Photovoltaic cell energy efficiency map detection method based on proximal gradient method

Abstract

This invention discloses a photovoltaic cell energy efficiency map detection method based on the proximal gradient method. The steps are as follows: determining a projection matrix based on the area of ​​the test region on the photovoltaic cell surface and the pixel size of the projector; collecting measurement vectors after combining and projecting the projection matrix; constructing a total variational perturbation compressed sensing model; representing the perturbation compressed sensing problem as a global least squares problem; reducing the solution difficulty by introducing convex relaxation; accelerating the iteration efficiency during the solution process by introducing the alternating direction multiplier method; and processing the obtained measurement vectors to obtain the energy efficiency map of the test region. This invention allows for arbitrary adjustment of the number of rows and columns of the projection matrix according to the area of ​​the test region, and uses the proximal gradient method to solve the perturbation compressed sensing model, effectively overcoming noise interference and obtaining more accurate and high-quality photovoltaic cell energy efficiency maps.

Description

Technical Field

[0001] This invention belongs to the field of electrical measurement technology, and more specifically relates to a photovoltaic cell energy efficiency map detection method based on the near-end gradient method within the field of photovoltaic cell testing technology. This invention can provide an energy efficiency map reflecting the health status and photoelectric conversion efficiency of photovoltaic cells by detecting defects in the cells. Background Technology

[0002] Photovoltaic cell energy efficiency testing results directly reflect the health status and photoelectric conversion efficiency of photovoltaic cells. Photovoltaic cells are extremely fragile in actual production and use, particularly prone to defects such as breakage, microcracks, grid breaks, chipping, and scratches. These defects severely affect the health status of photovoltaic cells and reduce their photoelectric conversion efficiency. An accurate, practical, and comprehensive photovoltaic cell defect detection method can provide valuable energy efficiency maps, playing a crucial role in improving the yield rate and quality classification of finished products in industrial production. Because the collected output current of a photovoltaic cell is the sum of the entire cell's output, it is difficult to determine the specific location of defects on the cell, nor is it easy to determine the output current value at each point on the cell. If the output current value at each point on the photovoltaic cell could be accurately obtained, the photoelectric conversion efficiency at that point could be calculated, and an energy efficiency map of the photovoltaic cell could be obtained, revealing the specific location of the defects.

[0003] Demirci et al. disclosed an electroluminescence (EL) detection method in their paper "Efficient deep feature extraction and classification for identifying defective photovoltaic module cells in Electroluminescence images." (2021. Expert Syst. Appl. 175, 114810.). This method mainly relies on a high-resolution infrared camera and image analysis software. By applying a voltage to the photovoltaic material, near-infrared images of the photovoltaic material are captured using an infrared camera. Then, image analysis software is used to process and analyze the images to discover internal defects and problems in the photovoltaic material, preventing potential safety hazards and enabling regular inspection and maintenance of the photovoltaic material. However, this method still has shortcomings: it requires the inspection of specific photovoltaic materials and necessitates infrared cameras and image analysis software, resulting in high construction costs. Furthermore, it may misjudge quality defects that are difficult to detect.

[0004] Xi'an University of Electronic Science and Technology disclosed a method for measuring the energy efficiency map of photovoltaic cells based on orthogonal modulation in its patent application "Method for Measuring Photovoltaic Cell Energy Efficiency Map Based on Orthogonal Modulation" (Application No.: 202110981312.5, Publication No.: CN 113824400 A, Application Date: 2021.08.25). The specific implementation steps of this method include: Step 1, determining the number of projection pixels based on the range of the photovoltaic cell to be measured and the size of the projected pixels, and determining the projection matrix based on the number of projection pixels; Step 2, constructing a sub-projection matrix, using the sub-projection matrix to control the structured light to illuminate the area to be measured to obtain measurement values; Step 3, arranging the measurement values ​​in column direction to form a measurement vector, and post-processing the measurement vector to obtain the energy efficiency map of the area to be measured. The drawback of this method is that the projection matrix is ​​a Hadamard matrix, the dimension of which is affected by an exponent, and the size of the measurement area cannot be arbitrarily adjusted, which limits its practical application. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the existing technology by providing a photovoltaic cell energy efficiency map detection method based on the near-end gradient method. This method aims to solve the problems associated with traditional EL detection methods, such as high construction costs, the inability to arbitrarily adjust the measurement area size due to the influence of dimensionality on the Hadamard matrix, and the impact of ambient light noise in the detection environment.

[0006] The technical approach to achieving the objectives of this invention is as follows: This invention employs a proximal gradient method based on total variational regularization to solve the constructed perturbation compressed sensing model. Because adaptive rules are used to select the step size during the iterative solution process, gradient fluctuations can be reduced, and the regularization part helps reduce overfitting and improve the model's generalization ability, thereby enhancing its noise resistance. This solves the problem of ambient light noise affecting actual detection in existing technologies, and enables efficient reconstruction of gradient-sparse images, thus providing more accurate and high-quality photovoltaic cell energy efficiency maps. This invention utilizes the alternating direction multiplier method to accelerate the solution of the perturbation compressed sensing model. This method decomposes the original problem into several simple sub-problems, solves these sub-problems in parallel, and adjusts the convergence speed by adjusting the penalty coefficient. Therefore, detection and energy efficiency map reconstruction can be completed without upgrading hardware, thus solving the problem of high construction costs in existing technologies. This invention can determine the dimension of the generated projection matrix based on the size of the test area, thus solving the problem that the Hadamard matrix cannot arbitrarily adjust the measurement area size due to dimensionality limitations.

[0007] The specific steps of this invention include the following:

[0008] Step 1: Determine the projection matrix based on the area of ​​the photovoltaic cell to be tested, and generate a projection matrix combination (Φ+E), where Φ represents the matrix combination and E represents noise, with a value range of [-20, -30] dB.

[0009] Step 2: Construct a fully variational perturbation compressed sensing model based on the generated measurement vectors;

[0010] Step 3: Solve the perturbation compressed sensing model using the proximal gradient method based on total variation regularization.

[0011] Step 4: Introduce convex relaxation to reduce the difficulty of solving the model;

[0012] Step 5: Accelerate the solution of the perturbation compressed sensing model by using the alternating direction multiplier method;

[0013] Step 6: Determine whether the current iteration meets the convergence condition. If yes, obtain an N*1 reconstructed vector and proceed to step 7; otherwise, proceed to step 4.

[0014] Step 7: Generate an energy efficiency map of the photovoltaic cell test area.

[0015] Compared with the prior art, the present invention has the following advantages:

[0016] First, this invention can determine the dimension of the generated projection matrix based on the size of the test area, overcoming the problem that when the Hadamard matrix is ​​used as the projection matrix, the matrix cannot be arbitrarily adjusted in size due to the influence of dimension. In practical photovoltaic cell testing applications, the detection can be carried out according to the size of the test area, without being affected by the dimension of the projection matrix.

[0017] Secondly, since the present invention uses the proximal gradient method based on total variational regularization to solve the constructed perturbation compressed sensing model, it overcomes the problem of ambient light noise affecting actual detection in the prior art. This allows the present invention to demonstrate stronger anti-noise performance in actual testing and efficiently realize the reconstruction of gradient sparse images, thereby providing more accurate and high-quality photovoltaic cell energy efficiency maps.

[0018] Third, this invention utilizes the alternating direction multiplier method to accelerate the solution of the perturbation compressed sensing model. It can complete the detection and solution of the model to recover the energy efficiency map without upgrading the hardware, overcoming the problem of high construction cost of existing technologies. It can realize the detection of photovoltaic cell energy efficiency map at low cost in actual industrial testing. Attached Figure Description

[0019] Figure 1 This is a flowchart of the present invention;

[0020] Figure 2This is a schematic diagram of the photovoltaic cell testing area according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the simplified 2*2 perturbation compressed sensing mathematical model of the present invention;

[0022] Figure 4 This is an energy efficiency diagram of the photovoltaic cell test area obtained from actual testing according to the present invention;

[0023] Figure 5 This is a simulation diagram of the present invention, wherein, Figure 5 (a) is a grayscale image of a photovoltaic cell simulated in the simulation experiment of this invention. Figure 5 (b) is a comparison diagram of the reconstructed images using PCLBIC and CLBIC of the present invention. Figure 5 (c) is a comparison chart of PSNR and SSIM of the output images of PCLBIC and CLBIC of the present invention. Detailed Implementation

[0024] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are only a part of the present invention, and not all of it. It is clear from the drawings that the present invention has significant advantages for the actual measurement of photovoltaic cells.

[0025] Reference Figure 1 The implementation steps of the embodiments of the present invention will be further described below.

[0026] Step 1: Determine the projection matrix based on the area to be measured.

[0027] The number of projection pixels is determined based on the area of ​​the photovoltaic cell to be tested and the pixel size of the projector. Then, the number of rows and columns m and n of the projection matrix and the total number N are constructed.

[0028] Reference Figure 2 The specific implementation steps of determining the projection matrix based on the region to be measured in this embodiment of the invention are described as follows:

[0029] The embodiments of the present invention are implemented on a hardware platform: an AMD Ryzen 7 3750H processor, 32GB of memory, a monochrome red light digital industrial projector, and a software platform: MATLAB R2023a and Windows 10 operating system.

[0030] The photovoltaic cell in the embodiments of the present invention includes grid lines, busbars, artificial defects, scratch defects, welding areas, and the photovoltaic cell test area marked within the red dashed box. The area to be tested is 6mm*6mm, and each projection pixel of the projector is 75μm*75μm. Based on this, the number of rows of the projection matrix is ​​determined to be 6mm / 75μm = 80, and the number of columns is also determined to be 6mm / 75μm = 80. Therefore, the number of rows and columns m and n of the projection matrix are 80 and 80 respectively, and the number of columns N = m*n = 6400.

[0031] Step 1.2: Use MATLAB R2023a to generate 6400 random matrices with 80*80 rows and columns, each containing only 0 and 1 elements. Combine all the random matrices into a matrix Φ and store it.

[0032] Step 1.3: Add -30dB noise E to each matrix in the generated matrix combination Φ to obtain the projection matrix combination (Φ+E), and store it.

[0033] Step 2: Collect measurement vectors.

[0034] This invention uses a computer-controlled projector to project the projection matrix combination (Φ+E) obtained in step 1 onto the test area sequentially using structured light. The photocurrent values ​​generated by the sequential projections (6400 times) are then used to form the measurement vector y. The specific process is as follows:

[0035] Step 2.1: Use a projector to project the stored projection matrix (Φ+E) onto the photovoltaic cell measurement area in sequence.

[0036] Step 2.2: Each projection generates a photocurrent value, which is recorded by the ADC.

[0037] Step 2.3: Record the photocurrent values ​​generated by 6400 projections as a 6400*1 measurement vector y.

[0038] The projector is a monochrome red light digital industrial projector.

[0039] Step 3, construct the fully variational perturbation compressed sensing model as follows:

[0040] min||Dx||1 sty+e=(Φ+E)x

[0041] Where min represents the minimum value operation, ||·||1 represents the L1 norm operation, D represents the total variation operator to be calculated using the number of rows and columns of the projection matrix, x represents the reconstructed vector to be determined, and e represents the interference noise in the measurement vector.

[0042] Reference Figure 3The process of simplifying the perturbation compressed sensing mathematical model to 2*2 is further described.

[0043] The total variation operator D is calculated based on the number of rows and columns of the projection matrix obtained in step 1. The perturbation compressed sensing model is constructed by combining the projection matrix group (Φ+E) and the measurement vector y obtained in step 2. Figure 3 In this example, Φ represents the randomly generated matrix Φ1, Φ2…Φ in the embodiment. 6400 The resulting random matrix, E, represents the -30dB noise E1, E2…E added to the random matrix in the example. 6400 (Φ+E) represents the final projection matrix obtained in the embodiment. This matrix is ​​projected onto the photovoltaic cell in the form of a beam through a projector to obtain a measurement vector y with noise e. The reconstructed vector x is obtained by solving the perturbation compressed sensing model, and the photovoltaic cell energy efficiency map can be obtained.

[0044] Step 4: Represent the perturbation compressed sensing problem as a total least squares problem.

[0045] Since the least squares problem provides an effective optimization framework and can significantly improve solution efficiency, considering projection matrix noise and measurement vector noise, the perturbation compressed sensing model in step 3 can be represented as a global least squares problem. Then, by using the proximal gradient method based on total variational regularization to solve the following equation, the system's interference noise can be effectively avoided.

[0046]

[0047] Where λ represents the regularization parameter, which controls the strength of the regularization term. λ>0, and an excessively large value may make the model too simple and affect the model performance. In this embodiment of the invention, it is taken as 0.02. ||·||2 is the L2 norm operation.

[0048] Step 5: Introduce convex relaxation to reduce model complexity according to the following formula:

[0049]

[0050] in, Indicates that x n The operation to obtain the minimum value involves d and p, which are auxiliary variables constructed using the alternating direction multiplier method to facilitate subsequent steps. Their expressions will be provided in step 6. δ is the Lagrange penalty term, where δ > 0, and in this embodiment, it is taken as 0.111. n+1 x n Let μ represent the reconstructed vectors after the (n+1)th and nth iterations, respectively. n This represents the iteration step size. A smaller step size results in better iteration optimization. In this embodiment of the invention, it is set to 0.0000005. nThis represents the gradient of the overall least squares problem after each iteration. In this embodiment of the invention, g needs to be calculated based on the updated overall least squares problem for each iteration. n .

[0051] The difficulty of solving the model in step 4 is reduced by introducing convex relaxation coefficients. Introducing convex relaxation can reduce the complexity of the problem, simplify the solution process, guarantee the global optimal solution, and also prepare for the subsequent introduction of the alternating direction multiplier method to accelerate the iteration speed.

[0052] Step 6: Using the alternating direction multiplier method described below, the iteration efficiency of solving the perturbation compressed sensing model is accelerated:

[0053]

[0054] p n+1 =p n +(Dx n+1 -d n+1 )

[0055] Where, d n+1 and d n Let d and p represent the intermediate variables d and p after the (n+1)th and nth iterations, respectively. n+1 p n Let p represent the intermediate variable after the (n+1)th and nth iterations, respectively.

[0056] The embodiments of the present invention introduce the alternating direction multiplier method to split the function in step 5 into three low-difficulty sub-functions, thereby accelerating the iteration to efficiently realize the total variational regularization reconstruction algorithm.

[0057] The alternating direction multiplier method breaks down the problem in step 5 into three simple subproblems, solves the subproblems in parallel, and can also adjust the convergence speed by adjusting the penalty coefficient, which greatly speeds up the iteration.

[0058] Step 7: Determine if the current iteration meets the convergence condition. If yes, obtain a 6400*1 reconstructed vector and proceed to Step 8. Otherwise, continue iterating from Step 5.

[0059] The convergence conditions are as follows:

[0060]

[0061] The superscript T indicates the transpose operation.

[0062] Step 8: Arrange the reconstructed vectors into a reconstructed matrix and represent them with pixel values ​​to obtain the energy efficiency map of the photovoltaic cell test area.

[0063] Using the `imadjust` function in MATLAB, the reconstructed vectors are arranged into an 80x80 reconstruction matrix in column-major order. Then, the `imshow` function in MATLAB is used to convert the reconstruction matrix into the final energy efficiency map of the photovoltaic cell test area. Figure 4 As shown in the figure, the grid lines, structural defects, scratches, and the defects of the photovoltaic cell itself (defects invisible to the naked eye) marked by the red circle in the upper right corner can be clearly observed.

[0064] The effects of this invention will be further illustrated below with simulation experiments:

[0065] 1. Simulation experimental conditions:

[0066] The simulation experiment hardware platform of this invention is: AMD Ryzen 7 3750H processor and 32GB of memory.

[0067] The simulation experiment software platform of this invention is MATLAB R2023a and Windows 10 operating system.

[0068] 2. Simulation content and result analysis.

[0069] The images used in the simulation experiments of this invention are simulated grayscale images of photovoltaic cells, such as... Figure 5 As shown in (a). The image includes simulated photovoltaic cell grid lines and simulated photovoltaic cell defects. It was imaged in July 2024, with a size of 64*64 pixels and a bmp format.

[0070] The simulation experiment of this invention uses the method of this invention, referred to as PCLBIC, and an existing technology (traditional algorithm to solve the compressed sensing model CLBIC) to reconstruct the input grayscale image of the simulated photovoltaic cell, and obtain the reconstructed image.

[0071] The existing technology CLBIC refers to a compressed light beam induced current sensing method disclosed by Quan Lei et al. in their paper "Compressive light beam induced current sensing for fast defect detection in photovoltaic cells." (2017.Sol.Energy 150,345–352.), abbreviated as CLBIC.

[0072] The simulation experiment of this invention adds three types of noise—-20dB, -25dB, and -30dB—to a simulated photovoltaic cell grayscale image, respectively. Then, the three noise-added images are input into the PCLBIC and CLBIC models of this invention for image reconstruction. The comparison of the reconstruction results is shown in the figure below. Figure 5 As shown in (b).

[0073] The effects of the present invention will be further described below with reference to the accompanying drawings.

[0074] Figure 5 The first column of (b) shows the reconstructed images output by the conventional CLBIC and the present invention's PCLBIC for grayscale images of simulated photovoltaic cells with -20dB noise added, respectively. Figure 5 As can be seen in column 1 of (b), the image output by the PCLBIC of this invention can roughly distinguish grid lines and defects, while the image output by the CLBIC is severely polluted by noise and is so blurry that grid lines and defects cannot be distinguished. Figure 5 The second column in (b) shows the reconstructed images output from the grayscale image of a simulated photovoltaic cell with added -25dB noise, represented by CLBIC and the PCLBIC of this invention, respectively. Figure 5 (b) Column 2 shows that the image output by the PCLBIC of the present invention can clearly distinguish grid lines and defects, while the image output by the CLBIC is blurry and cannot accurately distinguish grid lines and defects. Figure 5 Column 3 of (b) shows the reconstructed images output from the grayscale image of a simulated photovoltaic cell with -30dB noise added, obtained by CLBIC and the PCLBIC of this invention, respectively. Figure 5 As can be seen in column 3 of (b), the reconstructed image output by the PCLBIC of the present invention is almost unaffected by noise, while the reconstructed image output by the CLBIC still has a small number of blurred boundaries.

[0075] To objectively evaluate image quality, this invention introduces Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) as image quality evaluation metrics. The PSNR and SSIM values ​​of the output image can be directly calculated using the psnr and ssim functions in MATLAB.

[0076] Figure 5 (c) is a comparison chart of PSNR and SSIM of the reconstructed images output by PCLBIC and CLBIC of the present invention. Figure 5 (c) The horizontal axis represents the noise added to the grayscale image of the simulated photovoltaic cell by -20dB, -25dB, and -30dB respectively, the left vertical axis represents the PSNR value, and the right vertical axis represents the SSIM value. Figure 5In (c), the orange triangle curve represents the PSNR value of the reconstructed image output by the PCLBIC of this invention under different noise conditions, and the orange circle curve represents the PSNR value of the reconstructed image output by the CLBIC under different noise conditions. The blue triangle curve represents the SSIM value of the reconstructed image output by the PCLBIC of this invention under different noise conditions, and the blue circle curve represents the SSIM value of the reconstructed image output by the CLBIC under different noise conditions. By comparing PSNR and SSIM, it can be found that the PSNR and SSIM of the PCLBIC of this invention are superior to those of the traditional compressed sensing model CLBIC under all four noise conditions. Specifically, the PSNR of the PCLBIC of this invention is on average 3.5 dB higher than that of the traditional compressed sensing model CLBIC.

[0077] The simulation experiment of this invention uses a projector to project the generated random matrix combination onto the measurement area of ​​the photovoltaic cell. The collected photocurrent value is the measurement vector. The perturbation compressed sensing algorithm model is processed using the proximal gradient method based on total variational regularization to process the measurement vector. The proximal gradient method uses adaptive rules to select the step size during the solution iteration process, which can reduce gradient fluctuations. Furthermore, the regularization part can help reduce overfitting and improve the model's generalization ability, thereby improving noise resistance. This invention enables efficient reconstruction of gradient sparse images, thus providing more accurate and high-quality photovoltaic cell energy efficiency maps.

[0078] This invention accelerates the solution of a perturbation compressed sensing model using the alternating direction multiplier method. This method decomposes the original problem into several simple subproblems, solves these subproblems in parallel, and adjusts the convergence speed by regulating the penalty coefficient. Therefore, this invention can detect and recover energy efficiency maps without requiring hardware upgrades.

[0079] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for detecting the energy efficiency map of photovoltaic cells based on the proximal gradient method, characterized in that, The projection matrix is ​​determined based on the area of ​​the region to be tested. The constructed perturbation compressed sensing model is solved using the proximal gradient method based on total variational regularization. The alternating direction multiplier method is used to solve the perturbation compressed sensing model and accelerate the process. The detection steps of this method include the following: Step 1: Determine the projection matrix based on the area of ​​the photovoltaic cell to be tested, and generate a combination of projection matrices. ,in, Represents matrix combination, This represents noise, with a value ranging from [-20, -30] dB; Step 2: Construct a fully variational perturbation compressed sensing model based on the generated measurement vectors; Step 3: Solve the perturbation compressed sensing model using the proximal gradient method based on total variation regularization. Step 4: Introduce convex relaxation to reduce the difficulty of solving the model; Step 5: Accelerate the solution of the perturbation compressed sensing model by using the alternating direction multiplier method; Step 6: Determine whether the current iteration meets the convergence condition. If yes, obtain an N*1 reconstructed vector and proceed to step 7; otherwise, proceed to step 4. Step 7: Generate the energy efficiency map of the photovoltaic cell test area.

2. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 1, characterized in that, The generation of the projection matrix in step 1 refers to calculating the size of the projection matrix based on the length and width of the photovoltaic cell test area and the size of each pixel of the projector. The length of the test area is divided by the length of the pixel to obtain the number of columns n of the projection matrix, and the width of the test area is divided by the width of the pixel to obtain the number of rows m of the projection matrix. The number of projection matrices is equal to the product of m and n.

3. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 1, characterized in that, The generation of projection matrix combination mentioned in step 1 refers to generating N random matrices with m*n rows and columns, and whose elements are only 0 and 1, and then combining all the random matrices into a matrix combination. ; combine matrices Noise is added to each matrix to obtain a combination of projection matrices. .

4. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 3, characterized in that, The measurement vector mentioned in step 2 is a combination of projection matrices. The photovoltaic cell area to be measured is projected sequentially, and the photocurrent value generated by each projection is recorded. The photocurrent values ​​of all projections are combined to form a measurement vector y.

5. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 4, characterized in that, The total variational perturbation compressed sensing model described in step 2 is as follows: ， in, This indicates the operation of finding the minimum value. express The norm, D, represents the total variation operator to be calculated using the number of rows and columns of the projection matrix, x represents the reconstructed vector to be determined, and e represents the interference noise in the measurement vector.

6. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 5, characterized in that, The solution to the perturbation compressed sensing model using the proximal gradient method based on total variational regularization, as described in step 3, is accomplished by the following equation: ， in, This represents the regularization parameter, which controls the strength of the regularization term. , express Norm.

7. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 6, characterized in that, The convex relaxation introduced in step 4 to reduce the difficulty of solving the model is obtained from the following equation: ， in, Indicates to make The operation to obtain the minimum value uses d and p as auxiliary variables constructed by the alternating direction multiplier method to facilitate subsequent steps. Indicates the Lagrange penalty. , , Let these represent the reconstructed vectors after the (n+1)th and nth iterations, respectively. This indicates the iteration step size; a smaller step size results in better iteration optimization. This represents the gradient of the overall least squares problem after the nth iteration.

8. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 7, characterized in that, The acceleration of solving the perturbation compressed sensing model using the alternating direction multiplier method described in step 5 is accomplished by the following equation: ， ， ， in, and Let d represent the intermediate variable d after the (n+1)th and nth iterations, respectively. , Let p represent the intermediate variable after the (n+1)th and nth iterations, respectively.

9. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 8, characterized in that, The convergence condition described in step 6 is as follows: ， The superscript T indicates the transpose operation.

10. The photovoltaic cell energy efficiency map detection method based on the proximal gradient method according to claim 1, characterized in that, The step 7 of generating the photovoltaic cell test area energy efficiency map refers to using the imadjust function in MATLAB to arrange the reconstruction vector x into an m*n reconstruction matrix in column-major order, and then using the imshow function in MATLAB to convert the reconstruction matrix into an energy efficiency map of the photovoltaic cell test area.

Images