engineering value for equipment fault early warning and operation and maintenance optimization.

[0003] Currently, the mainstream technology for switchgear temperature monitoring is infrared thermometry. This involves using infrared sensors to collect infrared temperature images of the equipment surface, obtaining two-dimensional temperature distribution data. However, this method only reflects the temperature differences on the equipment surface and cannot present the temperature field distribution characteristics of internal components and three-dimensional space. It is difficult to accurately locate the three-dimensional position and diffusion path of localized overheating areas, thus failing to meet the needs of refined operation and maintenance. To address this issue, scholars both domestically and internationally have conducted extensive research on three-dimensional temperature field reconstruction. However, these studies generally suffer from low accuracy in matching the three-dimensional reconstruction model with the actual equipment structure and temperature distribution, and insufficient temperature characterization precision, making it difficult to meet the practical needs of equipment fault early warning and refined operation and maintenance. Summary of the Invention

[0004] This application aims to address the technical problem that the 3D reconstruction model of photovoltaic box-type switchgear has low conformity with the actual structure and temperature distribution of the equipment, and insufficient temperature characterization accuracy, making it difficult to meet the actual needs of equipment fault early warning and refined operation and maintenance.

[0005] To address the aforementioned technical problems, this application provides a method for three-dimensional reconstruction of the temperature field of a photovoltaic transformer switchgear, comprising: acquiring a first infrared temperature image of the photovoltaic transformer switchgear from multiple angles; performing image preprocessing on the first infrared temperature image to enhance image edge information and improve image contrast, thereby obtaining a preprocessed infrared temperature image; segmenting the preprocessed infrared temperature image using image segmentation technology to distinguish between target and background regions, and extracting temperature field feature points representing temperature distribution based on the target region; constructing an initial population based on the texture information of the preprocessed infrared temperature image, and constructing a three-dimensional geometric model of the photovoltaic transformer switchgear using finite element method and selecting a mesh type; performing thermal simulation on the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model; matching the temperature field feature points with the three-dimensional geometric model and the corresponding theoretical temperature distribution field using a feature matching method; and completing the three-dimensional reconstruction of the temperature field of the photovoltaic transformer switchgear based on the matching results.

[0006] Furthermore, the step of performing image preprocessing on the first infrared temperature image to enhance image edge information and improve image contrast to obtain a preprocessed infrared temperature image includes: performing multi-scale analysis on the first infrared temperature image, using filtering to enhance image edge information to obtain an intermediate infrared temperature image; decomposing the intermediate infrared temperature image into an illumination image and a reflection image, removing the reflection image, and then performing contrast stretching on the illumination image to obtain the preprocessed infrared temperature image.

[0007] Furthermore, the step of enhancing image edge information through filtering includes: employing a bilateral filter, controlling the spatial distance between pixels through a spatial Gaussian function, and retaining image edge information by combining a grayscale Gaussian function; the step of contrast stretching the illumination image includes: performing contrast stretching on the illumination image using histogram equalization technology.

[0008] Furthermore, the step of segmenting the preprocessed infrared temperature image using image segmentation technology to distinguish the target region from the background region includes: using the Tsallis entropy algorithm to extract the gray values ​​of the preprocessed infrared temperature image and constructing a two-dimensional histogram of the neighborhood average gray values ​​of the infrared temperature image. The entropy value of each pixel in the preprocessed infrared temperature image is calculated based on the two-dimensional histogram, and the grayscale segmentation threshold is determined according to the entropy value to segment the target area and the background area.

[0009] Furthermore, the step of extracting temperature field feature points representing temperature distribution based on the target region includes: obtaining a target region mask based on the segmented target region, and extracting the temperature field feature points within the corresponding region of the preprocessed infrared temperature image.

[0010] Furthermore, the construction of the initial population based on the texture information of the preprocessed infrared temperature image includes: constructing the initial population based on texture blocks containing edge and temperature correlation information in the preprocessed infrared temperature image, wherein the texture blocks are image units characterizing the local structure and temperature correlation features of the photovoltaic box transformer switch cabinet.

[0011] Furthermore, the mesh type is a hexahedral mesh, which is suitable for the computational requirements of finite element thermal simulation and improves the model's computational accuracy.

[0012] Furthermore, the step of performing thermal simulation on the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model includes: performing thermal simulation on the three-dimensional geometric model in combination with the energy balance equation to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model, wherein the energy balance equation is used to characterize the correlation between the heat flux rate and temperature of each node of the three-dimensional geometric model.

[0013] Furthermore, the step of using a feature matching method to match the temperature field feature points with the three-dimensional geometric model and the corresponding theoretical temperature distribution field includes: using a wavelet transform method to establish the correspondence between the temperature field feature points and various parts of the three-dimensional geometric model based on the theoretical temperature distribution field of the three-dimensional geometric model.

[0014] Furthermore, the temperature field feature points include at least the pixel points corresponding to the key heat-generating components of the photovoltaic box transformer switchgear and the pixel points corresponding to the connection points of the key heat-generating components. The key heat-generating components include components that are prone to local overheating, such as busbars, circuit breakers, and contactors, thereby improving the targeting and accuracy of temperature field reconstruction.

[0015] This application has the following beneficial effects: 1. This invention addresses the problems of excessive noise, blurred edges, and unclear temperature levels in the original infrared image by performing targeted preprocessing on the infrared temperature image, using bilateral filtering to achieve both noise reduction and edge enhancement, and removing redundant information such as reflected images, and then improving contrast through histogram equalization.

[0016] 2. The Tsallis entropy algorithm combined with a two-dimensional histogram of the average gray value of the neighborhood is used for image segmentation. Compared with traditional segmentation methods, it can more accurately capture the spatial gray-level distribution features of the image, achieve accurate separation of the target area and the background area of ​​the switch cabinet, and extract temperature field feature points by combining the target area mask, avoiding interference from redundant background information and improving the accuracy of feature point extraction.

[0017] 3. An initial population is constructed based on the preprocessed infrared image texture information. A three-dimensional geometric model is constructed by combining finite element technology and hexahedral mesh. This model not only fits the actual structure of the switch cabinet, but also establishes an early correlation with temperature information. This solves the problem of the disconnect between traditional drawing-based modeling and actual equipment, and lays the foundation for temperature field mapping.

[0018] 4. Thermal simulation is performed by combining the energy balance equation. Wavelet transform is used to achieve accurate matching between temperature field feature points and three-dimensional geometric models. The final constructed three-dimensional temperature field model can intuitively reflect the temperature distribution and spatial correlation of various parts of the switchgear. It can accurately locate the three-dimensional position of local overheated areas, providing reliable technical support for the status monitoring, fault early warning and maintenance of photovoltaic box-type switchgear. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a three-dimensional reconstruction method for the temperature field of a photovoltaic box-type switchgear provided in this application. Figure 2 This is a two-dimensional histogram of the neighborhood average gray value constructed in this application; Figure 3 This is a three-dimensional geometric model of the photovoltaic box-type switchgear constructed in this application; Figure 4 This is a comparison chart of temperature values ​​at 100 detection nodes between the method of this application and two control group methods. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0022] The method for three-dimensional reconstruction of the temperature field of a photovoltaic transformer switchgear provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios. This method 100 can be executed by a terminal device. In other words, the method can be executed by software or hardware installed on the terminal device, such as... Figure 1 As shown, the method may include the following steps.

[0023] S110: Acquires first infrared temperature images of photovoltaic box-type switchgear from multiple angles.

[0024] In this step, a high-precision infrared image sensor is used to collect data from multiple angles around the photovoltaic box-type switchgear. The collection range covers the front, sides, top, and bottom of the switchgear. It can be selected to collect data around the switchgear from 0° to 360°, or at least three different positions that do not obstruct each other can be selected for collection. This ensures that the first infrared temperature image collected is unobstructed and the temperature information is complete. It can fully cover the overall outline of the switchgear and key heat-generating components, such as busbars, circuit breakers, and contactors.

[0025] S120: Perform image preprocessing on the first infrared temperature image to enhance image edge information and improve image contrast, thereby obtaining a preprocessed infrared temperature image.

[0026] The first infrared temperature images acquired in step S110 are preprocessed one by one. Considering the multi-directional image acquisition characteristics in S110, multiple images acquired from the same direction are first merged into a single complete image using image stitching technology. This results in a comprehensive first infrared temperature image covering the front, sides, top, and bottom of the switchgear, avoiding image fragmentation from multiple directions. Then, a targeted preprocessing process is performed on each complete image to enhance image edge information and improve image contrast, resulting in a preprocessed infrared temperature image. Image edge information refers to the information presented by areas in the image where grayscale values ​​change abruptly. This corresponds to the cabinet outline, internal component boundaries, and key heat-generating components of the photovoltaic transformer switchgear, such as the outer boundaries of busbars, circuit breakers, or contactors. It is a core image feature for distinguishing equipment from the background and different components of the equipment. Image contrast refers to the brightness difference between bright and dark areas in the image, corresponding to the grayscale differences in different temperature regions of the switchgear. Higher contrast makes the distinction between high-temperature and low-temperature regions more obvious, facilitating the subsequent extraction of temperature-related features.

[0027] S130: Image segmentation technology is used to segment the preprocessed infrared temperature image, distinguishing the target area from the background area, and extracting temperature field feature points representing the temperature distribution based on the target area.

[0028] Image segmentation technology is used to segment each preprocessed infrared temperature image obtained in step S120. A reasonable grayscale segmentation threshold is set to divide the image into a target region (the image region corresponding to the photovoltaic box transformer switch cabinet body and internal components) and a background region (irrelevant areas such as the ground, support, and surrounding environment). Based on the segmented target region, temperature field feature points representing temperature distribution are extracted within the target region. A certain number of feature points can be extracted for each target region according to requirements. The focus is on selecting feature points in areas with obvious temperature gradient changes and on the surface of components. The temperature field feature point set is then compiled for subsequent matching.

[0029] In this application, the temperature field feature points include at least the pixel points corresponding to the key heat-generating components of the photovoltaic box transformer switchgear and the pixel points corresponding to the connection points of the key heat-generating components. For example, the pixel points corresponding to the key heat-generating components such as the busbar body, circuit breaker contacts, contactor terminals, and conductive parts of the disconnecting switch inside the switchgear, as well as the pixel points corresponding to the connection positions of key heat-generating components such as busbar overlap, circuit breaker and busbar connection, contactor and cable connection, and cable joint and switchgear terminal block connection.

[0030] S140: Based on the texture information of the preprocessed infrared temperature image, an initial population is constructed. Using finite element technology and selecting the mesh type, a three-dimensional geometric model of the photovoltaic box-type switchgear is constructed.

[0031] The texture information of the preprocessed infrared temperature image in step S120 is extracted, and each image is divided into multiple texture blocks containing edge and temperature correlation information. An initial population is constructed based on these texture blocks to ensure that the initial population covers the structural and temperature characteristics of the switchgear in different directions. Finite element analysis technology is used to select a mesh type that is suitable for the switchgear structure, and the initial population is meshed and structure-fitted. Combined with the standard structural parameters of the photovoltaic box transformer switchgear, a three-dimensional geometric model that matches the actual switchgear 1:1 is built. The model accurately restores the cabinet shape, internal component installation position and overall structural dimensions of the switchgear, providing a geometric carrier for temperature field mapping.

[0032] S150: Perform thermal simulation on the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model.

[0033] Import the three-dimensional geometric model constructed in step S140 into the finite element thermal simulation software, set simulation parameters that fit the actual operating conditions of the photovoltaic power station, the ambient temperature can be set to 25℃, the atmospheric pressure to standard atmospheric pressure, and the material parameters of the switchgear shell and internal components are entered according to the actual selected materials; perform thermal simulation calculations on the three-dimensional geometric model to simulate the temperature conduction process of the switchgear under rated load, obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model, clarify the temperature values ​​and temperature distribution laws of each node of the model, and form a theoretical temperature data set that corresponds one-to-one with the three-dimensional geometric model.

[0034] S160: The feature matching method is used to match the feature points of the temperature field with the three-dimensional geometric model and the corresponding theoretical temperature distribution field.

[0035] A feature matching method is used to accurately match the temperature field feature points extracted in step S130 with the three-dimensional geometric model constructed in step S140 and the theoretical temperature distribution field obtained in step S150. Based on the structural nodes of the three-dimensional geometric model, a correspondence between the temperature field feature points and the model nodes is established. Combined with the temperature data of the theoretical temperature distribution field, the spatial position and temperature mapping accuracy of the feature points are calibrated, and feature points with matching deviations greater than a preset threshold are eliminated. This ensures that each temperature field feature point can accurately correspond to the corresponding part of the three-dimensional geometric model and the theoretical temperature value, thereby achieving a precise association between temperature information and the three-dimensional geometric model.

[0036] S170: Based on the matching results, complete the three-dimensional reconstruction of the temperature field of the photovoltaic box-type switchgear.

[0037] Based on the matching results of step S160, the actual temperature information corresponding to the feature points of the full temperature field is mapped one by one to the corresponding nodes of the three-dimensional geometric model. Combined with the theoretical temperature distribution field, temperature calibration and completion are performed to construct a three-dimensional temperature field model of the photovoltaic box transformer switchgear. This model can intuitively present the three-dimensional temperature distribution of various parts of the switchgear, clearly show the temperature differences in different areas, and support viewing the temperature distribution of the switchgear surface and internal components from multiple angles, thus completing the three-dimensional reconstruction of the temperature field of the photovoltaic box transformer switchgear.

[0038] This application utilizes multi-angle infrared temperature image acquisition to ensure complete temperature information coverage of the entire switchgear and key heat-generating components. Image preprocessing enhances edge information and contrast, laying the foundation for subsequent image processing. Image segmentation and targeted extraction of temperature field feature points accurately focus on key heat-generating components and connections, eliminating background interference and improving the effectiveness of temperature features. An initial population is constructed based on texture information, and a three-dimensional geometric model is built using finite element technology, enabling a 1:1 reconstruction of the switchgear structure and providing a precise geometric carrier for temperature field mapping. Thermal simulation is conducted under actual operating conditions to obtain the theoretical temperature distribution field. Then, feature matching is used to achieve precise correlation between temperature field feature points and the three-dimensional model. The final three-dimensional reconstruction of the temperature field can intuitively present the three-dimensional temperature distribution of the switchgear, clearly showing the temperature differences in various parts, effectively improving the comprehensiveness and accuracy of switchgear thermal status monitoring. This provides reliable data support and intuitive visualization for early fault warning, safe operation and maintenance, and hidden danger investigation, helping to ensure the stable and safe operation of photovoltaic box-type switchgear.

[0039] In one possible implementation, step S120 may include: performing multi-scale analysis on the first infrared temperature image, using filtering to enhance the image edge information to obtain an intermediate infrared temperature image; decomposing the intermediate infrared temperature image into an illumination image and a reflection image, removing the reflection image, and performing contrast stretching on the illumination image to obtain a preprocessed infrared temperature image.

[0040] In this implementation, a multi-scale analysis is first performed on the complete first infrared temperature image, traversing the gray-level distribution characteristics at different scales. Filtering is then used to remove image noise and enhance regions of abrupt gray-level changes, thereby strengthening image edge information and obtaining an intermediate infrared temperature image. Subsequently, the intermediate infrared temperature image is decomposed into an illumination image and a reflection image. The reflection image is removed, and histogram equalization is applied to stretch the remaining portion of the image to enhance contrast, resulting in an infrared image for three-dimensional temperature field reconstruction. The illumination image carries the surface temperature radiation information of the switchgear, while the reflection image contains redundant interference information such as shell reflection and ambient light. After removing the reflection image, the illumination image undergoes contrast stretching to amplify the gray-level differences between different temperature regions, improving image contrast and ultimately obtaining the pre-processed infrared temperature image.

[0041] This preprocessing method employs multi-scale analysis and filtering to accurately remove image noise at different scales, while simultaneously enhancing regions of abrupt grayscale changes. This effectively strengthens the edge information of the switchgear cabinet outline, internal components, and key heat-generating parts, preventing edge feature loss and laying a clear feature foundation for subsequent image segmentation and temperature field feature point extraction. By decomposing the image into illumination and reflection images and removing the reflection image, redundant interference information such as shell reflection and strong ambient light is eliminated, retaining only the illumination image that carries the true temperature radiation information of the switchgear. This avoids temperature information distortion caused by interference factors and ensures the purity of temperature-related data. Furthermore, histogram equalization technology is used to contrast-stretch the illumination image, significantly expanding the grayscale difference between different temperature regions and greatly improving the distinction between high-temperature and low-temperature regions. This makes the temperature distribution characteristics more intuitive and prominent, further enhancing the accuracy and reliability of subsequent temperature field feature point extraction. This provides high-quality image data support for the entire switchgear temperature field 3D reconstruction process, effectively improving the accuracy and reliability of the final 3D reconstruction results.

[0042] Furthermore, enhancing image edge information through filtering can include: employing a bilateral filter, controlling the spatial distance between pixels using a spatial Gaussian function, and combining this with a grayscale Gaussian function to enhance image edge information. Contrast stretching of the illuminated image can include: using histogram equalization to perform contrast stretching on the illuminated image.

[0043] A bilateral filter is used for filtering, and the spatial distance between pixels is precisely controlled by a spatial Gaussian function. At the same time, the image edge information is enhanced by combining a grayscale Gaussian function. This can effectively filter out environmental noise, acquisition noise and other interference in the infrared image, while preserving the edge features of the photovoltaic box transformer switch cabinet outline, internal components and key heat-generating components to the greatest extent. This avoids the edge blurring and loss of details that are easy to cause by conventional filtering. It provides a clear image foundation with complete features for subsequent image segmentation and temperature field feature point extraction. Histogram equalization technology is used to perform contrast stretching on the illumination image after interference removal. It can specifically expand the grayscale difference of different temperature regions in the illumination image, significantly improve the image differentiation between high-temperature components and low-temperature regions, make the real temperature radiation distribution characteristics of the switch cabinet more prominent, and further enhance the recognition of temperature-related image features. The two work together to ensure the noise suppression effect and edge integrity of the preprocessed image, and greatly improve the visualization of temperature features.

[0044] Specifically, the enhanced image edge information can be obtained using the following formula: In the formula, The output image edge information represents the bilateral filter. Represents the first infrared temperature image. Represents the coefficients of the Gaussian function in space. The Gaussian number representing the grayscale value of an infrared image. Represents the set of the neighborhood of the center pixel. Represents the center pixel of the image. Represents neighboring pixels, Represents the standard deviation of the Gaussian function in space. The standard deviation of the Gaussian function representing the grayscale value. This represents the normalized weight value.

[0045] An illumination-reflection decomposition model is used to decompose the complete infrared image into an illumination image. The illumination-reflection decomposition model is as follows: in, For the mid-infrared temperature image in pixel coordinates The grayscale value at that location contains a mixture of information such as temperature radiation and ambient reflection. The image, representing the actual temperature radiation information on the surface of the photovoltaic box / variable switchgear, is the core and effective data for subsequent processing. As a reflected image, it contains redundant and interfering information such as reflections from the device casing and strong ambient light, which need to be removed to improve image quality. Represents the pixel coordinates of an infrared image.

[0046] Based on the obtained illumination image, contrast enhancement is performed using a histogram equalization grayscale stretching algorithm. The formula is as follows: This formula uses the offset coefficient. and grayscale stretching coefficient Adjusting the image's grayscale dynamic range expands the grayscale difference across different temperature regions, thereby enhancing image contrast.

[0047] In one possible implementation, image segmentation techniques are used to segment the preprocessed infrared temperature image to distinguish between target and background regions. This includes: using the Tsallis entropy algorithm to extract grayscale values ​​from the preprocessed infrared temperature image and constructing a two-dimensional histogram of the neighborhood average grayscale values ​​of the preprocessed infrared temperature image; calculating the entropy value of each pixel in the preprocessed infrared temperature image based on the two-dimensional histogram; determining the grayscale segmentation threshold based on the entropy value; and segmenting the target and background regions.

[0048] In this implementation, the grayscale value of each pixel in the preprocessed infrared temperature image is extracted. For each pixel Calculate the average gray value of its 3×3 neighborhood: Taking the pixel as the center, take the gray values ​​of the nine pixels above, below, left, right, and diagonally, and average them to obtain the neighborhood average gray value. ; based on its own grayscale value The horizontal threshold vector and the average gray value of the 3×3 neighborhood are given. Using the threshold vector on the vertical axis, construct a two-dimensional histogram of the neighborhood average gray value, as follows: Figure 2 As shown, this histogram clusters pixels into four feature clusters: Target area (lower left cluster): The fluctuation range of the pixel's own gray value and the average gray value of the neighborhood is small, corresponding to the uniform area of ​​the switch cabinet body (high gray value, high temperature); Background area (top right cluster): The gray value of the pixel itself fluctuates greatly with the average gray value of the neighborhood, corresponding to the environment, support, and other background areas (low gray value, low temperature); Noise region (top left cluster): The pixel itself has a low gray value but the average gray value of its neighborhood is high, corresponding to isolated interference points or image noise; Boundary points (bottom right cluster): pixels with high gray values ​​but low average gray values ​​in their neighborhood, corresponding to the edge transition area between the switch cabinet and the background.

[0049] Note: The “threshold vector” here is a feature dimension within the algorithm, representing the grayscale fluctuation range rather than the original grayscale value, and does not conflict with the physical characteristic that “the original grayscale value of the target area (switch cabinet) is high”.

[0050] Based on the constructed two-dimensional histogram, the information entropy of each pixel is calculated using the Tsallis entropy formula, which is as follows: Iterate through all possible grayscale thresholds, select the grayscale value that maximizes the sum of the target region entropy and the background region entropy, and determine it as the grayscale segmentation threshold Z.

[0051] Extract the grayscale value of each pixel in the image, denoted as . ,in These are the two-dimensional coordinates of a pixel, used to represent the grayscale information corresponding to the temperature at that point. For each pixel in the image... Select a 3×3 rectangular neighborhood, or other sizes such as 5×5 depending on the actual scene. Calculate the average grayscale value of all pixels within this neighborhood, denoted as . Based on the pixel's own grayscale value The horizontal axis represents the average gray value of the neighborhood. Using the vertical axis as the ordinate, statistically analyze different... The frequency of occurrence of combinations is used to construct a two-dimensional histogram of the neighborhood average gray value of the infrared temperature image. This histogram can simultaneously reflect the correlation characteristics between the gray value of the pixel itself and the gray value distribution of its local neighborhood, thus improving the robustness of subsequent segmentation.

[0052] Based on the above two-dimensional histogram, a pixel is defined as the target pixel, such as the probability of a pixel in the heat-generating area of ​​the switch cabinet. ,in The index of the target pixel. Let be the set of all pixels in the image. The information entropy of each pixel is calculated using the Tsallis entropy formula, which is as follows: in, The information entropy of a pixel reflects the uncertainty of whether the pixel belongs to the target region; The collection of all pixels in an infrared image; For the target pixel; This is the probability value that the pixel belongs to the target region (determined by the clustering frequency of the 2D histogram). Iterate through all possible grayscale thresholds, selecting the grayscale value that maximizes both the "target region entropy" and the "background region entropy," and determine this as the grayscale segmentation threshold. .

[0053] Based on the determined grayscale segmentation threshold Image segmentation is achieved using a binarization formula, which is as follows: in, The segmented binary image; Image grayscale levels; These are the original pixel grayscale values ​​of the preprocessed infrared temperature image. After segmentation: Grayscale values. Pixels are marked as target regions; grayscale values The pixels are marked as background areas. The final result is a binary image that clearly distinguishes the target from the background.

[0054] The Tsallis entropy algorithm used for image segmentation has the core advantage of adapting to the grayscale distribution characteristics of infrared temperature images of photovoltaic transformer switchgear, thus improving the segmentation accuracy and robustness of target and background regions. This algorithm extracts image grayscale values ​​and constructs a two-dimensional histogram of the average grayscale value of the neighborhood, taking into account both the grayscale features of a single pixel and the grayscale correlation with surrounding pixels. This avoids segmentation deviations caused by single-pixel grayscale fluctuations and adapts to the gradual grayscale changes caused by temperature distribution in infrared images. Based on the two-dimensional histogram, the entropy value of each pixel is calculated. By quantifying the disorder of grayscale distribution through entropy, a segmentation threshold that accurately matches the actual grayscale distribution of the image can be determined. This effectively solves the problems of conventional segmentation algorithms having a single threshold and being susceptible to environmental interference. It can clearly distinguish the target area corresponding to the switchgear body and internal components from background areas such as the ground and supports, reducing background redundancy interference. Meanwhile, it can accurately preserve the image areas of key heating components and connecting parts, laying the foundation for the accurate extraction of temperature field feature points in the subsequent process, avoiding interference from invalid background pixels on feature extraction, and further ensuring the accuracy and reliability of subsequent three-dimensional geometric model construction, feature matching and three-dimensional reconstruction of temperature field, thus adapting to the infrared image segmentation requirements of complex environments in photovoltaic power plants.

[0055] In one possible implementation, extracting temperature field feature points representing temperature distribution based on the target region may include: obtaining a target region mask based on the segmented target region, and extracting temperature field feature points within the corresponding region of the preprocessed infrared temperature image.

[0056] In this implementation, a target region mask that perfectly matches the size of the preprocessed infrared temperature image is generated based on the segmented binarized image. This mask uses a binary labeling rule to mark the segmented target region (grayscale value ≥ segmentation threshold Z) as a valid region (labeled as 1), and the background region (grayscale value < segmentation threshold Z) as an invalid region (labeled as 0), achieving precise localization of the target region and providing a region selection basis for subsequent feature point extraction. Image edge information supports the accurate generation of the mask, ensuring that the mask completely fits the outer boundary of the target region, avoiding the omission of key component areas or the inclusion of redundant background areas.

[0057] The generated target area mask is pixel-level aligned with the preprocessed infrared temperature image. By marking the effective area of ​​the mask, the pixel range corresponding to the target area in the preprocessed infrared temperature image is accurately located, while invalid pixels in the background area are masked, ensuring that subsequent feature point extraction focuses only on the area carrying effective temperature information. In the preprocessed infrared temperature image, feature points are extracted only from the effective area marked by the mask, i.e., the target area. Leveraging the advantages of improved contrast and high temperature region differentiation in the preprocessed image, pixels with significant temperature gradient changes are selected as temperature field feature points. These feature points correspond to the surfaces of key heat-generating components in the switchgear and component connections. These areas exhibit significant grayscale value changes, directly related to temperature distribution differences. Simultaneously, the uniformity and representativeness of the feature points are considered, avoiding overly dense or sparse feature points. After extraction, all temperature field feature points from the target area are summarized to form a complete feature point set. Invalid feature points caused by mask positioning deviations or residual noise are removed, providing accurate and reliable temperature feature data support for subsequent matching operations with the 3D geometric model and theoretical temperature distribution field.

[0058] In one possible implementation, constructing an initial population based on the texture information of the preprocessed infrared temperature image may include: constructing the initial population based on texture blocks containing edge and temperature correlation information in the preprocessed infrared temperature image, wherein the texture block is an image unit characterizing the local structure and temperature correlation features of the photovoltaic box converter switch cabinet.

[0059] In this implementation, based on the preprocessed infrared temperature image, texture blocks are first defined as image units representing the local structural and temperature-related features of the photovoltaic transformer switchgear. Each texture block must simultaneously contain edge information (local structural outline, gray-level abrupt change region) and temperature-related information (local gray-level distribution, corresponding temperature features). Then, the preprocessed omnidirectional infrared temperature image is divided into grids to ensure that the extraction range covers the target area in all directions of the switchgear. After extraction, invalid texture blocks with blurred edges or background noise are removed, while valid texture blocks with clear edges and significant temperature-related features are retained. Finally, the selected valid texture blocks are used as basic units to construct the initial population, as shown in the following formula: In the formula, The texture population representing the photovoltaic box-type switchgear is the initial population constructed in this application. Representing the Individual texture blocks (each individual corresponds to a 32×32 effective texture block, which is fused with edge contour parameters and temperature-related parameters); This is the texture block number.

[0060] In one possible implementation, the mesh type is a hexahedral mesh. The core advantage of using a hexahedral mesh lies in its ability to accurately fit the complex contours and delicate components of photovoltaic transformer switchgear, avoiding the generation of distorted meshes and restoring the installation details of key heat-generating components. Furthermore, its stable element stiffness matrix improves the accuracy of heat conduction simulation, precisely capturing heat conduction paths and temperature gradient changes, providing more reliable calculation results for the theoretical temperature field. Simultaneously, the regular topology reduces the number of elements, optimizes finite element solution efficiency, accelerates the iterative convergence speed of the energy balance equation, and the clear node distribution pattern supports the accuracy and stability of subsequent feature matching, laying the foundation for high-precision reconstruction of the three-dimensional temperature field.

[0061] In one possible implementation, thermal simulation of the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model may include: thermal simulation of the three-dimensional geometric model in combination with the energy balance equation to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model, wherein the energy balance equation is used to characterize the relationship between the heat flow rate and temperature of each node of the three-dimensional geometric model.

[0062] The following is the energy balance equation: In the formula, The node heat flux vector representing the photovoltaic box-type switchgear includes the heat flux input and output rates of each hexahedral mesh node; This represents the temperature vector, corresponding to the theoretical temperature value of each grid node; The energy conduction matrix is ​​determined by the thermal conductivity of the switchgear material, the size of the hexahedral mesh elements, and the structural contact relationships, reflecting the conduction law of heat in the model.

[0063] By substituting the hexahedral mesh node parameters of the 3D geometric model and the thermal properties of the switchgear materials (such as the thermal conductivity of copper busbars and the thermal conductivity of insulating components) into the energy balance equation, the theoretical temperature values ​​of each mesh node are obtained through finite element iterative solution, thereby generating a 3D theoretical temperature distribution field covering the entire switchgear. This distribution field accurately reproduces the theoretical temperature variation laws of key heat-generating components such as busbars, circuit breakers, and contactors, as well as the heat conduction path within the cabinet. The output theoretical temperature distribution field corresponds one-to-one with the hexahedral mesh elements of the 3D geometric model, providing benchmark data support for subsequent feature matching and error correction with the actual infrared temperature field.

[0064] In one possible implementation, a feature matching method is used to match temperature field feature points with a three-dimensional geometric model and the corresponding theoretical temperature distribution field. This includes: using wavelet transform to establish the correspondence between temperature field feature points and various parts of the three-dimensional geometric model based on the theoretical temperature distribution field of the three-dimensional geometric model.

[0065] Specifically, in this implementation, the extracted temperature field feature point information is converted into a one-dimensional temperature field signal. (The signal value is derived from the grayscale value and is positively correlated with temperature), and then multi-scale convolution is performed using the wavelet transform formula: In the formula, This represents the position coordinates of the temperature field signal in the three-dimensional model. These are wavelet coefficients, used to capture multi-scale local features of temperature field signals. These represent the extracted infrared image temperature field feature points. Through wavelet transform, the local temperature gradient and spatial location features of these feature points are accurately captured. A one-to-one mapping is established between the two-dimensional infrared temperature field feature points and the hexahedral mesh nodes of the three-dimensional geometric model, establishing a correspondence between the three-dimensional model coordinates and the theoretical temperature value for each feature point. This matching process effectively addresses the multi-scale changes in the temperature field under the complex structure of the switchgear, achieving accurate mapping from two-dimensional infrared temperature information to the three-dimensional model. The matched correspondence provides core support for subsequently mapping actual infrared temperature information to the three-dimensional model and generating a complete three-dimensional temperature field, ensuring the accuracy and reliability of the three-dimensional temperature field reconstruction. Ultimately, it provides accurate three-dimensional temperature data for the thermal status assessment of photovoltaic transformer switchgear.

[0066] To verify the feasibility of the three-dimensional temperature field reconstruction method for photovoltaic transformer switchgear provided in this application, this experiment constructed a three-dimensional temperature field reconstruction method for photovoltaic transformer switchgear provided in this application, as shown below. Figure 3 The three-dimensional temperature field model shown is illustrated. Analysis reveals that the temperature inside the photovoltaic transformer switchgear is highest at the button locations, and decreases outwards from the buttons. Therefore, the method presented in this paper can reconstruct the three-dimensional temperature field model of the photovoltaic transformer switchgear, and the constructed model reflects the actual temperature distribution inside the switchgear.

[0067] To verify the accuracy of the 3D temperature field model reconstructed by the proposed method, this experiment reconstructed the temperature field of the experimental object using a planar temperature field reconstruction method based on tensor decomposition and a closed container temperature field reconstruction method based on acoustic tomography, respectively, as a control group. Simultaneously, 100 locations within the photovoltaic box transformer switchgear were randomly selected as detection targets, and the actual temperatures at these locations were statistically analyzed. The temperature values ​​of the 3D temperature field model constructed in this application and the 3D temperature field model constructed in the control group at the same locations were then statistically analyzed, and the results were compared.

[0068] Depend on Figure 4The experimental results show that the temperature field constructed by the method provided in this application has a small difference between the temperature values ​​and the actual values ​​at 100 detection nodes, with a maximum difference of no more than 0.8℃. In contrast, the temperature field three-dimensional models constructed by the other two methods show a larger difference between the temperature values ​​and the actual values ​​at the same locations. This indicates that the three-dimensional temperature field model constructed by the method provided in this application has higher accuracy and better reflects the actual temperature changes inside the photovoltaic box transformer switchgear.

[0069] Finally, it should be noted that: The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear, characterized in that, include: The first infrared temperature image of the photovoltaic box variable switch cabinet was acquired from multiple angles; The first infrared temperature image is preprocessed to enhance image edge information and improve image contrast, resulting in a preprocessed infrared temperature image. The preprocessed infrared temperature image is segmented using image segmentation technology to distinguish between the target region and the background region, and temperature field feature points representing the temperature distribution are extracted based on the target region. An initial population is constructed based on the texture information of the preprocessed infrared temperature image. A three-dimensional geometric model of the photovoltaic box-type switch cabinet is constructed using finite element technology and a selected mesh type. Thermal simulation was performed on the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model; The feature matching method is used to match the temperature field feature points with the three-dimensional geometric model and the corresponding theoretical temperature distribution field; Based on the matching results, a three-dimensional reconstruction of the temperature field of the photovoltaic box-type switchgear is completed.

2. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The step of performing image preprocessing on the first infrared temperature image to enhance image edge information and improve image contrast, to obtain a preprocessed infrared temperature image, includes: Multi-scale analysis is performed on the first infrared temperature image, and filtering is used to enhance the image edge information to obtain the intermediate infrared temperature image. The intermediate infrared temperature image is decomposed into an illumination image and a reflection image. After removing the reflection image, the illumination image is subjected to contrast stretching to obtain the preprocessed infrared temperature image.

3. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 2, characterized in that, The method of enhancing image edge information through filtering includes: using a bilateral filter, controlling the spatial distance between pixels through a spatial Gaussian function, and combining a grayscale Gaussian function to enhance image edge information; The contrast stretching process for the illumination image includes: performing contrast stretching on the illumination image using histogram equalization technology.

4. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The step of segmenting the preprocessed infrared temperature image using image segmentation technology to distinguish between the target region and the background region includes: The Tsallis entropy algorithm is used to extract the gray values ​​of the preprocessed infrared temperature image and construct a two-dimensional histogram of the neighborhood average gray values ​​of the preprocessed infrared temperature image. The entropy value of each pixel in the preprocessed infrared temperature image is calculated based on the two-dimensional histogram, and the grayscale segmentation threshold is determined according to the entropy value to segment the target area and the background area.

5. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 4, characterized in that, The step of extracting temperature field feature points representing temperature distribution based on the target region includes: Based on the segmented target region, a target region mask is obtained, and the temperature field feature points are extracted within the corresponding region of the preprocessed infrared temperature image.

6. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The initial population is constructed based on the texture information of the preprocessed infrared temperature image, including: The initial population is constructed based on the texture blocks containing edge and temperature correlation information in the preprocessed infrared temperature image. The texture blocks are image units that characterize the local structure and temperature correlation features of the photovoltaic box converter switch cabinet.

7. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The mesh type is a hexahedral mesh.

8. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The thermal simulation of the three-dimensional geometric model to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model includes: Thermal simulation of the three-dimensional geometric model is performed using the energy balance equation to obtain the theoretical temperature distribution field of each part of the three-dimensional geometric model. The energy balance equation is used to characterize the relationship between heat flow rate and temperature at each node of the three-dimensional geometric model.

9. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The step of employing a feature matching method to match the temperature field feature points with the three-dimensional geometric model and the corresponding theoretical temperature distribution field includes: Using wavelet transform, based on the theoretical temperature distribution field of the three-dimensional geometric model, the correspondence between the temperature field feature points and various parts of the three-dimensional geometric model is established.

10. The method for three-dimensional reconstruction of the temperature field of a photovoltaic box-type switchgear according to claim 1, characterized in that, The temperature field feature points include at least the pixel points corresponding to the key heat-generating components of the photovoltaic box-type switchgear and the pixel points corresponding to the connection points of the key heat-generating components.

Images