Distributed photovoltaic geostationary satellite irradiance correction method, device, equipment and medium

Abstract

The invention discloses a distributed photovoltaic geostationary satellite irradiance correction method, device and equipment and a medium. The method comprises the following steps: acquiring irradiance prediction data of a stationary meteorological satellite in a target area and space-time auxiliary features corresponding to the irradiance prediction data; determining a clear sky irradiance reference corresponding to the target area at the prediction moment based on radiation transfer simulation; the method comprises the following steps: calculating sun geometry, stationary satellite observation geometry and corresponding grouping indexes based on geographic positions, prediction moments and stationary satellite sub-star point parameters of distributed photovoltaic sites, and constructing a geometric consistent absolute position deviation item based on the grouping indexes; inputting the irradiance prediction data, the space-time auxiliary features, the clear sky irradiance reference and the geometric consistent absolute position deviation item into a correction model, and outputting correction irradiance based on the model; and performing physical consistency reprojection constraint on the corrected irradiance to determine corrected irradiance prediction. And the effect of improving the stability and accuracy of irradiance prediction is achieved.

Description

Technical Field

[0001] This invention relates to the fields of photovoltaic power prediction and satellite remote sensing radiation processing technology, and in particular to a geostationary satellite irradiance correction method, apparatus, equipment and medium for distributed photovoltaics. Background Technology

[0002] Distributed photovoltaic (PV) sites are numerous, widely distributed, and have significant variations in underlying surface conditions. They are also significantly affected by rapid local cloud evolution, aerosols, and topographic shading, resulting in stronger spatial heterogeneity and time-varying characteristics in distributed PV power output. Irradiance is a key driver affecting PV power output, and the quality of irradiance prediction directly determines the upper limit of distributed PV power output prediction.

[0003] Geostationary meteorological satellites offer advantages such as high temporal resolution and continuous coverage, and are often used for short-term irradiance estimation and extrapolation. However, geostationary satellite irradiance predictions are prone to systematic biases at different latitudes and from different observation perspectives. Furthermore, factors such as rapid cloud formation and dissipation, parallax caused by cloud top height, and shadow propagation introduce non-stationary errors. Existing correction methods often employ moving averages, empirical piecewise regression, or simple bias corrections, which struggle to explicitly express the coupling relationship between "observation geometry and error morphology" and to constrain the correction results to meet physical consistency, resulting in insufficient generalization and stability in distributed scenarios. Summary of the Invention

[0004] This invention provides a geostationary satellite irradiance correction method, apparatus, equipment, and medium for distributed photovoltaic systems, to address the problems of existing correction methods being sensitive to observation geometry, lacking physical consistency constraints, and having insufficient generalization in distributed scenarios.

[0005] According to one aspect of the present invention, a geostationary satellite irradiance correction method for distributed photovoltaic systems is provided, the method comprising: Acquire irradiance prediction data from geostationary meteorological satellites in the target area and the corresponding spatiotemporal auxiliary features; The clear-sky irradiance benchmark for the target area at the predicted time is determined based on radiative transfer simulation. Based on the geographical location of the distributed photovoltaic site, the predicted time, and the parameters of the geostationary satellite sub-stars, the solar geometry and the geostationary satellite observation geometry are calculated, the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry is determined, and the geometrically consistent absolute position deviation term is constructed based on the grouping index. The irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term are input into the correction model, and the corrected irradiance is determined based on the model output; wherein, the correction model is a pre-trained hierarchical Transformer model based on window attention. Physical consistency back-feed constraints are applied to the corrected irradiance to determine the corrected irradiance prediction.

[0006] According to another aspect of the present invention, a geostationary satellite irradiance correction device for distributed photovoltaic systems is provided, the device comprising: The data acquisition module is used to acquire irradiance prediction data from geostationary meteorological satellites in the target area and spatiotemporal auxiliary features corresponding to the irradiance prediction data; The benchmark determination module is used to determine the clear-sky irradiance benchmark of the target area at the predicted time based on radiative transfer simulation. The deviation term construction module is used to calculate the solar geometry and geostationary satellite observation geometry based on the geographical location of the distributed photovoltaic site, the prediction time, and the geostationary satellite sub-star point parameters, determine the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry, and construct the geometrically consistent absolute position deviation term based on the grouping index; The correction module is used to input the irradiance prediction data, the spatiotemporal auxiliary features, the clear sky irradiance benchmark, and the geometrically consistent absolute position deviation term into the correction model, and determine the corrected irradiance based on the model output; wherein, the correction model is a pre-trained hierarchical Transformer model based on window attention. The constraint module is used to apply physical consistency back-feed constraints to the corrected irradiance in order to determine the corrected irradiance prediction.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores a computer program that can be executed by at least one processor, which enables the at least one processor to execute the geostationary satellite irradiance correction method for distributed photovoltaics according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions for causing a processor to execute and implement a geostationary satellite irradiance correction method for distributed photovoltaic systems according to any embodiment of the present invention.

[0009] The technical solution of this invention achieves accurate correction of geostationary satellite irradiance prediction by acquiring geostationary satellite irradiance prediction data and spatiotemporal auxiliary features, determining clear-sky irradiance benchmarks based on radiative transfer simulation, constructing a geometrically consistent absolute position deviation term and incorporating it into a window attention layer Transformer correction model, and applying physical consistency back-projection constraints. This solves the technical problems of large systematic errors and insufficient physical consistency in existing methods under different latitudes, observation perspectives, and scenarios of rapid cloud evolution. It achieves the technical effect of improving the stability and accuracy of irradiance prediction, thereby supporting accurate prediction of distributed photovoltaic power output.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart of a geostationary satellite irradiance correction method for distributed photovoltaic systems provided in this embodiment of the invention; Figure 2a A flowchart illustrating another geostationary satellite irradiance correction method for distributed photovoltaic systems provided in this embodiment of the invention; Figure 2b This is a schematic diagram illustrating the construction and injection of a geometrically consistent absolute position deviation term in another geostationary satellite irradiance correction method for distributed photovoltaics provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of a geostationary satellite irradiance correction device for distributed photovoltaic systems, provided in an embodiment of the present invention. Figure 4 A schematic diagram of the structure of an electronic device for implementing a geostationary satellite irradiance correction method for distributed photovoltaics according to an embodiment of the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] Figure 1 This is a flowchart illustrating a geostationary satellite irradiance correction method for distributed photovoltaic (PV) systems, provided as an embodiment of the present invention. This embodiment is applicable to geostationary satellite irradiance correction for distributed PV systems. The method can be executed by a geostationary satellite irradiance correction device for distributed PV systems, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method specifically includes the following steps: S110. Obtain the irradiance prediction data of geostationary meteorological satellites in the target area and the spatiotemporal auxiliary features corresponding to the irradiance prediction data.

[0016] The target area can be understood as the specific geographical region requiring irradiance correction and subsequent distributed photovoltaic power output prediction. A geostationary satellite can be understood as a satellite with a fixed position relative to the Earth's surface that continuously observes the target area. Irradiance prediction data can be understood as the irradiance energy data of the target area estimated using geostationary satellites. Spatiotemporal auxiliary features can be understood as the temporal and spatial dimensional auxiliary information related to the irradiance prediction data.

[0017] Specifically, the system collects irradiance prediction data for the target area from geostationary satellites, along with corresponding temporal and spatial auxiliary information. This collected data serves as the foundation for subsequent corrections. Preferably, the collected data can be preliminarily screened to remove obviously abnormal data, thereby improving the efficiency of subsequent processing.

[0018] Optionally, the irradiance prediction data includes at least one of total irradiance, direct irradiance, and module planar irradiance; the spatiotemporal auxiliary features include at least one of timestamp, location information of distributed photovoltaic sites, cloud-related products, quality identifiers, and historical error statistics, wherein the cloud-related products include at least one of cloud top height, cloud phase, and cloud optical thickness.

[0019] Total irradiance can be understood as all solar radiation energy received on a horizontal plane, including both direct and diffuse radiation. Direct irradiance can be understood as the radiation energy reaching the target surface directly from the direction of the sun. Module planar irradiance can be understood as the radiation energy received by the surface of a photovoltaic module. Timestamp can be understood as information marking the time when the data was generated. Location information can be understood as information identifying the specific location of a distributed photovoltaic site, such as latitude and longitude. Cloud-related products can be understood as various types of data related to clouds. Quality identifiers can be understood as information marking the reliability of the data. Historical error statistics can be understood as the error statistics of past irradiance prediction data. Cloud top height can be understood as the height of the top of the cloud layer above the ground. Cloud phase can be understood as the physical state of the cloud layer, such as liquid or solid. Cloud optical thickness can be understood as a parameter describing the degree to which the cloud layer blocks solar radiation.

[0020] S120. Determine the clear-sky irradiance benchmark for the target area at the predicted time based on radiative transfer simulation.

[0021] Among them, radiative transfer simulation can be understood as a technical method for simulating the propagation process of sunlight in the atmosphere. Clear-sky irradiance benchmark can be understood as the standard value of irradiance that a target area should have under cloudless conditions.

[0022] Specifically, using radiative transfer simulation technology or a pre-calculated lookup table, combined with information such as the absolute geographical location of the target area and the predicted time, the baseline irradiance value for the cloudless area is calculated, providing a reference standard for subsequent irradiance correction. Preferably, a radiative transfer simulation tool can be used (any existing simulation tool in the industry can be used; this embodiment is not limited to it), and typical atmospheric parameters are input to improve the accuracy of the baseline value. The pre-calculated lookup table is a structured data table that is pre-calculated using a radiative transfer simulation tool for combinations of different geographical locations, times, and typical atmospheric conditions, calculating the corresponding cloudless irradiance intensity in batches, and then organizing it according to the rule of geographical location + time + atmospheric condition. During use, real-time simulation is not required; simply look up matching items from the table based on the geographical location of the target area, the predicted time, and the current atmospheric condition to directly obtain the corresponding clear-sky irradiance baseline, adapting to the needs of fast-response business scenarios.

[0023] Optionally, determining the clear-sky irradiance benchmark for the target area at the predicted time based on radiative transfer simulation includes: selecting a radiative transfer simulation tool, inputting the geographical location of the target area, the predicted time, and typical atmospheric state parameters; wherein, the typical atmospheric state parameters include at least one of air pressure, aerosol content, water vapor content, and ozone concentration; by simulating the propagation path of sunlight after entering the Earth's atmosphere from outer space, the irradiance intensity reaching the ground of the target area under cloudless conditions is obtained, and the irradiance intensity is used as the clear-sky irradiance benchmark.

[0024] Among these, radiative transfer simulation tools can be understood as software or systems that simulate radiative transfer. Typical atmospheric state parameters can be understood as key parameters describing the basic state of the atmosphere. Atmospheric pressure can be understood as the pressure exerted by the atmosphere on the ground. Aerosol content can be understood as the amount of suspended particulate matter in the atmosphere. Water vapor content can be understood as the amount of water vapor in the atmosphere. Ozone concentration can be understood as the amount of ozone in the atmosphere.

[0025] Specifically, select a suitable radiative transfer simulation tool (such as the industry-common libradtran), input the specific geographical location of the target area, such as latitude and longitude, the time to be predicted, and typical parameters that reflect the basic state of the atmosphere, such as air pressure and aerosol content; the tool will simulate the propagation path of sunlight after entering the Earth's atmosphere from space, and after physical processes such as scattering and absorption, and specifically calculate the irradiance intensity that finally reaches the ground of the target area under cloudless weather, and use this intensity as the clear sky irradiance benchmark for subsequent irradiance correction.

[0026] S130. Based on the geographical location of the distributed photovoltaic site, the predicted time, and the parameters of the geostationary satellite sub-stars, calculate the solar geometry and the geostationary satellite observation geometry, determine the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry, and construct a geometrically consistent absolute position deviation term based on the grouping index.

[0027] Among these, solar geometry can be understood as a set of parameters describing the positional relationship of the sun relative to the target region. Geostationary satellite observation geometry can be understood as a set of parameters describing the spectral relationship of the target region observed by a geostationary satellite. The grouping index can be understood as an identifier after classifying the solar geometry and satellite observation geometry parameters. The geometrically consistent absolute position deviation term can be understood as a parameter term adapted to the observation geometry and used to correct prediction bias.

[0028] Specifically, based on the specific location of the distributed photovoltaic site, the predicted time, and relevant parameters of the geostationary satellite, solar geometric and satellite observation geometric parameters are calculated. These parameters are then categorized into grouped indexes, and bias correction terms adapted to the current observation scenario are constructed based on these indexes. Preferably, the division of the grouped indexes can be adjusted according to the geographical characteristics of different regions to improve the adaptability of the bias terms.

[0029] Optionally, the solar geometry includes the solar zenith angle and the solar azimuth angle; the geostationary satellite observation geometry includes the observation zenith angle and the observation azimuth angle; the grouping index includes a geometric grouping index and a latitude window index; determining the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry includes: obtaining the geometric grouping index by performing preset interval binning quantization on the solar zenith angle, the observation zenith angle, the difference between the solar azimuth angle and the observation azimuth angle respectively; and obtaining the latitude window index by dividing the target area into windows based on a preset latitude step size.

[0030] Among these, the solar zenith angle can be understood as the angle between the sun's rays and the perpendicular line to the target area. The solar azimuth angle can be understood as the angle between the projection of the sun's rays onto the horizontal plane and true south. The observation zenith angle can be understood as the angle between the satellite's observation direction and the perpendicular line to the target area. The observation azimuth angle can be understood as the angle between the projection of the satellite's observation direction onto the horizontal plane and true south. The geometric grouping index can be understood as the identifier after quantizing and combining the geometric parameters of the sun and satellite. The latitude window index can be understood as the identifier after dividing the window by latitude. The preset interval binning quantization can be understood as classifying and processing parameters according to a preset interval. The preset latitude step size can be understood as the interval distance set when dividing the latitude window.

[0031] Specifically, first, fixed intervals are set for the solar zenith angle, observed zenith angle, and the difference between the solar azimuth angle and the observed azimuth angle. The actual values ​​of these three parameters are then assigned to their corresponding intervals for binning and quantization. Finally, the quantization results of these three parameters are combined to form a geometric grouping index that identifies the current geometric scene. Simultaneously, according to pre-set latitude intervals, for example, every... or For each step size, the target region is divided into multiple dimensional windows. The identifier for each window is the dimensional window index. Both provide a basis for the subsequent construction of the adaptation deviation items.

[0032] Optionally, constructing a geometrically consistent absolute positional deviation term based on the grouping index includes: Based on the geometric grouping index and the latitude window index, the corresponding deviation submatrix is ​​matched from the pre-constructed deviation table to form a deviation item that is adapted to the current observation geometry.

[0033] The deviation table can be understood as a database storing various deviation submatrices. A deviation submatrix can be understood as a portion of the parameter matrix in the deviation table corresponding to a specific grouping index.

[0034] Specifically, the pre-constructed deviation table stores deviation sub-matrices corresponding to different geometric scenes and latitude windows. Based on the obtained geometric grouping index (identifying the current observational geometric relationship between the Sun and the satellite) and latitude window index (identifying the latitude window where the target area is located), the corresponding deviation sub-matrix is ​​accurately located in the deviation table. This sub-matrix is ​​used as the core to construct deviation terms, ensuring that the deviation terms are adapted to the current observational geometric scene, providing an accurate basis for deviation correction in subsequent irradiance predictions. Preferably, the deviation table can be continuously updated based on historical observation data to ensure the adaptability of the deviation sub-matrices.

[0035] S140. Input the irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term into the correction model, and determine the corrected irradiance based on the model output; wherein, the correction model is a pre-trained hierarchical Transformer model based on window attention.

[0036] The correction model can be understood as a computational model used to correct irradiance prediction data. Window attention can be understood as a mechanism by which the model focuses on local data features for calculation. The hierarchical Transformer model can be understood as a deep learning model with a multi-level structure that can capture data correlations. Corrected irradiance can be understood as irradiance data corrected by the model.

[0037] Specifically, the collected irradiance prediction data, spatiotemporal auxiliary features, clear-sky irradiance benchmark, and constructed bias terms are input into a pre-trained hierarchical Transformer model (which uses a window attention mechanism to improve computational accuracy). The corrected irradiance data is obtained through model computation. Preferably, the data can be standardized before input to the model to make the model computation more stable.

[0038] For example, the correction model (a hierarchical Transformer model based on window attention) adopts a progressive structure of input embedding, hierarchical window attention encoding, and fully connected output. The core layers are, in order of operation, the input embedding layer, the layer normalization layer, the window attention layer, the feedforward network layer, and the output mapping layer. Each layer has a clear function and is closely connected, which is suitable for the spatiotemporal data processing requirements of irradiance correction. The input embedding layer converts four types of input data—irradiance prediction data, spatiotemporal auxiliary features, clear-sky irradiance benchmark, and geometrically consistent absolute position deviation (GC-APE)—into a fixed-dimensional fusion feature tensor and feeds it into the layer normalization layer. The layer normalization layer normalizes the embedded feature tensor to improve model training stability, and its output is directly fed into the window attention layer. The window attention layer captures spatiotemporal feature correlations by dividing local attention windows, incorporates the GC-APE deviation submatrix to correct attention weights, and its output, after normalization, is fed into the feedforward network layer, forming a local closed loop of normalization, attention calculation, and normalization. The feedforward network layer performs nonlinear transformations on the features through a structure of linear transformation, activation function, and then linear transformation. Its output is residually connected to the input features of the window attention layer and then fed into the next level module or output mapping layer. The output mapping layer maps high-order features to the corrected clear-sky index tensor through a fully connected layer, which can then be combined with the clear-sky irradiance benchmark to obtain the corrected irradiance. The overall connection relationship of the model is as follows: the input data passes through the input embedding layer, layer normalization layer, window attention layer, layer normalization layer, feedforward network layer and residual connection, and then the encoding blocks are stacked according to a preset number of times. Finally, the clear sky index is output through the output mapping layer. The encoding block is composed of window attention layer and feedforward network layer. The residual connection can alleviate the gradient vanishing problem in deep model training. The input data needs to be spatiotemporally aligned in advance to ensure that the token sequence corresponds to the spatiotemporal position.

[0039] Optionally, before inputting the irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term into the correction model, the method further includes: acquiring historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, historical clear-sky irradiance benchmark obtained through radiative transfer simulation, and measured irradiance data from distributed photovoltaic sites; constructing a training dataset based on the historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, the historical clear-sky irradiance benchmark, and the measured irradiance data; preprocessing the training dataset, and combining the historical geostationary satellite irradiance prediction data with the measured irradiance data. The clear sky irradiance benchmark is normalized to obtain the clear sky index, which is then divided into training, validation, and test sets according to a preset ratio. A pre-constructed hierarchical Transformer model is iteratively trained based on the training set. In each training round, the input features are divided into attention windows according to window size, and a geometrically consistent absolute positional deviation term is incorporated into the window self-attention calculation. Backpropagation is used to optimize the model parameters and the deviation parameters in the deviation table. The model training effect is monitored using the validation set, and training stops when a preset number of training cycles is reached. The model's generalization ability is evaluated using the test set, and the model with the smallest error on the test set is selected as the correction model.

[0040] Historical geostationary satellite irradiance prediction data can be understood as irradiance prediction data obtained from geostationary satellites in the past. Historical clear-sky irradiance baseline can be understood as the clear-sky irradiance baseline value calculated in the past. Measured irradiance data can be understood as irradiance data obtained through actual measurement. Training dataset can be understood as the dataset set used to train the model. Preprocessing can be understood as the operations performed on the data beforehand. Clear-sky index can be understood as the ratio of actual irradiance to the clear-sky irradiance baseline. Training set can be understood as the portion of data used for training model parameters. Validation set can be understood as the portion of data used to monitor the model training effect. Test set can be understood as the portion of data used to evaluate the model's generalization ability. Hyperparameters can be understood as parameters set before model training (such as the number of training epochs). Backpropagation can be understood as the calculation method for adjusting parameters based on error in the model. Model parameters can be understood as the adjustable internal parameters during model operation.

[0041] Specifically, a training dataset is first constructed by collecting historical satellite irradiance prediction data, spatiotemporal auxiliary features, historical clear-sky irradiance benchmarks, and measured irradiance data. After preprocessing and dividing the dataset into training, validation, and test sets, model training hyperparameters are set, including the number of attention heads, window size for time / longitude / latitude dimensions, learning rate, number of training epochs, and batch size. The mean squared error loss function is used to measure the deviation between the model's predicted values ​​and the measured irradiance data. A hierarchical Transformer model is iteratively trained based on the training set, incorporating a geometrically consistent absolute positional deviation term and optimizing the parameters. The effect is monitored using the validation set, and the generalization ability is evaluated using the test set. Finally, the optimal model is selected as the correction model. Preferably, an adaptive learning rate adjustment strategy can be used to accelerate the model training convergence speed.

[0042] S150. Apply physical consistency back-feed constraints to the corrected irradiance to determine the corrected irradiance prediction.

[0043] Among them, the physical consistency re-implementation constraint can be understood as the restriction condition that ensures the corrected irradiance conforms to physical laws.

[0044] Specifically, constraints that conform to physical laws are imposed on the corrected irradiance output of the model, such as ensuring that the irradiance is non-negative and does not exceed a reasonable range of clear-sky baseline, to ultimately determine the accurate corrected irradiance prediction result. Preferably, the constraint parameters can be dynamically adjusted according to the climate characteristics of the target area to adapt to different scenario requirements.

[0045] Optionally, the physical consistency feedback constraint includes at least one of a non-negativity constraint, a clear sky upper bound constraint, a nighttime constraint, and a ramp rate constraint; wherein, the non-negativity constraint ensures that the corrected irradiance prediction is not less than zero; the clear sky upper bound constraint ensures that the corrected irradiance prediction does not exceed a preset multiple of the clear sky irradiance benchmark; the nighttime constraint sets the corrected irradiance prediction to zero or near zero when the solar altitude angle is lower than a preset altitude threshold; and the ramp rate constraint restricts or smooths the rate of change of the corrected irradiance prediction at adjacent times.

[0046] Among these constraints, the non-negativity constraint can be understood as a constraint that limits the data to be no less than zero. The clear-sky upper bound constraint can be understood as a constraint that limits the irradiance to no more than a preset multiple of the clear-sky benchmark. The nighttime constraint can be understood as an irradiance constraint set for nighttime scenes. The ramp rate constraint can be understood as a constraint that limits the rate of change of irradiance. The preset multiple can be understood as a pre-set multiple threshold of the clear-sky irradiance benchmark, which can be preset based on experience; this embodiment does not impose specific restrictions on it. The preset altitude threshold can be understood as a pre-set critical value for determining whether it is the solar altitude angle at night. The near-zero range can be understood as a numerical range close to zero. The rate of change can be understood as the ratio of the difference in irradiance between adjacent moments to the time interval.

[0047] The technical solution of this invention achieves accurate correction of geostationary satellite irradiance prediction by acquiring geostationary satellite irradiance prediction data and spatiotemporal auxiliary features, determining clear-sky irradiance benchmarks based on radiative transfer simulation, constructing a geometrically consistent absolute position deviation term and incorporating it into a window attention layer Transformer correction model, and applying physical consistency back-projection constraints. This solves the technical problems of large systematic errors and insufficient physical consistency in existing methods under different latitudes, observation perspectives, and scenarios of rapid cloud evolution. It achieves the technical effect of improving the stability and accuracy of irradiance prediction, thereby supporting accurate prediction of distributed photovoltaic power output.

[0048] Figure 2a This is a flowchart of another geostationary satellite irradiance correction method for distributed photovoltaic systems provided by an embodiment of the present invention. Based on the above embodiments, this embodiment is a further refinement of the previous embodiments, and its specific implementation can be found in the technical solution of this embodiment. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here. Figure 2a As shown, the correction method provided in this embodiment of the invention includes S101-S105, and can optionally be extended to distributed photovoltaic power output prediction (S106-S107). Specifically, S102 introduces a clear-sky irradiance physical prior, and S103 constructs and injects a geometrically consistent absolute position deviation term. S105 ensures that the corrected output meets physical consistency through the back-feedback constraint.

[0049] S101. Obtain geostationary satellite irradiance prediction data and auxiliary features.

[0050] Specifically, obtain geostationary satellite irradiance prediction data for the target area during the prediction period. The irradiance prediction data can be at least one of Total Irradiance (GHI), Direct Irradiance (DNI), and Planar Radiance (POA) of the module. Simultaneously, spatiotemporal auxiliary features corresponding to the irradiance prediction data are acquired. These auxiliary features include at least one of timestamps, site / grid location information, cloud-related products (cloud top height, cloud phase, cloud optical thickness, etc.), quality identifiers, and historical error statistics. In distributed scenarios, the irradiance prediction data can be interpolated, resampled, or aggregated to the location of the distributed photovoltaic site or the operational grid.

[0051] S102. Generate clear-sky irradiance benchmark.

[0052] Specifically, clear-sky irradiance benchmarks are generated based on radiative transfer simulation tools or their pre-calculated lookup tables (LUTs). Preferably, radiative transfer simulation software (such as libradtran) is used to perform offline simulations of typical atmospheric conditions and establish a LUT (Local Underlying Test). During the online phase, the results are obtained by quickly looking up a table based on the station location and the prediction time. .pass Irradiance can be normalized to a clear sky index. This reduces the interference of strong periodic factors such as seasons and solar altitude angle on the correction learning and provides an upper bound reference for subsequent reinvestment constraints.

[0053] S103. Construct and inject the geometrically consistent absolute positional deviation term GC-APE (B_GC).

[0054] Specifically, unlike directly superimposing the position vector onto the token, this invention injects "absolute position-related information" into the window self-attention computation as an attention bias term.

[0055] Figure 2b This is a schematic diagram illustrating the construction and injection of the geometrically consistent absolute position deviation term in another geostationary satellite irradiance correction method for distributed photovoltaics provided in an embodiment of the present invention; as shown. Figure 2b As shown, the solar geometry is first calculated based on the absolute position of the station / grid and the predicted time: solar zenith angle. With solar azimuth ; Calculate satellite observation geometry based on geostationary satellite sub-satellite point parameters and absolute positions: observation zenith angle With the observed azimuth angle And calculate the relative geometric quantities. and its expanded terms (such as , The above geometric quantities are quantized into geometric grouping indexes according to preset binning rules. Simultaneously, a latitude window index is obtained by dividing the window according to the latitude direction. .

[0056] For example, the correction model input feature X is divided according to the window size. Divide into multiple attention windows. Maintain a learnable bias table. It is based on the number of attention heads H and the number of geometric groups. Number of latitude windows Organization, example shape: For any window, based on Selecting the deviation submatrix And for any token pair within the window The deviation was obtained by looking up the table. .

[0057] In the window self-attention process, the aforementioned deviations are added to the attention logits pairwise to ensure that the attention calculation satisfies the following: Where Q, K, and V are the query, key, and value vectors, respectively, and D is the feature dimension. Obtained from the above table lookup It is composed within a window. The longitude direction can employ a period-sharing strategy, ensuring that the window longitude index does not participate in submatrix selection, but is determined solely by relative displacement. Construct an index to reduce the number of parameters and improve generalization across longitudes.

[0058] S104. Correction model inference yields corrected irradiance or correction coefficients.

[0059] Specifically, geostationary satellite irradiance prediction data Spatiotemporal auxiliary features, clear-sky irradiance benchmark and Common input correction model, output corrected irradiance Or a correction factor. Preferably, the output of the correction model is the clear sky index. Correction results: First calculate Model output or Then throw it back to get The preferred correction model is a hierarchical Transformer network, which uses window attention to model the local cloud evolution and spatiotemporal correlation, and injects [something] into the attention. It learns the systematic error morphology of "observation geometric consistency". This structure does not rely on U-Net-style symmetric upsampling paths and is suitable for short-term, multi-time, multi-site / grid parallel inference.

[0060] S105, Physical Consistency Return Constraints and Output.

[0061] Specifically, a physical consistency backfeed constraint is applied to the correction output to obtain the final corrected irradiance prediction. And output. Constraints include, but are not limited to: (1) Non-negativity constraints: (2) Clear sky upper bound constraint: , where α is a preset upper bound multiple; (3) Nighttime constraint: when the solar altitude angle is lower than the threshold, Set to zero or near-zero range; (4) Climb rate constraint: for the difference between adjacent time points Set a threshold or apply smoothing to suppress anomalous jumps.

[0062] Optionally, the method also includes S106-S107 distributed photovoltaic power output prediction.

[0063] Specifically, the corrected irradiance prediction The distributed photovoltaic (PV) power output prediction model is input along with historical power data from distributed sites, module / inverter parameters, and meteorological factors such as temperature and wind speed, and outputs power prediction results. In distributed scenarios, deviation terms or absolute location features related to the absolute location of the sites can be further introduced into the power prediction model to explicitly model the correlation between geographical location and power output differences.

[0064] The technical solution of this invention introduces a geometric grouping index based on solar geometry and geostationary satellite observation geometry, enabling the model to learn systematic error patterns consistent with the observation geometry, thereby enhancing cross-latitude and cross-view generalization. Through a closed loop of "clear-sky physical prior—normalization—correction—re-implementation," the correction output has a clear physical meaning and is controllable. Unreasonable jumps are suppressed by constraints such as non-negativity, upper bound, nighttime, and ramp rate, improving physical consistency and operational availability. Corrected irradiance, as a key input for distributed photovoltaic power output prediction, can further improve power prediction accuracy.

[0065] Figure 3 This is a schematic diagram of a geostationary satellite irradiance correction device for distributed photovoltaic systems, provided as an embodiment of the present invention. Figure 3 As shown, the device includes: a data acquisition module 310, a benchmark determination module 320, a deviation term construction module 330, a correction module 340, and a constraint module 350.

[0066] The system includes the following modules: a data acquisition module 310, which acquires irradiance prediction data from geostationary meteorological satellites for the target area and corresponding spatiotemporal auxiliary features; a benchmark determination module 320, which determines the clear-sky irradiance benchmark for the target area at the prediction time based on radiative transfer simulation; a deviation term construction module 330, which calculates solar geometry and geostationary satellite observation geometry based on the geographical location of the distributed photovoltaic site, the prediction time, and geostationary satellite sub-satellite point parameters, determines the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry, and constructs a geometrically consistent absolute position deviation term based on the grouping index; a correction module 340, which inputs the irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term into a correction model, and determines the corrected irradiance based on the model output; wherein the correction model is a pre-trained hierarchical Transformer model based on window attention; and a constraint module 350, which applies physical consistency back-feedback constraints to the corrected irradiance to determine the corrected irradiance prediction.

[0067] The technical solution of this invention achieves accurate correction of geostationary satellite irradiance prediction by acquiring geostationary satellite irradiance prediction data and spatiotemporal auxiliary features, determining clear-sky irradiance benchmarks based on radiative transfer simulation, constructing a geometrically consistent absolute position deviation term and incorporating it into a window attention layer Transformer correction model, and applying physical consistency back-projection constraints. This solves the technical problems of large systematic errors and insufficient physical consistency in existing methods under different latitudes, observation perspectives, and scenarios of rapid cloud evolution. It achieves the technical effect of improving the stability and accuracy of irradiance prediction, thereby supporting accurate prediction of distributed photovoltaic power output.

[0068] Optionally, the irradiance prediction data includes at least one of total irradiance, direct irradiance, and module planar irradiance; the spatiotemporal auxiliary features include at least one of timestamp, location information of distributed photovoltaic sites, cloud-related products, quality identifiers, and historical error statistics, wherein the cloud-related products include at least one of cloud top height, cloud phase, and cloud optical thickness.

[0069] Optionally, the benchmark determination module includes: The parameter input unit is used to select a radiative transfer simulation tool and input the geographical location of the target area, the prediction time, and typical atmospheric state parameters; wherein, the typical atmospheric state parameters include at least one of air pressure, aerosol content, water vapor content, and ozone concentration. The path simulation unit is used to simulate the propagation path of sunlight after it enters the Earth's atmosphere from outer space, so as to obtain the irradiance reaching the ground of the target area under cloudless conditions, and use the irradiance as the clear sky irradiance benchmark.

[0070] Optionally, the solar geometry includes the solar zenith angle and the solar azimuth angle; the geostationary satellite observation geometry includes the observation zenith angle and the observation azimuth angle; the grouping index includes a geometric grouping index and a latitude window index; correspondingly, the deviation term construction module includes: The first index determination unit is used to obtain the geometric grouping index by performing preset interval binning quantization on the solar zenith angle, the observed zenith angle, the difference between the solar azimuth angle and the observed azimuth angle; The second index determination unit is used to divide the target area into windows based on a preset latitude step size to obtain the latitude window index.

[0071] Optionally, the deviation term construction module includes: The deviation term construction unit is used to match the corresponding deviation submatrix from the pre-constructed deviation table based on the geometric grouping index and the latitude window index to form a deviation term that is adapted to the current observation geometry.

[0072] Optionally, the device further includes: The dataset acquisition module is used to acquire historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, historical clear-sky irradiance benchmark obtained through radiative transfer simulation, and measured irradiance data of distributed photovoltaic sites before inputting the irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark and the geometrically consistent absolute position deviation term into the correction model. The dataset construction module is used to construct a training dataset based on the historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, the historical clear-sky irradiance benchmark, and the measured irradiance data. The dataset partitioning module is used to preprocess the training dataset, normalize the historical geostationary satellite irradiance prediction data with the clear sky irradiance benchmark to obtain the clear sky index, and divide it into training set, validation set and test set according to a preset ratio. The training module is used to iteratively train a pre-built hierarchical Transformer model based on the training set. In each training round, the input features are divided into attention windows according to the window size. A geometrically consistent absolute position deviation term is incorporated into the window self-attention calculation. The model parameters and the deviation parameters in the deviation table are optimized through backpropagation. The model determination module is used to monitor the model training effect through the validation set, stop training when the preset number of training cycles is reached, evaluate the model's generalization ability through the test set, and select the model with the smallest error on the test set as the correction model.

[0073] Optionally, the physical consistency feedback constraint includes at least one of a non-negativity constraint, a clear sky upper bound constraint, a nighttime constraint, and a ramp rate constraint; wherein, the non-negativity constraint ensures that the corrected irradiance prediction is not less than zero; the clear sky upper bound constraint ensures that the corrected irradiance prediction does not exceed a preset multiple of the clear sky irradiance benchmark; the nighttime constraint sets the corrected irradiance prediction to zero or near zero when the solar altitude angle is lower than a preset altitude threshold; and the ramp rate constraint restricts or smooths the rate of change of the corrected irradiance prediction at adjacent times.

[0074] The geostationary satellite irradiance correction device for distributed photovoltaic power generation provided in this embodiment of the invention can execute the geostationary satellite irradiance correction method for distributed photovoltaic power generation provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0075] Figure 4This is a schematic diagram of an electronic device for implementing the geostationary satellite irradiance correction method for distributed photovoltaic systems according to embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0076] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0077] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0078] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the method for geostationary satellite irradiance correction for distributed photovoltaic systems.

[0079] In some embodiments, the method for geostationary satellite irradiance correction for distributed photovoltaic systems can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for geostationary satellite irradiance correction for distributed photovoltaic systems described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the method for geostationary satellite irradiance correction for distributed photovoltaic systems by any other suitable means (e.g., by means of firmware).

[0080] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0081] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0082] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0083] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0084] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0085] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0086] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0087] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A geostationary satellite irradiance correction method for distributed photovoltaic systems, characterized in that, include: Acquire irradiance prediction data from geostationary meteorological satellites in the target area and the corresponding spatiotemporal auxiliary features; The clear-sky irradiance benchmark for the target area at the predicted time is determined based on radiative transfer simulation. Based on the geographical location of the distributed photovoltaic site, the predicted time, and the parameters of the geostationary satellite sub-stars, the solar geometry and the geostationary satellite observation geometry are calculated, the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry is determined, and the geometrically consistent absolute position deviation term is constructed based on the grouping index. The irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term are input into the correction model, and the corrected irradiance is determined based on the model output; wherein, the correction model is a pre-trained hierarchical Transformer model based on window attention. Physical consistency back-feed constraints are applied to the corrected irradiance to determine the corrected irradiance prediction.

2. The method according to claim 1, characterized in that, The irradiance prediction data includes at least one of total irradiance, direct irradiance, and module planar irradiance; the spatiotemporal auxiliary features include at least one of timestamp, location information of distributed photovoltaic sites, cloud-related products, quality identifiers, and historical error statistics, and the cloud-related products include at least one of cloud top height, cloud phase, and cloud optical thickness.

3. The method according to claim 1, characterized in that, The determination of the clear-sky irradiance benchmark for the target area at the predicted time based on radiative transfer simulation includes: A radiative transfer simulation tool is selected, and the geographical location of the target area, the prediction time, and typical atmospheric state parameters are input; wherein, the typical atmospheric state parameters include at least one of air pressure, aerosol content, water vapor content, and ozone concentration; By simulating the propagation path of sunlight from outer space into the Earth's atmosphere, the irradiance reaching the ground of the target area under cloudless conditions is obtained, and the irradiance is used as the clear-sky irradiance benchmark.

4. The method according to claim 1, characterized in that, The solar geometry includes the solar zenith angle and the solar azimuth angle; the geostationary satellite observation geometry includes the observation zenith angle and the observation azimuth angle; the grouping index includes the geometric grouping index and the latitude window index; The determination of the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry includes: The geometric grouping index is obtained by combining the solar zenith angle, the observed zenith angle, the difference between the solar azimuth angle and the observed azimuth angle after performing preset interval binning quantization on each of them. The latitude window index is obtained by dividing the target area into windows based on a preset latitude step size.

5. The method according to claim 4, characterized in that, The construction of the geometrically consistent absolute position deviation term based on the grouping index includes: Based on the geometric grouping index and the latitude window index, the corresponding deviation submatrix is ​​matched from the pre-constructed deviation table to form a deviation item that is adapted to the current observation geometry.

6. The method according to claim 1, characterized in that, Before inputting the irradiance prediction data, the spatiotemporal auxiliary features, the clear-sky irradiance benchmark, and the geometrically consistent absolute position deviation term into the correction model, the method further includes: Acquire historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, historical clear-sky irradiance benchmarks obtained through radiative transfer simulation, and measured irradiance data from distributed photovoltaic sites; A training dataset is constructed based on the historical geostationary satellite irradiance prediction data and corresponding spatiotemporal auxiliary features, the historical clear-sky irradiance benchmark, and the measured irradiance data. The training dataset is preprocessed by normalizing the historical geostationary satellite irradiance prediction data with the clear sky irradiance benchmark to obtain the clear sky index, and then dividing it into training set, validation set and test set according to a preset ratio. The pre-built hierarchical Transformer model is iteratively trained based on the training set. In each training round, the input features are divided into attention windows according to the window size. A geometrically consistent absolute position deviation term is incorporated into the window self-attention calculation. The model parameters and the deviation parameters in the deviation table are optimized through backpropagation. The training effect of the model is monitored through the validation set. Training is stopped when the preset number of training cycles is reached. The generalization ability of the model is evaluated through the test set, and the model with the smallest error in the test set is selected as the correction model.

7. The method according to claim 1, characterized in that, The physical consistency feedback constraint includes at least one of the following: non-negativity constraint, clear sky upper bound constraint, nighttime constraint, and ramp rate constraint; wherein, the non-negativity constraint ensures that the corrected irradiance prediction is not less than zero; the clear sky upper bound constraint ensures that the corrected irradiance prediction does not exceed a preset multiple of the clear sky irradiance benchmark; the nighttime constraint sets the corrected irradiance prediction to zero or near zero when the solar altitude angle is lower than a preset altitude threshold; and the ramp rate constraint restricts or smooths the rate of change of the corrected irradiance prediction at adjacent times.

8. A geostationary satellite irradiance correction device for distributed photovoltaic systems, characterized in that, include: The data acquisition module is used to acquire irradiance prediction data from geostationary meteorological satellites in the target area and spatiotemporal auxiliary features corresponding to the irradiance prediction data; The benchmark determination module is used to determine the clear-sky irradiance benchmark of the target area at the predicted time based on radiative transfer simulation. The deviation term construction module is used to calculate the solar geometry and geostationary satellite observation geometry based on the geographical location of the distributed photovoltaic site, the prediction time, and the geostationary satellite sub-star point parameters, determine the grouping index corresponding to the solar geometry and the geostationary satellite observation geometry, and construct the geometrically consistent absolute position deviation term based on the grouping index; The correction module is used to input the irradiance prediction data, the spatiotemporal auxiliary features, the clear sky irradiance benchmark, and the geometrically consistent absolute position deviation term into the correction model, and determine the corrected irradiance based on the model output; wherein, the correction model is a pre-trained hierarchical Transformer model based on window attention. The constraint module is used to apply physical consistency back-feed constraints to the corrected irradiance in order to determine the corrected irradiance prediction.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform the geostationary satellite irradiance correction method for distributed photovoltaics as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the geostationary satellite irradiance correction method for distributed photovoltaics as described in any one of claims 1-7.

Images