Photovoltaic output prediction method and device integrating satellite and ground data, equipment and medium

Abstract

The invention relates to the technical field of power system load prediction, in particular to a photovoltaic output prediction method and device integrating satellite and ground data, equipment and a medium, and the method comprises the steps: carrying out the preprocessing of data, which comprises remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data; space grid division is carried out based on the spatial continuous remote sensing inversion radiation data, grid division is carried out based on geographic positions and meteorological characteristics, areas with different meteorological characteristics are divided into different grid units, and each grid unit at least comprises one ground observation station; establishing a solar radiation prediction model for each grid unit; according to the trained solar radiation prediction model, calculating a solar radiation prediction value of each grid in combination with a real-time satellite cloud picture and ground observation radiation data; the limitation that a traditional foundation method cannot accurately predict radiation fluctuation caused by cloud layer changes is overcome, and therefore the prediction precision of solar radiation and photovoltaic output is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system load prediction, and in particular to a prediction method, device, equipment and medium for photovoltaic output integrating satellite and ground data. Background Art

[0002] Due to the uncertainty of solar radiation, solar photovoltaic power generation often exhibits randomness, volatility and intermittency. In order to ensure the stability of power grid system operation with a high penetration rate of photovoltaic power generation, reliable and accurate photovoltaic output forecasting is crucial, which can provide key support for the optimal dispatch, safe operation and reasonable pricing of power system.

[0003] Photovoltaic output forecast usually includes two core links: meteorological parameter prediction and photovoltaic power generation simulation, among which surface solar radiation prediction is the most critical. Solar radiation is the most important factor affecting solar power generation and the main source of uncertainty in power generation forecasting. Common methods for solar radiation prediction include statistical methods, physical methods, and hybrid methods. The main input of the above methods is the historical solar radiation and its attribute information observed on the ground, which can be collectively referred to as ground-based methods. Such methods are usually unable to output random fluctuations in solar radiation caused by changes in cloud cover, resulting in reduced prediction performance. Due to the rapid movement of clouds, clouds far away from the point of interest will also affect local weather within tens of minutes to hours, which makes local observations insufficient to predict future changes in solar radiation, and the scope of spatiotemporal observations must be expanded.

[0004] The cloud image influence of high-frequency meteorological satellite observations can be used to infer the movement of clouds, thereby improving the accuracy of solar radiation and photovoltaic output forecasts. The disadvantage of this method is that the calculation of satellite images is time- and space-intensive, and faces high computing costs in regional applications, resulting in low service efficiency.

[0005] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0006] The present invention provides a prediction method, device, equipment and medium for photovoltaic output integrating satellite and ground data, thereby effectively solving the problems in the background technology.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is: a photovoltaic output prediction method integrating satellite and ground data, comprising the following steps:

[0008] S10: preprocessing the data, wherein the data includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data;

[0009] S20: performing spatial grid division based on the spatially continuous remote sensing inversion radiation data, wherein the grid division is based on geographical location and meteorological characteristics, and regions with different meteorological characteristics are divided into different grid units, each grid unit including at least one ground observation station;

[0010] S30: Establishing a solar radiation prediction model for each grid unit, and training it by combining the ground observation radiation data and the satellite remote sensing data;

[0011] S40: Calculating the solar radiation prediction value of each grid according to the trained solar radiation prediction model, combined with the real-time satellite cloud image and the ground observation radiation data;

[0012] S50: Based on the solar radiation prediction value, using a photovoltaic system simulation tool, calculate the power output of the photovoltaic system within a target time period.

[0013] Furthermore, in step S10, the data is preprocessed, including:

[0014] S11: Performing a quality control check on the remote sensing inverted radiation data, comparing it with the real ground observation data, eliminating data records with an absolute difference ratio greater than 20%, and filling the eliminated data with the inverted multi-year average value;

[0015] S12: Performing quality inspection on the ground-based radiation observation data, eliminating noise and missing values ​​through preprocessing techniques such as filtering and interpolation, and detecting and eliminating outliers using statistical methods and physical models; finally, calibrating and verifying the data to ensure its accuracy and consistency;

[0016] S13: Perform geometric and radiometric corrections on satellite remote sensing data to eliminate errors caused by instruments and observation angles, apply quality assessment indicators to screen contaminated or distorted data, and fill in contaminated areas based on adjacent observation images;

[0017] S14: Matching the processed remote sensing inversion radiation data, the ground observation radiation data and the satellite remote sensing data by using time and space information to obtain coordinates and time information, so as to form a pairing sequence of satellite data and ground data.

[0018] Furthermore, in step S20, spatial grid division is performed based on the spatially continuous remote sensing inversion radiation data, and the steps include:

[0019] S21: Construct an infinite Gaussian mixture model, initialize the mean and covariance matrix of the infinite Gaussian mixture model, perform parameter estimation through Gibbs sampling, calculate the posterior probability of each data point belonging to each Gaussian distribution, and update the parameter value until convergence;

[0020] S22: using an adaptive method to determine the optimal number of clusters in the data based on the output result of the infinite Gaussian mixture model;

[0021] S23: inputting the preprocessed data into an infinite Gaussian mixture model fitted with an optimal number of clusters, predicting the partition to which each pixel belongs, and forming a spatially continuous partition prediction result;

[0022] S24: Perform edge detection on the partition prediction results, extract boundary lines and perform smoothing, determine the precise boundary and range of each partition, and convert them into spatial coordinates for storage.

[0023] Further, in step S22, the adaptive method is used to determine the optimal number of clusters in the data, and the adaptive method is implemented by the Bayesian information criterion, and the model of the Bayesian information criterion includes:

[0024]

[0025] In the formula, BIC is the Bayesian information criterion, n is the number of samples, k is the number of model parameters, is the maximum likelihood estimate of the model.

[0026] Further, in step S30, a solar radiation prediction model is established for each grid unit, the steps comprising:

[0027] S31: taking each of the spatial grid unit partitions as a unit, matching the ground observation radiation data with the satellite remote sensing data according to the geographic coordinates and timestamp of the partition to form paired data;

[0028] S32: Based on the paired data, construct a solar radiation prediction model integrating the ground observation radiation data and the satellite remote sensing data;

[0029] S33: extracting training samples from historical observation data, including input features and target variables, dividing the data set into a training set, a validation set, and a test set according to a ratio of 3:1:1 using the training samples, and training the solar radiation prediction model using the training set to optimize model parameters;

[0030] S34: Using the validation set and the test set to evaluate the solar radiation prediction model, calculating the performance index of the solar radiation prediction model, and ensuring that the performance of the solar radiation prediction model on the validation set and the test set is optimal.

[0031] Further, in step S32, a solar radiation prediction model integrating the ground observation radiation data and the satellite remote sensing data is constructed, including constructing the solar radiation prediction model as follows:

[0032]

[0033] In the formula, represents the solar radiation to be predicted; t represents the time of the most recent historical observation; m represents the maximum prediction duration; f represents the deep learning prediction model; D represents the model input; x i is the solar radiation observation data at the i-th time point; is a visible light satellite image block of size h×w at the i-th time point; T is the length of the input historical sequence.

[0034] Furthermore, in step S40, the solar radiation prediction value of each grid is calculated, and the steps include:

[0035] S41: Determine the geographical coordinates and time period of the point to be predicted, the geographical coordinates include longitude and latitude, and the time period includes year, month, day, hour, and minute information;

[0036] S42: Mapping the point to be predicted to the divided grid unit according to the geographic coordinates, and matching it to a solar radiation prediction model corresponding to the grid unit;

[0037] S43: extracting historical observation data of the ground observation station closest to the geographical location of the point to be predicted in the grid unit corresponding to the point to be predicted;

[0038] S44: Taking the point to be predicted as the center, extracting a visible light cloud image sequence T hours before the time period to be predicted from the satellite image, and pairing it with the historical observation data to form a data pair that meets the input format requirements of the prediction model;

[0039] S45: Inputting the paired data pair into the solar radiation prediction model to generate a solar radiation prediction result for the point to be predicted.

[0040] Further, in step S50, a photovoltaic system simulation tool is used to calculate the power output of the photovoltaic system in the target time period, the steps comprising:

[0041] S51: Select a photovoltaic power generation simulator; configure the operating environment of the photovoltaic power generation simulator, including dependent libraries and initial parameters; set the basic configuration of the photovoltaic power generation simulator by calling an initialization function;

[0042] S52: configuring various parameters of the photovoltaic system, the parameters including component parameters, installation parameters and geographical location;

[0043] S53: inputting meteorological data of the location of the photovoltaic power station, wherein the meteorological data at least includes predicted solar radiation and near real-time observed temperature;

[0044] S54: using the configured photovoltaic system parameters and the input meteorological data, calculating the DC power output of the photovoltaic assembly within the target time period through the photovoltaic power generation simulator;

[0045] S55: configuring parameters of a photovoltaic power station inverter model, wherein the parameters of the photovoltaic power station inverter model at least include maximum DC input power, maximum AC output power, and efficiency;

[0046] S56: inputting the calculated DC power generation of the photovoltaic assembly into the inverter model, performing a DC to AC conversion calculation through the inverter model, and obtaining the AC power output of the photovoltaic system within the target time period;

[0047] S57: Output the calculated AC power output, and perform necessary verification and evaluation.

[0048] The present invention also includes a photovoltaic output prediction device integrating satellite and ground data, using the above method, including:

[0049] A data preprocessing unit, used for preprocessing data, wherein the data includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data;

[0050] A spatial grid division unit, which performs spatial grid division based on the spatially continuous remote sensing inversion radiation data, wherein the grid division is based on geographical location and meteorological characteristics, and divides areas with different meteorological characteristics into different grid units, and each grid unit contains at least one ground observation station;

[0051] A partition prediction modeling unit, used to establish a solar radiation prediction model for each grid unit, and to train the model by combining the ground observation radiation data and the satellite remote sensing data;

[0052] A solar radiation prediction unit, used to calculate the solar radiation prediction value of each grid according to the trained solar radiation prediction model, combined with the real-time satellite cloud image and the ground observation radiation data;

[0053] The photovoltaic output calculation unit calculates the power output of the photovoltaic system within a target time period based on the solar radiation prediction value and using a photovoltaic system simulation tool.

[0054] The present invention also includes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described above is implemented.

[0055] The present invention also includes a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method described above is implemented.

[0056] The beneficial effects of the present invention are:

[0057] By analyzing the spatiotemporal heterogeneity of meteorological factors that affect photovoltaic output, the region is divided into grids. Only one solar radiation prediction model needs to be established for each grid, and then the photovoltaic output is predicted based on the specific location and parameters of the photovoltaic station, which greatly reduces the computational cost of the solar radiation prediction link. By taking satellite and ground observation sequences as input at the same time, extracting spatiotemporal features using convolutional neural networks, and performing time series reasoning using LSTM, the spatiotemporal dynamic changes of the cloud layer can be effectively inferred and their impact on photovoltaic output can be quantified, thereby improving the accuracy of photovoltaic output prediction under cloudy weather conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0059] Figure 1 A flowchart of a photovoltaic power output prediction method integrating satellite and ground data;

[0060] Figure 2 A technology roadmap for PV output prediction methods integrating satellite and ground-based data;

[0061] Figure 3 An architectural diagram for integrating satellite and ground prediction models;

[0062] Figure 4 The grid division results of Zhenjiang, Jiangsu obtained by the infinite Gaussian mixture (IGMM) model;

[0063] Figure 5 This is a sample graph for prediction at 2023-06-20 12:00;

[0064] Figure 6 This is a comparison chart of the prediction results with different lead times;

[0065] Figure 7 Schematic diagram of the structure of the photovoltaic output prediction device integrating satellite and ground data;

[0066] Figure 8 A schematic diagram of the structure of a computer device. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0068] Embodiment 1:

[0069] like Figures 1 to 3 As shown: A method for predicting photovoltaic output by integrating satellite and ground data includes the following steps:

[0070] S10: preprocessing the data, which includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data;

[0071] S20: Based on spatially continuous remote sensing inversion radiation data, spatial grid division is performed. The purpose of grid division is to gather locations with similar spatiotemporal variation characteristics of solar radiation and to separate areas with large differences in variation characteristics, thereby guiding the ground data collection process for calibrating photovoltaic output prediction results. Grid division is based on geographical location and meteorological characteristics, and regions with different meteorological characteristics are divided into different grid units, each of which contains at least one ground observation station. Due to differences in meteorological conditions in different regions, directly using a national or large-region prediction model may not be accurate enough. Dividing the region into multiple different grid units to ensure that the meteorological characteristics in each unit are as consistent as possible can improve the prediction accuracy of each unit, and ensure that each grid unit contains at least one ground observation station to ensure that the model can obtain sufficient calibration data.

[0072] S30: Establish a solar radiation prediction model for each grid unit, and train it by combining ground observation radiation data and satellite remote sensing data. The advantages of this step are: on the one hand, since the temporal and spatial variations of solar radiation in each sub-area have similar characteristics, it can be ensured that the solar radiation prediction results of the model can effectively support the output prediction of all photovoltaic stations in the sub-area; on the other hand, it is only necessary to establish a prediction model for each sub-area instead of each station, which avoids the high computational cost of intensive prediction and improves the computational efficiency of regional applications.

[0073] S40: Based on the trained solar radiation prediction model, combined with real-time satellite cloud images and ground observation radiation data, calculate the solar radiation prediction value of each grid; use real-time data to ensure the timeliness of the prediction results and adapt to the dynamic changes in meteorological conditions. This step provides specific radiation input data for subsequent photovoltaic output prediction;

[0074] S50: Based on the predicted solar radiation value, using a photovoltaic system simulation tool, calculate the power output of the photovoltaic system within a target time period.

[0075] By analyzing the temporal and spatial heterogeneity of meteorological factors that affect photovoltaic output, the region is divided into grids, and at least one solar radiation prediction model is established for each grid. Then, photovoltaic output is predicted based on the specific location and parameters of the photovoltaic stations, which greatly reduces the computational consumption of the solar radiation prediction link.

[0076] By combining remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data, it is possible to more comprehensively capture the spatiotemporal variation characteristics of solar radiation, especially the impact of cloud cover and its rapid movement on solar radiation. This method overcomes the limitation of traditional ground-based methods that cannot accurately predict radiation fluctuations caused by cloud changes, thereby significantly improving the prediction accuracy of solar radiation and photovoltaic output.

[0077] By combining satellite remote sensing data and ground-based radiation data and integrating radiation information from different sources, the limitations of a single data source can be effectively supplemented, which helps to improve the accuracy of photovoltaic output forecasts, especially under complex meteorological conditions such as cloud cover.

[0078] Spatial grid division is based on geographical location and meteorological characteristics. Areas with similar meteorological characteristics are clustered in the same grid, and each grid is ensured to have at least one ground observation station for data calibration. This can ensure that the prediction model can accurately reflect the meteorological changes in each grid and enhance the reliability of photovoltaic output prediction in the region.

[0079] By building a solar radiation prediction model for each grid cell instead of modeling it separately for each observation station, the amount of calculation is greatly reduced, especially in large-scale regional applications, avoiding the high cost of repeated calculations while ensuring an improvement in overall computing efficiency.

[0080] As a preference of the above embodiment, in step S10, preprocessing the data includes:

[0081] S11: Quality control check of remote sensing inverted radiation data, compare with real ground observation data, remove data records with absolute difference ratio greater than 20%, and use the inverted multi-year average value to fill the removed data;

[0082] S12: Perform quality inspection on ground-based radiation data, eliminate noise and missing values ​​through pre-processing techniques such as filtering and interpolation, and use statistical methods and physical models to detect and remove outliers; finally, calibrate and verify the data to ensure its accuracy and consistency;

[0083] S13: Perform geometric and radiometric corrections on satellite remote sensing data to eliminate errors caused by instruments and observation angles, apply quality assessment indicators to screen contaminated or distorted data, and fill in contaminated areas based on adjacent observation images;

[0084] S14: Match the processed remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data by using time and space information to perform coordinate and time information, so as to form a pairing sequence of satellite data and ground data.

[0085] Through step-by-step quality control checks and preprocessing, the accuracy and consistency of the data are ensured, thereby improving the reliability of subsequent prediction models; by eliminating outliers, filling missing data and correcting image data, data errors are reduced and the impact of inaccurate data on prediction results is avoided; through data pairing in time and space, consistency between different data sources is ensured, which is crucial for model training and prediction accuracy; the processed high-quality data enables the model training to better fit the actual situation, improving the prediction accuracy and reliability; through effective data preprocessing and quality control, the interference of erroneous data in subsequent model training is reduced, thereby improving computational efficiency.

[0086] In this embodiment, in step S20, spatial grid division is performed based on spatially continuous remote sensing inversion radiation data, and the steps include:

[0087] S21: Construct an infinite Gaussian mixture (IGMM) model. IGMM is a model based on the Bayesian non-parametric method. Its main advantage is that there is no need to pre-set the number of clusters. Initialize the mean and covariance matrix of the infinite Gaussian mixture (IGMM) model, perform parameter estimation through Gibbs sampling, calculate the posterior probability of each data point belonging to each Gaussian distribution, and update the parameter value until convergence; this step is the training process of the infinite Gaussian mixture (IGMM) model to ensure that the model can correctly capture the distribution of the data; update the parameter value until convergence to ensure that the parameters of the model obtain stable values ​​through repeated iterations, thereby improving the accuracy of the model;

[0088] S22: Using the output of the infinite Gaussian mixture (IGMM) model, an adaptive method is used to determine the optimal number of clusters in the data;

[0089] S23: inputting the preprocessed data into an infinite Gaussian mixture (IGMM) model fitted with the optimal cluster number to predict the partition to which each pixel belongs, thereby forming a spatially continuous partition prediction result; ensuring that the data used is quality controlled, thereby improving the accuracy of the prediction;

[0090] S24: Perform edge detection on the partition prediction results, extract boundary lines and perform smoothing, determine the precise boundary and range of each partition, and convert them into spatial coordinates for storage.

[0091] Through the unsupervised learning characteristics of the infinite Gaussian mixture (IGMM) model, the optimal partitioning of the data can be automatically determined without manually setting the number of clusters, thereby improving processing efficiency; through edge detection and smoothing processing, the boundary of each partition can be determined more accurately, improving the spatial resolution of the model; accurate spatial grid division and refined partitioning can make subsequent solar radiation predictions more accurate, thereby improving the reliability and practicality of photovoltaic output predictions.

[0092] As a preferred embodiment of the above, in step S22, an adaptive method is used to determine the optimal number of clusters in the data, and the adaptive method is implemented by Bayesian Information Criterion (BIC). The model of Bayesian Information Criterion (BIC) includes:

[0093]

[0094] In the formula, BIC is the Bayesian information criterion, n is the number of samples, k is the number of model parameters, is the maximum likelihood estimate of the model. BIC is calculated for different cluster numbers k, and the k with the smallest BIC is selected as the optimal cluster number.

[0095] In the model selection process, BIC selects the best model by weighing the accuracy of data fitting and the complexity of the model; specifically, BIC adds a penalty term based on the maximum likelihood estimation, and the penalty term is related to the number of model parameters. Therefore, when selecting a model, BIC can avoid selecting an overly complex model, that is, avoid overfitting.

[0096] BIC provides a scientific method to automatically select the optimal number of clusters. By calculating the BIC values ​​of different cluster numbers, the model with the best balance can be objectively selected, avoiding the deviation that may be caused by manually setting the number of clusters. By selecting the optimal number of clusters, the spatial area can be divided more accurately, so that the subsequent solar radiation prediction and photovoltaic output calculation can reflect the real spatial distribution characteristics. The optimal number of clusters improves the prediction accuracy of the model and further enhances the reliability and practicality of photovoltaic output prediction.

[0097] Wherein, in step S30, a solar radiation prediction model is established for each grid unit, and the steps include:

[0098] S31: Based on the geographic coordinates and timestamps of each spatial grid unit partition, ground observation radiation data and satellite remote sensing data are matched to form paired data; this step is the basis for the entire model training. First, the ground observation radiation data and satellite observation data must be matched based on each spatial grid unit. This step ensures that the input data for model training contains both actual ground observations and visible light images in satellite data, providing a complete data perspective;

[0099] S32: Based on paired data, a solar radiation prediction model integrating ground observation radiation data and satellite remote sensing data is constructed; this step integrates multi-source data, fully utilizes the characteristics and spatiotemporal information of different data, and further provides a model architecture for subsequent training;

[0100] S33: Extract training samples from historical observation data, including input features and target variables, divide the data set into training set, validation set and test set in a ratio of 3:1:1, and use the training set to train the solar radiation prediction model to optimize the model parameters; select the root mean square error as the loss function during the training process, and select the Adam optimizer for gradient calculation and parameter optimization; this step enables the model to learn the characteristics of the data and optimize the parameters, thereby improving the prediction accuracy, and is the core step to improve model performance.

[0101] S34: Use the validation set and test set to evaluate the solar radiation prediction model and calculate the performance indicators of the solar radiation prediction model (such as mean square error, R 2 , accuracy, etc.) to ensure that the solar radiation prediction model has the best performance on the validation set and test set. The optimal model is saved as an .h5 file for subsequent application and deployment. Through dual verification of the validation set and test set, problems such as model overfitting can be avoided, and the optimal model can be finally used in practical applications.

[0102] By matching ground-based radiation data with satellite data, data from different sources are fully utilized. This fusion of multi-source data improves the model's input diversity and comprehensiveness, and can better capture the spatiotemporal variation characteristics of solar radiation, thereby improving prediction accuracy. The use of deep learning models can effectively capture complex nonlinear relationships, especially the extraction of spatiotemporal features. The model can learn from satellite images and historical data of ground observations and predict future trends in solar radiation. By refining the prediction process to each spatial grid unit, the model can adapt to the specific radiation characteristics of different geographical regions. This targeted processing improves the accuracy of large-area predictions and helps improve prediction results under different geographical and climatic conditions. Through the evaluation of validation and test sets, the performance of the model on different data sets is ensured to be stable and overfitting is avoided. This training, validation, and testing structure can ensure that the model not only performs well on historical data, but also maintains a high prediction accuracy in future practical applications.

[0103] As a preferred embodiment of the above, in step S32, a solar radiation prediction model integrating ground observation radiation data and satellite remote sensing data is constructed, including constructing the solar radiation prediction model as follows:

[0104]

[0105] In the formula, represents the solar radiation to be predicted; t represents the time of the most recent historical observation; m represents the maximum prediction duration; f represents the deep learning prediction model; D represents the model input; x i is the solar radiation observation data at the i-th time point; is a visible light satellite image block of size h×w at the i-th time point; T is the length of the input historical sequence.

[0106] By building a solar radiation prediction model that integrates ground-based radiation data and satellite remote sensing data, the advantages of both types of data are fully utilized, combining actual ground observations with the large-scale spatial coverage of satellite remote sensing to more accurately predict future solar radiation changes; by introducing visible light image blocks from satellite remote sensing, the model can capture information such as cloud changes and weather fluctuations, improving the ability to predict photovoltaic output under complex weather conditions.

[0107] The structural architecture of the deep learning model is as follows Figure 3 As shown in the figure, 3D convolution is used to extract the spatiotemporal features of the satellite image sequence, and 1D convolution is used to extract the fluctuation features of the ground observation sequence. Then, all features are aligned in the time dimension and input into the LSTM structure for time series reasoning. Finally, MLP is used to establish a nonlinear relationship between the reasoning information and subsequent solar radiation.

[0108] In this embodiment, in step S40, the solar radiation prediction value of each grid is calculated, and the steps include:

[0109] S41: Determine the geographic coordinates and time period of the point to be predicted, where the geographic coordinates include longitude and latitude, and the time period includes year, month, day, hour, and minute information;

[0110] S42: Mapping the predicted point to the divided grid unit according to the geographic coordinates, and matching it to the solar radiation prediction model corresponding to the grid unit; this step ensures the pertinence and accuracy of the prediction model;

[0111] S43: extracting historical observation data of the ground observation station that is closest to the geographical location of the predicted point in the grid cell corresponding to the predicted point; using the principle of spatial proximity, selecting the data of the nearest ground observation station as a reference is helpful to capture the meteorological characteristics of the local area;

[0112] S44: Taking the predicted point as the center, extract the visible light cloud image sequence T hours before the predicted time period from the satellite image, and pair it with the historical observation data to form a data pair that meets the input format requirements of the prediction model;

[0113] S45: Input the paired data pair into the solar radiation prediction model to generate a solar radiation prediction result for the point to be predicted.

[0114] By accurately matching the prediction model of the grid where the predicted point is located and combining the historical data of the nearest ground observation station and real-time satellite cloud images, the temporal and spatial variation characteristics of solar radiation can be more accurately captured, thereby improving the prediction accuracy;

[0115] Wherein, in step S50, the photovoltaic system simulation tool is used to calculate the power output of the photovoltaic system within the target time period, and the steps include:

[0116] S51: Select a suitable photovoltaic power generation simulator, such as PVlib or other professional software; configure the operating environment of the photovoltaic power generation simulator, including dependent libraries and initial parameters; set the basic configuration of the photovoltaic power generation simulator by calling the initialization function;

[0117] S52: configure various parameters of the photovoltaic system, including component parameters, installation parameters and geographical location;

[0118] The parameters can be divided into three groups: ① Module parameters: set the electrical parameters of the PV modules, such as open circuit voltage (Voc), short circuit current (Isc), maximum power point voltage (Vmp), maximum power point current (Imp), etc.; ② Installation parameters: set the installation angle and azimuth of the PV modules, etc.; ③ Geographic location: set the geographical location (longitude, latitude, altitude) and prediction time period (year-month-day-hour-minute) of the PV power station;

[0119] S53: inputting meteorological data of the location of the photovoltaic power station, the meteorological data at least including predicted solar radiation and near-real-time observed temperature; using the configured photovoltaic system parameters, simulating the DC power output of the photovoltaic module;

[0120] S54: using the configured photovoltaic system parameters and the input meteorological data, calculating the DC power output of the photovoltaic module within the target time period through the photovoltaic power generation simulator;

[0121] S55: configuring parameters of a photovoltaic power station inverter model, where the parameters of the photovoltaic power station inverter model at least include maximum DC input power, maximum AC output power, and efficiency;

[0122] S56: inputting the calculated DC power generation of the photovoltaic module into the inverter model, performing a DC to AC conversion calculation through the inverter model, and obtaining the AC power output of the photovoltaic system within the target time period;

[0123] S57: Output the calculated AC output and conduct necessary verification and evaluation to confirm the accuracy and reliability of the simulation results and provide a scientific basis for decisions such as optimal dispatch, safe operation and reasonable pricing of the power system.

[0124] By selecting a suitable photovoltaic power generation simulator and configuring its operating environment, dependent libraries and initial parameters in detail, it is possible to ensure that the simulation environment is as close as possible to the actual photovoltaic system, thereby improving the accuracy of power output prediction, helping to more accurately evaluate the power generation capacity of the photovoltaic system, and providing a reliable basis for the scheduling and planning of the power system; the solution requires the configuration of various parameters of the photovoltaic system, including component parameters, installation parameters and geographical location, etc. Comprehensive parameter configuration can more realistically reflect the actual operation of the photovoltaic system, including the impact of different installation angles, azimuths and geographical locations on power generation efficiency, thereby improving the comprehensiveness and accuracy of the prediction; inputting meteorological data at the location of the photovoltaic power station, including predicted solar radiation and near-real-time observed temperature, etc., can fully consider the impact of meteorological conditions on the power generation efficiency of the photovoltaic system. The fusion of multi-source data can more accurately simulate the power generation performance of photovoltaic components under different meteorological conditions and improve the accuracy of the prediction results.

[0125] Embodiment 2:

[0126] In order to verify the performance of the present invention, the photovoltaic output prediction of a station in Zhenjiang, Jiangsu Province is taken as an example to verify the present invention.

[0127] 1) Data collection and processing: From the Pangaea website ( https: / / doi.org / 10.1594 / PANGAEA.904136 ) collect monthly solar radiation data with a resolution of 5 km, collect observation data of solar radiation and photovoltaic output from ground stations, download satellite cloud image data from the Fengyun Meteorological Remote Sensing Data Service Network, and pre-process and pair the data in time and space.

[0128] 2) Spatial grid division: An infinite Gaussian mixture (IGMM) model was constructed, and the optimal number of partitions in Zhenjiang area was inferred to be 14 through the BIC criterion adaptively. Then, the spatially continuous monthly solar radiation data was input into the model to predict the partition of each location, and the spatial grid division results were obtained as follows: Figure 4 As shown, pixels of the same color in the figure belong to the same partition.

[0129] 3) Zoning prediction modeling: The sites for ground observation radiation data collected in this application example are spatially Figure 4 The light green zone is in the middle, so a prediction model is built for this area. The observation data covers the period from January 1, 2022 to December 31, 2023. We build a prediction model based on the data from 2022, and the data from 2023 is used as an independent test of the PV output prediction performance. The samples are organized in the form of a solar radiation prediction model, divided into a training set, a validation set, and a test set in a ratio of 3:1:1, and the model is trained using the training data set to obtain optimized model parameters. The model with the smallest mean square error on the validation set is retained for subsequent application and deployment.

[0130] 4) Solar radiation prediction: With the observation station as the center, extract the visible light cloud image sequence of the 6 hours before the predicted time period from the satellite image, and pair it with the historical observation data to form a data pair that meets the input format requirements of the prediction model; input the paired ground and satellite data into the prediction model to predict the solar radiation. Figure 5 As shown in the figure, the prediction result at 2023-06-2012:00 is shown. The model predicts the radiation changes in the next 6 hours based on the real observations in the past 6 hours. By continuously rolling the time to be predicted in a loop, the prediction results for the whole year of 2023 can be obtained.

[0131] 5) Photovoltaic output calculation: The photovoltaic power generation process is simulated with PVLIB professional software to obtain the power generation of a photovoltaic station with a specific installed capacity under the predicted solar radiation level. The photovoltaic system of the ground photovoltaic station selected in the example of the present invention is tilted at an angle of 25° to the south, using polycrystalline silicon photovoltaic modules, with an installed capacity of 12.8MW, a module temperature difference ΔT of about 3°C, and an inverter model of "PVP2500 240V [CEC 2006]". Figure 6 The comparison results between the photovoltaic output prediction results and the actual recorded values ​​are shown. When the lead time increases from 1 hour to 6 hours, the prediction error gradually increases and the accuracy gradually decreases; the average prediction error 1 hour in advance is 0.77MW and the accuracy is 93.96%.

[0132] In view of the insufficient prediction accuracy of commonly used photovoltaic output prediction technologies under cloud conditions and the defects of satellite-based prediction methods that require intensive calculations, the present invention proposes a method for predicting photovoltaic output by integrating satellite and ground data based on spatial similarity grid division. By analyzing the spatiotemporal heterogeneity characteristics of meteorological factors that affect photovoltaic output, the region is gridded. Only one solar radiation prediction model needs to be established for each grid, and then photovoltaic output is predicted in combination with the specific location and parameters of the photovoltaic station, which greatly reduces the computational consumption of the solar radiation prediction link. By taking satellite and ground observation sequences as inputs at the same time, and using convolutional neural networks to extract spatiotemporal features, and using LSTM for time series reasoning, the spatiotemporal dynamic changes of the cloud layer can be effectively inferred and their impact on photovoltaic output can be quantified, thereby improving the accuracy of photovoltaic output prediction under cloudy weather conditions. Therefore, the present invention has the advantages of efficient regional application and high model prediction accuracy.

[0133] The present invention also includes a photovoltaic output prediction device integrating satellite and ground data, using the above method, such as Figure 7 As shown, including:

[0134] A data preprocessing unit is used to preprocess the data, the data includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data;

[0135] Spatial grid division unit: spatial grid division is performed based on spatially continuous remote sensing inversion radiation data. The grid division is based on geographical location and meteorological characteristics. Regions with different meteorological characteristics are divided into different grid units. Each grid unit contains at least one ground observation station.

[0136] The partition prediction modeling unit is used to establish a solar radiation prediction model for each grid cell, which is trained by combining ground observation radiation data and satellite remote sensing data;

[0137] The solar radiation prediction unit is used to calculate the solar radiation prediction value of each grid based on the trained solar radiation prediction model, combined with real-time satellite cloud images and ground observation radiation data;

[0138] The photovoltaic output calculation unit calculates the power output of the photovoltaic system within the target time period based on the predicted solar radiation value and using the photovoltaic system simulation tool.

[0139] See also Figure 8 A computer device 400 provided in an embodiment of the present application includes: a processor 410 and a memory 420, wherein the memory 420 stores a computer program executable by the processor 410, and when the computer program is executed by the processor 410, the above method is executed.

[0140] The embodiment of the present application further provides a storage medium 430 on which a computer program is stored. When the computer program is run by the processor 410, the above method is executed.

[0141] Among them, the storage medium 430 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0142] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. "Multiple" means two or more, unless otherwise clearly and specifically defined.

[0143] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0144] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0145] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0146] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0147] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0148] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0149] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A photovoltaic output prediction method integrating satellite and ground data, characterized in that: The steps include: S10: preprocessing the data, wherein the data includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data; S20: performing spatial grid division based on the spatially continuous remote sensing inversion radiation data, wherein the grid division is based on geographical location and meteorological characteristics, and regions with different meteorological characteristics are divided into different grid units, each grid unit including at least one ground observation station; S30: Establishing a solar radiation prediction model for each grid unit, and training it by combining the ground observation radiation data and the satellite remote sensing data; S40: Calculating the solar radiation prediction value of each grid according to the trained solar radiation prediction model, combined with the real-time satellite cloud image and the ground observation radiation data; S50: Based on the solar radiation prediction value, using a photovoltaic system simulation tool, calculate the power output of the photovoltaic system within a target time period.

2. The photovoltaic output prediction method integrating satellite and ground data according to claim 1 is characterized in that: In step S10, the data is preprocessed, including: S11: Performing a quality control check on the remote sensing inverted radiation data, comparing it with the real ground observation data, eliminating data records with an absolute difference ratio greater than 20%, and filling the eliminated data with the inverted multi-year average value; S12: Performing quality inspection on the ground-based radiation observation data, eliminating noise and missing values ​​through preprocessing techniques such as filtering and interpolation, and detecting and eliminating outliers using statistical methods and physical models; finally, calibrating and verifying the data to ensure its accuracy and consistency; S13: Perform geometric and radiometric corrections on satellite remote sensing data to eliminate errors caused by instruments and observation angles, apply quality assessment indicators to screen contaminated or distorted data, and fill in contaminated areas based on adjacent observation images; S14: Matching the processed remote sensing inversion radiation data, the ground observation radiation data and the satellite remote sensing data by using time and space information to obtain coordinates and time information, so as to form a pairing sequence of satellite data and ground data.

3. The photovoltaic output prediction method integrating satellite and ground data according to claim 1 is characterized in that: In step S20, spatial grid division is performed based on the spatially continuous remote sensing inversion radiation data, and the steps include: S21: Construct an infinite Gaussian mixture model, initialize the mean and covariance matrix of the infinite Gaussian mixture model, perform parameter estimation through Gibbs sampling, calculate the posterior probability of each data point belonging to each Gaussian distribution, and update the parameter value until convergence; S22: using an adaptive method to determine the optimal number of clusters in the data based on the output result of the infinite Gaussian mixture model; S23: inputting the preprocessed data into an infinite Gaussian mixture model fitted with an optimal number of clusters, predicting the partition to which each pixel belongs, and forming a spatially continuous partition prediction result; S24: Perform edge detection on the partition prediction results, extract boundary lines and perform smoothing, determine the precise boundary and range of each partition, and convert them into spatial coordinates for storage.

4. The photovoltaic output prediction method integrating satellite and ground data according to claim 3 is characterized in that: In step S22, the adaptive method is used to determine the optimal number of clusters in the data, and the adaptive method is implemented by the Bayesian Information Criterion. The model of the Bayesian Information Criterion includes: In the formula, BIC is the Bayesian information criterion, n is the number of samples, k is the number of model parameters, is the maximum likelihood estimate of the model.

5. The photovoltaic output prediction method integrating satellite and ground data according to claim 1 is characterized in that: In step S30, a solar radiation prediction model is established for each grid unit, the steps comprising: S31: taking each of the spatial grid unit partitions as a unit, matching the ground observation radiation data with the satellite remote sensing data according to the geographic coordinates and timestamp of the partition to form paired data; S32: Based on the paired data, construct a solar radiation prediction model integrating the ground observation radiation data and the satellite remote sensing data; S33: extracting training samples from historical observation data, including input features and target variables, dividing the data set into a training set, a validation set, and a test set according to a ratio of 3:1:1 using the training samples, and training the solar radiation prediction model using the training set to optimize model parameters; S34: Using the validation set and the test set to evaluate the solar radiation prediction model, calculating the performance index of the solar radiation prediction model, and ensuring that the performance of the solar radiation prediction model on the validation set and the test set is optimal.

6. The photovoltaic output prediction method integrating satellite and ground data according to claim 5 is characterized in that: In step S32, a solar radiation prediction model integrating the ground observation radiation data and the satellite remote sensing data is constructed, including constructing the solar radiation prediction model as follows: In the formula, represents the solar radiation to be predicted; t represents the time of the most recent historical observation; m represents the maximum prediction duration; f represents the deep learning prediction model; D represents the model input; x i is the solar radiation observation data at the i-th time point; is a visible light satellite image block of size h×w at the i-th time point; T is the length of the input historical sequence.

7. The photovoltaic output prediction method integrating satellite and ground data according to claim 1, characterized in that: In step S40, the solar radiation prediction value of each grid is calculated, and the steps include: S41: Determine the geographical coordinates and time period of the point to be predicted, the geographical coordinates include longitude and latitude, and the time period includes year, month, day, hour, and minute information; S42: mapping the point to be predicted to the divided grid unit according to the geographic coordinates, and matching it to a solar radiation prediction model corresponding to the grid unit; S43: extracting historical observation data of the ground observation station closest to the geographical location of the point to be predicted in the grid unit corresponding to the point to be predicted; S44: Taking the point to be predicted as the center, extracting a visible light cloud image sequence T hours before the time period to be predicted from the satellite image, and pairing it with the historical observation data to form a data pair that meets the input format requirements of the prediction model; S45: Inputting the paired data pair into the solar radiation prediction model to generate a solar radiation prediction result for the point to be predicted.

8. The photovoltaic output prediction method integrating satellite and ground data according to claim 1, characterized in that: In step S50, a photovoltaic system simulation tool is used to calculate the power output of the photovoltaic system within a target time period, and the steps include: S51: Select a photovoltaic power generation simulator; configure the operating environment of the photovoltaic power generation simulator, including dependent libraries and initial parameters; set the basic configuration of the photovoltaic power generation simulator by calling an initialization function; S52: configuring various parameters of the photovoltaic system, the parameters including component parameters, installation parameters and geographical location; S53: inputting meteorological data of the location of the photovoltaic power station, wherein the meteorological data at least includes predicted solar radiation and near real-time observed temperature; S54: using the configured parameters of the photovoltaic system and the input meteorological data, calculating the DC power output of the photovoltaic assembly within the target time period through the photovoltaic power generation simulator; S55: configuring parameters of a photovoltaic power station inverter model, wherein the parameters of the photovoltaic power station inverter model at least include maximum DC input power, maximum AC output power, and efficiency; S56: inputting the calculated DC power generation of the photovoltaic assembly into the inverter model, performing a DC to AC conversion calculation through the inverter model, and obtaining the AC power output of the photovoltaic system within the target time period; S57: Output the calculated AC power output, and perform necessary verification and evaluation.

9. A photovoltaic output prediction device integrating satellite and ground data, characterized in that: Use of the method according to any one of claims 1 to 8, comprising: A data preprocessing unit, used for preprocessing data, wherein the data includes remote sensing inversion radiation data, ground observation radiation data and satellite remote sensing data; A spatial grid division unit, which performs spatial grid division based on the spatially continuous remote sensing inversion radiation data, wherein the grid division is based on geographical location and meteorological characteristics, and divides areas with different meteorological characteristics into different grid units, and each grid unit contains at least one ground observation station; A partition prediction modeling unit, used to establish a solar radiation prediction model for each grid unit, and to train the model by combining the ground observation radiation data and the satellite remote sensing data; A solar radiation prediction unit, used to calculate the solar radiation prediction value of each grid according to the trained solar radiation prediction model, combined with the real-time satellite cloud image and the ground observation radiation data; The photovoltaic output calculation unit calculates the power output of the photovoltaic system within a target time period based on the solar radiation prediction value and using a photovoltaic system simulation tool.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

11. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images