Automatic grading method and system for solar photovoltaic panel products

Abstract

The invention discloses an automatic grading method and system for solar photovoltaic panel products, and relates to the technical field of automatic grading of photovoltaic panels. The automatic grading method for the solar photovoltaic panel product comprises the following steps: acquiring IV curve data and ultraviolet fluorescence image time sequence data of the solar photovoltaic panel product; performing electrical analysis processing on the IV curve data, and extracting an electrical attenuation characteristic value; and based on a pre-trained photovoltaic ultraviolet aging mapping model, performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data to obtain an aging significant feature value, and performing comprehensive analysis on the aging significant feature value and an electrical attenuation feature value to obtain a recovery residual value score of the solar photovoltaic panel product. According to the invention, automatic grading processing is carried out on the solar photovoltaic panel product through grading of the recovery residual value, so that comprehensive evaluation of the recovery value of the photovoltaic panel is realized, and it is ensured that the grading result can truly reflect the actual state of the photovoltaic panel.

Description

Technical Field

[0001] This invention relates to the field of automatic grading technology for photovoltaic panels, specifically to an automatic grading method and system for solar photovoltaic panel products. Background Technology

[0002] The installed capacity of solar photovoltaic panels is growing explosively. The photovoltaic panels in the early distributed power stations, centralized power stations and residential photovoltaic systems have gradually entered the core retirement cycle. The market size of photovoltaic panel recycling continues to expand. The efficient recycling of retired photovoltaic panels can not only realize the recycling of scarce resources such as silicon materials and metal frames and reduce dependence on the mining of primary resources, but also avoid the environmental pollution risks caused by heavy metals and polymer materials in waste components.

[0003] However, current grading technologies in the field of solar photovoltaic panel recycling still have shortcomings that make it difficult to adapt to the needs of large-scale and precise recycling. For example, the grading process is highly dependent on manual operation. Existing technologies mostly rely on manual observation of the appearance of photovoltaic panels to observe aging, combined with basic parameters such as open circuit voltage tested by simple multimeters, and then judging the grading level based on experience. This method is not only extremely inefficient, but also greatly affected by the experience and subjective judgment of the inspectors. The grading consistency of photovoltaic panels in the same batch is poor, making it difficult to meet the needs of batch recycling scenarios with thousands of panels per day.

[0004] The limitations of existing technologies include at least the following problems: Existing technologies struggle to accurately assess residual value based on the inherent correlation between the electrical performance degradation characteristics of recycled solar photovoltaic panels and the aging characteristics of encapsulation materials. This makes it difficult to avoid misjudging high-power, high-risk modules when faced with complex retirement scenarios involving intertwined electrical performance and material conditions. Furthermore, existing technologies neglect the coupling relationship between electrical degradation and aging, resulting in a lack of integrated basis for residual value assessment. This can easily lead to problems such as inflated values ​​and missed risks in grading results. In large-scale, highly heterogeneous recycling scenarios, this can easily reduce resource recovery efficiency. Moreover, existing technologies fail to combine electrical and optical multimodal data to deconstruct and integrate module conditions, which can easily lead to inaccurate grading standards. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an automatic grading method and system for solar photovoltaic panel products, which solves the problem that existing technologies, due to the lack of coordinated evaluation of electrical performance and encapsulation aging, are prone to inaccurate grading standards.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an automatic grading method for solar photovoltaic panel products, comprising the following steps: acquiring IV curve data and ultraviolet fluorescence image time-series data of the solar photovoltaic panel products; performing electrical analysis processing on the IV curve data of the solar photovoltaic panel products to extract electrical degradation feature values ​​of the solar photovoltaic panel products; performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of the solar photovoltaic panel products based on a pre-trained photovoltaic ultraviolet aging mapping model to obtain significant aging feature values ​​of the solar photovoltaic panel products; analyzing the residual value score of the solar photovoltaic panel products based on the electrical degradation feature values ​​and significant aging feature values ​​of the solar photovoltaic panel products; and automatically grading the solar photovoltaic panel products based on the residual value score.

[0007] Furthermore, the IV curve data specifically refers to IV data pairs at each time point. The specific steps for extracting the electrical degradation characteristic values ​​of solar photovoltaic panel products are as follows: Based on the IV curve data of solar photovoltaic panel products, analyze the electrical evaluation dataset of solar photovoltaic panel products, including power degradation gradient values, electrical degradation toughness values, and curve distortion degradation values; perform fusion processing on the electrical evaluation dataset of solar photovoltaic panel products to obtain the electrical degradation characteristic values ​​of solar photovoltaic panel products.

[0008] Furthermore, the specific steps for analyzing the electrical performance evaluation dataset of solar photovoltaic panel products are as follows: initial processing of the IV curve data of the solar photovoltaic panel products; parsing and integrating the initially processed IV curve data of the solar photovoltaic panel products to obtain the power attenuation gradient value, electrical attenuation toughness value, and curve distortion attenuation value of the solar photovoltaic panel products.

[0009] Furthermore, the ultraviolet fluorescence image time-series data includes several frames of ultraviolet fluorescence image data, and the ultraviolet fluorescence image data specifically includes the gray value of each pixel in the ultraviolet fluorescence image and the corresponding two-dimensional coordinates. The photovoltaic ultraviolet aging mapping model includes an input layer, a recognition and segmentation layer, an aging deconstruction layer, and an output layer.

[0010] Further, the specific steps for obtaining the significant aging characteristic values ​​of solar photovoltaic panel products are as follows: Input the ultraviolet fluorescence image data of the solar photovoltaic panel products into a pre-trained photovoltaic ultraviolet aging mapping model; analyze the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel products, including connectivity entropy feature values, gray-level heterogeneity feature values, and power-law distribution feature values; based on the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel products, extract the aging contribution ratio of the corresponding frame of ultraviolet fluorescence image; based on the ultraviolet aging mapping feature set and aging contribution ratio of each frame of ultraviolet fluorescence image of the solar photovoltaic panel products, analyze the significant aging characteristic values ​​of the solar photovoltaic panel products.

[0011] Further, the specific steps for analyzing the UV aging mapping feature set of each frame of UV fluorescence image of a solar photovoltaic panel product are as follows: In the input layer of the photovoltaic UV aging mapping model, the UV fluorescence image data of each frame of the solar photovoltaic panel product is received and preprocessed; in the recognition and segmentation layer of the photovoltaic UV aging mapping model, pixel segmentation processing is performed on the preprocessed UV fluorescence image data of each frame of the solar photovoltaic panel product to obtain the aging region set of the corresponding frame of UV fluorescence image; in the aging deconstruction layer of the photovoltaic UV aging mapping model, aging mapping processing is performed on the aging region set of each frame of UV fluorescence image of the solar photovoltaic panel product to obtain the aging feature vector of the corresponding frame of UV fluorescence image; in the output layer of the photovoltaic UV aging mapping model, based on the aging feature vector of each frame of UV fluorescence image of the solar photovoltaic panel product, the UV aging mapping feature set of the corresponding frame of UV fluorescence image is output.

[0012] Furthermore, the specific steps for extracting the aging contribution ratio of each frame of ultraviolet fluorescence image of a solar photovoltaic panel product are as follows: Based on the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, extract the aging mapping geometric mean of the corresponding frame of ultraviolet fluorescence image; based on the aging mapping geometric mean of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, analyze the aging contribution ratio of the corresponding frame of ultraviolet fluorescence image.

[0013] Further, the specific steps for analyzing the residual value scoring of solar photovoltaic panel products are as follows: read the electrical degradation characteristic value and aging significant characteristic value of the solar photovoltaic panel product, and perform normalization processing; input the normalized electrical degradation characteristic value and aging significant characteristic value of the solar photovoltaic panel product into the preset residual value assessment model to obtain the residual value score of the solar photovoltaic panel product.

[0014] Furthermore, the specific steps for automatically grading solar photovoltaic panel products based on recycling residual value scores are as follows: compare the recycling residual value scores of solar photovoltaic panel products with preset recycling residual value score ranges; and automatically grade solar photovoltaic panel products based on the comparison results.

[0015] An automatic grading system for solar photovoltaic (PV) panel products includes: a data acquisition module for acquiring IV curve data and ultraviolet fluorescence image time-series data of the solar PV panel products; an electrical degradation analysis module for performing electrical analysis processing on the IV curve data of the solar PV panel products to extract electrical degradation feature values; an ultraviolet aging analysis module for performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of the solar PV panel products based on a pre-trained photovoltaic ultraviolet aging mapping model to obtain significant aging feature values ​​of the solar PV panel products; a residual value assessment module for analyzing the recycling residual value score of the solar PV panel products based on the electrical degradation feature values ​​and significant aging feature values; and an automatic grading module for automatically grading the solar PV panel products based on the recycling residual value score.

[0016] The present invention has the following beneficial effects: (1) The automatic grading method for solar photovoltaic panel products obtains the IV curve data and ultraviolet fluorescence image time series data of the photovoltaic panel, and uses them to analyze the electrical decay characteristic value and the significant aging characteristic value. After normalization, the two types of characteristics are deeply integrated through the residual value evaluation model, which not only covers the core electrical performance indicators, but also takes into account the aging state of the packaging, so as to realize the comprehensive evaluation of the recycling value of the photovoltaic panel, ensure that the grading results can truly reflect the actual state of the photovoltaic panel, greatly improve the reliability of grading, and make the recycling residual value score of the photovoltaic panel product and the final grading result more stable, avoid the evaluation error caused by the disconnect between electrical and aging characteristics, and enhance the accuracy of the evaluation of the quality of the photovoltaic panel in the actual recycling process.

[0017] (2) The automatic grading method for solar photovoltaic products introduces a photovoltaic ultraviolet aging mapping model to process the ultraviolet fluorescence image of photovoltaic products. Based on multi-level feature extraction and analysis, it significantly improves the accuracy of aging area identification and aging feature extraction. For example, in the identification segmentation layer, the aging area is accurately identified through pixel segmentation, and the aging feature vector is extracted in the aging deconstruction layer. This enables the accurate capture of the aging changes of the photovoltaic panel and the formation of comprehensive aging salient features, thereby effectively improving the ability to assess the degree of aging of the photovoltaic panel and thus improving the accuracy of automatic grading.

[0018] (3) The automatic grading method for solar photovoltaic products significantly improves the accuracy of the assessment of photovoltaic performance degradation by deeply mining the IV curve data of photovoltaic products. First, the IV curve data is initially processed and then analyzed and integrated to extract the electrical performance assessment dataset, which can intuitively reflect the electrical performance degradation of photovoltaic panels during use. By fusing these electrical performance assessment data, more accurate electrical attenuation characteristic values ​​are obtained, which can more comprehensively depict the overall electrical performance degradation of photovoltaic panels. While improving the grading accuracy, it also significantly improves the efficiency of large-scale recycling and processing operations.

[0019] (4) The automatic grading system for solar photovoltaic panels realizes intelligent processing of the entire chain of grading assessment of retired photovoltaic panels through hierarchical collaboration between modules. The data acquisition module acquires the IV curve data and ultraviolet fluorescence image time series data of the photovoltaic panels; the electrical degradation analysis module and the ultraviolet aging analysis module work in parallel to perform deep feature extraction on the two types of data to obtain electrical degradation feature values ​​and aging significant feature values; the residual value assessment module receives the above features and performs normalization and fusion analysis to generate a recycling residual value score; finally, the automatic grading module automatically executes the grading decision based on the comparison result of the score and the preset interval, which significantly improves the response efficiency of the system.

[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0021] Figure 1 This is a flowchart of an automatic grading method for solar photovoltaic panel products according to the present invention.

[0022] Figure 2 This is a flowchart illustrating the specific steps involved in obtaining significant aging characteristic values ​​of solar photovoltaic panel products in an automatic grading method for solar photovoltaic panel products according to the present invention.

[0023] Figure 3 This is a schematic diagram of the UV aging mapping time series feature set data of solar photovoltaic panel products in an automatic grading method for solar photovoltaic panel products according to the present invention.

[0024] Figure 4 This is a block diagram of an automatic sorting system for solar photovoltaic panel products according to the present invention. Detailed Implementation

[0025] Please see Figure 1This invention provides a technical solution: an automatic grading method for solar photovoltaic panel products, comprising the following steps: within a preset recycling grading assessment period (e.g., 30-60 seconds), acquiring IV curve data and ultraviolet fluorescence image time-series data of solar photovoltaic panel products (which may be recyclable crystalline silicon encapsulated solar photovoltaic panels); performing electrical analysis processing on the IV curve data of the solar photovoltaic panel products to extract electrical degradation characteristic values ​​of the solar photovoltaic panel products; based on a pre-trained photovoltaic ultraviolet aging mapping model, performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of the solar photovoltaic panel products to obtain significant aging characteristic values ​​of the solar photovoltaic panel products; analyzing the recycling residual value score of the solar photovoltaic panel products based on the electrical degradation characteristic values ​​and significant aging characteristic values ​​of the solar photovoltaic panel products; and performing automatic (recycling) grading processing on the solar photovoltaic panel products based on the recycling residual value score.

[0026] Specifically, the IV curve data refers to IV data pairs (i.e., current and voltage) at each time point. The specific steps for extracting the electrical degradation characteristic values ​​of solar photovoltaic (PV) panel products are as follows: Based on the IV curve data of solar PV panel products, analyze the electrical evaluation dataset of solar PV panel products, including power degradation gradient values, electrical degradation toughness values, and curve distortion degradation values; fuse the electrical evaluation dataset of solar PV panel products to obtain the electrical degradation characteristic values ​​of solar PV panel products. Specifically, the power degradation gradient values, electrical degradation toughness values, and curve distortion degradation values ​​of solar PV panel products are weighted. In this weighting process, the electrical degradation toughness value needs to be transformed using the reciprocal suppression mapping function f(x) = 1 / (1+x) to obtain the electrical degradation characteristic values ​​of solar PV panel products. It should be noted that the weighting coefficients corresponding to each parameter in this weighting process can be obtained through the following steps: Obtain the power attenuation gradient value, electrical attenuation toughness value, and curve distortion attenuation value of several solar photovoltaic panel products (the same type of solar photovoltaic panel products as in this implementation example), and extract the information entropy value of the corresponding parameters. Then, transform the corresponding information entropy value using the reciprocal suppression mapping function f(x)=1 / (1+x), such as 1 / (1+information entropy value of power attenuation gradient value), and sum them to obtain the information entropy sum value. Ratio the information entropy value of the corresponding parameter after transformation with the information entropy sum value to obtain the weighting coefficient corresponding to each parameter.

[0027] The specific steps for analyzing the electrical performance evaluation dataset of solar photovoltaic panel products are as follows: Initial processing of the IV curve data of the solar photovoltaic panel products is performed, namely: screening IV data pairs on the curve where the voltage value is close to 0 (e.g., voltage ≤ 0.01V), the corresponding current value of this IV data pair is the short-circuit current; and screening IV data pairs on the curve where the current value is close to 0 (e.g., current ≤ 0.01A), the corresponding voltage value of this IV data pair is the open-circuit voltage. All IV data pairs on the IV curve are iterated to extract the power value corresponding to each IV data pair at each time point, and the maximum power value is calculated. This maximum power value is then compared with the short-circuit current and open-circuit voltage, i.e., maximum power value / (short-circuit current × open-circuit voltage), to obtain the fill factor. The IV curve data of the initially processed solar photovoltaic (PV) panel products are analyzed and integrated to obtain the power attenuation gradient value, electrical attenuation toughness value, and curve distortion attenuation value. Specifically, for each IV data pair, the voltage values ​​are sorted from smallest to largest, and the power value corresponding to each IV data pair in the sequence is read. The power values ​​of two adjacent IV data pairs in the sequence are differenced (absolute value is taken) and compared with the rated maximum power value of the solar PV panel products (in the technical manual) stored in the database to obtain a relative difference sequence. All relative differences are accumulated to obtain the power attenuation gradient accumulation value. At the same time, the voltage values ​​of the IV data pairs corresponding to the open-circuit voltage and short-circuit current are differenced, and the reciprocal is taken. This reciprocal is then multiplied with the power attenuation gradient accumulation value to extract the power attenuation gradient value (used to characterize the uniformity of the output power attenuation distribution in the entire voltage range from the short-circuit state to the open-circuit state of the PV panel; the smaller the value, the more severe the power fluctuation, the more uneven the attenuation distribution, and the more serious the electrical attenuation). The short-circuit current, open-circuit voltage, and fill factor are compared with the rated values ​​of short-circuit current, open-circuit voltage, and fill factor stored in the database (it should be noted that the above rated values ​​can be obtained from the technical manual of the solar photovoltaic panel product) to obtain the relative attenuation ratio of short-circuit current, relative attenuation ratio of open-circuit voltage, and relative attenuation ratio of fill factor. These are then combined and calculated as: relative attenuation ratio of short-circuit current × √(relative attenuation ratio of open-circuit voltage × relative attenuation ratio of fill factor) to extract the electrical degradation toughness value (used to characterize the output stability of the photovoltaic panel's short-circuit current; the smaller the value, the more severe the short-circuit current attenuation and the more significant the electrical degradation). The IV curve data are sorted in ascending order of voltage value to obtain an ordered data point sequence. The second derivative of each data point on the IV curve is calculated using the central difference method. This second derivative is used to characterize the curve curvature change of the corresponding data point. Abrupt curvature changes can directly reflect the inflection point of the curve. In the calculation, three consecutive data points are selected: the current data point, the previous data point, and the next data point. The current data point is numbered starting from 2 and ending with the total number of data points minus 1 to avoid calculation errors caused by boundary data points. The voltage and current values ​​of the three data points are substituted into the calculation. That is, the voltage value of the next data point is subtracted by twice the voltage value of the current data point, and then the voltage value of the previous data point is added to obtain the voltage calculation result. The current value of the next data point is subtracted from the current value of the current data point to obtain the current difference. Then, the current difference is squared to obtain the current calculation result. Finally, the voltage calculation result is divided by the current calculation result to obtain the second derivative of the current data point. The average of the second derivatives of all valid data points on the IV curve is calculated by summing the second derivatives of all valid data points. The range of valid data points is the same as the range used when calculating the second derivative of a single data point, i.e., all data points from index 2 to the total number of data points minus 1. The summation result is then divided by the number of valid data points, which is equal to the total number of data points minus 2. Finally, the average of the second derivatives of the entire curve is obtained. Calculate the absolute value of the deviation between the second derivative of each valid data point and the average value of the second derivative of the entire curve. This involves first calculating the difference between the second derivative of a single valid data point and the average value of the second derivative of the entire curve, and then taking the absolute value of this difference. Sum the absolute values ​​of the deviations of all valid data points, and then divide the sum by the number of valid data points to obtain the average value of the absolute values ​​of the deviations of all valid data points. Divide this average deviation value by the average value of the rated second derivative to obtain the curve distortion attenuation value (the larger the value, the more severe the curve distortion, the more prominent the latent fault, and the more significant the electrical attenuation).

[0028] In this implementation plan, by performing an electrical evaluation on the IV curve data of solar photovoltaic (PV) panels, the power attenuation gradient value, electrical toughness value, and curve distortion attenuation value are extracted in a refined manner. This provides detailed and accurate indicators for the analysis of electrical attenuation characteristics. For example, the extraction of the power attenuation gradient value can effectively reflect the power change characteristics of the PV panel from a short-circuit to an open-circuit state, revealing the uniformity of power fluctuations. The more severe the power fluctuations and the more uneven the attenuation distribution, the more severe the electrical degradation of the PV panel. The extraction of the electrical toughness value focuses on the stability of the short-circuit current of the PV panel. The smaller the value, the more significant the short-circuit current attenuation, thus revealing potential problems in the electrical degradation of the PV panel. Furthermore, by calculating the curve distortion attenuation value, the inflection point change of the PV panel's IV curve can be accurately captured, reflecting potential hidden faults inside the PV panel, which helps to comprehensively assess the degree of electrical degradation of the PV panel.

[0029] Specifically, such as Figure 2 As shown, the ultraviolet fluorescence image time series data includes several frames of ultraviolet fluorescence image data, and the ultraviolet fluorescence image data specifically consists of the gray value of each pixel in the ultraviolet fluorescence image and its corresponding two-dimensional coordinates. The photovoltaic ultraviolet aging mapping model includes an input layer, a recognition and segmentation layer, an aging deconstruction layer, and an output layer.

[0030] The specific steps for obtaining the significant aging characteristic values ​​of solar photovoltaic (PV) panel products are as follows: Input the UV fluorescence image data of the solar PV panel products into a pre-trained PV UV aging mapping model. Analyze the UV aging mapping feature set of each frame of the UV fluorescence image of the solar PV panel products, including connectivity entropy feature values, gray-level heterogeneity feature values, and power-law distribution feature values. Based on the UV aging mapping feature set of each frame of the UV fluorescence image of the solar PV panel products, extract the aging contribution ratio of the corresponding frame of the UV fluorescence image. Based on the UV aging mapping feature set and aging contribution ratio of each frame of the UV fluorescence image of the solar PV panel products, analyze the significant aging characteristic values ​​of the solar PV panel products, specifically as follows: The connectivity correlation entropy feature value, gray-level heterogeneity feature value, and distribution power law feature value of each frame of ultraviolet fluorescence image of solar photovoltaic panel product are fused to obtain the aging fusion feature value of each frame of ultraviolet fluorescence image of solar photovoltaic panel product. The aging contribution ratio of the corresponding frame of ultraviolet fluorescence image is then weighted (the aging contribution ratio is used as the weighting coefficient of the aging fusion feature value of the corresponding frame of ultraviolet fluorescence image) to obtain the significant aging feature value of solar photovoltaic panel product.

[0031] The specific formula for calculating the aging fusion characteristic value of a certain frame of ultraviolet fluorescence image of a solar photovoltaic panel product is as follows: ;in, This represents the aging and melting characteristic value of a specific frame of an ultraviolet fluorescence image of a solar photovoltaic panel product. This represents the connectivity entropy feature value of a specific frame of an ultraviolet fluorescence image of a solar photovoltaic panel product. The connectivity adjustment coefficients stored in the database. This refers to the grayscale heterogeneous feature values ​​of a specific frame of an ultraviolet fluorescence image of a solar photovoltaic panel product. The grayscale heterogeneity adjustment coefficient is stored in the database. This represents the power-law characteristic value of the distribution of a certain frame of ultraviolet fluorescence image of a solar photovoltaic panel product. The power-law adjustment coefficients of the distribution are stored in the database. , These are the nonlinear adjustment coefficients stored in the database.

[0032] It should be noted that the connectivity adjustment coefficients stored in the database Gray-scale heterogeneity adjustment coefficient Power-law adjustment coefficient The acquisition steps are as follows: Read the connectivity entropy feature value, gray-level heterogeneity feature value, and power-law distribution feature value of each frame of the ultraviolet fluorescence image of the solar photovoltaic panel product. Extract the mean of the connectivity entropy feature, the mean of the gray-level heterogeneity feature, and the mean of the power-law distribution feature, and sum them to obtain the aging fusion value. Ratio the mean of the connectivity entropy feature, the mean of the gray-level heterogeneity feature, and the mean of the power-law distribution feature with the aging fusion value, and use the corresponding results as the connectivity adjustment coefficient. Gray-scale heterogeneity adjustment coefficient Power-law adjustment coefficient .

[0033] Nonlinear adjustment coefficients stored in the database The acquisition steps are as follows: Read the connectivity entropy feature value, gray-level heterogeneity feature value, and power-law distribution feature value of each frame of the ultraviolet fluorescence image of the solar photovoltaic panel product. Square each of these features and sum them to obtain the aging value of each frame. Extract the maximum, minimum, and average aging values, and perform a comprehensive processing, i.e., (maximum aging value - average aging value) / (average aging value - minimum aging value). Use this result as a non-linear adjustment coefficient. .

[0034] The following is a specific implementation example for calculating the aging fusion characteristic value of a frame of ultraviolet fluorescence image of a solar photovoltaic panel product. Given the following data, the connectivity correlation entropy characteristic value, gray-level heterogeneity characteristic value, and power-law distribution characteristic value of any 5 frames of ultraviolet fluorescence images of the selected solar photovoltaic panel product are shown in Table 1 and... Figure 3 As shown:

[0035] Connectivity adjustment coefficients stored in the database Approximately 0.293; Gray-scale heterogeneity adjustment coefficients stored in the database Approximately 0.397; The distributed power-law adjustment coefficients stored in the database Approximately 0.310; Nonlinear adjustment coefficients stored in the database Approximately 1.508; Substituting the data from Table 1 and the aforementioned coefficients into the specific formula for calculating the aging fusion characteristic value of a certain frame of ultraviolet fluorescence image of a solar photovoltaic panel product, we obtain: The aging fusion characteristic value of the first frame of the ultraviolet fluorescence image of the solar photovoltaic panel product = (0.293×0.483+0.397×0.678+0.310×0.512)×1.508≈0.859; The aging fusion characteristic value of the second frame of the ultraviolet fluorescence image of the solar photovoltaic panel product = (0.293×0.502+0.397×0.672+0.310×0.518)×1.508≈0.866; The aging fusion characteristic value of the third frame of the ultraviolet fluorescence image of the solar photovoltaic panel product = (0.293×0.494+0.397×0.667+0.310×0.523)×1.508≈0.862; The aging fusion characteristic value of the fourth frame of the ultraviolet fluorescence image of the solar photovoltaic panel product is (0.293×0.498+0.397×0.663+0.310×0.527)×1.508≈0.863; The aging fusion characteristic value of the fifth frame of the ultraviolet fluorescence image of the solar photovoltaic panel product is (0.293×0.487+0.397×0.662+0.310×0.530)×1.508≈0.859.

[0036] The specific steps for analyzing the UV aging mapping feature set of each frame of UV fluorescence image of a solar photovoltaic panel product are as follows: In the input layer of the photovoltaic UV aging mapping model, the UV fluorescence image data of each frame of the solar photovoltaic panel product is received and preprocessed. Specifically, the gray value distribution range of all pixels in a single frame image is statistically analyzed, and abnormal pixels (such as sensor fault points) with gray values ​​exceeding [0,255] (8-bit image) are removed. The gray values ​​of abnormal pixels are filled using linear interpolation (taking the average gray value of the 8 surrounding pixels). Then, the gray values ​​of all effective pixels are mapped to the standardized range of [0,255] to eliminate the gray value scale difference caused by the fluctuation of UV irradiation intensity in different frames of images. Gaussian filtering (with the convolution kernel size set to 3×3 and the standard deviation σ=0.8) is used to smooth the image and filter out isolated noise points caused by sensor electronic noise. In the recognition and segmentation layer of the photovoltaic ultraviolet aging mapping model, pixel segmentation is performed on each frame of ultraviolet fluorescence image data of the preprocessed solar photovoltaic panel product to obtain the aging region set of the corresponding frame of ultraviolet fluorescence image, specifically as follows: For each pixel in each frame of the ultraviolet fluorescence image, the mean gray value and standard deviation of all pixels are calculated. Candidate aging pixels whose gray values ​​exceed the range of [mean gray value - standard deviation gray value, mean gray value + standard deviation gray value] are selected. Then, the optimal segmentation threshold is solved using the Otsu method. The gray value range of the image pixels (0~255) is used as the threshold traversal interval. For each candidate threshold, the image pixels are divided into two categories: foreground (gray value ≥ candidate threshold) and background (gray value < candidate threshold). The proportion of pixels in each category and the mean gray value are calculated. Then, the inter-class variance is calculated using the inter-class variance formula (inter-class variance = background pixel proportion × (background mean gray value - global mean gray value)² + foreground pixel proportion × (foreground mean gray value - global mean gray value)²). After traversal, the gray value corresponding to the largest inter-class variance is selected as the optimal segmentation threshold. This threshold can maximize the differentiation of gray value differences between aging and non-aging regions. The grayscale value of each pixel is compared with the optimal segmentation threshold. Pixels with a grayscale value greater than or equal to the optimal segmentation threshold are identified as aging-related regions and marked as 1, while pixels with a grayscale value less than the optimal segmentation threshold are identified as non-aging regions and marked as 0. A binary segmentation mask map with the same size as the original preprocessed image is generated. Isolated pseudo-aging regions with an area of ​​less than 8 pixels are removed by 8-neighborhood connectivity analysis. Then, Gaussian filtering with a standard deviation of σ=0.5 is used to smooth the edge pixels of the aging regions, thus achieving accurate differentiation between aging regions and non-aging regions. This yields several aging regions, i.e., the aging region set. In the aging deconstruction layer of the photovoltaic ultraviolet aging mapping model, the aging region set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product is subjected to aging mapping processing to obtain the aging feature vector of the corresponding frame of ultraviolet fluorescence image, which is as follows: The two-dimensional coordinates of each pixel in all aging regions of the aging region set are averaged to extract the geometric center coordinates of each aging region. The spatial straight-line distance between any two regions is defined as the connectivity distance between the two regions using the Euclidean distance formula. For any two pairs of aging regions (which can be called a pair of aging regions), if the connectivity distance between them is ≤ a preset connectivity mean threshold (e.g., 5mm), then a potential connectivity association is determined to exist, and the corresponding value is assigned 1. If the connectivity distance is > the preset connectivity mean threshold, then no potential connectivity association is determined to exist, and the corresponding value is assigned 0. All pairs of aging regions are traversed to generate a connectivity association matrix. The proportion of elements with a value of 1 in this matrix to the total number of elements in the matrix is ​​counted to obtain the connectivity probability between regions. Based on this probability, a probability distribution is constructed, and the information entropy value of this matrix is ​​calculated using the information entropy formula. This is the connectivity association entropy feature, which characterizes the complexity of the spatial association between aging regions. The higher the entropy value of this feature, the more complex the spatial association between aging regions and the more significant the potential connectivity diffusion trend. For each aging region in the aging region set, the grayscale values ​​of all pixels within it are counted, and the grayscale mean of a single region is calculated. Simultaneously, the global aging grayscale mean (arithmetic mean of grayscale values ​​of pixels in all aging regions) is calculated. The grayscale deviation rate of each aging region is analyzed using the formula: (Grayscale mean of a single region - Global aging grayscale mean) / Global aging grayscale mean, resulting in a deviation rate set. The total number of pixels in each aging region is counted, and this is used as the area of ​​each aging region. The area percentage of each aging region is calculated, i.e., the area of ​​a single region / the sum of the areas of all aging regions. This area percentage is used as a weight, and a weighted calculation is performed with the deviation rate of the corresponding region (Weighted deviation rate = Deviation rate × Area percentage), yielding the weighted deviation rate of each aging region. The deviation rate probability distribution function is constructed by sorting the values ​​from smallest to largest. The fractal dimension of the distribution is calculated using the box counting method, which divides the range of weighted deviation rate values ​​into several equally spaced intervals (i.e., boxes). The number of boxes containing the deviation rate is counted. The interval intervals are continuously reduced and the count is repeated. The fractal dimension of the probability distribution is obtained by the slope of the linear fitting of "interval interval - number of boxes" under double logarithmic coordinates. This is the gray-scale heterogeneous feature, which is used to characterize the local aging state of the EVA encapsulation layer of the photovoltaic panel. The higher the value, the greater the difference in fluorescence response in different aging areas and the more significant the local specificity of aging. Read the area of ​​each aging region, sort them in descending order of area, and assign a corresponding ranking to each region (the region with the largest area is ranked 1, the second largest is ranked 2, and so on); perform a logarithmic operation to the base 10 on the ranking of each region to obtain the logarithm of the ranking of each aging region, and at the same time perform a logarithmic operation to the base 10 on the area of ​​each aging region to obtain the logarithm of the area of ​​each aging region. A scatter plot was drawn in a double logarithmic coordinate system with the logarithm of the rank as the x-axis and the logarithm of the area as the y-axis, forming a scatter plot of rank and area logarithmic distribution. The scatter plot was fitted with a linear regression algorithm to obtain the optimal fitted line equation. The logarithm of the rank of each aging region was substituted into the equation to obtain the fitted logarithm of the area corresponding to each aging region. The mean of the logarithms of the area of ​​each aging region was then calculated to obtain the mean logarithm of the area. For each aging region, calculate the difference between the logarithm of the area and the fitted logarithm of the area, then square this difference. Finally, sum the squared results of all aging regions and use this as the numerator. Simultaneously, for each aging region, calculate the difference between the logarithm of the area and the mean of the actual logarithms of the area, then square this difference. Finally, sum the squared results of all aging regions and use this as the denominator. Divide the numerator by the denominator to obtain the deviation percentage. Subtract this deviation percentage from 1. The final result is the power-law distribution feature. The larger the value, the more the area distribution of the aging region conforms to the power-law law, that is, the coexistence of a few large-area aging regions and a large number of small-area aging regions. This indicates that aging has entered a stage where the dominant region dominates the diffusion, and the overall aging degree is relatively heavy. The connectivity entropy feature, gray-level heterogeneity feature, and power-law distribution feature are concatenated into an aging feature vector. In the output layer of the photovoltaic ultraviolet aging mapping model, based on the aging feature vector of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, the ultraviolet aging mapping feature set of the corresponding frame of ultraviolet fluorescence image is output. Specifically, the connectivity entropy feature, gray-level heterogeneity feature, and distribution power law feature in the aging feature vector of each frame of ultraviolet fluorescence image are activated by the Sigmoid function to obtain the connectivity entropy feature value, gray-level heterogeneity feature value, and distribution power law feature value between 0 and 1.

[0037] The pre-training steps of the photovoltaic ultraviolet aging mapping model are as follows: Training dataset construction: Collect 1000+ sets of solar photovoltaic panel samples consistent with the target type (covering different service years and different aging levels: mild aging - no obvious encapsulation yellowing / microcracks, moderate aging - local EVA yellowing / a few microcracks, severe aging - large area yellowing / multi-region microcracks). For each sample, acquire 20-50 consecutive frames of ultraviolet fluorescence images (each frame contains pixel grayscale values ​​and two-dimensional coordinates). At the same time, through manual inspection and professional equipment testing, label the actual aging level (level 1-5, level 1 is the mildest and level 5 is the most severe) and the ground truth of the aging area for each sample (manually select the aging area using an image annotation tool to generate a binary ground truth mask).

[0038] Dataset preprocessing: For each frame of ultraviolet fluorescence image of all training samples, preprocessing operations are performed (such as removing abnormal pixels outside [0,255] and filling with linear interpolation, standardizing gray values ​​to [0,255], etc.) to obtain a standardized training image set; at the same time, the aging level labels are converted into one-hot encoding, and the ground truth mask is used for training supervision of the segmentation layer.

[0039] Model initialization: Initialize the parameters of the recognition and segmentation layer (you can set the threshold traversal step size of Otsu's method, the minimum region area threshold of 8-neighborhood connectivity analysis to 8 pixels), the feature calculation parameters of the aging and deconstruction layer (preset connectivity mean threshold to 5mm, box counting interval division number to 20, etc.), and the Sigmoid activation function parameters of the output layer; set the optimizer to Adam optimizer, the initial learning rate to 0.001, and the loss function to a joint loss function of "segmentation loss + feature loss + classification loss".

[0040] Phased training: Phase 1 (Segmentation layer training): Input the standardized training image set into the model and train the segmentation logic of the recognition segmentation layer using the ground truth mask image as supervision. By adjusting the Otsu method threshold solution strategy and connected region selection parameters, minimize the segmentation loss (using the intersection-union ratio IoU loss) to ensure that the model can accurately segment aging regions. When IoU ≥ 0.85, the segmentation layer training converges.

[0041] The second stage (aging deconstruction layer + output layer training): Based on the aging region set output by the segmentation layer, the feature extraction logic of the aging deconstruction layer is trained. By calculating the connectivity entropy, gray-level heterogeneity, and power-law distribution features, a mapping relationship is established with the manually labeled aging levels. Taking the correlation between the three feature values ​​of the output layer (after Sigmoid activation) and the aging level as the optimization objective, the feature loss (mean squared error loss) and classification loss (cross-entropy loss) are minimized. The feature calculation parameters of the aging deconstruction layer (such as connectivity distance threshold, box counting interval, etc.) are iteratively adjusted until the joint loss function value tends to stabilize (loss fluctuation ≤ 0.0001 for 100 consecutive iterations).

[0042] Model validation and tuning: Divide 20% of the samples as a validation set, input the validation set images into the trained model, output the UV aging mapping feature set, calculate the Pearson correlation coefficient between the feature values ​​and the actual aging level, and require the correlation coefficient to be ≥0.9; if it does not meet the standard, adjust the optimizer learning rate (down to 0.0001), add 50 rounds of fine-tuning training until the validation criteria are met.

[0043] Model saving: Save the model parameters after training convergence (including the threshold strategy of the segmentation layer, the feature calculation parameters of the aging deconstruction layer, and the activation function parameters of the output layer) to the database to form a pre-trained photovoltaic ultraviolet aging mapping model, which will be used for subsequent ultraviolet feature decoding processing of photovoltaic panels to be graded.

[0044] The specific steps for extracting the aging contribution ratio of each frame of ultraviolet fluorescence image of a solar photovoltaic panel product are as follows: Based on the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, extract the aging mapping geometric mean of the corresponding frame of ultraviolet fluorescence image. That is, perform geometric mean processing on the connectivity correlation entropy feature value, gray-level heterogeneity feature value, and distribution power law feature value of each frame of ultraviolet fluorescence image to obtain the aging mapping geometric mean of each frame of ultraviolet fluorescence image. Based on the aging mapping geometric mean of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, analyze the aging contribution ratio of the corresponding frame of ultraviolet fluorescence image. That is, sum the aging mapping geometric mean of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product to obtain the aging sum value of the solar photovoltaic panel product. Ratio the aging mapping geometric mean of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product with the aging sum value to obtain the aging contribution ratio of the corresponding frame of ultraviolet fluorescence image.

[0045] In this implementation scheme, a standardized process from image preprocessing to feature output is formed through a four-layer architecture of the photovoltaic ultraviolet aging mapping model: the preprocessing stage removes abnormal pixels, standardizes gray values, and filters noise to ensure the quality of input data; the recognition and segmentation layer combines statistical screening and Otsu's method to accurately segment aging regions and remove pseudo regions, laying the foundation for feature extraction; the aging deconstruction layer extracts corresponding features from three dimensions: spatial correlation, gray value difference, and area distribution, comprehensively representing the complex state of aging; the output layer unifies the feature scale through Sigmoid activation to improve fusion adaptability; secondly, the adjustment coefficient is dynamically obtained based on the feature mean ratio, and the nonlinear adjustment coefficient is quantized through feature distribution skewness to ensure the rationality of feature fusion; finally, the aging contribution ratio of each frame image is combined with temporal weighting to highlight the aging impact of key frames, so that the significant aging feature values ​​can truly reflect the overall aging state of the photovoltaic panel, realizing a deep transformation from image data to aging features, thereby improving the accuracy of aging state assessment.

[0046] Specifically, the steps for analyzing the residual value rating of solar photovoltaic panel products are as follows: The system reads the electrical degradation characteristic values ​​and aging significant characteristic values ​​of solar photovoltaic (PV) panels and performs normalization processing. Specifically, it extracts the global maximum and minimum values ​​of electrical degradation characteristic values ​​and aging significant characteristic values ​​of similar solar PV panels from the database. Normalization operations are then performed on the electrical degradation characteristic values ​​and aging significant characteristic values ​​of the current PV panel to obtain values ​​between 0 and 1. These normalized values ​​are then input into a preset residual value assessment model to obtain the residual value score of the solar PV panel.

[0047] The residual value assessment model is as follows: ;in, Assess the residual value of solar photovoltaic panels. The electrical degradation characteristic value of the solar photovoltaic panel product after normalization. The electrical attenuation adjustment coefficient is stored in the database. These are the normalized aging significant characteristic values ​​of the solar photovoltaic panel products. The aging significant adjustment coefficients are stored in the database. , These are the coordination coefficients stored in the database.

[0048] It should be noted that the electrical attenuation adjustment coefficients stored in the database Significant adjustment coefficient for aging The acquisition steps are as follows: Obtain the electrical degradation characteristic values ​​and aging significant characteristic values ​​of several solar photovoltaic panel products (of the same type as the solar photovoltaic panel products in this implementation example), and transform them respectively using the reciprocal suppression mapping function f(x)=1 / (1+x), such as 1 / (1+electrical degradation characteristic value). Based on the electrical degradation characteristic values ​​and aging significant characteristic values ​​of each solar photovoltaic panel product after transformation, extract the mean of the transformed electrical degradation characteristic value and the mean of the aging significant characteristic value, and sum them to obtain the residual sum value. Ratio the mean of the transformed electrical degradation characteristic value and the mean of the aging significant characteristic value with the residual sum value, and use the corresponding results as the electrical degradation adjustment coefficient. Significant adjustment coefficient for aging .

[0049] Coordination coefficients stored in the database The acquisition steps are as follows: Obtain the electrical degradation characteristic values ​​and aging significance characteristic values ​​of several solar photovoltaic panel products (of the same type as the solar photovoltaic panel products in this implementation example), and extract the correlation coefficient (absolute value) between the two based on the Pearson correlation coefficient, using the result as the co-regulation coefficient. .

[0050] Used to characterize the synergistic deterioration of electrical degradation and aging, when a photovoltaic panel is simultaneously in a state of low electrical degradation and low aging, the product of the magnitudes of these two indicators will be very small. However, after passing through the exponential function, it will amplify the previously calculated result, allowing this type of photovoltaic panel with both advantages to obtain a higher residual value score. If the photovoltaic panel shows obvious electrical degradation or obvious aging (even if only one of them is present), the product of the magnitudes of these two indicators will become larger. After passing through the exponential function, it will not amplify the previous basic health level, but only retain the residual value corresponding to the basic health level, because this type of panel can only be dismantled and recycled, without any additional value premium.

[0051] The specific steps for automatically grading solar photovoltaic panel products based on recycling residual value scores are as follows: The recycling residual value scores of the solar photovoltaic panel products are compared with preset recycling residual value score ranges; based on the comparison results, the solar photovoltaic panel products are automatically graded, specifically as follows: If the residual value score of a solar photovoltaic (PV) panel is lower than the lower limit of the preset residual value score range, the PV panel is classified as a material-level recycling product. The PV panel is then completely disassembled to recover basic raw materials such as silicon and metal frames. If the residual value score is within the preset residual value score range, the PV panel is classified as a module-level recycling product. The PV panel is disassembled into functional components such as cells and junction boxes for recycling. If the residual value score is higher than the upper limit of the preset residual value score range, the PV panel is classified as a secondary utilization product. After cleaning and electrical performance retesting, the PV panel can be reused in low-power photovoltaic scenarios.

[0052] It should be noted that the preset steps for the residual value scoring range are as follows: Obtain the residual value scores of several solar photovoltaic panel products (the same type of solar photovoltaic panel products as in this implementation example), extract the mean and standard deviation of the residual value scores, and use these to set the residual value scoring range. That is, the multiple of the mean and standard deviation of the residual value scores is used as the upper limit of the residual value scoring range, and the multiple of the mean and standard deviation of the residual value scores is used as the lower limit of the residual value scoring range. The value of this multiple can be in the range of 1.000 to 2.000.

[0053] In this implementation plan, the significant adjustment coefficients for electrical degradation and aging are obtained through the reciprocal suppression mapping function and the mean proportion method. Combined with the Pearson correlation coefficient, the synergistic adjustment coefficient is determined, allowing the residual value assessment model to dynamically adapt to the contribution weights and synergistic relationships of the two characteristics, accurately quantifying the comprehensive impact of electrical properties and aging on residual value. Secondly, the innovative exponential synergistic term in the model realizes the premium for dual-excellent products and the removal of premiums for deteriorated products. This highlights the secondary utilization value of low-degradation and low-aging photovoltaic panels and avoids misjudging the value of products with excellent single indicators but synergistic deterioration, making the scoring more in line with the actual recycling scenario. Finally, the grading threshold is set based on the multiple range of the mean and standard deviation to adapt to different batch recycling needs. The grading rules are clear and explicit, and the gradient processing from material-level recycling, component-level classification recycling to secondary utilization achieves efficient resource matching without manual intervention throughout the process, improving grading efficiency and ensuring grading reliability.

[0054] Please see Figure 4 This invention provides a technical solution: an automatic grading system for solar photovoltaic (PV) panel products, comprising: a data acquisition module for acquiring IV curve data and ultraviolet fluorescence image time-series data of solar PV panel products; an electrical degradation analysis module for performing electrical analysis processing on the IV curve data of solar PV panel products to extract electrical degradation feature values ​​of solar PV panel products; an ultraviolet aging analysis module for performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of solar PV panel products based on a pre-trained photovoltaic ultraviolet aging mapping model to obtain significant aging feature values ​​of solar PV panel products; a residual value assessment module for analyzing the recycling residual value score of solar PV panel products based on the electrical degradation feature values ​​and significant aging feature values ​​of solar PV panel products; and an automatic grading module for automatically grading solar PV panel products based on the recycling residual value score.

[0055] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0056] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An automatic grading method for solar photovoltaic panel products, characterized in that, The method comprises the following steps: Obtaining IV curve data and ultraviolet fluorescence image time series data of a solar photovoltaic panel product; Performing electrical analysis processing on the IV curve data of the solar photovoltaic panel product to extract electrical attenuation characteristic values of the solar photovoltaic panel product; Based on a pre-trained photovoltaic ultraviolet aging mapping model, performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of the solar photovoltaic panel product to obtain aging significant characteristic values of the solar photovoltaic panel product; Based on the electrical attenuation characteristic values and the aging significant characteristic values of the solar photovoltaic panel product, analyzing the recycling residual value score of the solar photovoltaic panel product; Based on the recycling residual value score, automatically grading the solar photovoltaic panel product.

2. The automatic grading method for solar photovoltaic panel products as claimed in claim 1, wherein, The IV curve data is specifically IV data pairs at each time point, and the specific steps of extracting the electrical attenuation characteristic values of the solar photovoltaic panel product are as follows: Based on the IV curve data of the solar photovoltaic panel product, analyzing the electrical evaluation data set of the solar photovoltaic panel product, including power attenuation gradient values, electrical decay toughness values, and curve distortion attenuation values; Performing fusion processing on the electrical evaluation data set of the solar photovoltaic panel product to obtain the electrical attenuation characteristic values of the solar photovoltaic panel product.

3. The automatic grading method for solar photovoltaic panel products according to claim 2, characterized in that, The specific steps of analyzing the electrical evaluation data set of the solar photovoltaic panel product are as follows: Performing initial processing on the IV curve data of the solar photovoltaic panel product; Performing analysis and integration processing on the IV curve data of the solar photovoltaic panel product after initial processing to obtain power attenuation gradient values, electrical decay toughness values, and curve distortion attenuation values of the solar photovoltaic panel product.

4. The automatic grading method for solar photovoltaic panel products as claimed in claim 1, wherein, The ultraviolet fluorescence image time series data includes a plurality of frames of ultraviolet fluorescence image data, and the ultraviolet fluorescence image data is specifically the gray value and corresponding two-dimensional coordinates of each pixel point in the ultraviolet fluorescence image, and the photovoltaic ultraviolet aging mapping model includes an input layer, a recognition segmentation layer, an aging deconstruction layer, and an output layer.

5. The automatic grading method for solar photovoltaic panel products as claimed in claim 4, wherein, The specific steps of obtaining the aging significant characteristic values of the solar photovoltaic panel product are as follows: Inputting the ultraviolet fluorescence image data of the solar photovoltaic panel product into the pre-trained photovoltaic ultraviolet aging mapping model, analyzing the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, including connected correlation entropy characteristic values, gray heterogeneity characteristic values, and distribution power law characteristic values; Based on the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, extracting the aging contribution proportion value of the corresponding frame of ultraviolet fluorescence image; Based on the ultraviolet aging mapping feature set and the aging contribution proportion value of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, analyzing the aging significant characteristic values of the solar photovoltaic panel product.

6. The automatic grading method for solar photovoltaic panel products as claimed in claim 5, wherein, The specific steps of analyzing the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product are as follows: In the input layer of the photovoltaic ultraviolet aging mapping model, receiving the ultraviolet fluorescence image data of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product and performing preprocessing; In the recognition segmentation layer of the photovoltaic ultraviolet aging mapping model, performing pixel segmentation processing on the preprocessed ultraviolet fluorescence image data of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product to obtain the aging region set of the corresponding frame of ultraviolet fluorescence image; In the aging decomposition layer of the photovoltaic ultraviolet aging mapping model, the aging area set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product is subjected to aging mapping processing, and the aging feature vector of the corresponding frame of ultraviolet fluorescence image is obtained. In the output layer of the photovoltaic ultraviolet aging mapping model, based on the aging feature vector of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, the ultraviolet aging mapping feature set of the corresponding frame of ultraviolet fluorescence image is output.

7. The automatic grading method for solar photovoltaic panel products as claimed in claim 5, wherein, The specific steps of extracting the aging contribution proportion value of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product are as follows: Based on the ultraviolet aging mapping feature set of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, the aging mapping geometric mean of the corresponding frame of ultraviolet fluorescence image is extracted; Based on the aging mapping geometric mean of each frame of ultraviolet fluorescence image of the solar photovoltaic panel product, the aging contribution proportion value of the corresponding frame of ultraviolet fluorescence image is analyzed.

8. The automatic grading method for solar photovoltaic panel products as claimed in claim 1, wherein, The specific steps of analyzing the recycling residual value score of the solar photovoltaic panel product are as follows: Read the electrical attenuation characteristic value and aging significant feature value of the solar photovoltaic panel product, and perform normalization processing; The electrical attenuation characteristic value and aging significant feature value of the solar photovoltaic panel product after normalization processing are input into the preset residual value evaluation model, and the recycling residual value score of the solar photovoltaic panel product is obtained.

9. The automatic grading method for solar photovoltaic panel products of claim 8, wherein, The specific steps of automatically grading the solar photovoltaic panel product based on the recycling residual value score are as follows: Compare the recycling residual value score of the solar photovoltaic panel product with the preset recycling residual value score interval; Based on the comparison result, the solar photovoltaic panel product is automatically graded.

10. An automatic grading system for solar photovoltaic panel products, applying the automatic grading method for solar photovoltaic panel products according to any one of claims 1-9, characterized in that, It includes: A data acquisition module for acquiring IV curve data and ultraviolet fluorescence image time series data of a solar photovoltaic panel product; An electrical attenuation analysis module for performing electrical analysis processing on the IV curve data of the solar photovoltaic panel product, and extracting the electrical attenuation characteristic value of the solar photovoltaic panel product; An ultraviolet aging analysis module for performing ultraviolet feature decoding processing on the ultraviolet fluorescence image data of the solar photovoltaic panel product based on a pre-trained photovoltaic ultraviolet aging mapping model, to obtain the aging significant feature value of the solar photovoltaic panel product; A residual value evaluation module for analyzing the recycling residual value score of the solar photovoltaic panel product based on the electrical attenuation characteristic value and the aging significant feature value of the solar photovoltaic panel product; An automatic grading module for automatically grading the solar photovoltaic panel product based on the recycling residual value score.

Images