Method and device for rapid detection of hot spot failure on photovoltaic module surface

Abstract

This invention provides a rapid detection method and device for hot spot faults on the surface of photovoltaic modules, relating to the field of photovoltaic power generation technology. The invention applies a uniform planar thermal pulse to the surface of the photovoltaic module and acquires an infrared image sequence of the entire temperature relaxation process. It then uses the temperature decay curve of the healthy region to fit a standard thermal diffusivity, constructing a two-dimensional heat conduction model to generate a theoretical thermal wave propagation image sequence. The measured sequence is compared frame-by-frame with the theoretical sequence to identify continuous negative deviations and generate a thermal wave shadow image through time-domain integration. After threshold segmentation of this image, candidate regions are determined by morphological screening based on the theoretical area of ​​the solar cell, rectangular features, and grid line direction. Finally, the candidate regions are mapped back to the image at the moment the thermal pulse ends, and high-temperature characteristics are checked to confirm the actual hot spot and mark its location and shadow intensity. This dual verification of morphology and heat source effectively eliminates interference, significantly improving detection accuracy and reliability, and providing a quantitative assessment basis for photovoltaic module operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, specifically to a method and apparatus for rapid detection of hot spot faults on the surface of photovoltaic modules. Background Technology

[0002] During long-term outdoor operation, photovoltaic modules are prone to forming localized overheating areas inside the module, known as hot spot faults, due to factors such as cell cracking, solder ribbon aging, bypass diode failure, or partial shading. Hot spots not only reduce the module's power generation efficiency but also accelerate the aging of encapsulation materials, and in severe cases, can cause backsheet burn-through or even fire accidents. Therefore, developing a method for rapid, accurate, and early detection of hot spot faults on the surface of photovoltaic modules has significant engineering application value for ensuring the safe operation of power plants, reducing operation and maintenance costs, and extending the service life of modules.

[0003] The current mainstream detection method for hot spot faults in photovoltaic modules mainly relies on infrared thermal imaging technology. This method uses an infrared thermal imager to capture images of the surface temperature distribution of the module under normal power generation conditions. The fault is identified by utilizing the physical characteristic that the temperature of the hot spot area is significantly higher than that of the surrounding normal area. A typical detection process includes acquiring infrared images, performing preprocessing such as grayscale conversion and filtering, extracting high-temperature areas using global or local threshold segmentation algorithms, and finally determining whether it is a hot spot based on the area, shape, and relative temperature difference of the high-temperature area. Some improved methods introduce the fusion registration of visible light and infrared images, and eliminate interference from normal heat-generating components such as busbars and junction boxes by comparing the spatial correspondence between the cell outline and the high-temperature area. However, existing technologies still have several limitations in engineering practice that they cannot overcome. Traditional infrared detection methods are essentially passive, and their detection capability depends entirely on the temperature rise caused by the hot spot fault itself. When the hot spot is in its early stages of formation, or when the heating power of the hot spot is low due to insufficient lighting conditions or low ambient temperature, its surface temperature rise is often weak or even not yet apparent. In this case, passive infrared imaging is difficult to effectively identify, resulting in a large number of early hot spot faults being missed. Secondly, existing methods mainly rely on the static temperature distribution at a single moment to determine the hot spot, which is subject to rapid temperature changes. The traditional detection mode loses the dynamic information of the heat conduction process in the hot spot area. In fact, the impact of hot spots on module performance is not only reflected in the steady-state temperature rise, but also in its obstruction effect on the heat wave propagation path. The latter is precisely a sensitive indicator reflecting hidden damage such as delamination of the internal interface of the cell and poor solder joint. Existing methods cannot capture and quantify this dynamic physical process. In addition, existing technical solutions lack an effective mechanism to distinguish between hot spots and cold spots. Non-heating defects also show low-temperature characteristics in infrared images, but are often misjudged as hot spots due to similar shapes, resulting in a high false alarm rate.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for rapid detection of hot spot faults on the surface of photovoltaic modules, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A rapid detection method for hot spot faults on the surface of photovoltaic modules, comprising the following steps: Step 1: Apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and use a high frame rate infrared thermal imager to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied. Step 2: Select a known fault-free reference area on the component surface from the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use this model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. Step 3: Perform pixel-by-pixel and frame-by-frame difference calculations between the infrared image sequence and the theoretical thermal wave propagation image sequence. Identify the negative deviation region in the difference calculation. Perform time-domain integration on all difference calculation results during the temperature relaxation process to generate an integrated thermal wave shadow image. Step 4: Threshold segmentation is performed on the integral thermal wave shadow image to obtain the preliminary abnormal region. Morphological screening is applied to set the regions in the preliminary abnormal region that meet the triple geometric judgment criteria as candidate regions. Step 5: Map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface.

[0007] Furthermore, a uniform planar thermal pulse is applied, specifically by using a portable flexible planar heat source that matches the surface size of the photovoltaic module to be tested, ensuring that its working surface is parallel to the module surface and in zero-gap contact, and applying a rectangular wave thermal pulse with a duration between 0.3 seconds and 0.8 seconds. The infrared image sequence of the temperature relaxation process is acquired by using an uncooled infrared thermal imager with a frame rate of not less than 90Hz, taking the moment when the thermal pulse is turned off as the zero point of timing, and continuously acquiring the evolution process of the component surface temperature field for a duration of not less than 5 seconds to ensure that the complete temperature relaxation process of the heat wave transitioning from the surface convection heat transfer-dominated stage to the internal heat diffusion-dominated stage is recorded.

[0008] Furthermore, fault-free reference areas are selected and temperature decay curves are analyzed. Specifically, in the first frame of the infrared image sequence, an automatic recognition algorithm based on backplane texture features is used to delineate at least three rectangular areas located within the component border neighborhood that have never experienced hot spot faults according to historical maintenance records, as fault-free reference areas. The size of each reference area is set to be no less than 30 pixels × 30 pixels. All frames of the entire infrared image sequence are traversed, and the instantaneous temperature values ​​of all pixels within each reference area in each frame are extracted and their arithmetic mean is calculated to obtain the average temperature of each reference area in that frame. After arranging all frames in chronological order, each fault-free reference area obtains a temperature decay data sequence composed of the average temperatures of each frame. The standard thermal diffusivity is then obtained through curve fitting, as follows: Using a fault-free reference area's temperature decay data sequence during the temperature relaxation process, and taking the ambient temperature as a baseline, the difference between the average temperature of the reference area in the first frame of the infrared image sequence and the ambient temperature is defined as the initial temperature rise. The temperature decay over time is expressed as an exponential decay function with the ambient temperature as the baseline, the initial temperature rise as the decay amplitude, and the thermal time constant to be solved as the decay rate control parameter. A nonlinear least squares fitting algorithm is used to numerically fit the temperature decay data sequence of the reference area, minimizing the sum of squared errors between the exponential decay function and the temperature decay data sequence to obtain the best-fit exponential function for the reference area. A decay function is used to construct the temperature decay curve of the reference region during the temperature relaxation process. The thermal time constant corresponding to the reference region is obtained from the fitting parameters of the best-fit exponential decay function. Based on the classical analytical solution of the one-dimensional thermal diffusion process in the thickness direction of the flat layered packaging structure, the thermal time constant is divided by the square of the total packaging thickness of the component, and the quotient is multiplied by the reciprocal of the square of pi to calculate the standard thermal diffusivity corresponding to the reference region. The above operation is repeated for all selected fault-free reference regions to obtain multiple standard thermal diffusivity calculation results. The arithmetic mean of these calculation results is taken as the standard thermal diffusivity of the component in a healthy state.

[0009] Furthermore, a two-dimensional heat conduction digital model is established and a theoretical thermal wave propagation image sequence is generated. Specifically, the standard thermal diffusivity is used as the only material thermal property input parameter to establish a two-dimensional heat conduction digital model to describe the evolution of the surface temperature field of the component after the planar thermal pulse ends. Based on this two-dimensional heat conduction digital model, the measured temperature values ​​of each pixel position on the component surface recorded in the first frame of the infrared image sequence are used to form the initial temperature distribution field, and the average temperature of the background area far from the thermal excitation center and whose temperature tends to be stable in the infrared image sequence is used as a constant environmental thermal boundary condition. Based on the Green's function method in two-dimensional unsteady heat conduction theory, the theoretical temperature at any pixel location at any time is expressed as the result of spatially weighted integration of the initial temperature distribution field and superimposed with environmental thermal boundary conditions. The weighting function used in this spatial weighted integration is a Gaussian weighted distribution function with the Euclidean distance between any pixel location in the initial temperature distribution field and the target pixel location for calculating the theoretical temperature as the independent variable, and determined by the standard thermal diffusivity and the time corresponding to the current calculation frame. The above-mentioned theoretical temperature calculation method based on the Green's function corresponds to an analytical solution in continuous integral form. In actual numerical calculations, this analytical solution in continuous integral form is discretized into a two-dimensional convolution operation that matches the pixel grid coordinates of the infrared image sequence, and a fast Fourier transform algorithm is used to achieve a highly efficient numerical solution. By performing this convolution operation frame by frame, a theoretical heat wave propagation image sequence with a spatiotemporal resolution completely consistent with the infrared image sequence is generated.

[0010] Furthermore, pixel-by-pixel and frame-by-frame difference calculations are performed to identify negative deviation regions. Specifically, for corresponding frames with identical timestamps in the infrared image sequence and the theoretical heat wave propagation image sequence, the measured infrared temperature value at the same pixel position within the same frame is subtracted from the theoretical temperature value predicted by the model. The summed differences constitute the difference image for that frame. In this difference image, if the measured temperature of a pixel is lower than the theoretical temperature, the difference calculation result for that pixel is negative, and that pixel is marked as a negative deviation pixel. In this process, to eliminate false negative deviations caused by random noise and instantaneous disturbances in a single frame, a continuity criterion in the time dimension is introduced. That is, for any pixel, if it presents a negative deviation pixel in three or more consecutive difference images, it is determined that the pixel is indeed in a state of heat wave propagation obstruction caused by internal defects during the duration corresponding to the three or more consecutive frames. The negative deviation amplitude of the pixel in each frame during the duration is recorded as the heat wave shadow intensity value of the pixel in the corresponding frame. The integral thermal shadow image is generated by performing time-domain integration. Specifically, for each pixel, the first frame of the infrared image sequence is taken as the integration starting point, and the last frame corresponding to the end of the temperature relaxation process is taken as the integration ending point. All frames from the starting point to the ending point are traversed. For each frame traversed, it is determined whether the pixel has a thermal shadow intensity value in the thermal shadow image of that frame. If it does, the intensity value is extracted and added to the sum of the pixels. If it does not, the contribution value of the pixel in that frame is recorded as zero and is not accumulated. After traversing all frames from the starting point to the ending point, each pixel obtains a single accumulated result value. The accumulated result values ​​of all pixels together constitute the integral thermal shadow image.

[0011] Further, threshold segmentation and morphological screening are performed, specifically: the global segmentation threshold is automatically calculated using the maximum inter-class variance method on the integral thermal wave shadow image; pixels with values ​​higher than the global segmentation threshold are marked as foreground, and pixels with values ​​lower than or equal to the global segmentation threshold are marked as background, to generate a preliminary binarized abnormal region image; morphological opening and closing operations are sequentially performed on the preliminary binarized abnormal region image, wherein the opening operation is erosion followed by dilation, and the closing operation is dilation followed by erosion. After morphological cleansing, the foreground pixels in the binarized preliminary anomalous region image form several interconnected components with independent boundaries in space. Each connected component is a preliminary anomalous region. Geometric feature parameters of each preliminary anomalous region are extracted, including the region area, minimum bounding rectangle size, and principal axis direction. A triple geometric judgment criterion is set for candidate region screening: First, the region area of ​​the preliminary anomalous region is between 0.8 and 1.5 times the theoretical area of ​​a single photovoltaic cell of the photovoltaic module to be detected; second, the aspect ratio of the minimum bounding rectangle of the preliminary anomalous region is less than 2.0; third, the global dominant direction of the main grid line on the surface of the module is extracted from the first frame of the module infrared image sequence using the Hough transform algorithm, and the angle between the principal axis direction of the preliminary anomalous region and the global dominant direction is compared, requiring the absolute value of the angle between the two to be less than 5 degrees. Preliminary anomalous regions that simultaneously meet the above triple geometric judgment criteria are judged as candidate regions.

[0012] Furthermore, it is verified whether the candidate region has already shown high temperature at the end of the thermal pulse. Specifically, the spatial pixel position of the candidate region in the integrated thermal wave shadow image is mapped to the first frame of the infrared image sequence, that is, the initial temperature field image acquired at the moment the thermal pulse is turned off; the initial temperature value of all pixels in the candidate region in this frame is extracted and its arithmetic mean is calculated as the initial characteristic temperature of the candidate region. Using the minimum bounding rectangle of the candidate region as a reference, an annular neighborhood is formed by extending outward at equal intervals to enclose the region. The width of the annular neighborhood is set to be consistent with the short side size of the candidate region. The initial temperature values ​​of all pixels in the annular neighborhood in the first frame are extracted and their arithmetic mean is calculated as the initial feature temperature of the background. The initial feature temperature of the candidate region is subtracted from the initial feature temperature of the background to obtain the relative temperature rise value of the candidate region at the instant the thermal pulse ends. If the relative temperature rise value reaches or exceeds the preset temperature difference judgment threshold, it is determined whether the candidate region has shown a high temperature area at the end of the planar thermal pulse, and it is determined to be a real hot spot caused by internal electrical defects; otherwise, it is determined to be a non-heating defect and is excluded. The cumulative intensity value of thermal wave shadow is labeled as follows: For each candidate area identified as a real hot spot, all pixels covered by the area are located in the integrated thermal wave shadow image. The integrated thermal wave shadow values ​​of these pixels are extracted and their arithmetic mean is calculated. This average value is used as the cumulative intensity value of the thermal wave shadow of the real hot spot. Based on the numerical range of the cumulative intensity value of the thermal wave shadow, three fault levels—mild, moderate, and severe—are pre-divided. On the first frame of the infrared image sequence, the outer position of each real hot spot is marked with a bright colored rectangle, and the cumulative intensity value of the thermal wave shadow of the real hot spot and the corresponding fault level are simultaneously presented in the form of text labels in the area adjacent to the rectangle.

[0013] The present invention also provides a rapid detection device for hot spot faults on the surface of photovoltaic modules. The device is used to perform the above-described rapid detection method for hot spot faults on the surface of photovoltaic modules, comprising: The image acquisition module is used to apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied using a high frame rate infrared thermal imager. The theoretical calculation module is used to select a known fault-free reference area on the surface of the component in the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use the model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. The integration calculation module is used to perform pixel-by-pixel and frame-by-frame difference calculation between the infrared image sequence and the theoretical thermal wave propagation image sequence. In the difference calculation, the negative deviation area is identified, and all difference calculation results during the temperature relaxation process are integrated in the time domain to generate an integrated thermal wave shadow image. The candidate screening module is used to perform threshold segmentation on the integral thermal wave shadow image to obtain the preliminary abnormal area, and apply morphological screening to set the region in the preliminary abnormal area that meets the triple geometric judgment criterion as the candidate area. The hot spot detection module is used to map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse, so as to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention fundamentally breaks through the dependence of traditional passive infrared detection on the heating amplitude of the hot spot itself by actively exciting planar thermal pulses and acquiring high frame rate temperature relaxation processes. This method does not wait for the hot spot to generate a significant temperature rise under normal operating conditions, but actively injects a controllable thermal signal into the component and fully records the complete dynamic process of heat wave propagation and dissipation inside the component. This active excitation mechanism enables even early latent hot spots that are in the early stage of formation, have extremely low heating power, or have not yet started to actively heat up to be effectively captured due to their obstruction effect on the heat wave propagation path. This significantly shifts the hot spot detection window from the high temperature stage after the fault occurs to the thermal resistance anomaly stage during the fault incubation period, significantly improving detection sensitivity and early warning capability. This invention achieves a paradigm shift in hot spot fault analysis, moving from static temperature identification to dynamic process anomaly analysis, through the construction of a health status digital model and thermal wave shadow integral imaging. It uses temperature decay data from the healthy region of the component to fit a standard thermal diffusivity coefficient and establishes a two-dimensional thermal conduction digital model strictly bound to the measured initial conditions. This model can predict the theoretical temperature evolution process of the component under a fully healthy state pixel by pixel and frame by frame. By comparing the measured temperature field with this theoretical benchmark in time and space, the location of thermal wave obstruction is locked using a continuous negative deviation criterion. Furthermore, by integrating the negative deviation over the entire relaxation period in the time domain, the originally weak, dispersed, and noise-prone thermal wave propagation anomaly signal is accumulated and enhanced into a spatially significant integral thermal wave shadow image with a very high signal-to-noise ratio. This process not only eliminates the interference of external factors such as ambient temperature and light intensity on the detection results but also quantifies the degree of damage to the component's internal heat transport capacity by the hot spot into a temperature cumulative deficit value with clear physical meaning, providing an objective and continuous quantitative indicator for fault severity assessment. This invention constructs a rigorous pseudo-defect suppression mechanism through multi-dimensional geometric morphology screening and reverse verification of heat source attributes. Utilizing the inherent triple geometric constraints of photovoltaic cells—area, aspect ratio, and grid line direction—it accurately locates candidate regions from the integrated thermal wave shadow image that highly match the physical characteristics of the actual hot spot, effectively eliminating interference sources with inconsistent shapes such as busbars, borders, and dirt. By mapping the candidate regions back to the initial temperature field image at the moment the thermal pulse ends, it rigorously verifies whether they have actively generated heat as active heat sources during the excitation phase, thereby achieving an essential distinction between hot and cold spots. This dual verification mechanism enables the invention to maintain extremely high detection confidence in complex background environments. The final output hot spot location label and shadow intensity quantification value can directly support the assessment of the remaining life of the module and differentiated operation and maintenance decisions, achieving a fundamental improvement from binary judgment of whether there is a fault to predictive diagnostic capabilities that determine the severity of the fault and when it needs to be addressed. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a scatter plot of the time-measured temperature of this invention. Figure 3 This is a time-measured temperature fitting curve for the present invention; Figure 4 This is a flowchart of the overall device structure of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: Please see Figures 1-3 The present invention provides a technical solution: A rapid detection method for hot spot faults on the surface of photovoltaic modules, comprising the following steps: Step 1: Apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and use a high frame rate infrared thermal imager to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied. To achieve uniform and controllable thermal excitation, this invention uses a portable flexible planar electrothermal film that matches the overall size of the surface of the photovoltaic module to be tested as a heat source. The flexible planar electrothermal film is chosen because it can achieve good conformal contact with the glass cover surface of the photovoltaic module, thereby ensuring that the heat flow is uniformly injected into the module surface and avoiding local overheating or underheating due to uneven contact. In actual operation, the working surface of the flexible planar electrothermal film is placed parallel to the glass cover surface of the photovoltaic module. To ensure maximum heat conduction efficiency, the air gap between the two must be eliminated. Therefore, this invention uses a mechanical pressing device (such as an adjustable pressure clamp) or vacuum adsorption method to keep the working surface of the heat film and the module surface in close contact with zero gap. Air is a poor conductor of heat. If there is a gap, it will significantly reduce the transmission efficiency of the heat pulse and cause uneven heat distribution, which will affect the accuracy of subsequent detection. Apply a uniform planar thermal pulse, specifically by using a portable flexible planar heat source that matches the surface size of the photovoltaic module to be tested, ensuring that its working surface is parallel to the module surface and in zero-gap contact, and applying a rectangular wave thermal pulse with a duration between 0.3 seconds and 0.8 seconds. A flexible electric heating film capable of bending and deformation is selected as the planar heat source. The size of its heating surface is determined according to the detection target. If a single battery cell is used as the detection unit, the size of the heat source should be consistent with the size of the battery cell. If the entire module is used as the detection unit, the size of the heat source should cover the entire surface area of ​​the module. After the flexible electric heating film is laid flat on the surface of the module, a counterweight is placed on the back of the heat source or a vacuum adsorption device is used to apply uniform pressure to ensure that there are no visible air bubbles or local lifting between the working surface of the heat source and the glass surface of the module, thus achieving zero-gap bonding. It should be noted that zero-gap bonding is necessary because the thermal conductivity of air is much lower than that of solid materials. Any tiny air gap will significantly reduce the heat conduction efficiency, resulting in the actual heat flux density entering the module being lower than the preset value, which will destroy the uniformity and repeatability of the heat pulse. The purpose of achieving zero-gap bonding is to ensure the quantitative controllability of thermal excitation, so that subsequent theoretical modeling can be based on the known input heat flux. The thermal pulse waveform is selected as a rectangular wave, which has a steep rise edge, enabling an approximately step-like instantaneous thermal excitation on the module surface. In this embodiment, the rise time of the rectangular wave is controlled within 0.05 seconds, the peak heating power density is stabilized between 800 W / m² and 1200 W / m², and the pulse duration is set to 0.3 to 0.8 seconds. These parameters are chosen based on the thermal diffusion characteristics of the photovoltaic module encapsulation material: too short a duration will not generate sufficient temperature rise on the module surface, resulting in a low signal-to-noise ratio; too long a duration will cause excessive heat diffusion into the depth, resulting in an insufficiently sharp thermal wave front and reduced sensitivity for subsequent defect detection; a power density range of 800 W / m² to 1200 W / m² can generate a temperature rise of 5 to 10 degrees Celsius on the module surface, which is sufficient for clear differentiation by an infrared thermal imager without causing thermal stress damage to the module due to overheating. The infrared image sequence of the temperature relaxation process is acquired by using an uncooled infrared thermal imager with a frame rate of not less than 90Hz, taking the moment when the thermal pulse is turned off as the zero point of timing, and continuously acquiring the evolution process of the component surface temperature field for a duration of not less than 5 seconds to ensure that the complete temperature relaxation process of the heat wave transitioning from the surface convection heat transfer-dominated stage to the internal heat diffusion-dominated stage is recorded. At the instant the thermal pulse is turned off, the surface of the module begins a cooling process that gradually recovers from a high temperature to the ambient temperature. This process is called the temperature relaxation process in physics. Whether there are hot spot defects inside the module will significantly change the propagation path and speed of the heat wave during the relaxation process. These tiny changes will be reflected in the spatiotemporal evolution of the surface temperature field. Therefore, it is crucial to record this process accurately and completely. Therefore, this invention uses an uncooled infrared thermal imager with a frame rate of not less than 90 Hz as the temperature field acquisition device. The reason for choosing a frame rate of 90 Hz or higher is that in the initial stage after the thermal pulse is turned off, the temperature drops at an extremely fast rate. Low frame rate devices cannot capture sufficiently dense time series information, which will lead to the loss of key details of heat wave propagation. The lens optical axis of the infrared thermal imager is arranged perpendicular to the surface of the photovoltaic module, and its field of view is adjusted to ensure that the imaging range can completely cover the entire module under test without leaving any blind spots. A key synchronization operation is defining the starting point of the acquisition. This invention takes the instant when the thermal pulse is turned off, that is, the moment when the rectangular wave electrical signal ends, as the zero point of timing. At the same time as the thermal pulse is turned off, the infrared thermal imager is triggered to start continuous acquisition. This precise synchronization is the basis for subsequent frame-by-frame comparison between the measured temperature and the temperature predicted by the theoretical model. The acquisition duration was set to be no less than 5 seconds. This duration was also based on physical principles: For a typical crystalline silicon photovoltaic module encapsulation structure (composed of laminated glass cover, EVA film, solar cells, backsheet, etc.), in the first 1 to 2 seconds after the thermal pulse excitation ends, heat is mainly controlled by surface convection heat transfer, and the temperature drops rapidly. From about the 3rd second onwards, heat transfer enters a stage dominated by internal thermal diffusion. Within the time window of 3 to 5 seconds after the thermal pulse ends, the leading edge of the heat wave has fully penetrated the glass cover and exchanged heat with the solar cell layer. At this time, if there are defects such as cracks, low bypass resistance, or aging of the solder ribbon inside the solar cell, it will significantly disturb the normal propagation of the heat wave and leave a detectable shadow on the surface temperature field. If the total acquisition time is too short (e.g., less than 3 seconds), the most significant characteristic period of this defect signal will be missed. Conversely, if the acquisition time is too long (e.g., more than 10 seconds), the surface temperature of the module will approach the ambient thermal equilibrium, and the signal will be extremely weak. This will not only fail to provide more effective information but will also increase the burden of data processing. During the entire acquisition process, which lasts no less than 5 seconds, the infrared thermal imager operates continuously at a constant frame rate of 90 Hz or higher, generating a series of consecutive image frames. Each frame is a grayscale thermal image that records the two-dimensional temperature distribution on the surface of the component, where the grayscale value of a pixel has a known linear relationship with the temperature value at that point. More importantly, the infrared thermal imager or its associated data acquisition system synchronously generates a high-precision timestamp for each frame. These timestamps are unique and increment over time, precisely marking the time point corresponding to that frame, calculated from the zero point of timing. All image frames carrying unique timestamps, arranged in chronological order, constitute the infrared image sequence described in this invention. This sequence has a clear structure: the first frame precisely corresponds to the initial temperature field on the surface of the component at the timing zero point (i.e., the instant the thermal pulse is turned off); the second frame corresponds to the temperature field of the first time interval thereafter (e.g., 1 / 90 second); and so on, until the last frame corresponds to the temperature field at the end of the acquisition. This image sequence with a clear structure and a well-defined timeline provides a high-quality and standardized data foundation for building theoretical models, performing frame-by-frame comparisons, and time-domain integration in subsequent steps.

[0019] Step 2: Select a known fault-free reference area on the component surface from the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use this model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. Selecting fault-free reference areas and analyzing temperature decay curves involves the following steps: In the first frame of the infrared image sequence, an automatic identification algorithm based on backplane texture features is used to delineate at least three rectangular areas located within the component's border neighborhood that have never experienced hot spot faults according to historical maintenance records. These areas are then designated as fault-free reference areas, with each reference area having a size of at least 30 pixels × 30 pixels. All frames of the entire infrared image sequence are traversed, and the instantaneous temperature values ​​of all pixels within each reference area in each frame are extracted and their arithmetic mean is calculated to obtain the average temperature of each reference area in that frame. After arranging all frames in chronological order, each fault-free reference area obtains a temperature decay data sequence composed of the average temperatures of each frame. To accurately characterize the thermal properties of the module's health status, a typical region unaffected by any defects needs to be selected from the infrared image as an analysis sample. In the first frame of the infrared image sequence, i.e., the temperature field image recorded at the instant the thermal pulse is turned off, a fault-free reference area is selected. This first frame contains the initial temperature distribution information of the module surface after the thermal excitation stops. At this time, the temperature difference between the defective and non-defective areas is not significant, which is beneficial for selecting areas using non-temperature features as an auxiliary method. This invention uses an automatic recognition algorithm based on backsheet texture features to assist in positioning. The backsheet of the photovoltaic module has specific texture or color features and is located near the module frame. Due to the heat dissipation effect and physical isolation of the frame, the area below... The photovoltaic cells operate in a more stable environment, making them a low-risk area for failures. Therefore, the algorithm prioritizes searching within the neighborhood of the module's frame, such as a strip area 5 to 10 centimeters from the inner edge of the frame. The algorithm then verifies the candidate area by calling the historical maintenance database of the photovoltaic module to ensure that there has never been any record of hot spot failures or electrical performance abnormalities since the installation. Based on the above conditions, at least three independent rectangular areas with a size of not less than 30 pixels × 30 pixels are finally delineated as fault-free reference areas. The minimum size of 30 pixels × 30 pixels is set to ensure that each reference area contains a sufficient number of pixels, so that the subsequent calculation of the average temperature of the area has good statistical representativeness and can effectively suppress the interference of single-point noise. After selecting the reference area, it is necessary to extract its temperature change information during the entire temperature relaxation process. To do this, all frames in the infrared image sequence obtained in step 1 are traversed. For each frame, the pixel positions of the three reference areas are located, the instantaneous temperature values ​​of all pixels in each reference area are extracted, and their arithmetic mean is calculated. Thus, for each fault-free reference area, as all frames are traversed (from the first frame to the last frame), a set of average temperature values ​​arranged in chronological order can be obtained. This set of data completely records the temperature drop trajectory of the healthy area from the moment the thermal pulse is turned off until the end of the temperature relaxation process. This invention defines it as the temperature decay data sequence of the reference area. The standard thermal diffusivity is then obtained through curve fitting, specifically as follows: A fault-free reference area's temperature decay data sequence during temperature relaxation is used. Using the ambient temperature as a baseline, the difference between the average temperature of the reference area in the first frame of the infrared image sequence and the ambient temperature is defined as the initial temperature rise. The temperature decay over time is expressed as an exponential decay function with the ambient temperature as the baseline, the initial temperature rise as the decay amplitude, and the thermal time constant to be solved as the decay rate control parameter. A nonlinear least squares fitting algorithm is used to numerically fit the temperature decay data sequence of the reference area, minimizing the sum of squared errors between the exponential decay function and the temperature decay data sequence to obtain the standard thermal diffusivity. The best-fit exponential decay function of the reference region is used to construct the temperature decay curve of the reference region during the temperature relaxation process. The thermal time constant corresponding to the reference region is obtained from the fitting parameters of the best-fit exponential decay function. Based on the classical analytical solution of the one-dimensional thermal diffusion process in the thickness direction of the flat layered packaging structure, the thermal time constant is divided by the square of the total packaging thickness of the component, and the quotient is multiplied by the reciprocal of the square of pi to calculate the standard thermal diffusivity corresponding to the reference region. The above operation is repeated for all selected fault-free reference regions to obtain the calculation results of multiple standard thermal diffusivity. The arithmetic mean of these calculation results is taken as the standard thermal diffusivity of the component in a healthy state. The exponential decay function is: , The ambient temperature is the average stable temperature value of the reference area over several consecutive frames (10-20 frames) before the thermal pulse is applied. It represents the baseline for temperature decay and the equilibrium temperature to which the component will eventually recover. The initial temperature rise is the average temperature of the reference area at the instant the thermal pulse is turned off (first frame image) minus... The resulting difference represents the initial degree of overheating formed by the thermal pulse on the component surface, and is used as the attenuation amplitude in the formula; The thermal time constant is the parameter to be solved obtained by nonlinear least squares fitting, which is the temperature from... decay to Time required; s represents the time point, which is the instant the thermal pulse is turned off as the zero point of timing. The timestamps corresponding to each frame of the image represent the time index of the temperature relaxation process. For a moment The average temperature of the reference region at time t is the temperature of the reference region at time t. The average measured temperature value represents the target variable for fitting; this formula describes the physical law that the temperature of the fault-free reference area decays exponentially over time after the thermal pulse is turned off. The measured temperature is based on the theoretical temperature, with an average value of 0 and a range of [missing value]. Random fluctuations within the ℃ range; specific data on some frame numbers and temperatures are shown in Table 1.

[0020] Table 1 Data Statistics Table Analysis of the first 15 sets of simulation data clearly shows the exponential decay law of the reference area temperature with relaxation time. The time in the data represents the number of seconds after the thermal pulse is turned off, and the measured temperature is the average temperature of the reference area measured by the infrared thermal imager at that time. Overall, there is a significant negative correlation between the measured temperature and the relaxation time. As the time increases, the temperature in the reference area exhibits a nonlinear decay characteristic, gradually transitioning from a rapid decrease to a slow decrease. For example, at 0.213 seconds corresponding to frame number 1, the measured temperature is relatively high at 31.741℃; while at 2.956 seconds corresponding to frame number 15, the measured temperature has dropped to 25.709℃. This indicates that in the initial stage after the thermal pulse is turned off, the temperature decreases rapidly, and then gradually approaches the environmental reference as time progresses. Further analysis reveals that the rate of temperature decay is not constant, but gradually slows down over time. Comparing frame number 3 (time 1.562 seconds, temperature 27.250℃) and frame number 7 (time 1.238 seconds, temperature 27.886℃), it can be seen that although the time interval is not large, the rate of temperature decrease has begun to decrease. This decay characteristic of being fast at first and then slow is consistent with the physical law that heat diffuses from the surface to the inside during heat conduction and gradually reaches thermal equilibrium with the environment. Furthermore, comparing the measured temperature with the theoretical temperature reveals that they are very close, but not completely identical. For example, at frame number 5, the time is 0.895 seconds, the theoretical temperature is 28.795℃, and the measured temperature is 28.890℃, a difference of approximately 0.1℃. Similarly, at frame number 12, the time is 1.769 seconds, the theoretical temperature is 26.833℃, and the measured temperature is 26.790℃, also showing a slight deviation. This deviation is generally stable within ±0.1℃, reflecting the normal detection noise present in the infrared thermal imager during actual measurement, and making the data closer to the real detection scenario. The aforementioned data characteristics of exponential temperature decay over time with random fluctuations provide reliable raw data support for subsequent nonlinear least squares fitting to solve for the thermal time constant. It is precisely based on this clear decay logic and reasonable noise level that the core thermophysical parameters characterizing the health status of the component can be accurately derived from the measured data.

[0021] After obtaining the temperature decay data sequence of the reference area, it is necessary to inversely derive the standard thermal diffusivity that can characterize the thermal conductivity of the component's health status. Before performing data fitting, it is first necessary to establish a mathematical model that conforms to physical laws. This invention introduces the one-dimensional semi-infinite body model from classical thermal conduction theory to approximate the local thermal behavior of the reference area. It should be noted that the one-dimensional semi-infinite body is an idealized model in thermophysics. It regards the object under study as extending to infinity in one dimension (here, the thickness direction of the component), while there is no heat exchange or the heat exchange is negligible in other dimensions perpendicular to it. For the application scenario of this invention, when the size of a reference area is much larger than the thermal penetration depth of the heat pulse during the acquisition time, it can be reasonably assumed that the lateral thermal diffusion effect inside the area is very small, and its temperature decay behavior in the thickness direction conforms to the law of the one-dimensional semi-infinite body. Based on this theory, the temperature decay of the reference region over time can be expressed as a standard exponential decay function. Several key parameters need to be clearly defined in this function: First, the ambient temperature, which serves as the final limit of temperature decay, can be obtained by calculating the average temperature of the background region in the infrared image sequence that is far from the thermal excitation center and whose temperature has stabilized. Second, the initial temperature rise, defined as the difference between the average temperature of the reference region in the first frame of the infrared image sequence at the moment the thermal pulse is turned off and the ambient temperature; this initial temperature rise constitutes the initial amplitude of temperature decay. Finally, a core parameter to be solved is called the thermal time constant, which controls the rate at which the temperature decays from the initial temperature rise to the ambient temperature. In summary, the exponential decay function to be fitted describes the temperature decay law with the ambient temperature as the baseline, the initial temperature rise as the decay amplitude, and the thermal time constant as the decay rate control parameter. After establishing the function form, a numerical optimization algorithm is used for fitting. This invention uses a nonlinear least squares fitting algorithm, the goal of which is to find an optimal thermal time constant value that minimizes the sum of squared errors between the theoretical temperature value calculated by the above exponential decay function and the measured temperature decay data sequence in the reference area. When the sum of squared errors is minimized, the best-fit exponential decay function that best matches the measured data in the reference area is obtained. At the same time, the thermal time constant corresponding to the reference area is calculated from the fitting parameters of the function. Since the above fitting was performed independently on the three reference regions, preliminary results for three thermal time constants are obtained. However, the thermal time constants obtained at this time are quantities related to specific geometric dimensions. In order to obtain a universal parameter that is only related to the thermal properties of the material itself, it is necessary to convert it according to the classical analytical solution of the one-dimensional thermal diffusion process in the thickness direction of the flat layered encapsulation structure. According to this analytical solution, there is a definite mathematical relationship between the thermal diffusivity, the thermal time constant, and the characteristic length of the thermal diffusion path (which is the total encapsulation thickness of the component from the surface of the glass cover to the back plate). The specific conversion method is as follows: divide the thermal time constant obtained from fitting each reference region by... The standard thermal diffusivity of each reference region is calculated by multiplying the square of the total package thickness of the component by the reciprocal of the square of pi. This conversion yields an intrinsic parameter characterizing the thermal conductivity of the material at that reference region. To obtain a comprehensive parameter representing the overall health of the component, the arithmetic mean of the standard thermal diffusivity of the three reference regions is taken. This mean is used as the final standard thermal diffusivity of the component. The magnitude of this standard thermal diffusivity directly determines the speed of heat propagation within the component and is an indispensable core input for constructing a two-dimensional thermal conduction digital model. The specific formula for the standard thermal diffusivity is as follows: ,in The standard thermal diffusivity is derived from the thermal time constant. and component thickness The calculations reflect the physical ability of the module to diffuse heat from the surface to the interior and backsheet under fault-free conditions, and characterize the rate at which heat is conducted along the thickness direction in the module encapsulation material system. The total encapsulation thickness of the module is the total thickness of the photovoltaic module's glass layer, EVA layer, and backsheet layer. It can be obtained from the module's datasheet or measured in practice, and refers to the total path length of heat waves propagating in the thickness direction. The thermal time constant is the reference thermal time constant obtained from the fitting in the previous section, which is the temperature decay rate parameter of the healthy region. Pi is a mathematical constant, approximately 3.14159, which originates from the eigenvalues ​​in the analytical solutions of the one-dimensional heat conduction equation. This formula originates from the analytical solution of the heat conduction equation for a one-dimensional semi-infinite macrobe after instantaneous surface excitation by a heat source, for a thickness of... For a flat plate with an adiabatic or isothermal boundary on its back, the time constant corresponding to the dominant mode of temperature decay satisfies The inverse solution yields the result. This conversion formula will be able to measure macroscopic parameters ( , The thermal diffusivity, which characterizes the intrinsic properties of a material, is converted into this coefficient. This allows subsequent two-dimensional heat conduction digital models to be built in a personalized manner based on the actual health status of the current components, significantly improving the accuracy of theoretical predictions.

[0022] A two-dimensional heat conduction digital model was established and a theoretical thermal wave propagation image sequence was generated. Specifically, the standard thermal diffusivity coefficient was used as the only material thermal property input parameter to establish a two-dimensional heat conduction digital model to describe the evolution of the surface temperature field of the component after the end of the planar thermal pulse. Based on this two-dimensional heat conduction digital model, the measured temperature values ​​of each pixel position on the component surface recorded in the first frame of the infrared image sequence were used as the initial temperature distribution field, and the average temperature of the background area far from the thermal excitation center and whose temperature tends to be stable in the infrared image sequence was used as a constant environmental thermal boundary condition. Using the standard thermal diffusivity obtained from the fitting calculation in step two as the sole input parameter characterizing the thermal properties of the material, a partial differential equation model describing the spatiotemporal evolution of the two-dimensional temperature field on the component surface after the end of the planar thermal pulse is established, namely, a two-dimensional heat conduction digital model. The core of this model is a physical law, namely the diffusion behavior of heat in two-dimensional space. In order for this model to be able to perform calculations for the specific component, it must be given two key initial conditions and boundary conditions. The first is the initial condition, which precisely tells the model where to start. This invention directly uses the first frame of the infrared image sequence acquired in step 1 as the initial condition. In this frame, the measured temperature value of each pixel position on the component surface constitutes the initial temperature distribution field of the model. This means that the starting point of the model's evolution is completely faithful to the true initial state of this detection. The second is the boundary condition, which tells the model what the environment is like. This invention selects pixels in the infrared image sequence that are located at the edge of the component, far from the thermal excitation center, and whose temperature is relatively stable and changes very little throughout the relaxation process. The average temperature of these pixels throughout the process is calculated and used as a constant environmental thermal boundary condition input into the model. This is equivalent to setting an environmental reference value that the model's temperature will eventually approach. Based on the Green's function method in two-dimensional unsteady heat conduction theory, the theoretical temperature at any pixel location at any time is expressed as the result of spatially weighted integration of the initial temperature distribution field and superimposed with environmental thermal boundary conditions. The weighting function used in this spatial weighted integration is a Gaussian weighted distribution function with the Euclidean distance between any pixel location in the initial temperature distribution field and the target pixel location for calculating the theoretical temperature as the independent variable, and determined by the standard thermal diffusivity and the time corresponding to the current calculation frame. The above-mentioned theoretical temperature calculation method based on the Green's function corresponds to an analytical solution in continuous integral form. In actual numerical calculations, this analytical solution in continuous integral form is discretized into a two-dimensional convolution operation that matches the pixel grid coordinates of the infrared image sequence, and a fast Fourier transform algorithm is used to achieve a highly efficient numerical solution. By performing this convolution operation frame by frame, a theoretical heat wave propagation image sequence with a spatiotemporal resolution completely consistent with the infrared image sequence is generated. This model needs to be solved to obtain the theoretical temperature value at any location on the component surface at any time. This invention is based on the Green's function method in two-dimensional unsteady-state heat conduction theory. The core idea of ​​the Green's function method can be understood as a spatial weighted superposition effect. Specifically, for any target pixel on the component surface, the theoretical temperature at any target time is calculated as follows: The model traverses all pixel positions in the initial temperature distribution field. For each initial pixel, the model calculates the planar Euclidean distance between the initial pixel and the target pixel. Then, based on a Gaussian weighted distribution function determined by the standard thermal diffusivity and the current target time, the theoretical temperature is calculated for this... The initial temperature value of each initial pixel is assigned a weight. This weighting function is designed so that the closer the initial point is to the target pixel, the greater its influence on the current temperature of the target point. At the same time, as time passes, this influence spreads smoothly in the form of a Gaussian function. By multiplying the initial temperature values ​​of all initial pixels by their corresponding weights and integrating (i.e., summing them up), we obtain the total contribution of the initial temperature distribution field to the current temperature of the target point. Finally, by superimposing a constant environmental thermal boundary condition, we obtain the final theoretical temperature value of the target pixel at that target moment. The above process corresponds to an analytical solution in the form of a continuous integral in mathematics, which is physically rigorous and precise. In practical computer numerical calculations, continuous integrals cannot be directly processed. Therefore, this invention discretizes the continuous integral form and transforms it into a two-dimensional convolution operation that perfectly matches the pixel grid coordinates of the infrared image. Convolution operations have mature and fast algorithms in the fields of computer image processing and numerical calculation. This invention uses the Fast Fourier Transform technique to achieve high-efficiency convolution solution for each frame of the image. By performing this two-dimensional convolution operation frame by frame, a completely new image sequence can be generated. Each frame in this image sequence corresponds strictly one-to-one with the measured infrared image sequence acquired in step 1 in terms of timestamp and is also completely identical in terms of spatial resolution. It accurately depicts the evolution process of the surface temperature field that should be presented under the ideal condition that the internal material of the photovoltaic module is absolutely homogeneous and there are no hot spots or other thermal anomalies. This invention defines this set of virtual image sequences, which serve as a comparison benchmark, as a theoretical heat wave propagation image sequence. Thus, a precise and ideal benchmark is prepared for subsequent defect detection.

[0023] Step 3: Perform pixel-by-pixel and frame-by-frame difference calculations between the infrared image sequence and the theoretical thermal wave propagation image sequence. Identify the negative deviation region in the difference calculation. Perform time-domain integration on all difference calculation results during the temperature relaxation process to generate an integrated thermal wave shadow image. The process involves pixel-by-pixel and frame-by-frame difference calculation and identification of negative deviation regions. Specifically, for corresponding frames with identical timestamps in the infrared image sequence and the theoretical heat wave propagation image sequence, the measured infrared temperature value at the same pixel location within the same frame is subtracted from the theoretical temperature value predicted by the model. The summed differences form the difference image for that frame. In this difference image, if the measured temperature of a pixel is lower than the theoretical temperature, the difference calculation result for that pixel is negative, and that pixel is marked as a negative deviation pixel. To eliminate false negative deviations caused by random noise and instantaneous disturbances in a single frame, a continuity criterion in the time dimension is introduced. That is, for any pixel, if it presents a negative deviation pixel in three or more consecutive difference images, it is determined that the pixel is indeed in a state of heat wave propagation obstruction caused by internal defects during the duration corresponding to the three or more consecutive frames. The negative deviation amplitude of the pixel in each frame during the duration is recorded as the heat wave shadow intensity value of the pixel in the corresponding frame. The purpose of difference calculation is to quantify the degree of deviation between the measured temperature field and the ideal health model, and to filter out anomalous signals with physical significance. First, the infrared image sequence and the theoretical thermal wave propagation image sequence are strictly time-aligned. Since both sequences are generated using the same timestamp marking method, corresponding frames with the same timestamp can be accurately found. For each pair of such corresponding frames, a pixel-by-pixel subtraction operation is performed: the measured infrared temperature value at a certain pixel coordinate position in the infrared image sequence is subtracted from the theoretical predicted temperature value at the same pixel coordinate position in the same frame of the theoretical thermal wave propagation image sequence. After performing this operation on all pixels, the set of calculation results constitutes the difference image of that frame. In the difference image, the calculation result is presented as a positive or negative value. When the measured temperature value of a certain pixel is lower than the theoretical predicted temperature value, its difference calculation result is negative. This invention marks such pixels as negative deviation pixels. The appearance of these negative deviation pixels means that at the pixel location, the actual temperature drop rate is faster than predicted by the health model, or that its temperature is always lower than expected. This is because there is a defect below the location, which hinders the normal replenishment of heat or accelerates the loss of heat, thus forming a temperature deficit on the surface. However, isolated negative deviation pixels in a single frame are unreliable, as they are caused by non-defect factors such as transient electronic noise from the infrared thermal imager, weak airflow disturbances, or slight dust accumulation on component surfaces. To eliminate these false signals and ensure the physical persistence of detected anomalies, this invention introduces a continuity criterion in the time dimension. The specific execution method of this criterion is as follows: For any pixel in the image, starting from the first frame of the infrared image sequence, the state of that pixel in all difference images is scanned frame by frame; if the pixel is marked as a negative deviation pixel in three or more consecutive difference images, then... The pixel was determined not to be caused by random noise, but to be in a state of heat wave propagation obstruction caused by internal defects during this period. The present invention defines the entire time interval from the first frame that satisfies the condition of three consecutive frames to the last frame that breaks the continuous state as the heat wave propagation obstruction duration of the pixel. For a pixel in this state, the absolute value of the negative deviation value in the difference image of each frame during the obstruction duration is extracted, and this absolute value is recorded as the heat wave shadow intensity value of the pixel in that frame. The physical meaning of this intensity value is the amount of temperature loss caused by the defect at the current moment. After the above processing, in each frame of the difference image, those negative deviation pixels that simultaneously meet the continuity criterion and are assigned a thermal wave shadow intensity value together constitute a new image. This image eliminates random noise and retains only the abnormal signal with physical continuity. This invention defines it as the thermal wave shadow image of this frame. Thus, the original difference image sequence is transformed into a purer and more physically meaningful thermal wave shadow image sequence. This sequence still maintains strict alignment with the original infrared image sequence in terms of timestamps. The integral thermal shadow image is generated by performing time-domain integration. Specifically, for each pixel, the first frame of the infrared image sequence is taken as the integration start point, and the last frame corresponding to the end of the temperature relaxation process is taken as the integration end point. All frames from the start point to the end point are traversed. For each frame traversed, it is determined whether the pixel has a thermal shadow intensity value in the thermal shadow image of that frame. If it does, the intensity value is extracted and added to the sum of the pixels. If it does not, the contribution value of the pixel in that frame is recorded as zero and no accumulation is performed. After traversing all frames from the start point to the end point, each pixel obtains a single accumulated result value. The accumulated result values ​​of all pixels together constitute the integral thermal shadow image. This step compares measured and theoretical temperatures frame by frame to identify time-dependent negative biases, which are then integrated in the time domain to generate an integral image that quantifies the severity of hotspots. The formula for defining the negative bias is as follows: ,in The instantaneous negative deviation is the difference between the theoretical temperature and the measured temperature, indicating that the pixel... The degree of obstruction to heat wave propagation at any given moment; a positive value indicates that the heat wave is obstructed, while a negative value or zero indicates that it is not obstructed. The theoretical temperature, calculated from the two-dimensional heat conduction digital model in step 2, represents the temperature at which the pixel is in a healthy state. The temperature that should be reached at any time; The measured temperature is from the infrared image sequence. time The pixel value of the location represents the actual surface temperature of the component being captured, which is obtained directly from the thermal imager; Pixel coordinates are spatial indexes in the image, representing tiny areas on the surface of the corresponding component, determined by the spatial resolution of the thermal imager; when At that time, the pixel was marked as a negative deviation pixel in that frame, and its value This refers to the instantaneous intensity of the heat wave shadow in that frame; Continuity criterion: and , The consecutive frame count, defined by the number of consecutive frames where pixels are marked with negative deviation, characterizes the persistence of heat wave obstruction. That is, three or more consecutive frames; The starting frame, which corresponds to the first frame at the beginning of the continuous negative deviation state, is the starting time of the heat wave being blocked. The end frame, corresponding to the last frame marking the end of the continuous negative deviation state, is the end time of heat wave obstruction. The index of the intraframe from the start frame to the end frame; only if At that time, the pixel is in the time period Negative deviations within the range are considered to be caused by real defects and participate in subsequent integration; Isolated negative biases are considered noise and do not participate in the integration. Definition of heatwave shadow intensity value: If the pixel When a frame satisfies the continuity criterion, In other cases, ; The intensity of the heatwave shadow is the negative deviation value after continuous filtering; in the continuous time domain, the integral heatwave shadow image is defined as: Under actual discrete sampling conditions, the integral is transformed into a summation: ,in The integral thermal wave shadow image is derived from the final generated two-dimensional image. The pixel value at a location represents the total temperature loss accumulated by that pixel during the entire relaxation process; the larger the value, the more severe the defect. The starting point of integration is the instant the thermal pulse is turned off, corresponding to the moment of the first frame of the infrared sequence, which represents the starting point of the temperature relaxation process. The endpoint of the integration is the time when the temperature relaxation process ends, which is usually 5 seconds, indicating the end point of the temperature relaxation process. The intensity of the heatwave shadow is in pixels. In the The intensity value of the heat wave shadow of the frame represents the degree of instantaneous obstruction after continuous filtering, and i is the frame index from the start point to the end point of integration; The sampling interval is the reciprocal of the infrared thermal imager's frame rate. For example, 90Hz corresponds to approximately 11.11ms, which represents the time interval between adjacent frames. The total number of frames, from the start point to the end point, represents the number of time sampling points involved in the integration. The core innovation of this integral formula lies in its selective accumulation mechanism, which accumulates only negative deviations that satisfy the continuity criterion, and only positive values ​​are accumulated (because...). Defined only when And satisfying the condition of non-zero continuity), this mechanism achieves triple signal enhancement: the original weak negative deviation signal (only 0.1-0.5℃) in a single frame is amplified to the order of tens of℃·second by accumulating it over tens of frames (e.g., 90Hz × 5 seconds = 450 frames), thus improving the signal-to-noise ratio. The continuity criterion eliminates single-frame random noise and transient disturbances, ensuring that only persistent anomalies are retained; the cumulative result... The physical meaning of this is the total amount of temperature deficit, which directly corresponds to the degree to which the defect hinders the propagation of heat waves and is positively correlated with the severity of the defect. Although noise has been removed from the thermal wave shadow image, the intensity value of the thermal wave shadow caused by defects is still very weak in a single frame image, resulting in a low signal-to-noise ratio, making it difficult to directly use for subsequent morphological analysis and fault determination. To converge these weak signals scattered along the time axis into a significant feature, this invention introduces time-domain integration. The core idea of ​​time-domain integration is to accumulate the thermal wave shadow intensity of each pixel throughout the entire temperature relaxation process, thereby converting the instantaneous defect amount into the cumulative defect amount. The specific operation process is as follows: The first frame of the infrared image sequence (corresponding to the instant the thermal pulse is turned off) is used as the starting point for the integration operation, and the last frame corresponding to the end of the temperature relaxation process is used as the ending point for the integration operation. For each pixel in the image, starting from the starting frame, the process is repeated frame by frame until the ending frame. For each frame, the following is determined: whether the pixel is assigned a thermal wave shadow intensity value in the thermal wave shadow image of that frame. If it is assigned, the intensity value is extracted and added to the sum of the pixels. If the pixel does not exist in the thermal wave shadow image of that frame, that is, it is not marked as a negative deviation pixel or does not meet the continuity criterion, the contribution value of the pixel in that frame is recorded as zero, and no accumulation operation is performed. After traversing all frames from the start point to the end point, each pixel obtains a single accumulated result value. This value is the sum of the intensity values ​​of all moments during the entire temperature relaxation process when the pixel is determined to be a valid thermal wave shadow. These accumulated result values ​​of all pixels are arranged according to their original pixel positions to form a completely new image.

[0024] Step 4: Threshold segmentation is performed on the integral thermal wave shadow image to obtain the preliminary abnormal region. Morphological screening is applied to set the regions in the preliminary abnormal region that meet the triple geometric judgment criteria as candidate regions. Threshold segmentation and morphological screening are performed, specifically as follows: the global segmentation threshold is automatically calculated using the maximum inter-class variance method on the integral thermal wave shadow image; pixels with values ​​higher than the global segmentation threshold are marked as foreground, and pixels with values ​​lower than or equal to the global segmentation threshold are marked as background, to generate a preliminary binarized anomalous region image; morphological opening and closing operations are sequentially performed on the preliminary binarized anomalous region image, wherein the opening operation is erosion followed by dilation, and the closing operation is dilation followed by erosion. The purpose of threshold segmentation is to distinguish pixels with high cumulative intensity in the integral heatwave shadow image from the low-intensity background, forming binary region labels. This invention uses the maximum inter-class variance method to automatically calculate the global segmentation threshold. The principle of this method is to iterate through all grayscale thresholds and calculate the inter-class variance between the foreground and background parts after the image is segmented at that threshold. When the inter-class variance reaches its maximum value, the corresponding threshold is the optimal segmentation threshold. After calculating the global segmentation threshold, the integral heatwave shadow image is judged pixel by pixel: if the integral heatwave shadow value of a pixel is higher than the global segmentation threshold, the pixel is marked as foreground and assigned a white value in the binary image; if the pixel value is lower than or equal to the global segmentation threshold, it is marked as background and assigned a black value. After this operation, the original grayscale image is converted into a black and white binary image, which this invention defines as a preliminary binary abnormal region image. All white pixels in this image constitute the set of regions initially judged to have anomalies. Binary images obtained from direct thresholding often suffer from two types of problems: isolated white noise caused by noise or subtle texture interference; and tiny black holes or breaks within the region caused by uneven temperature distribution within the real abnormal region. Both of these problems affect the accurate extraction of the region's geometric features. Therefore, morphological operations are needed for purification. This invention performs morphological opening and closing operations sequentially. First, the morphological opening operation is performed, which consists of two steps: erosion and dilation. The erosion operation uses a predefined structuring element (a 3-pixel × 3-pixel rectangular structuring element is used in this invention) to scan the entire image. Only when the area covered by the structuring element is entirely foreground pixels is the center pixel retained as foreground; otherwise, it is changed to background. This operation can peel away isolated pixels attached to the edge of the region, thereby eliminating tiny noise. The subsequent dilation operation uses the same structuring element. As long as there is one foreground pixel in the area covered by the structuring element, the center pixel is changed to foreground. This operation can restore the main shape of the region slightly weakened by the erosion operation. The overall effect of the opening operation is to eliminate isolated tiny noise and smooth the edges of the region without significantly changing the area and basic shape of the region. After the opening operation, the morphological closing operation is performed. The closing operation consists of two steps: dilation and erosion. The dilation can bridge the small fractures and gaps inside the region and connect the sub-regions that originally belonged to the same hot spot but were broken due to uneven strength. The erosion is used to restore the region boundary that was slightly enlarged due to the dilation, shrinking it back to close to the original size. After the sequential processing of the opening and closing operations, the noise in the binarized preliminary abnormal region image is effectively suppressed, the internal holes of the real abnormal region are filled, and the edge contours are more complete and continuous. After morphological cleansing, the foreground pixels in the binarized preliminary anomalous region image form several interconnected components with independent boundaries in space. Each connected component is a preliminary anomalous region. Geometric feature parameters of each preliminary anomalous region are extracted, including its area, minimum bounding rectangle size, and principal axis direction. A triple geometric criterion is established for candidate region selection: First, the area of ​​the preliminary anomalous region is between 0.8 and 1.5 times the theoretical area of ​​a single photovoltaic cell in the photovoltaic module to be tested. Second, the aspect ratio of the minimum bounding rectangle of the preliminary anomalous region is less than 2.0. Third, the global dominant direction of the main grid lines on the module surface is extracted from the first frame of the module's infrared image sequence using the Hough transform algorithm, and the angle between the principal axis direction of the preliminary anomalous region and this global dominant direction is compared, requiring the absolute value of the angle to be less than 5 degrees. Preliminary anomalous regions that simultaneously meet the above triple geometric criterion are determined as candidate regions. After morphological cleanup, the foreground pixels in the binarized preliminary abnormal region image form several regions that are not connected to each other and have independent boundaries. In the field of image processing, these connected sets composed of adjacent foreground pixels are called connected components. This invention defines each such connected component as a preliminary abnormal region, which represents the suspected defect regions that are retained after thresholding and morphological cleanup and need to be further identified. To perform subsequent geometric screening on each preliminary anomaly region, it is necessary to extract its key geometric feature parameters. This invention extracts the following three parameters: The first parameter is the area of ​​the preliminary anomaly region. This area is calculated by counting the total number of pixels covered by the connected component and combining it with the imaging resolution and shooting distance of the infrared thermal imager to obtain the actual physical area value. The second parameter is the minimum bounding rectangle size of the preliminary anomaly region. Specifically, this involves calculating the rectangle with the smallest area that can completely contain the connected component, obtaining the length of the long side and the length of the short side of the rectangle, and then calculating the ratio of the long side to the short side. The third parameter is the principal axis direction of the preliminary anomaly region. Specifically, this involves fitting an ellipse to the connected component and using the major axis direction of the fitted ellipse as the principal axis direction of the region; or by calculating the covariance matrix of all pixels in the region and taking the direction of the eigenvector corresponding to the largest eigenvalue as the principal axis direction, which is represented by the tilt angle relative to the horizontal axis of the image. Hot spot faults in photovoltaic modules have specific physical forms. Hot spots originate inside individual cells. Due to local defects, this area becomes a load during power generation, generating Joule heat. Therefore, the morphology of the actual hot spot on the thermal image is often highly correlated with the geometric characteristics of the cell: its area is comparable to that of a single cell; its shape tends to be rectangular due to the constraints of the cell's rectangular outline; and its edge extension direction is often parallel to the grid line direction due to the directional arrangement of the main grid lines and fine grid lines within the cell. Based on the above physical characteristics, this invention sets the following three geometric judgment criteria to screen each preliminary abnormal area one by one. The first criterion is the area criterion, which compares the actual physical area of ​​the initial anomaly area with the theoretical area of ​​a single photovoltaic cell in the photovoltaic module to be tested. Since the hot spot is in the early stage of development or affected by the surrounding heat diffusion, its area may not be exactly equal to that of a complete cell. Therefore, this invention sets a reasonable tolerance range: if the area of ​​the initial anomaly area is between 0.8 and 1.5 times the theoretical area of ​​a single cell, it is considered to meet the area criterion. Areas with too small an area are noise or tiny stains, while areas with too large an area are the fusion of multiple adjacent defects or interference such as heat reflection from the frame, and are therefore excluded. The second criterion is the shape criterion, which judges the rectangular similarity of the initial anomaly area. This invention uses the aspect ratio of the minimum bounding rectangle as a quantitative indicator. If the ratio of the long side to the short side of the minimum bounding rectangle is less than 2.0, the area is considered to have good rectangular characteristics and conforms to the typical shape of a hot spot. Areas with too large an aspect ratio (i.e., too thin and long) correspond to the thermal response of the solder strip, busbar, or grid line itself, rather than cell defects, and need to be excluded. The third criterion is the direction criterion. The main axis direction of the initial anomaly area is compared with the global dominant direction of the main grid lines on the component surface. In order to obtain the global direction of the main grid lines of the component, this invention adopts the Hough transform line detection algorithm to extract all obvious line features from the first frame of the infrared image sequence, and determines the main direction of these lines, i.e., the global dominant direction, through statistical analysis. The absolute value of the angle between the main axis direction of the initial anomaly area and the global dominant direction is calculated. If the absolute value of the angle is less than 5 degrees, it is considered that the direction of the area is parallel to the grid line direction, which conforms to the physical characteristics of the real hot spot being constrained by the grid line structure. This criterion can effectively eliminate stains, scratches or non-defect thermal reflection interference with random directions.

[0025] Step 5: Map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse, to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface. To verify whether the candidate region has active heating characteristics, it is necessary to examine its initial temperature response at the moment the heat pulse ends. Since the active heating source absorbs external heat and generates Joule heat due to defects when it is excited by the heat pulse, its temperature will be significantly higher than the surrounding healthy area that is only heated by the external heat pulse at the moment the heat pulse is turned off. Cold spot defects do not generate heat actively, and may even absorb less heat due to shading, so their temperature is often lower than or equal to the surrounding area. Based on the above physical principles, the present invention performs the following operations: First, the spatial pixel coordinates of each candidate region determined in step 4 in the integrated thermal wave shadow image are precisely mapped to the first frame image of the infrared image sequence. The first frame image is the initial temperature field image acquired at the moment the thermal pulse is turned off. It completely records the initial temperature value of each position on the surface of the component at this moment. For each candidate region, the temperature values ​​of all pixels covered by the region in the first frame image are extracted, and the arithmetic mean of these values ​​is calculated. The present invention defines the average value as the initial characteristic temperature of the candidate region, which represents the overall thermal state of the region just after the thermal excitation ends. To accurately assess whether the temperature in a region is truly too high, an objective reference benchmark is needed, namely the temperature level of the surrounding healthy region. Therefore, this invention uses the minimum bounding rectangle of the candidate region as a benchmark, extending outwards at equal intervals to form a ring-shaped neighborhood that completely surrounds the region. The width of this ring-shaped neighborhood is set to be consistent with the short side dimension of the candidate region. This design aims to ensure that the collected surrounding pixels are sufficiently close to the target region to reflect the local background temperature, without over-expansion that could introduce other distant interference sources or independent defects. After determining the ring-shaped neighborhood, the temperature values ​​of all pixels within the ring-shaped neighborhood in the first frame image are extracted, and their arithmetic mean is calculated. This invention defines this average value as the initial background characteristic temperature of the candidate region. Subtracting the initial characteristic temperature of the background from the initial characteristic temperature of the candidate region yields the relative temperature rise of the candidate region relative to the surrounding background at the moment the heat pulse ends. This relative temperature rise directly quantifies whether the region exhibits autonomous heating beyond the background. To verify whether the candidate region has already exhibited high temperature at the end of the thermal pulse, the following steps are taken: map the spatial pixel position of the candidate region in the integrated thermal wave shadow image to the first frame of the infrared image sequence, i.e., the initial temperature field image acquired at the moment the thermal pulse is turned off; extract the initial temperature value of all pixels in the candidate region in this frame and calculate their arithmetic mean, which is used as the initial characteristic temperature of the candidate region. To determine the relative temperature rise, a reasonable temperature difference threshold needs to be set. This threshold is based on extensive experimental data and field application experience. The core idea is that only when the temperature rise caused by self-heating reaches a certain level can it be confirmed as a hot spot caused by an electrical defect, thus avoiding misjudging slight temperature differences caused by measurement errors or minor environmental fluctuations as defects. This invention sets the preset temperature difference threshold range to 1.5 degrees Celsius to 2.5 degrees Celsius. The specific selection of this range can be adjusted according to the field environmental conditions: when the ambient temperature is low and the thermal imager noise is low, a lower threshold (such as 1.5 degrees Celsius) is selected to improve detection sensitivity; when the ambient temperature is high or there is strong airflow disturbance, a higher threshold (such as 2.5 degrees Celsius) is selected to ensure detection reliability. For each candidate area, its calculated relative temperature rise value is compared with the preset temperature difference threshold. If the relative temperature rise of the candidate area reaches or exceeds the preset temperature difference judgment threshold, it is determined that the candidate area has actively generated heat as an active heat source during the thermal pulse excitation stage. This heat generation behavior originates from the Joule heating effect generated when the module is subjected to external excitation or generates its own electricity due to electrical defects such as cell cracking, aging, low bypass resistance, or poor soldering of the solder strip. Such defects are the real hot spot faults that this invention aims to detect. Therefore, the candidate area is finally confirmed as a real hot spot. Conversely, if the relative temperature rise of the candidate area is lower than the preset temperature difference judgment threshold, or even negative (i.e., the temperature of the area is lower than the surrounding background), it is determined that the area is a non-heating defect, including but not limited to dust accumulation on the backsheet surface, deposits on the glass surface, and cold spots with microcracks in the cell that have not yet formed a continuous heating channel. Although such defects leave faint traces in the integrated thermal wave shadow image, they are not active heat sources and do not belong to the hot spot faults defined in this invention. They are excluded and not included in the final detection result statistics. Using the minimum bounding rectangle of the candidate region as a reference, an annular neighborhood is formed by extending outward at equal intervals to enclose the region. The width of the annular neighborhood is set to be consistent with the short side size of the candidate region. The initial temperature values ​​of all pixels in the annular neighborhood in the first frame are extracted and their arithmetic mean is calculated as the initial feature temperature of the background. The initial feature temperature of the candidate region is subtracted from the initial feature temperature of the background to obtain the relative temperature rise value of the candidate region at the instant the thermal pulse ends. If the relative temperature rise value reaches or exceeds the preset temperature difference judgment threshold, it is determined whether the candidate region has shown a high temperature area at the end of the planar thermal pulse, and it is determined to be a real hot spot caused by internal electrical defects; otherwise, it is determined to be a non-heating defect and is excluded. The cumulative intensity value of thermal wave shadow is labeled as follows: For each candidate area identified as a real hot spot, all pixels covered by the area are located in the integrated thermal wave shadow image. The integrated thermal wave shadow values ​​of these pixels are extracted and their arithmetic mean is calculated. This average value is used as the cumulative intensity value of the thermal wave shadow of the real hot spot. Based on the numerical range of the cumulative intensity value of the thermal wave shadow, three fault levels—mild, moderate, and severe—are pre-divided. On the first frame of the infrared image sequence, the outer position of each real hot spot is marked with a bright colored rectangle, and the cumulative intensity value of the thermal wave shadow of the real hot spot and the corresponding fault level are simultaneously presented in the form of text labels in the area adjacent to the rectangle. This invention uses the cumulative intensity value of thermal wave shadow as the core indicator for quantifying the severity of real hot spots. The calculation method of this indicator is as follows: In the integrated thermal wave shadow image, all pixels covered by the real hot spot are repositioned, the integrated thermal wave shadow values ​​of each of these pixels are extracted, and the arithmetic mean of these values ​​is calculated. The average value is the cumulative intensity value of the thermal wave shadow of the real hot spot. The physical meaning of this value is that it directly measures the total amount of temperature deficit accumulated in the hot spot area during the entire temperature relaxation process. The higher the value, the stronger the obstruction effect of the location on the propagation of heat waves, and the more severe the degree of degradation of the electrical performance or damage to the physical structure of the corresponding solar cell. To facilitate maintenance personnel in quickly grasping the urgency of a fault, this invention pre-classifies three fault levels based on the distribution range of the cumulative intensity of thermal wave shadows. Specifically, the classification method is as follows: through statistical analysis of a large amount of historical fault data, threshold ranges for three levels—mild, moderate, and severe—are determined. For example, faults below a certain empirical threshold L1 are classified as mild faults, those between L1 and L2 are classified as moderate faults, and those above L2 are classified as severe faults. These thresholds can be calibrated and adjusted according to different photovoltaic module models and on-site application requirements. Finally, this invention performs visual annotation on the first frame of the infrared image sequence. For each real hot spot, its outer position is marked with a bright colored rectangle (based on its smallest outer rectangle). In the vicinity of the rectangle, the cumulative intensity value of the thermal wave shadow of the real hot spot and the corresponding fault level are presented in the form of text labels. For example, hot spot intensity: 8.5℃·s level: moderate. This annotation method allows maintenance personnel to obtain the precise location and severity information of each hot spot in the first instance.

[0026] Please see Figure 4 The present invention also provides a rapid detection device for hot spot faults on the surface of photovoltaic modules. The device is used to perform the above-described rapid detection method for hot spot faults on the surface of photovoltaic modules, comprising: The image acquisition module is used to apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied using a high frame rate infrared thermal imager. The theoretical calculation module is used to select a known fault-free reference area on the surface of the component in the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use the model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. The integration calculation module is used to perform pixel-by-pixel and frame-by-frame difference calculation between the infrared image sequence and the theoretical thermal wave propagation image sequence. In the difference calculation, the negative deviation area is identified, and all difference calculation results during the temperature relaxation process are integrated in the time domain to generate an integrated thermal wave shadow image. The candidate screening module is used to perform threshold segmentation on the integral thermal wave shadow image to obtain the preliminary abnormal area, and apply morphological screening to set the region in the preliminary abnormal area that meets the triple geometric judgment criterion as the candidate area. The hot spot detection module is used to map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse, so as to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface.

[0027] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0028] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0029] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0030] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for rapid detection of hot spot faults on the surface of photovoltaic modules, characterized in that, The specific steps include: Step 1: Apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and use a high frame rate infrared thermal imager to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied. Step 2: Select a known fault-free reference area on the component surface from the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use this model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. Step 3: Perform pixel-by-pixel and frame-by-frame difference calculations between the infrared image sequence and the theoretical thermal wave propagation image sequence. Identify the negative deviation region in the difference calculation. Perform time-domain integration on all difference calculation results during the temperature relaxation process to generate an integrated thermal wave shadow image. Step 4: Threshold segmentation is performed on the integral thermal wave shadow image to obtain the preliminary abnormal region. Morphological screening is applied to set the regions in the preliminary abnormal region that meet the triple geometric judgment criteria as candidate regions. Step 5: Map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface.

2. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 1, characterized in that: Apply a uniform planar thermal pulse, specifically by using a portable flexible planar heat source that matches the surface size of the photovoltaic module to be tested, ensuring that its working surface is parallel to the module surface and in zero-gap contact, and applying a rectangular wave thermal pulse with a duration between 0.3 seconds and 0.8 seconds. The infrared image sequence of the temperature relaxation process is acquired by using an uncooled infrared thermal imager with a frame rate of not less than 90Hz, taking the moment when the thermal pulse is turned off as the zero point of timing, and continuously acquiring the evolution process of the component surface temperature field for a duration of not less than 5 seconds to ensure that the complete temperature relaxation process of the heat wave transitioning from the surface convection heat transfer-dominated stage to the internal heat diffusion-dominated stage is recorded.

3. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 2, characterized in that: The process involves selecting fault-free reference areas and analyzing their temperature decay curves. Specifically, in the first frame of the infrared image sequence, an automatic identification algorithm based on backplane texture features is used to delineate at least three rectangular areas within the component's border neighborhood that have never experienced hot spot faults according to historical maintenance records. These areas are designated as fault-free reference areas, with each reference area having a size of at least 30 pixels × 30 pixels. All frames of the entire infrared image sequence are traversed, and the instantaneous temperature values ​​of all pixels within each reference area in each frame are extracted and their arithmetic mean is calculated to obtain the average temperature of each reference area in that frame. After arranging all frames in chronological order, each fault-free reference area obtains a temperature decay data sequence composed of the average temperatures of each frame. The standard thermal diffusivity is then obtained through curve fitting, as detailed below: Using a fault-free reference area's temperature decay data sequence during the temperature relaxation process, and taking the ambient temperature as a baseline, the difference between the average temperature of the reference area in the first frame of the infrared image sequence and the ambient temperature is defined as the initial temperature rise. The temperature decay over time is expressed as an exponential decay function with the ambient temperature as the baseline, the initial temperature rise as the decay amplitude, and the thermal time constant to be solved as the decay rate control parameter. A nonlinear least squares fitting algorithm is used to numerically fit the temperature decay data sequence of the reference area, minimizing the sum of squared errors between the exponential decay function and the temperature decay data sequence to obtain the best-fit exponential function for the reference area. A decay function is used to construct the temperature decay curve of the reference region during the temperature relaxation process. The thermal time constant corresponding to the reference region is obtained from the fitting parameters of the best-fit exponential decay function. Based on the classical analytical solution of the one-dimensional thermal diffusion process in the thickness direction of the flat layered packaging structure, the thermal time constant is divided by the square of the total packaging thickness of the component, and the quotient is multiplied by the reciprocal of the square of pi to calculate the standard thermal diffusivity corresponding to the reference region. The above operation is repeated for all selected fault-free reference regions to obtain multiple standard thermal diffusivity calculation results. The arithmetic mean of these calculation results is taken as the standard thermal diffusivity of the component in a healthy state.

4. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 3, characterized in that: A two-dimensional heat conduction digital model was established and a theoretical thermal wave propagation image sequence was generated. Specifically, the standard thermal diffusivity coefficient was used as the only material thermal property input parameter to establish a two-dimensional heat conduction digital model to describe the evolution of the surface temperature field of the component after the end of the planar thermal pulse. Based on this two-dimensional heat conduction digital model, the measured temperature values ​​of each pixel position on the component surface recorded in the first frame of the infrared image sequence were used as the initial temperature distribution field, and the average temperature of the background area far from the thermal excitation center and whose temperature tends to be stable in the infrared image sequence was used as a constant environmental thermal boundary condition. Based on the Green's function method in two-dimensional unsteady heat conduction theory, the theoretical temperature at any pixel location at any time is expressed as the result of spatially weighted integration of the initial temperature distribution field and superimposed with environmental thermal boundary conditions. The weighting function used in this spatial weighted integration is a Gaussian weighted distribution function with the Euclidean distance between any pixel location in the initial temperature distribution field and the target pixel location for calculating the theoretical temperature as the independent variable, and determined by the standard thermal diffusivity and the time corresponding to the current calculation frame. The above-mentioned theoretical temperature calculation method based on the Green's function corresponds to an analytical solution in continuous integral form. In actual numerical calculations, this analytical solution in continuous integral form is discretized into a two-dimensional convolution operation that matches the pixel grid coordinates of the infrared image sequence, and a fast Fourier transform algorithm is used to achieve a highly efficient numerical solution. By performing this convolution operation frame by frame, a theoretical heat wave propagation image sequence with a spatiotemporal resolution completely consistent with the infrared image sequence is generated.

5. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 4, characterized in that: The process involves pixel-by-pixel and frame-by-frame difference calculation and identification of negative deviation regions. Specifically, for corresponding frames with identical timestamps in the infrared image sequence and the theoretical heat wave propagation image sequence, the measured infrared temperature value at the same pixel location within the same frame is subtracted from the theoretical temperature value predicted by the model. The summed differences form the difference image for that frame. In this difference image, if the measured temperature of a pixel is lower than the theoretical temperature, the difference calculation result for that pixel is negative, and that pixel is marked as a negative deviation pixel. To eliminate false negative deviations caused by random noise and instantaneous disturbances in a single frame, a continuity criterion in the time dimension is introduced. That is, for any pixel, if it presents a negative deviation pixel in three or more consecutive difference images, it is determined that the pixel is indeed in a state of heat wave propagation obstruction caused by internal defects during the duration corresponding to the three or more consecutive frames. The negative deviation amplitude of the pixel in each frame during the duration is recorded as the heat wave shadow intensity value of the pixel in the corresponding frame. The integral thermal shadow image is generated by performing time-domain integration. Specifically, for each pixel, the first frame of the infrared image sequence is taken as the integration start point, and the last frame corresponding to the end of the temperature relaxation process is taken as the integration end point. All frames from the start point to the end point are traversed. For each frame traversed, it is determined whether the pixel has a thermal shadow intensity value in the thermal shadow image of that frame. If it does, the intensity value is extracted and added to the sum of the pixels. If it does not have a value, the contribution value of the pixel in the frame is recorded as zero and is not accumulated. After traversing all frames from the start point to the end point, each pixel obtains a single accumulated result value. The accumulated result values ​​of all pixels together constitute the integral thermal wave shadow image.

6. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 5, characterized in that: Threshold segmentation and morphological screening are performed, specifically: the global segmentation threshold is automatically calculated using the maximum inter-class variance method on the integral heat wave shadow image, pixels with pixel values ​​higher than the global segmentation threshold are marked as foreground, and pixels with values ​​lower than or equal to the global segmentation threshold are marked as background, so as to generate a preliminary binarized abnormal region image. Morphological opening and closing operations are sequentially performed on the binarized preliminary abnormal region map. The opening operation is erosion followed by dilation, and the closing operation is dilation followed by erosion. After morphological cleansing, the foreground pixels in the binarized preliminary anomalous region image form several interconnected components with independent boundaries in space. Each connected component is a preliminary anomalous region. Geometric feature parameters of each preliminary anomalous region are extracted, including the region area, minimum bounding rectangle size, and principal axis direction. A triple geometric judgment criterion is set for candidate region screening: First, the region area of ​​the preliminary anomalous region is between 0.8 and 1.5 times the theoretical area of ​​a single photovoltaic cell of the photovoltaic module to be detected; second, the aspect ratio of the minimum bounding rectangle of the preliminary anomalous region is less than 2.0; third, the global dominant direction of the main grid line on the surface of the module is extracted from the first frame of the module infrared image sequence using the Hough transform algorithm, and the angle between the principal axis direction of the preliminary anomalous region and the global dominant direction is compared, requiring the absolute value of the angle between the two to be less than 5 degrees. Preliminary anomalous regions that simultaneously meet the above triple geometric judgment criteria are judged as candidate regions.

7. The method for rapid detection of hot spot faults on the surface of a photovoltaic module according to claim 6, characterized in that: To verify whether the candidate region has already exhibited high temperature at the end of the thermal pulse, the following steps are taken: map the spatial pixel position of the candidate region in the integrated thermal wave shadow image to the first frame of the infrared image sequence, i.e., the initial temperature field image acquired at the moment the thermal pulse is turned off; extract the initial temperature value of all pixels in the candidate region in this frame and calculate their arithmetic mean, which is used as the initial characteristic temperature of the candidate region. Using the minimum bounding rectangle of the candidate region as a reference, an annular neighborhood is formed by extending outward at equal intervals to enclose the region. The width of the annular neighborhood is set to be consistent with the short side size of the candidate region. The initial temperature values ​​of all pixels in the annular neighborhood in the first frame are extracted and their arithmetic mean is calculated as the initial feature temperature of the background. The initial feature temperature of the candidate region is subtracted from the initial feature temperature of the background to obtain the relative temperature rise value of the candidate region at the instant of the end of the thermal pulse. If the relative temperature rise value reaches or exceeds the preset temperature difference judgment threshold, it is determined whether the candidate region has shown a high temperature area at the end of the planar thermal pulse, and it is determined to be a real hot spot caused by internal electrical defects. Conversely, if the defect is not exothermic, it is determined to be a non-exothermic defect and is excluded. The cumulative intensity value of thermal wave shadow is labeled as follows: For each candidate area identified as a real hot spot, all pixels covered by the area are located in the integrated thermal wave shadow image. The integrated thermal wave shadow values ​​of these pixels are extracted and their arithmetic mean is calculated. This average value is used as the cumulative intensity value of the thermal wave shadow of the real hot spot. Based on the numerical range of the cumulative intensity value of the thermal wave shadow, three fault levels—mild, moderate, and severe—are pre-divided. On the first frame of the infrared image sequence, the outer position of each real hot spot is marked with a bright colored rectangle, and the cumulative intensity value of the thermal wave shadow of the real hot spot and the corresponding fault level are simultaneously presented in the form of text labels in the area adjacent to the rectangle.

8. A rapid detection device for hot spot faults on the surface of photovoltaic modules, characterized in that: The device is used to perform a rapid detection method for hot spot faults on the surface of a photovoltaic module as described in any one of claims 1-7, comprising: The image acquisition module is used to apply a uniform planar thermal pulse to the surface of the photovoltaic module to be tested, and to acquire an infrared image sequence of the module surface during the temperature relaxation process after the thermal pulse is applied using a high frame rate infrared thermal imager. The theoretical calculation module is used to select a known fault-free reference area on the surface of the component in the infrared image sequence, analyze its temperature decay curve during the temperature relaxation process, fit the standard thermal diffusivity of the component under healthy conditions, establish a two-dimensional thermal conduction digital model for the propagation of heat waves from the excitation surface outward, use the model to predict the theoretical temperature distribution of the component surface under healthy conditions, and generate a theoretical heat wave propagation image sequence. The integration calculation module is used to perform pixel-by-pixel and frame-by-frame difference calculation between the infrared image sequence and the theoretical thermal wave propagation image sequence. In the difference calculation, the negative deviation area is identified, and all difference calculation results during the temperature relaxation process are integrated in the time domain to generate an integrated thermal wave shadow image. The candidate screening module is used to perform threshold segmentation on the integral thermal wave shadow image to obtain the preliminary abnormal area, and apply morphological screening to set the region in the preliminary abnormal area that meets the triple geometric judgment criterion as the candidate area. The hot spot detection module is used to map the candidate area back to the infrared image sequence, check whether the candidate area has shown a high temperature area at the end of the planar thermal pulse, so as to confirm that it is a heat source rather than a cold spot, output the candidate area that passes the check as the real hot spot, and mark its position and shadow intensity on the component surface.

Images