Method and device for identifying local shielding of photovoltaic module

Abstract

The invention belongs to the technical field of photovoltaic power generation and machine learning, and relates to a photovoltaic module local shielding identification method and device. Obtaining an IV curvilinear equation of each battery piece in the photovoltaic module, and obtaining a circuit characteristic equation of the photovoltaic module based on the IV curvilinear equations of all the battery pieces and the connection relation; changing the irradiance proportion of different battery pieces in the photovoltaic module to simulate different shielding states; generating a grey-scale map of the photovoltaic module in each shielding state based on the irradiance proportion of each battery piece in the photovoltaic module in each shielding state, and obtaining an IV curve of the photovoltaic module in each shielding state based on the circuit characteristic equation; and taking the IV curve and the grey-scale map of the photovoltaic module in each shielding state as a sample and a label thereof, constructing a training set and training a deep learning model so as to obtain a real-time grey-scale map based on the real-time IV curve of the photovoltaic module and identify the shielded battery piece in the photovoltaic module. According to the scheme provided by the invention, the detection efficiency is improved, and the detection cost is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation and machine learning, and in particular to a method and device for identifying partial shading of a photovoltaic component. Background Art

[0002] The photovoltaic modules in a photovoltaic power station are composed of multiple connected cells. When the cells are blocked by dust, fallen leaves, snow and other materials, the light intensity reaching their surface is reduced, causing the cells to be in a dark state. The current generated by them will be significantly smaller than that of the unblocked cells, causing cell mismatch problems and affecting the power generation of the photovoltaic modules. At the same time, the voltage of the unblocked cells will form a reverse bias at both ends of the blocked cells, thereby increasing the local temperature of the photovoltaic modules, generating a hot spot effect, and even causing a fire. In addition, cell mismatch will accelerate the aging of photovoltaic modules and reduce the life of photovoltaic power stations.

[0003] There are two methods for detecting partial shading of photovoltaic modules in the prior art. One is to monitor the current and voltage of the photovoltaic module in real time, and compare it with the voltage and current in a fault state, so as to determine whether there is a problem of cell shading in the photovoltaic module. However, this method can only detect the entire photovoltaic module and cannot accurately obtain the specific location of the blocked cell. Further investigation of the entire photovoltaic module is required to determine the blocking location. If the voltage and current of each cell in the photovoltaic module are collected in real time, and the voltage and current of each cell are compared with the voltage and current in its fault state to specifically determine whether each cell is blocked, a large amount of data acquisition equipment is required, and a large amount of data comparison and analysis is required for each detection, which not only has high hardware costs but also low detection efficiency. Another method is to use high-definition cameras or drones combined with image processing algorithms to obtain infrared thermal imaging images of photovoltaic modules, and identify the infrared thermal imaging images to determine whether there are local obstruction problems. Although this method can locate the obstruction position in the photovoltaic module, for large-scale photovoltaic power stations, due to the large number of solar cells, a large number of high-definition cameras or drone arrays need to be set up for detection. The detection is time-consuming and energy-intensive, and the maintenance cost of a large number of camera equipment is also high.

[0004] In summary, the existing photovoltaic module partial occlusion identification method has the problems of high detection cost and low detection efficiency. Summary of the invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problems of high detection cost and low detection efficiency in the photovoltaic component partial occlusion identification method in the prior art.

[0006] In order to solve the above technical problems, the present invention provides a method for identifying partial shading of a photovoltaic module, comprising: Obtaining the IV curve equation of each cell in the photovoltaic module, and obtaining the circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module; By changing the irradiance ratio of different cells in the photovoltaic module, different shading states of the photovoltaic module can be simulated; Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, a grayscale image of the photovoltaic module under each shading state is generated, and the IV curve of the photovoltaic module under each shading state is obtained based on the circuit characteristic equation; The IV curve and grayscale image of the photovoltaic module under each shading state are used as samples and their labels respectively, and a training set is constructed and a deep learning model is trained so that a real-time grayscale image can be obtained based on the real-time IV curve of the photovoltaic module, thereby identifying the obscured cells in the photovoltaic module.

[0007] Preferably, obtaining the IV curve equation of each cell in the photovoltaic module includes: Based on the number of series-connected cells and the number of parallel-connected cells in the photovoltaic module, the electrical parameters of the photovoltaic module are normalized to obtain the electrical parameters of a single cell; Based on the electrical parameters of a single cell, the dual-diode equivalent port current equation of the cell and its constraints, a residual matrix is ​​constructed; Solving the residual matrix, obtaining the first saturated dark current, the second saturated dark current, the series resistance and the parallel resistance of the double-diode equivalent model of a single cell; After inputting the first saturated dark current, the second saturated dark current, the series resistance, the parallel resistance and the reverse breakdown voltage of the photovoltaic module into the reverse breakdown curve equation of the cell, they are respectively fitted with the reverse breakdown curve of each cell in the photovoltaic module to obtain the reverse breakdown coefficient of each cell; The IV curve equation of each battery cell is obtained based on the reverse breakdown coefficient of each battery cell.

[0008] Preferably, the residual matrix is ​​expressed as: , in, represents the residual matrix; represents the photogenerated current; Indicates the short-circuit current of a single battery cell; represents the first saturation dark current; represents the series resistance; represents thermal voltage; represents the second saturation dark current; represents parallel resistance; Indicates the open circuit voltage of a single cell; Indicates the double-diode equivalent port current of the battery cell; Indicates the double diode equivalent port voltage of the battery cell; Indicates the dual-diode equivalent port power of the battery cell; Indicates the maximum power of the battery cell; The reverse breakdown curve equation of the battery cell is expressed as: , in, Represents the reverse breakdown curve equation of the battery cell; , , Respectively represent the first reverse breakdown coefficient, the second reverse breakdown coefficient and the third reverse breakdown coefficient of the battery cell; Indicates the equivalent saturation dark current when the battery cell is reversely broken down; Indicates the reverse breakdown voltage of a photovoltaic module.

[0009] Preferably, based on the irradiance ratio of each cell in the photovoltaic module under each shading state, generating a grayscale image of the photovoltaic module under each shading state includes: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the light intensity data of each cell under each shading state is obtained; Based on the light intensity data of all cells in each shading state, a light intensity map of the photovoltaic module in each shading state is generated; The light intensity map of the photovoltaic module under each shading state is converted to obtain a grayscale map of the photovoltaic module under each shading state.

[0010] Preferably, based on the irradiance ratio of each cell in the photovoltaic module under each shading state, generating a grayscale image of the photovoltaic module under each shading state includes: Calculate the irradiance ratio of each cell in the photovoltaic module under each shading state, and multiply the product of the preset irradiance ratio and the gray value mapping coefficient to obtain the gray value of each cell under each shading state; Based on the position of each cell in the photovoltaic module, the grayscale value of each cell in each shading condition is filled into a two-dimensional array to obtain a grayscale image of the photovoltaic module in each shading state.

[0011] Preferably, obtaining the IV curve of the photovoltaic module under each shading state based on the circuit characteristic equation includes: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the photocurrent of each cell under each shading state is calculated; The electrical parameters of the photovoltaic module and the photocurrent of all the cells under each shading state are input into the circuit characteristic equation, the voltage value in the circuit characteristic equation is changed, and the current value corresponding to each voltage value is calculated; Based on each voltage value and its corresponding current value, the IV curve of the photovoltaic module is obtained.

[0012] Preferably, the deep learning model includes a variational autoencoder and an adaptive pooling layer.

[0013] Preferably, obtaining a real-time grayscale image based on a real-time IV curve of a photovoltaic module, thereby identifying a blocked cell in the photovoltaic module comprises: Input the real-time IV curve into the variational autoencoder and output the predicted grayscale image; Input the predicted grayscale image into the adaptive pooling layer for upsampling or downsampling, and output a real-time grayscale image; The grayscale value of each pixel in the real-time grayscale image is obtained, and the obscured battery cell is identified based on the grayscale value of the pixel corresponding to each battery cell.

[0014] Preferably, inputting the predicted grayscale image into the adaptive pooling layer for upsampling or downsampling includes: If the size of the predicted grayscale image is smaller than the preset size, the predicted grayscale image is upsampled using bilinear interpolation; If the size of the predicted grayscale image is larger than the preset size, the predicted grayscale image is downsampled using the average pooling method.

[0015] The present invention also provides a photovoltaic module partial shading identification device, comprising: A circuit characteristic equation acquisition module is used to acquire the IV curve equation of each cell in the photovoltaic module, and obtain the circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module; The shading simulation module is used to simulate the shading state of different cells in the photovoltaic module by changing the irradiance ratio of different cells in the photovoltaic module; A training data generation module is used to generate a grayscale image of the photovoltaic module under each shading state based on the irradiance ratio of each cell in the photovoltaic module under each shading state, and obtain an IV curve of the photovoltaic module under each shading state based on a circuit characteristic equation; The model training and occlusion recognition module is used to take the IV curve and grayscale image of the photovoltaic module under each occlusion state as samples and their labels, build a training set and train a deep learning model, so as to obtain a real-time grayscale image based on the real-time IV curve of the photovoltaic module, thereby identifying the obscured cells in the photovoltaic module.

[0016] The photovoltaic module partial shading identification method provided in this application has the following beneficial effects: Since the irradiance ratio is 1 when the cell is not blocked, as the blocked area increases, its irradiance ratio gradually decreases. Therefore, the blocking condition of the cell in the photovoltaic module can be measured by the irradiance ratio of the cell. Based on this, the present application simulates different blocking states of the photovoltaic module by changing the irradiance ratio of different cells in the photovoltaic module; since the change in irradiance ratio will cause the grayscale value and current of the cell to change, that is, the grayscale image and IV curve of the photovoltaic module under different blocking states are also different, therefore, the present application first obtains the circuit characteristic equation of the photovoltaic module based on the IV curve equation and connection relationship of each cell, generates grayscale images under different blocking states based on the irradiance ratio mapping of each cell in the photovoltaic module under different blocking states, and obtains the IV curve of the photovoltaic module under different blocking states based on the circuit characteristic equation; constructs a training set based on the IV curves and grayscale images under various blocking states and trains a deep learning model to learn the relationship between the grayscale image and the IV curve, so that a real-time grayscale image can be obtained based on the real-time IV curve of the photovoltaic module, and then the blocked cell is identified. During detection, the present application only needs to collect the IV curve of the photovoltaic module to identify the specific location of the obstructed cell. There is no need to collect and analyze the voltage and current of each cell, nor is there any need to rely on hardware equipment such as cameras and drones to obtain an image of the entire photovoltaic module. This not only improves the detection efficiency, but also greatly reduces the detection cost. The detection results obtained based on the present application can be directly used to guide cleaning robots or manual inspections of photovoltaic module cells, thereby reducing the power generation loss of photovoltaic power stations and extending their service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 A flow chart of the method for identifying partial shading of photovoltaic modules provided in this application; Figure 2 A schematic diagram of the deep learning model structure provided for this application; Figure 3 A photovoltaic module composed of multiple connected solar cells provided in this application, and a schematic diagram of a double-diode equivalent model of a single solar cell; Figure 4 A schematic diagram of the structure of a photovoltaic module partial shading identification device provided in this application; Figure 5 It is a schematic diagram comparing the predicted results and measured results of multiple groups of photovoltaic modules with a small shading ratio using the method provided by the present application; wherein, Figure 5 (a) is a schematic diagram of the actual IV curve of the first group of photovoltaic modules with a smaller shading ratio. Figure 5(b) is a schematic diagram of the actual IV curve of the second group of photovoltaic modules with a smaller shading ratio. Figure 5 (c) is a schematic diagram of the actual IV curve of the third group of photovoltaic modules with a smaller shading ratio. Figure 5 (d) is the grayscale image of the prediction result of local occlusion identification of the first group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (e) is the grayscale image of the prediction result of local occlusion identification of the second group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (f) is the grayscale image of the prediction result of local occlusion identification of the third group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (g) is the grayscale image of the measured results of partial shading of the first group of photovoltaic modules with a smaller shading ratio. Figure 5 (h) is the grayscale image of the measured results of partial shading of the second group of photovoltaic modules with a smaller shading ratio. Figure 5 (i) is a grayscale image of the measured results of partial shading of the third group of photovoltaic modules with a smaller shading ratio; Figure 6 It is a schematic diagram comparing the predicted results and measured results of multiple groups of photovoltaic modules with large shading ratios using the method provided by the present application; wherein, Figure 6 (a) is a schematic diagram of the actual IV curve of the first group of photovoltaic modules with a large shading ratio. Figure 6 (b) is a schematic diagram of the actual IV curve of the second group of photovoltaic modules with a larger shading ratio. Figure 6 (c) is a schematic diagram of the actual IV curve of the third group of photovoltaic modules with a larger shading ratio. Figure 6 (d) is the grayscale image of the prediction result of local occlusion identification of the first group of photovoltaic modules with a large occlusion ratio. Figure 6 (e) is the grayscale image of the prediction result of local occlusion identification of the second group of photovoltaic modules with a larger occlusion ratio. Figure 6 (f) is the grayscale image of the prediction result of local occlusion identification of the third group of photovoltaic modules with a larger occlusion ratio. Figure 6 (g) is the grayscale image of the measured results of partial shading of the first group of photovoltaic modules with a large shading ratio. Figure 6 (h) is the grayscale image of the measured results of partial shading of the second group of photovoltaic modules with a larger shading ratio. Figure 6 (i) is a grayscale image of the measured results of local shading of the third group of photovoltaic modules with a larger shading ratio. DETAILED DESCRIPTION

[0018] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0019] See also Figure 1 , Figure 1 The flowchart of the photovoltaic module partial shading identification method provided by the present application is shown, and the method specifically includes: S10: Obtaining an IV curve equation of each cell in the photovoltaic module, and obtaining a circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module.

[0020] Specifically, based on the IV curve equations of all battery cells and the series-parallel relationship between each battery cell, combined with Kirchhoff's law and Ohm's law, the circuit characteristic equation of the photovoltaic module can be obtained. Furthermore, based on the series-parallel relationship between each photovoltaic module in the photovoltaic power station and the circuit characteristic equation of each photovoltaic module, the circuit characteristic equation of the photovoltaic power station can be obtained.

[0021] S20: Simulate different shading states of the photovoltaic module by changing the irradiance ratio of different cells in the photovoltaic module.

[0022] Specifically, when the cell is not shaded, the value of its irradiance ratio E is 1. As the shaded area gradually increases, the value of its irradiance ratio E gradually decreases. Therefore, the shade condition of the photovoltaic module can be characterized by the irradiance ratio E of each cell. By setting the irradiance ratio of cells at different positions, photovoltaic modules under different local shade conditions can be simulated.

[0023] For example, optical simulation software such as Solar Pro, PVsyst, and COMSOL Multiphysics can all build models of photovoltaic modules, and simulate various shading scenarios by setting the irradiance ratio of different cells in the photovoltaic module.

[0024] S30: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, a grayscale image of the photovoltaic module under each shading state is generated, and an IV curve of the photovoltaic module under each shading state is obtained based on a circuit characteristic equation.

[0025] Specifically, the grayscale value of the cell with a larger irradiance ratio is smaller, and the grayscale value of the cell with a smaller irradiance ratio is larger. Therefore, the irradiance ratio of all cells in the photovoltaic module under each shielding state can be mapped into the grayscale values ​​of different pixels in the grayscale image.

[0026] Optionally, in some embodiments of the present application, a grayscale image can be randomly generated and inversely mapped into the irradiance ratio of each cell in the photovoltaic module, thereby calculating the IV curve of the photovoltaic module so as to quickly establish a training set. It is worth noting that this method takes an average of 15 minutes to calculate 10,000 IV curve data.

[0027] S40: The IV curve and grayscale image of the photovoltaic module under each shading state are used as samples and their labels respectively, a training set is constructed and a deep learning model is trained, so as to obtain a real-time grayscale image based on the real-time IV curve of the photovoltaic module, thereby identifying the shaded cells in the photovoltaic module.

[0028] Optionally, in some embodiments of the present application, a series of voltage values ​​and current values ​​of the battery cell can be collected, the collected voltage values ​​and current values ​​can be plotted on a two-dimensional coordinate plane, and the data points on the two-dimensional coordinate plane can be connected into a smooth curve through data fitting methods, thereby obtaining the IV curve equation of the battery cell.

[0029] However, the method of drawing the IV curve and obtaining the IV curve equation through data collection has a large error. Therefore, another method for obtaining the IV curve equation of the battery cell is also provided in the embodiment of the present application, which specifically includes: Step 1: Based on the number of series-connected cells and the number of parallel-connected cells in the photovoltaic module, the electrical parameters of the photovoltaic module are normalized to obtain the electrical parameters of a single cell.

[0030] Specifically, the electrical parameters of the photovoltaic module are measured under standard test conditions (STC).

[0031] In some embodiments of the present application, the electrical parameters include short-circuit current, open-circuit voltage, maximum power point voltage, and maximum power point current.

[0032] Specifically, the calculation formula of the electrical parameters of a single cell is expressed as: , , , , in, Indicates the short-circuit current of a single battery cell; Indicates the short-circuit current of the photovoltaic module; Indicates the number of parallel connected cells; Indicates the open circuit voltage of a single cell; Indicates the open circuit voltage of the photovoltaic module; Indicates the number of battery cells connected in series; Indicates the maximum power point current of a single battery cell; Indicates the maximum power point current of the photovoltaic module; Indicates the maximum power point voltage of a single battery cell; Indicates the maximum power point voltage of the PV module.

[0033] Step 2: Based on the electrical parameters of a single cell, the dual-diode equivalent port current equation of the cell and its constraints, a residual matrix is ​​constructed.

[0034] Specifically, the double-diode equivalent port current equation of the battery cell is: , in, Indicates the double-diode equivalent port current of the battery cell; Indicates the double diode equivalent port voltage of the battery cell; represents the photogenerated current; represents the first saturation dark current; represents the series resistance; represents thermal voltage; represents the second saturation dark current; represents parallel resistance; , is the dual diode ideality factor, , .

[0035] The constraints include: 1. Short circuit constraints: , ; 2. Open circuit constraints: , ; 3. ;in, Indicates the dual-diode equivalent port power of the battery cell; Indicates the maximum power of the battery cell; 4. .

[0036] Furthermore, the residual matrix is ​​expressed as: .

[0037] Step 3: Solve the residual matrix to obtain the first saturated dark current, the second saturated dark current, the series resistance and the parallel resistance of the double-diode equivalent model of a single cell.

[0038] In a specific example of the present application, the Powell method is used to solve the above residual matrix, and the first saturated dark current, the second saturated dark current, the series resistance and the parallel resistance obtained by the solution are processed so that the order of magnitude of these parameters is equal. Specifically, the formula for processing these parameters is expressed as follows: .

[0039] Step 4: After inputting the first saturated dark current, the second saturated dark current, the series resistance, the parallel resistance and the reverse breakdown voltage of the photovoltaic module into the reverse breakdown curve equation of the cell, they are respectively fitted with the reverse breakdown curve of each cell in the photovoltaic module to obtain the reverse breakdown coefficient of each cell.

[0040] Specifically, the reverse breakdown curve equation of the battery cell is expressed as: , in, Represents the reverse breakdown curve equation of the battery cell; , , Respectively represent the first reverse breakdown coefficient, the second reverse breakdown coefficient and the third reverse breakdown coefficient of the battery cell; Indicates the equivalent saturation dark current when the battery cell is reversely broken down; Indicates the reverse breakdown voltage of a photovoltaic module.

[0041] Step 5: Obtain the IV curve equation of each battery cell based on the reverse breakdown coefficient of each battery cell.

[0042] Further, in some embodiments of the present application, the three-dimensional model of the photovoltaic module can be directly established by using optical simulation software, by simulating different shading states, and simulating the light intensity data of each cell under different shading states, the simulated light intensity data is converted into grayscale values, thereby obtaining a grayscale image of the photovoltaic module under different local shading conditions, which specifically includes: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the light intensity data of each cell under each shading state is obtained; Based on the light intensity data of all cells in each shading state, a light intensity map of the photovoltaic module in each shading state is generated; The light intensity map of the photovoltaic module under each shading state is converted to obtain a grayscale map of the photovoltaic module under each shading state.

[0043] Optionally, in other embodiments of the present application, the irradiance ratio of each cell under each shading state may be directly mapped to a grayscale value, so as to construct a grayscale map of the photovoltaic module under each shading state based on the grayscale values ​​of all cells, which specifically includes: Calculate the irradiance ratio of each cell in the photovoltaic module under each shading state, and multiply the product of the preset irradiance ratio and the gray value mapping coefficient to obtain the gray value of each cell under each shading state; Based on the position of each cell in the photovoltaic module, the grayscale value of each cell in each shading condition is filled into a two-dimensional array to obtain a grayscale image of the photovoltaic module in each shading state.

[0044] For example, the grayscale value G is usually between 0 and 255, and the irradiance ratio E of the cell is usually between 0 and 1. Therefore, the grayscale value of the pixel at the location of the cell where the irradiance ratio E is 1 can be set to 255. The calculation formula for the grayscale value G of the pixel at each cell can be: , so that the grayscale image of the photovoltaic module can be obtained based on the grayscale values ​​of the pixels where all the cells are located.

[0045] Furthermore, obtaining the IV curve of the photovoltaic module under each shading state based on the circuit characteristic equation includes: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the photocurrent of each cell under each shading state is calculated; The electrical parameters of the photovoltaic module and the photocurrent of all the cells under each shading state are input into the circuit characteristic equation, the voltage value in the circuit characteristic equation is changed, and the current value corresponding to each voltage value is calculated; Based on each voltage value and its corresponding current value, the IV curve of the photovoltaic module is obtained.

[0046] Further, in some embodiments of the present application, Figure 2 As shown in Figure 1, the deep learning model includes a variational auto encoder (VAE) and an adaptive pooling layer.

[0047] Specifically, since the present application uses transfer learning technology, that is, using a pre-trained variational autoencoder to train a new training set, and the image size in the new training set may be inconsistent with the image size during pre-training, therefore, in order to ensure that the image size output by the model on the new training set is consistent with the preset size, the present application connects an adaptive pooling layer after the variational autoencoder to upsample or downsample the grayscale image output by the variational autoencoder, so that the grayscale image size output by the model on different data sets can be equal to the preset size, thereby improving the accuracy of battery cell occlusion recognition. At the same time, since the variational autoencoder has been pre-trained, the model only needs to be trained for a dozen rounds on different training sets to achieve high-precision recognition results, shortening the training time of the model.

[0048] It is worth noting that during the model training process, the preset size refers to the size of the labels in the training set. During the model prediction process, the preset size is set based on the correspondence between the cells in the photovoltaic module and the pixels in the grayscale image. For example: Figure 3Shown is a 72*2 photovoltaic module provided by the present application, in which each cell can be equivalent to a dual-diode equivalent model, and it is preset that every three pixels correspond to one cell, and the pixel position in the grayscale image corresponds to the relative position of the cell and the photovoltaic module. When occlusion identification is performed on the photovoltaic module, the preset size of the grayscale image is 216*432.

[0049] Further, in some embodiments of the present application, training the deep learning model in step S40 specifically includes: Get the samples and labels of the samples in the training set, input the samples into the variational autoencoder, and output the initial predicted grayscale image.

[0050] The initial predicted grayscale image is input into the adaptive pooling layer, and based on the size of the initial predicted grayscale image and the size of the label, the initial predicted grayscale image is upsampled or downsampled to output the target predicted grayscale image.

[0051] A loss function is constructed based on the target prediction grayscale image and label, and the variational autoencoder and adaptive pooling layer are iteratively trained using samples in the training set until the value of the loss function converges or reaches the preset number of iterations to obtain a trained deep learning model.

[0052] In a specific example of the present application, the preset number of iterations during model training is 10 to 30.

[0053] Furthermore, in step S40, a specific implementation method of obtaining a real-time grayscale image based on the real-time IV curve of the photovoltaic module to identify the obscured cells in the photovoltaic module includes: The real-time IV curve is input into the variational autoencoder and the predicted grayscale image is output.

[0054] The predicted grayscale image is input into the adaptive pooling layer for upsampling or downsampling, and a real-time grayscale image is output.

[0055] The grayscale value of each pixel in the real-time grayscale image is obtained, and the obscured battery cell is identified based on the grayscale value of the pixel corresponding to each battery cell.

[0056] Specifically, inputting the predicted grayscale image into the adaptive pooling layer for upsampling or downsampling includes: If the size of the predicted grayscale image is smaller than the preset size, the predicted grayscale image is upsampled using bilinear interpolation; If the size of the predicted grayscale image is larger than the preset size, the average pooling method is used to downsample the predicted grayscale image.

[0057] For example, assume that the size of the predicted grayscale image is , the default size is ,like , , then the predicted grayscale image is upsampled. Specifically, the real-time grayscale image obtained after upsampling is expressed as: , in, Represents the real-time grayscale image obtained after upsampling; Represents the row and column values ​​of pixels in the real-time grayscale image; Represents the predicted grayscale image; Represents the row and column values ​​of the pixel points in the predicted grayscale image; Represents the interpolation weight.

[0058] like , , then the predicted grayscale image is downsampled. Specifically, the real-time grayscale image obtained after downsampling is expressed as: , in, Represents the real-time grayscale image obtained after downsampling; Represents the row and column values ​​of pixels in the real-time grayscale image; Indicates the size of the pooling kernel.

[0059] It is worth noting that the interpolation weights during upsampling can be automatically calculated by the interpolation method, and the size of the pooling kernel during downsampling can be automatically adjusted according to the scaling ratio.

[0060] Based on the photovoltaic module partial shading identification method provided in the above embodiment, the embodiment of the present application also provides a photovoltaic module partial shading identification device, such as Figure 4 As shown, the device specifically includes: The circuit characteristic equation acquisition module 10 is used to acquire the IV curve equation of each cell in the photovoltaic module, and obtain the circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module; The shielding simulation module 20 is used to simulate the shielding state of different cells in the photovoltaic module by changing the irradiance ratio of different cells in the photovoltaic module; The training data generation module 30 is used to generate a grayscale image of the photovoltaic module under each shading state based on the irradiance ratio of each cell in the photovoltaic module under each shading state, and obtain an IV curve of the photovoltaic module under each shading state based on a circuit characteristic equation; The model training and occlusion recognition module 40 is used to use the IV curve and grayscale image of the photovoltaic module under each occlusion state as samples and their labels, respectively, to construct a training set and train a deep learning model, so as to obtain a real-time grayscale image based on the real-time IV curve of the photovoltaic module, thereby identifying the obscured cells in the photovoltaic module.

[0061] like Figure 5The figure shows a comparison diagram of multiple groups of prediction results and measured results for local shading identification of photovoltaic modules with a small shading ratio using the method provided by the present application, wherein: Figure 5 (a) is a schematic diagram of the actual IV curve of the first group of photovoltaic modules with a smaller shading ratio. Figure 5 (b) is a schematic diagram of the actual IV curve of the second group of photovoltaic modules with a smaller shading ratio. Figure 5 (c) is a schematic diagram of the actual IV curve of the third group of photovoltaic modules with a smaller shading ratio. Figure 5 (d) is the grayscale image of the prediction result of local occlusion identification of the first group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (e) is the grayscale image of the prediction result of local occlusion identification of the second group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (f) is the grayscale image of the prediction result of local occlusion identification of the third group of photovoltaic modules with a smaller occlusion ratio. Figure 5 (g) is the grayscale image of the measured results of partial shading of the first group of photovoltaic modules with a smaller shading ratio. Figure 5 (h) is the grayscale image of the measured results of partial shading of the second group of photovoltaic modules with a smaller shading ratio. Figure 5 (i) is a grayscale image of the measured results of local shading of the third group of photovoltaic modules with a smaller shading ratio.

[0062] like Figure 6 The figure shows a comparison diagram of multiple groups of prediction results and measured results for local shading identification of photovoltaic modules with a large shading ratio using the method provided by the present application, wherein: Figure 6 (a) is a schematic diagram of the actual IV curve of the first group of photovoltaic modules with a large shading ratio. Figure 6 (b) is a schematic diagram of the actual IV curve of the second group of photovoltaic modules with a larger shading ratio. Figure 6 (c) is a schematic diagram of the actual IV curve of the third group of photovoltaic modules with a larger shading ratio. Figure 6 (d) is the grayscale image of the prediction result of local occlusion identification of the first group of photovoltaic modules with a large occlusion ratio. Figure 6 (e) is the grayscale image of the prediction result of local occlusion identification of the second group of photovoltaic modules with a larger occlusion ratio. Figure 6 (f) is the grayscale image of the prediction result of local occlusion identification of the third group of photovoltaic modules with a larger occlusion ratio. Figure 6 (g) is the grayscale image of the measured results of partial shading of the first group of photovoltaic modules with a large shading ratio. Figure 6 (h) is the grayscale image of the measured results of partial shading of the second group of photovoltaic modules with a larger shading ratio. Figure 6 (i) is a grayscale image of the measured results of local shading of the third group of photovoltaic modules with a larger shading ratio.

[0063] from Figure 5 and Figure 6 It can be seen from the multiple groups of comparison results given that the local occlusion identification method based on the IV curve of the photovoltaic module provided in the present application has good identification performance on photovoltaic modules with a small occlusion ratio and a large occlusion ratio.

[0064] The photovoltaic module partial occlusion recognition method provided in the present application generates a grayscale image and combines it with electrical simulation to quickly build a training set when training a deep learning model, so that a large amount of training data can be generated in a short time, reducing the labor cost of data collection and annotation; when using the trained deep learning model for detection, it is only necessary to collect the IV curve data of the photovoltaic module, and the occlusion situation can be automatically analyzed by the algorithm, without the need to set up additional high-definition cameras or drone arrays, which significantly reduces hardware costs and time costs; at the same time, the occlusion data obtained in the present application can be directly used to guide cleaning robots or manual inspections, help reduce power generation losses, ensure the efficient and stable operation of the cells in the photovoltaic module, and extend the service life of the photovoltaic module, which has high practical value.

[0065] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0066] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for identifying partial shading of a photovoltaic module, characterized in that: include: Obtaining the IV curve equation of each cell in the photovoltaic module, and obtaining the circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module; By changing the irradiance ratio of different cells in the photovoltaic module, different shading states of the photovoltaic module can be simulated; Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, a grayscale image of the photovoltaic module under each shading state is generated, and the IV curve of the photovoltaic module under each shading state is obtained based on the circuit characteristic equation; The IV curve and grayscale image of the photovoltaic module under each shading state are used as samples and their labels respectively, and a training set is constructed and a deep learning model is trained so that a real-time grayscale image can be obtained based on the real-time IV curve of the photovoltaic module, thereby identifying the obscured cells in the photovoltaic module.

2. The method for identifying partial shading of photovoltaic modules according to claim 1, characterized in that: The IV curve equation for each cell in a photovoltaic module includes: Based on the number of series-connected cells and the number of parallel-connected cells in the photovoltaic module, the electrical parameters of the photovoltaic module are normalized to obtain the electrical parameters of a single cell; Based on the electrical parameters of a single cell, the dual-diode equivalent port current equation of the cell and its constraints, a residual matrix is ​​constructed; Solving the residual matrix, obtaining the first saturated dark current, the second saturated dark current, the series resistance and the parallel resistance of the double-diode equivalent model of a single cell; After inputting the first saturated dark current, the second saturated dark current, the series resistance, the parallel resistance and the reverse breakdown voltage of the photovoltaic module into the reverse breakdown curve equation of the cell, they are respectively fitted with the reverse breakdown curve of each cell in the photovoltaic module to obtain the reverse breakdown coefficient of each cell; The IV curve equation of each battery cell is obtained based on the reverse breakdown coefficient of each battery cell.

3. The method for identifying partial shading of a photovoltaic module according to claim 2, characterized in that: The residual matrix is ​​expressed as: ， in, represents the residual matrix; represents the photogenerated current; Indicates the short-circuit current of a single cell; represents the first saturation dark current; represents the series resistance; represents thermal voltage; represents the second saturation dark current; represents parallel resistance; Indicates the open circuit voltage of a single cell; Indicates the double-diode equivalent port current of the battery cell; Indicates the double diode equivalent port voltage of the battery cell; Indicates the dual-diode equivalent port power of the battery cell; Indicates the maximum power of the battery cell; The reverse breakdown curve equation of the battery cell is expressed as: ， in, Represents the reverse breakdown curve equation of the battery cell; , , Respectively represent the first reverse breakdown coefficient, the second reverse breakdown coefficient and the third reverse breakdown coefficient of the battery cell; Indicates the equivalent saturation dark current when the battery cell is reversely broken down; Indicates the reverse breakdown voltage of a photovoltaic module.

4. The method for identifying partial shading of a photovoltaic module according to claim 1, characterized in that: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the grayscale image of the photovoltaic module under each shading state is generated, including: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the light intensity data of each cell under each shading state is obtained; Based on the light intensity data of all cells in each shading state, a light intensity map of the photovoltaic module in each shading state is generated; The light intensity map of the photovoltaic module under each shading state is converted to obtain a grayscale map of the photovoltaic module under each shading state.

5. The method for identifying partial shading of a photovoltaic module according to claim 1, characterized in that: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the grayscale image of the photovoltaic module under each shading state is generated, including: Calculate the irradiance ratio of each cell in the photovoltaic module under each shading state, and multiply the product of the preset irradiance ratio and the gray value mapping coefficient to obtain the gray value of each cell under each shading state; Based on the position of each cell in the photovoltaic module, the grayscale value of each cell in each shading condition is filled into a two-dimensional array to obtain a grayscale image of the photovoltaic module in each shading state.

6. The method for identifying partial shading of a photovoltaic module according to claim 1, characterized in that: The IV curve of the photovoltaic module under each shading state is obtained based on the circuit characteristic equation, including: Based on the irradiance ratio of each cell in the photovoltaic module under each shading state, the photocurrent of each cell under each shading state is calculated; The electrical parameters of the photovoltaic module and the photocurrent of all the cells under each shading state are input into the circuit characteristic equation, the voltage value in the circuit characteristic equation is changed, and the current value corresponding to each voltage value is calculated; Based on each voltage value and its corresponding current value, the IV curve of the photovoltaic module is obtained.

7. The method for identifying partial shading of a photovoltaic module according to claim 1, characterized in that: The deep learning model includes a variational autoencoder and an adaptive pooling layer.

8. The method for identifying partial shading of a photovoltaic module according to claim 7, characterized in that: Based on the real-time IV curve of the photovoltaic module, a real-time grayscale image is obtained to identify the obscured cells in the photovoltaic module, including: Input the real-time IV curve into the variational autoencoder and output the predicted grayscale image; Input the predicted grayscale image into the adaptive pooling layer for upsampling or downsampling, and output a real-time grayscale image; The grayscale value of each pixel in the real-time grayscale image is obtained, and the obscured battery cell is identified based on the grayscale value of the pixel corresponding to each battery cell.

9. The method for identifying partial shading of a photovoltaic module according to claim 8, characterized in that: Inputting the predicted grayscale image into the adaptive pooling layer for upsampling or downsampling includes: If the size of the predicted grayscale image is smaller than the preset size, the predicted grayscale image is upsampled using bilinear interpolation; If the size of the predicted grayscale image is larger than the preset size, the predicted grayscale image is downsampled using the average pooling method.

10. A photovoltaic module partial shading identification device, characterized in that: include: A circuit characteristic equation acquisition module is used to acquire the IV curve equation of each cell in the photovoltaic module, and obtain the circuit characteristic equation of the photovoltaic module based on the IV curve equations and connection relationships of all cells in the photovoltaic module; The shading simulation module is used to simulate the shading state of different cells in the photovoltaic module by changing the irradiance ratio of different cells in the photovoltaic module; A training data generation module is used to generate a grayscale image of the photovoltaic module under each shading state based on the irradiance ratio of each cell in the photovoltaic module under each shading state, and obtain an IV curve of the photovoltaic module under each shading state based on a circuit characteristic equation; The model training and occlusion recognition module is used to take the IV curve and grayscale image of the photovoltaic module under each occlusion state as samples and their labels, build a training set and train a deep learning model, so as to obtain a real-time grayscale image based on the real-time IV curve of the photovoltaic module, thereby identifying the obscured cells in the photovoltaic module.

Images