Method for predicting power generation of photovoltaic power generation system and related device

Abstract

The application provides a power generation amount prediction method based on a photovoltaic power generation system and related devices, which comprises the following steps: first, acquiring multi-source data of the photovoltaic power generation system, including photovoltaic component-level electrical data, photovoltaic panel terrain data, and photovoltaic panel shelter data; then, performing local shelter identification on the photovoltaic components according to the multi-source data, determining a corrected gray map and sheltered cell positioning data; subsequently, combining the sheltered cell positioning data with the multi-source data of the photovoltaic power generation system, calculating the power generation amount of the locally sheltered cells, and obtaining a power station-level power prediction value; finally, performing error evaluation on the power station-level power prediction value, and outputting a final power generation amount prediction result when the error threshold is met. By introducing the photovoltaic panel terrain data and the photovoltaic panel shelter data, and combining the local shelter identification to obtain accurate positioning information of the sheltered cells, the problem of overestimation or underestimation of output caused by simplification or neglect of local sheltering in the existing method can be avoided.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power generation technology, specifically to a method and related apparatus for predicting power generation based on a photovoltaic power generation system. Background Technology

[0002] With the rapid development of the global new energy industry, photovoltaic (PV) power generation, as a clean and sustainable energy form, is constantly expanding its installed capacity and application scenarios, extending from large-scale ground-mounted PV power plants to complex scenarios such as distributed rooftop power plants and mountain power plants. Power generation forecasting, as a core supporting technology for the efficient operation of PV systems and grid dispatch, directly affects the new energy absorption rate and grid stability. Currently, neural network-based forecasting methods have become one of the mainstream technologies in the field of PV power generation forecasting. These methods typically use global irradiance, temperature, wind speed, and other meteorological data provided by numerical weather prediction (NWP), combined with historical power output data of the PV system and nominal component parameters as input. By training models such as Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNN), the mapping relationship between meteorological conditions and power generation is learned, thereby achieving power generation forecasting for future periods.

[0003] However, existing neural network-based photovoltaic power generation prediction methods still have shortcomings in complex application scenarios. For example, photovoltaic panels often face dynamic or static shading such as tree shadows, temporary construction equipment, low-altitude cumulus clouds, and birds during operation. Such shading only affects some cells or modules, but due to the series-parallel circuit characteristics of photovoltaic modules, it will reduce the current output of the entire series branch, thus overestimating or underestimating the output of some photovoltaic panels. Ultimately, this leads to a large deviation between the model's prediction results and the actual power generation, making it difficult to meet the grid dispatch's demand for high-precision prediction. Summary of the Invention

[0004] This application provides a method and related apparatus for predicting the power generation of a photovoltaic power generation system. It can obtain the accurate location information of the shaded solar cells by introducing photovoltaic panel terrain data and photovoltaic panel shading data, combined with local shading identification, so as to avoid the problem of overestimation or underestimation of power output caused by simplification or neglect of local shading in existing methods.

[0005] A first aspect of this application provides a method for predicting power generation based on a photovoltaic power generation system, the method comprising:

[0006] Acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data;

[0007] Based on the multi-source data of the photovoltaic power generation system, the photovoltaic modules are partially shaded to identify the corrected grayscale image and the location data of the shaded cells.

[0008] Based on the location data of the shaded solar cells and the multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is determined, and the power station-level power prediction value is obtained.

[0009] The predicted power generation value at the power plant level is evaluated. If the predicted power generation value at the power plant level meets the error threshold, the predicted power generation value at the power plant level is output as the final power generation prediction result.

[0010] In one possible implementation, the step of identifying partial shading of photovoltaic modules based on multi-source data from the photovoltaic power generation system and determining the corrected grayscale image includes:

[0011] Extract photovoltaic module cell curve data and cell irradiance ratio data from the photovoltaic module-level electrical data;

[0012] Based on the photovoltaic panel topography data and photovoltaic module cell curve data, determine the comprehensive value of the micro-topography features for each cell;

[0013] The micro-topography correction coefficient is determined based on the type of shading object in the photovoltaic panel shading object data;

[0014] The grayscale value of the micro-topography correction cell is determined based on the battery irradiance ratio data, the comprehensive value of micro-topography features, and the micro-topography correction coefficient.

[0015] The grayscale image update cycle is determined based on the moving speed of the obstruction in the photovoltaic panel obstruction data;

[0016] The grayscale value of the battery cell is corrected based on the micro-topography and the update cycle, and the corrected grayscale image is determined in combination with the location of the battery cell.

[0017] In one possible implementation, the step of identifying partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system and determining the location data of the shaded cells includes:

[0018] The photovoltaic module cell curve data is used as training samples, and the corrected grayscale image is used as training labels to train a variational autoencoder neural network with an attention mechanism.

[0019] Based on the type of shading object in the photovoltaic panel shading object data, optimize the attention weights of the variational autoencoder neural network;

[0020] The real-time collected photovoltaic module cell curve data is input into the variational autoencoder neural network, and the output is a real-time corrected grayscale image.

[0021] Based on the real-time corrected grayscale image, the location data of the obscured battery cell is determined.

[0022] In one possible implementation, based on the location data of the shaded solar cells and multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is determined, and a power plant-level power prediction value is obtained, including:

[0023] Based on the photovoltaic module-level electrical data and the location data of the shaded solar cells, the local shading factor and the mismatch factor of the solar cells are determined.

[0024] The actual irradiance of each solar cell is determined based on the local shading factor of the solar cell.

[0025] The effective area and working efficiency of the solar cells are extracted from the photovoltaic module-level electrical data.

[0026] The actual output of a single solar cell is determined based on the effective area of ​​the solar cell, the working efficiency of the solar cell, and the mismatch factor of the solar cell.

[0027] The module-level output is obtained by summing the actual output of each individual battery cell.

[0028] The summation of the component-level outputs yields the predicted power generation value at the power station level.

[0029] In one possible implementation, determining the local shading factor of the solar cell based on the photovoltaic module-level electrical data and the location data of the shaded solar cell includes:

[0030] Reference irradiance and surface albedo data are extracted from the photovoltaic module-level electrical data;

[0031] Based on the reference irradiance, surface albedo data, and the location data of the obscured solar cells, the initial local shading factor of the solar cells is determined.

[0032] Based on the moving speed of the obstruction in the photovoltaic panel obstruction data, the update range of the initial cell local shading factor is determined, and the cell local shading factor is obtained.

[0033] In one possible implementation, determining the cell mismatch factor based on the photovoltaic module-level electrical data and the location data of the shaded cells includes:

[0034] Extract the series connection data of the solar cells within the photovoltaic module from the photovoltaic module-level electrical data;

[0035] Based on the series connection data, determine the series branch number k where the nth battery cell is located;

[0036] Based on the location data of the obscured battery cells, determine the total number of battery cells, the number of obscured battery cells, and the degree of obstruction of each obscured battery cell in the series branch number k.

[0037] Based on the total number of solar cells, the number of solar cells that are blocked, and the degree of solar cell blocking, determine the branch-level blocking impact coefficient;

[0038] The initial cell mismatch factor is determined based on the branch-level shading influence coefficient.

[0039] The cell mismatch factor is obtained by correcting the cell mismatch factor based on the type of shading object in the photovoltaic panel shading object data.

[0040] In one possible implementation, the evaluation of the predicted power generation value at the power plant level, and the output of the predicted power generation value as the final power generation prediction result when the predicted power generation value at the power plant level meets an error threshold, includes:

[0041] The root mean square error and mean absolute error are used to evaluate the prediction results of the power station-level power prediction values.

[0042] If the root mean square error is less than 50W and the mean absolute error is less than 8%, then the power plant-level power prediction value is output as the final power generation prediction result.

[0043] This example provides a method for predicting power generation based on a photovoltaic (PV) power generation system. First, it acquires multi-source data of the PV power generation system, including electrical data at the PV module level, PV panel topographic data, and PV panel shading data. Then, based on this multi-source data, it identifies local shading of the PV modules, determines the corrected grayscale image, and obtains the location data of the shaded cells. Subsequently, it combines the location data of the shaded cells with the multi-source data of the PV power generation system to calculate the power generation of the locally shaded cells and obtain a power plant-level power prediction value. Finally, it evaluates the error of this power plant-level power prediction value, and outputs the final power generation prediction result when the error threshold is met. By introducing PV panel topographic data and PV panel shading data, and combining them with local shading identification, it obtains accurate location information of the shaded cells, avoiding the overestimation or underestimation of output caused by simplification or neglect of local shading in existing methods. Simultaneously, it captures the local irradiance differences of PV panels under different terrain conditions, reducing prediction bias caused by the inability of global terrain data to reflect microscopic differences, thus meeting the grid dispatching requirements for high-precision prediction, thereby improving the renewable energy absorption rate and ensuring the stability of grid operation.

[0044] A second aspect of this application provides a power generation prediction device for a photovoltaic power generation system, the device comprising:

[0045] The first acquisition unit is used to acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data;

[0046] The first processing unit is used to identify partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system, and to determine the corrected grayscale image and the location data of the shaded battery cells.

[0047] The second processing unit is used to determine the power generation of the partially shaded battery based on the location data of the shaded battery cell and the multi-source data of the photovoltaic power generation system, and to obtain the power station-level power prediction value.

[0048] The third processing unit is used to evaluate the prediction results of the power plant-level power prediction value. When the predicted value of the power plant-level power prediction value meets the error threshold, the power plant-level power prediction value is output as the final power generation prediction result.

[0049] A third aspect of this application provides a terminal including a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, and the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the steps of the power generation prediction method based on a photovoltaic power generation system as described in the first aspect of this application.

[0050] A fourth aspect of this application provides a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the power generation prediction method based on a photovoltaic power generation system in the first aspect of this application.

[0051] A fifth aspect of this application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the power generation prediction method based on a photovoltaic power generation system in the first aspect of this application. The computer program product may be a software installation package. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This application provides a schematic diagram of the overall process for a photovoltaic power generation prediction method based on a photovoltaic power generation system.

[0054] Figure 2 This application provides a schematic diagram of the overall structure of a photovoltaic power generation prediction device for an embodiment of the present application;

[0055] Figure 3 This application provides a schematic diagram of the structure of a terminal.

[0056] Figure label:

[0057] First acquisition unit-1, first processing unit-2, second processing unit-3, third processing unit-4. Detailed Implementation

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0060] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0061] To better understand the power generation prediction method based on photovoltaic power generation systems provided in this application embodiment, the following is a brief introduction to the application scenarios of the power generation prediction method based on photovoltaic power generation systems. In the prediction method used for photovoltaic power generation, local shading is a key factor leading to a sharp increase in prediction error. During operation, photovoltaic panels often face dynamic or static shading such as tree shadows, temporary construction equipment, low-altitude cumulus clouds, and birds. Such shading only affects some cells or modules, but due to the series and parallel circuit characteristics of photovoltaic modules, the shaded cells will cause the current output of the entire series branch to be reduced. Existing prediction methods either do not specifically identify the shading situation and directly attribute the power output reduction caused by shading to changes in weather conditions, or they only correct it with a simple shading ratio coefficient without distinguishing the dynamic impact of the type of shading object and its movement speed on the shading range, and without accurately locating the specific location of the shaded cells. Ultimately, the model cannot accurately quantify the power output loss caused by shading, and the prediction results deviate significantly from the actual power generation, especially in scenarios such as cloudy weather and complex terrain, making it difficult to meet the grid dispatching requirements for high-precision prediction.

[0062] The power generation prediction method based on photovoltaic power generation systems is applied to power generation prediction devices based on photovoltaic power generation systems. Figure 1 A schematic diagram of the overall process for a method to predict power generation based on a photovoltaic power generation system is shown. Figure 1 As shown, it includes:

[0063] S1. Obtain multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data.

[0064] The photovoltaic module-level electrical data includes photovoltaic module cell curve data and cell irradiance ratio data. The photovoltaic module cell curve data can be obtained through a module-level IV tester deployed at the photovoltaic module junction box, and the cell irradiance ratio data can be collected by a miniature irradiance sensor to collect the irradiance ratio of the cells. The photovoltaic module-level electrical data also includes basic cell parameters, including the effective area of ​​the cell, the cell efficiency, and the series connection method of the cells within the module, which can be obtained from the design drawings of the photovoltaic power station. The photovoltaic module-level electrical data also includes irradiance and albedo correlation data, which includes baseline irradiance intensity and surface albedo data. The baseline irradiance intensity can be obtained from meteorological stations around the photovoltaic power station, and the surface albedo data can be obtained by using a multispectral camera mounted on a drone and then determining the image pixel grayscale value.

[0065] The photovoltaic panel terrain data includes slope data, orientation data, altitude data, and surface albedo type data. Slope and orientation data can be obtained through drone aerial photography combined with 3D modeling. Specifically, a drone equipped with a lidar scans the power station area to generate a digital elevation model, and then GIS software is used to extract the slope and orientation corresponding to each photovoltaic panel. Altitude data can be extracted from high-precision GIS maps and combined with on-site GPS positioning calibration. Surface albedo type data (such as grassland, concrete, snow, etc.) can be obtained through pixel classification algorithms of drone visible light aerial images.

[0066] The photovoltaic panel shading data includes shading object type data and shading object movement speed data. Shading object type data is used to distinguish between fixed shading objects (such as surrounding buildings and long-term trees) and dynamic shading objects (such as low-altitude cumulus clouds and birds), and can be obtained in real time by high-definition smart cameras deployed around the photovoltaic array with the help of a pre-trained target detection model. Shading object movement speed data is for dynamic shading objects and can be obtained by measuring with millimeter-wave radar paired with the cameras, while the movement speed of fixed shading objects is uniformly set to 0.

[0067] S2. Based on the multi-source data of the photovoltaic power generation system, identify the partial shading of the photovoltaic modules, and determine the corrected grayscale image and the positioning data of the shaded cells.

[0068] Specifically, step S2 includes the following steps:

[0069] S201. Extract the photovoltaic module cell curve data and cell irradiance ratio data from the photovoltaic module-level electrical data.

[0070] S202. Based on the photovoltaic panel topography data and photovoltaic module cell curve data, determine the comprehensive value of the micro-topography features of each cell.

[0071] The formula for calculating the comprehensive value of micro-topographic features is as follows:

[0072]

[0073] in, This represents the comprehensive value of the micro-topographic features of the nth solar cell. This is the slope weighting coefficient for the solar cells. The orientation weighting factor for the solar cells. This is the weighting coefficient for surface albedo. Let n be the slope factor of the nth solar cell. Let n be the orientation factor of the nth solar cell. Let be the surface albedo factor of the nth solar cell.

[0074] S203. Determine the micro-topography correction coefficient based on the type of shading object in the photovoltaic panel shading object data.

[0075] Specifically, when the "shading object type" in the photovoltaic panel shading object data obtained in step S1 is a fixed shading object such as a building or a long-term tree, the micro-topography correction coefficient is... Set to 0.1 when the occlusion type is dynamic occlusion. Take 0.05.

[0076] S204. Determine the grayscale value of the micro-topography correction battery cell based on the battery irradiance ratio data, the comprehensive value of micro-topography features, and the micro-topography correction coefficient.

[0077] The formula for calculating the grayscale value of the micro-topography correction battery cell is as follows:

[0078]

[0079] in, Here, represents the grayscale value of the nth solar cell after micro-topography correction, with 255 being the maximum grayscale value. The irradiance ratio data for the nth solar cell is given. This is the micro-topography correction factor. This is the comprehensive value of the micro-topographic features of the nth battery cell.

[0080] S205. Determine the grayscale image update cycle based on the moving speed of the obstruction in the photovoltaic panel obstruction data.

[0081] Specifically, when the moving speed of the shading object in the photovoltaic panel shading object data obtained in step S1 is greater than 5 km / h, the grayscale image update cycle is set to 10 seconds; when the moving speed is less than or equal to 5 km / h and greater than 0, the update cycle is set to 60 seconds; when the moving speed is 0, the update cycle is set to 300 seconds. The new cycle starts counting from the time the shading object speed data is updated, and the next grayscale image generation is triggered after the cycle is reached.

[0082] S206. Based on the micro-topography, the grayscale value of the battery cell is corrected and the update cycle is updated. Combined with the location of the battery cell, the corrected grayscale image is determined.

[0083] First, based on the physical structure of the photovoltaic module, the coordinate position of each cell in the grayscale image is determined; then, the grayscale value Gn of the cell calculated in step S204 is filled into the corresponding coordinates; finally, the grayscale image data is refreshed periodically according to the update cycle determined in step S205.

[0084] In this example, the size of the corrected grayscale image is consistent with the arrangement of the photovoltaic module cells (e.g., 60 cells correspond to a 6×10 pixel image). The grayscale value distribution intuitively reflects the degree of shading of each cell - dark areas (low grayscale values) indicate severe shading, and light areas (high grayscale values) indicate slight or no shading. This provides clear image input for subsequent positioning of shaded cells, and combines the aforementioned results to generate the corrected grayscale image, thereby realizing the visualization and quantification of the shading status.

[0085] S207. Using the photovoltaic module cell curve data as training samples and the corrected grayscale image as training labels, train a variational autoencoder neural network with an attention mechanism.

[0086] The system employs a variational autoencoder (VAE) with an attention mechanism to map battery curve data to occlusion grayscale images. Specifically, the photovoltaic module battery curve data extracted in step S201 is used as training samples (input layer), with each curve containing 200 voltage-current sampling points. The corrected grayscale image generated in step S206 is used as the training label (output layer, with the same size as the battery cell arrangement). The network structure adopts an encoder-decoder architecture. The encoder compresses the battery curve data into latent vectors using a 3-layer convolutional neural network (CNN), and the decoder reconstructs the latent vectors into grayscale images using a 3-layer deconvolutional network. An attention mechanism module is embedded between the encoder and decoder. This module enhances feature extraction of occlusion-sensitive areas by calculating the correlation between battery curve features and grayscale image regions. During training, the Adam optimizer is used, with the weighted sum of reconstruction loss and KL divergence loss as the total loss function. The training period is set to 50 epochs until the validation set loss stabilizes.

[0087] S208. Optimize the attention weights of the variational autoencoder neural network based on the type of shading object in the photovoltaic panel shading object data.

[0088] Specifically, when the occlusion type obtained in step S1 is a fixed occlusion (such as buildings or trees), the recognition ability for fixed contour occlusions is enhanced by increasing the attention weight corresponding to the edge features. When the occlusion type is a dynamic occlusion (such as cumulus clouds or birds), the capture ability for dynamically changing occlusions is enhanced by increasing the attention weight corresponding to the grayscale change rate feature. Weight optimization is achieved through an "occlusion type-weight mapping table." The weight parameters in the mapping table are calibrated offline based on historical data from 100,000 different occlusion scenarios to ensure that weight adjustments for a certain type of occlusion can improve recognition accuracy.

[0089] S209. The real-time collected photovoltaic module cell curve data is input into the variational autoencoder neural network, and the real-time corrected grayscale image is output.

[0090] Specifically, photovoltaic module cell curve data can be acquired in real time using a module-level IV tester (the acquisition frequency is consistent with the grayscale image update cycle in step S205, such as 10 seconds / time). After acquisition, the data is standardized (the current value is scaled to the range of 0-1), and then input into the variational autoencoder neural network trained in step S207 and optimized in step S208. The network calculates through forward propagation and outputs a real-time corrected grayscale image corresponding to the current moment (the size is consistent with step S206, such as 6×10 pixels). This grayscale image can reflect the shading status of each cell in real time (e.g., when a bird flies over a certain area, the corresponding pixel grayscale value drops instantly). To ensure real-time performance, the network inference process is accelerated by GPU, and the generation time of a single grayscale image is controlled within 50ms, meeting the update cycle requirements.

[0091] S2010. Determine the location data of the obscured battery cell based on the real-time corrected grayscale image.

[0092] The process involves setting a grayscale threshold and identifying pixels in the real-time corrected grayscale image with grayscale values ​​less than the threshold as occluded areas. Then, based on the "cell-pixel coordinate" mapping established in step S206, the pixel coordinates of the occluded areas are converted into the corresponding cell numbers (e.g., in a 6×10 pixel image, the pixel in the 2nd row and 3rd column corresponds to cell number 13). Finally, the location data of the occluded cells is output as a list containing all occluded cell numbers (e.g., [5,6,13,14]), and the occlusion time of each cell is recorded to provide a timestamp for dynamic occlusion processing in subsequent power generation calculations. For ambiguous areas with grayscale values ​​near the threshold, the trend of three consecutive grayscale images is used to assist in the judgment: if the grayscale values ​​of three consecutive frames are all less than the threshold, it is determined to be occluded; if only a single frame is less than the threshold, it is determined to be noise and excluded from the location results to reduce misjudgments.

[0093] S3. Based on the location data of the shaded battery cells and the multi-source data of the photovoltaic power generation system, determine the power generation of the partially shaded battery and obtain the power station-level power prediction value.

[0094] Specifically, step S3 includes the following steps:

[0095] S301. Based on the photovoltaic module-level electrical data and the location data of the shaded solar cells, determine the local shading factor and the mismatch factor of the solar cells.

[0096] Among them, the local shading factor of the solar cell can be determined by comparing the reference irradiance intensity with the actual received irradiance of the shaded solar cell, and dynamically updated in combination with the moving speed of the shading object to reflect the real-time degree of shading; the solar cell mismatch factor can be determined by the distribution characteristics of the shaded solar cells in the series branch (such as the number of shaded cells and the degree of shading), and corrected according to the type of shading object to reflect the output loss caused by the barrel effect in the series circuit.

[0097] S302. Determine the actual irradiance of each solar cell based on the local shading factor of the solar cell.

[0098] The formula for calculating the actual irradiance of each solar cell is as follows:

[0099]

[0100] in, Let n be the actual irradiance of the nth solar cell. As the reference irradiance, Let be the local shading factor of the nth solar cell. It is the surface reflection gain factor.

[0101] S303. Extract the effective area and working efficiency of the solar cells from the photovoltaic module-level electrical data.

[0102] S304. Determine the actual output of a single solar cell based on the effective area of ​​the solar cell, the working efficiency of the solar cell, and the mismatch factor of the solar cell.

[0103] The formula for calculating the actual output of a single battery cell is as follows:

[0104]

[0105] in, For the actual output of the nth solar cell, Let n be the effective area of ​​the nth solar cell. The actual irradiation intensity determined in step S302. For the working efficiency of the battery cells, This represents the cell mismatch factor.

[0106] S305. Sum the actual output of each individual cell to obtain the module-level output.

[0107] The formula for calculating the component-level output is as follows:

[0108]

[0109] in, To contribute to the component level, The total number of solar cells in a single component. This represents the actual output of the nth battery cell.

[0110] S306. Sum the output of the component level to obtain the predicted power generation value of the power station level.

[0111] The power station-level power prediction value is obtained by summing the output of all components, realizing the aggregation of power generation from the solar cells to the entire power station. First, the total number of photovoltaic modules in the power station is counted, and the output of each module calculated in step S305 is summed to obtain the total power of the power station. Then, the power prediction value can be obtained by multiplying it by the prediction duration. The calculation formula for the power station-level power prediction value is as follows:

[0112]

[0113] in, This is the predicted power generation value at the power plant level. This represents the total number of components within the power station. For the output of the m-th component, For the predicted duration.

[0114] In one possible implementation, the local shading factor of the solar cell is determined based on the photovoltaic module-level electrical data and the location data of the shaded solar cell, including:

[0115] S3011. Extract the reference irradiance and surface albedo data from the photovoltaic module-level electrical data.

[0116] S3012. Determine the initial local shading factor of the solar cells based on the reference irradiance, surface albedo data and the location data of the shaded solar cells.

[0117] The initial local shading factor of the solar cell is used to initially quantify the degree of shading of a single solar cell. The calculation can be combined with cross-validation of multi-source data. The calculation formula for the initial local shading factor of the solar cell is as follows:

[0118]

[0119] in, Let be the initial local shading factor for the nth solar cell. The calculated irradiance of the nth solar cell is... As the reference irradiance, Let be the surface reflection gain factor of the nth solar cell.

[0120] S3013. Based on the moving speed of the obstruction in the photovoltaic panel obstruction data, determine the update range of the initial cell local shading factor to obtain the cell local shading factor.

[0121] Among these, the update magnitude of the initial factor can be determined. The update magnitude is related to the moving speed of the obstruction in the photovoltaic panel obstruction data obtained in step S1. A speed classification rule can be adopted: when the moving speed V > 10 km / h, the update magnitude is 0.3; when 5 km / h < V ≤ 10 km / h, the update magnitude is 0.15; when 0 km / h < V ≤ 5 km / h, the update magnitude is 0.05; and when V = 0, the update magnitude is 0.

[0122] Furthermore, the final local occlusion factor is calculated by combining the update amplitude. The formula for calculating the local occlusion factor of the solar cell is as follows:

[0123]

[0124] in, Let be the final local shading factor of the nth solar cell. The initial local occlusion factor, For the update range, The velocity direction coefficient can be set to 1 when the obstruction moves toward the battery cell and -1 when it moves away from it, and is determined by the direction of the obstruction's movement monitored by millimeter-wave radar.

[0125] In one possible implementation, the cell mismatch factor is determined based on the photovoltaic module-level electrical data and the location data of the shaded cells, including:

[0126] S3014. Extract the series connection data of the cells within the photovoltaic module from the photovoltaic module-level electrical data.

[0127] S3015. Based on the series connection data, determine the series branch number k where the nth battery cell is located.

[0128] S3016. Based on the location data of the obscured battery cells, determine the total number of battery cells, the number of obscured battery cells, and the degree of obstruction of each obscured battery cell in the series branch number k.

[0129] S3017. Determine the branch-level shading impact coefficient based on the total number of battery cells, the number of shaded battery cells, and the degree of shading of battery cells.

[0130] The formula for calculating the branch-level shading impact coefficient is as follows:

[0131]

[0132] in, Let be the shading influence coefficient of the k-th series branch. This is the sum of the degree of shading of all the blocked solar cells within that branch. This represents the total number of battery cells in this branch.

[0133] S3018. Determine the initial cell mismatch factor based on the branch-level shading influence coefficient.

[0134] The formula for calculating the initial cell mismatch factor is as follows:

[0135]

[0136] in, Let be the initial mismatch factor of the nth solar cell. The mismatch amplification factor is set at 2.5 for polycrystalline silicon solar cells and 2.0 for monocrystalline silicon solar cells. This is because the current uniformity of polycrystalline silicon solar cells is more sensitive to shading, resulting in a larger amplification factor. Let be the shading influence coefficient of the branch k where the nth battery cell is located.

[0137] S3019. Correct the cell mismatch factor according to the type of shading object in the photovoltaic panel shading object data to obtain the cell mismatch factor.

[0138] The shading type data from the photovoltaic panel shading data obtained in step S1 can be incorporated to enable the factor to match the mismatch characteristics of different shading scenarios. The formula for calculating the cell mismatch factor is as follows:

[0139]

[0140] in, Let be the final mismatch factor of the nth solar cell. The initial mismatch factor, The shading object type coefficient is set according to the type of shading object: 0.8 for fixed shading objects such as buildings and long-term trees, because the shadow position of fixed shading objects is stable, the current fluctuation of the battery cells in the branch is small, and the mismatch effect is relatively weak; 1.2 for dynamic shading objects such as cumulus clouds and birds, because dynamic shading objects cause frequent changes in the current of the battery cells in the branch, which can easily cause transient mismatch and requires stronger correction.

[0141] S4. Evaluate the prediction results of the power plant-level power generation prediction value. If the predicted value of the power plant-level power generation prediction value meets the error threshold, then output the power plant-level power generation prediction value as the final power generation prediction result.

[0142] Specifically, the root mean square error and mean absolute error can be used to evaluate the predicted power generation value of the power station. The predicted power generation value of the power station can be obtained simultaneously and then stored according to the "prediction time - prediction duration". The actual power generation data of the power station during the same period can be obtained directly from the grid-connected inverter or meter of the photovoltaic power station. When extracting, it is necessary to ensure that the "time dimension" of the predicted value is completely matched. If the predicted value is the power generation of a 1-hour period (such as 10:00-11:00), the actual power generation should also be the cumulative metered value of the same period. At the same time, outliers caused by metering equipment failures should be removed to ensure the accuracy of the evaluation data.

[0143] The root mean square error (RMSE) is calculated using the following formula:

[0144]

[0145] in, The root mean square error, Let J be the actual power generation of the power station during the j-th prediction period. Let j be the predicted power generation value for the power plant level during the j-th prediction period. The number of prediction samples to be evaluated (set according to actual needs, such as 24 predictions per day at 1-hour intervals, then N=24, or 168 predictions per week, then N=168).

[0146] The formula for calculating the mean absolute error (MAE) is as follows:

[0147]

[0148] in, The mean absolute error, Let be the absolute difference between the actual value and the predicted value for the j-th prediction period. Let J be the actual power generation of the power station during the j-th prediction period. Let j be the predicted power generation value for the power plant level during the j-th prediction period. The number of predicted samples to be evaluated.

[0149] Furthermore, a dual error threshold is set. If the root mean square error is <50W and the mean absolute error is <8%, the power station-level power prediction value is output as the final power generation prediction result. If either threshold is not met (e.g., RMSE=60W or MAE=9%), the error feedback optimization mechanism is triggered, returning to step S3 to readjust the update magnitude of the local shading factor of the battery cell or the correction coefficient of the mismatch factor of the battery cell. For example, the mismatch factor correction coefficient of the dynamic shading object can be increased to 1.3, or the attention weight of the variational autoencoder can be optimized by returning to step S2. For example, the feature extraction of low grayscale areas can be strengthened. After re-executing the calculation of steps S2-S3, the system re-enters step S4 for evaluation until the error meets the threshold requirement. If the error still does not meet the threshold after three consecutive optimizations, an early warning signal is triggered, prompting maintenance personnel to check whether the shading recognition sensor (e.g., camera, millimeter-wave radar) or metering equipment is abnormal to ensure stable system operation.

[0150] This example provides a method for predicting power generation based on a photovoltaic (PV) power generation system. First, it acquires multi-source data of the PV power generation system, including electrical data at the PV module level, PV panel topographic data, and PV panel shading data. Then, based on this multi-source data, it identifies local shading of the PV modules, determines the corrected grayscale image, and obtains the location data of the shaded cells. Subsequently, it combines the location data of the shaded cells with the multi-source data of the PV power generation system to calculate the power generation of the locally shaded cells and obtain a power plant-level power prediction value. Finally, it evaluates the error of this power plant-level power prediction value, and outputs the final power generation prediction result when the error threshold is met. By introducing PV panel topographic data and PV panel shading data, and combining them with local shading identification, it obtains accurate location information of the shaded cells, avoiding the overestimation or underestimation of output caused by simplification or neglect of local shading in existing methods. Simultaneously, it captures the local irradiance differences of PV panels under different terrain conditions, reducing prediction bias caused by the inability of global terrain data to reflect microscopic differences, thus meeting the grid dispatching requirements for high-precision prediction, thereby improving the renewable energy absorption rate and ensuring the stability of grid operation.

[0151] For those consistent with the above, please refer to Figure 2 , Figure 2 This application provides a schematic diagram of the structure of a photovoltaic power generation prediction device for an embodiment of the present application. For example... Figure 2 As shown, the device includes:

[0152] The first acquisition unit 1 is used to acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data.

[0153] The first processing unit 2 is used to identify partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system, and to determine the corrected grayscale image and the positioning data of the shaded battery cells.

[0154] The second processing unit 3 is used to determine the power generation of the partially shaded battery based on the location data of the shaded battery cell and the multi-source data of the photovoltaic power generation system, and to obtain the power station-level power prediction value.

[0155] The third processing unit 4 is used to evaluate the prediction result of the power plant-level power prediction value. When the predicted value of the power plant-level power prediction value meets the error threshold, the power plant-level power prediction value is output as the final power generation prediction result.

[0156] In one possible implementation, in the aspect of identifying partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system and determining the corrected grayscale image, the first processing unit 2 is configured to:

[0157] Extract photovoltaic module cell curve data and cell irradiance ratio data from the photovoltaic module-level electrical data;

[0158] Based on the photovoltaic panel topography data and photovoltaic module cell curve data, determine the comprehensive value of the micro-topography features for each cell;

[0159] The micro-topography correction coefficient is determined based on the type of shading object in the photovoltaic panel shading object data;

[0160] The grayscale value of the micro-topography correction cell is determined based on the battery irradiance ratio data, the comprehensive value of micro-topography features, and the micro-topography correction coefficient.

[0161] The grayscale image update cycle is determined based on the moving speed of the obstruction in the photovoltaic panel obstruction data;

[0162] The grayscale value of the battery cell is corrected based on the micro-topography and the update cycle, and the corrected grayscale image is determined in combination with the location of the battery cell.

[0163] In one possible implementation, in the aspect of identifying partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system and determining the location data of the shaded cells, the first processing unit 2 is configured to:

[0164] The photovoltaic module cell curve data is used as training samples, and the corrected grayscale image is used as training labels to train a variational autoencoder neural network with an attention mechanism.

[0165] Based on the type of shading object in the photovoltaic panel shading object data, optimize the attention weights of the variational autoencoder neural network;

[0166] The real-time collected photovoltaic module cell curve data is input into the variational autoencoder neural network, and the output is a real-time corrected grayscale image.

[0167] Based on the real-time corrected grayscale image, the location data of the obscured battery cell is determined.

[0168] In one possible implementation, in determining the power generation of partially shaded cells based on the location data of the shaded cells and multi-source data of the photovoltaic power generation system to obtain a power plant-level power prediction value, the second processing unit 3 is configured to:

[0169] Based on the photovoltaic module-level electrical data and the location data of the shaded solar cells, the local shading factor and the mismatch factor of the solar cells are determined.

[0170] The actual irradiance of each solar cell is determined based on the local shading factor of the solar cell.

[0171] The effective area and working efficiency of the solar cells are extracted from the photovoltaic module-level electrical data.

[0172] The actual output of a single solar cell is determined based on the effective area of ​​the solar cell, the working efficiency of the solar cell, and the mismatch factor of the solar cell.

[0173] The module-level output is obtained by summing the actual output of each individual battery cell.

[0174] The summation of the component-level outputs yields the predicted power generation value at the power station level.

[0175] In one possible implementation, in determining the local shading factor of a solar cell based on the photovoltaic module-level electrical data and the location data of the shaded solar cell, the second processing unit 3 is configured to:

[0176] Reference irradiance and surface albedo data are extracted from the photovoltaic module-level electrical data;

[0177] Based on the reference irradiance, surface albedo data, and the location data of the obscured solar cells, the initial local shading factor of the solar cells is determined.

[0178] Based on the moving speed of the obstruction in the photovoltaic panel obstruction data, the update range of the initial cell local shading factor is determined, and the cell local shading factor is obtained.

[0179] In one possible implementation, in determining the cell mismatch factor based on the photovoltaic module-level electrical data and the location data of the shaded cells, the second processing unit 3 is configured to:

[0180] Extract the series connection data of the solar cells within the photovoltaic module from the photovoltaic module-level electrical data;

[0181] Based on the series connection data, determine the series branch number k where the nth battery cell is located;

[0182] Based on the location data of the obscured battery cells, determine the total number of battery cells, the number of obscured battery cells, and the degree of obstruction of each obscured battery cell in the series branch number k.

[0183] Based on the total number of solar cells, the number of solar cells that are blocked, and the degree of solar cell blocking, determine the branch-level blocking impact coefficient;

[0184] The initial cell mismatch factor is determined based on the branch-level shading influence coefficient.

[0185] The cell mismatch factor is obtained by correcting the cell mismatch factor based on the type of shading object in the photovoltaic panel shading object data.

[0186] In one possible implementation, in the aspect of evaluating the prediction result of the power plant-level power generation prediction value, and outputting the power plant-level power generation prediction value as the final power generation prediction result when the predicted value meets an error threshold, the third processing unit 4 is used to:

[0187] The root mean square error and mean absolute error are used to evaluate the prediction results of the power station-level power prediction values.

[0188] If the root mean square error is less than 50W and the mean absolute error is less than 8%, then the power plant-level power prediction value is output as the final power generation prediction result.

[0189] For examples consistent with the above embodiments, please refer to... Figure 3 , Figure 3 A schematic diagram of a terminal structure provided in an embodiment of this application is shown in the figure. It includes a processor, an input device, an output device, and a memory. The processor, input device, output device, and memory are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to call the program instructions. The program includes instructions for performing the following steps.

[0190] Acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data;

[0191] Based on the multi-source data of the photovoltaic power generation system, the photovoltaic modules are partially shaded to identify the corrected grayscale image and the location data of the shaded cells.

[0192] Based on the location data of the shaded solar cells and the multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is determined, and the power station-level power prediction value is obtained.

[0193] The predicted power generation value at the power plant level is evaluated. If the predicted power generation value at the power plant level meets the error threshold, the predicted power generation value at the power plant level is output as the final power generation prediction result.

[0194] In this example, multi-source data of the photovoltaic power generation system, including photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data, is first acquired. Then, based on this multi-source data, partial shading of the photovoltaic modules is identified, and the corrected grayscale image and the location data of the shaded cells are determined. Subsequently, by combining the location data of the shaded cells with the multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is calculated, and a power plant-level power prediction value is obtained. Finally, the error of the power plant-level power prediction value is evaluated, and the final power generation prediction result is output when the error threshold is met. By introducing photovoltaic panel topographic data and photovoltaic panel shading data, and combining partial shading identification, the accurate location information of the shaded cells can be obtained, avoiding the overestimation or underestimation of output caused by simplification or neglect of partial shading in existing methods. At the same time, it can capture the local irradiance differences of photovoltaic panels under different terrain conditions, reduce the prediction deviation caused by the inability of global terrain data to reflect micro-differences, and thus meet the grid dispatching requirements for high-precision prediction, thereby improving the renewable energy absorption rate and ensuring the stability of grid operation.

[0195] The above mainly describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the terminal includes the corresponding hardware structure and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0196] This application embodiment can divide the terminal into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0197] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the power generation prediction methods based on a photovoltaic power generation system as described in the above method embodiments.

[0198] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps of any of the power generation prediction methods based on a photovoltaic power generation system as described in the above method embodiments.

[0199] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0200] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0201] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0202] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0203] Furthermore, the functional units in the various embodiments of the application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software program module.

[0204] If the integrated unit is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0205] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk, etc.

[0206] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting power generation based on a photovoltaic power generation system, characterized in that, include: Acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data; Based on the multi-source data of the photovoltaic power generation system, the photovoltaic modules are partially shaded to identify the corrected grayscale image and the location data of the shaded cells. Based on the location data of the shaded battery cells and the multi-source data of the photovoltaic power generation system, the power generation of the partially shaded battery cells is determined, and the power station-level power prediction value is obtained. The prediction results of the power plant-level power generation prediction value are evaluated. If the predicted value of the power plant-level power generation prediction value meets the error threshold, the predicted value of the power plant-level power generation prediction value is output as the final power generation prediction result. The step of identifying partial shading of photovoltaic modules based on multi-source data from the photovoltaic power generation system and determining the corrected grayscale image includes: Extract photovoltaic module cell curve data and cell irradiance ratio data from the photovoltaic module-level electrical data; Based on the photovoltaic panel topography data and photovoltaic module cell curve data, determine the comprehensive value of the micro-topography features for each cell; The micro-topography correction coefficient is determined based on the type of shading object in the photovoltaic panel shading object data; The grayscale value of the micro-topography correction cell is determined based on the battery irradiance ratio data, the comprehensive value of micro-topography features, and the micro-topography correction coefficient. The grayscale image update cycle is determined based on the moving speed of the obstruction in the photovoltaic panel obstruction data; Based on the micro-topography, the grayscale value and update cycle of the battery cell are corrected, and the corrected grayscale image is determined in combination with the location of the battery cell. The step of identifying partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system and determining the location data of the shaded cells includes: The photovoltaic module cell curve data is used as training samples, and the corrected grayscale image is used as training labels to train a variational autoencoder neural network with an attention mechanism. Based on the type of shading object in the photovoltaic panel shading object data, optimize the attention weights of the variational autoencoder neural network; The real-time collected photovoltaic module cell curve data is input into the variational autoencoder neural network, and the output is a real-time corrected grayscale image. Based on the real-time corrected grayscale image, determine the location data of the obscured battery cell; Based on the location data of the shaded solar cells and multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is determined, and a power station-level power prediction value is obtained, including: Based on the photovoltaic module-level electrical data and the location data of the shaded solar cells, the local shading factor and the mismatch factor of the solar cells are determined. The actual irradiance of each solar cell is determined based on the local shading factor of the solar cell. The effective area and working efficiency of the solar cells are extracted from the photovoltaic module-level electrical data. The actual output of a single solar cell is determined based on the effective area of ​​the solar cell, the working efficiency of the solar cell, and the mismatch factor of the solar cell. The module-level output is obtained by summing the actual output of each individual battery cell. The summation of the component-level outputs yields the predicted power generation value at the power station level.

2. The power generation prediction method based on a photovoltaic power generation system according to claim 1, characterized in that, The step of determining the local shading factor of the solar cell based on the photovoltaic module-level electrical data and the location data of the shaded solar cell includes: Reference irradiance and surface albedo data are extracted from the photovoltaic module-level electrical data; Based on the reference irradiance, surface albedo data, and the location data of the obscured solar cells, the initial local shading factor of the solar cells is determined. Based on the moving speed of the obstruction in the photovoltaic panel obstruction data, the update range of the initial cell local shading factor is determined, and the cell local shading factor is obtained.

3. The power generation prediction method based on a photovoltaic power generation system according to claim 1, characterized in that, The step of determining the cell mismatch factor based on the photovoltaic module-level electrical data and the location data of the shaded cells includes: Extract the series connection data of the solar cells within the photovoltaic module from the photovoltaic module-level electrical data; Based on the series connection data, determine the series branch number k where the nth battery cell is located; Based on the location data of the obscured battery cells, determine the total number of battery cells, the number of obscured battery cells, and the degree of obstruction of each obscured battery cell in the series branch number k. Based on the total number of solar cells, the number of solar cells that are blocked, and the degree of solar cell blocking, determine the branch-level blocking impact coefficient; The initial cell mismatch factor is determined based on the branch-level shading influence coefficient. The cell mismatch factor is obtained by correcting the cell mismatch factor based on the type of shading object in the photovoltaic panel shading object data.

4. The power generation prediction method based on a photovoltaic power generation system according to claim 1, characterized in that, The evaluation of the predicted power generation value at the power plant level, wherein if the predicted power generation value at the power plant level meets the error threshold, the predicted power generation value at the power plant level is output as the final power generation prediction result, includes: The root mean square error and mean absolute error are used to evaluate the prediction results of the power station-level power prediction values. If the root mean square error is less than 50W and the mean absolute error is less than 8%, then the power plant-level power prediction value is output as the final power generation prediction result.

5. A power generation prediction device for a photovoltaic power generation system, characterized in that, include: The first acquisition unit is used to acquire multi-source data of the photovoltaic power generation system, wherein the multi-source data of the photovoltaic power generation system includes photovoltaic module-level electrical data, photovoltaic panel topographic data, and photovoltaic panel shading data; The first processing unit is used to identify partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system, and to determine the corrected grayscale image and the location data of the shaded battery cells. The second processing unit is used to determine the power generation of the partially shaded battery based on the location data of the shaded battery cell and the multi-source data of the photovoltaic power generation system, and to obtain the power station-level power prediction value. The third processing unit is used to evaluate the prediction result of the power plant-level power prediction value. When the predicted value of the power plant-level power prediction value meets the error threshold, the power plant-level power prediction value is output as the final power generation prediction result. The step of identifying partial shading of photovoltaic modules based on multi-source data from the photovoltaic power generation system and determining the corrected grayscale image includes: Extract photovoltaic module cell curve data and cell irradiance ratio data from the photovoltaic module-level electrical data; Based on the photovoltaic panel topography data and photovoltaic module cell curve data, determine the comprehensive value of the micro-topography features for each cell; The micro-topography correction coefficient is determined based on the type of shading object in the photovoltaic panel shading object data; The grayscale value of the micro-topography correction cell is determined based on the battery irradiance ratio data, the comprehensive value of micro-topography features, and the micro-topography correction coefficient. The grayscale image update cycle is determined based on the moving speed of the obstruction in the photovoltaic panel obstruction data; Based on the micro-topography, the grayscale value and update cycle of the battery cell are corrected, and the corrected grayscale image is determined in combination with the location of the battery cell. The step of identifying partial shading of photovoltaic modules based on multi-source data of the photovoltaic power generation system and determining the location data of the shaded cells includes: The photovoltaic module cell curve data is used as training samples, and the corrected grayscale image is used as training labels to train a variational autoencoder neural network with an attention mechanism. Based on the type of shading object in the photovoltaic panel shading object data, optimize the attention weights of the variational autoencoder neural network; The real-time collected photovoltaic module cell curve data is input into the variational autoencoder neural network, and the output is a real-time corrected grayscale image. Based on the real-time corrected grayscale image, determine the location data of the obscured battery cell; Based on the location data of the shaded solar cells and multi-source data of the photovoltaic power generation system, the power generation of the partially shaded cells is determined, and a power station-level power prediction value is obtained, including: Based on the photovoltaic module-level electrical data and the location data of the shaded solar cells, the local shading factor and the mismatch factor of the solar cells are determined. The actual irradiance of each solar cell is determined based on the local shading factor of the solar cell. The effective area and working efficiency of the solar cells are extracted from the photovoltaic module-level electrical data. The actual output of a single solar cell is determined based on the effective area of ​​the solar cell, the working efficiency of the solar cell, and the mismatch factor of the solar cell. The module-level output is obtained by summing the actual output of each individual battery cell. The summation of the component-level outputs yields the predicted power generation value at the power station level.

6. A terminal, characterized in that, The system includes a processor, an input device, an output device, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute the power generation prediction method based on a photovoltaic power generation system as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the power generation prediction method based on a photovoltaic power generation system as described in any one of claims 1-4.

Images