Ultra-short-term solar irradiance prediction method and device based on satellite cloud picture

Abstract

The invention relates to the technical field of new energy power generation prediction, and particularly provides an ultra-short-term solar irradiance prediction method and device based on a satellite cloud atlas, and the method comprises the steps: obtaining the prediction time period of a prediction result and an error evaluation index of a parameter type relative to each pre-trained multi-layer perceptron model; selecting a pre-trained multi-layer perceptron model with the minimum error evaluation index to predict the prediction result; wherein the parameter types comprise global level irradiance, direct normal irradiance and diffusion level irradiance. According to the technical scheme provided by the invention, model optimization selection is carried out for different irradiance types, and the adaptability and accuracy of prediction are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy power generation prediction, and in particular to a method and device for predicting ultra-short-term solar irradiance based on satellite cloud images. Background Art

[0002] Since 2011, China has focused on developing photovoltaic power generation and related industries. Global installed photovoltaic capacity has skyrocketed from 135 GW in 2013 to 480 GW in 2018, a nearly fourfold increase in just five years. Photovoltaic power generation is playing an increasingly important role in our energy mix.

[0003] With the widespread use of solar energy in power systems, accurate prediction of solar irradiance is crucial for the efficient operation of photovoltaic power generation systems and the stable dispatch of power grids. This allows power dispatch departments to adjust dispatch plans in a timely manner, improve the economy and stability of power grid operation, and promote the integration of new energy sources.

[0004] Existing solar irradiance prediction methods include:

[0005] Statistical methods are primarily based on data-driven formulas. They analyze the temporal variations of historical solar irradiance data and construct relationships between inputs and outputs to generate forecasts. For example, the autoregressive moving average (ARMA) model leverages the autocorrelation and moving average properties of time series to fit and extrapolate historical data. Specifically, the ARMA model establishes a mathematical model by determining the autoregressive and moving average orders to predict future irradiance based on past irradiance values.

[0006] Traditional AI methods are widely used in solar irradiance prediction, such as BP neural networks and support vector machines. BP neural networks construct a multi-layered neural network structure and utilize a back-propagation algorithm to continuously adjust network weights, learning the complex nonlinear relationship between input data (such as historical irradiance and meteorological data) and predicted irradiance values. Support vector machines, based on the principle of structural risk minimization, map input data into a high-dimensional space and search for an optimal classification hyperplane to predict solar irradiance.

[0007] Prediction methods based on ground imagery: A common method for solar irradiance prediction is using ground cloud maps. Because ground cloud maps have high spatial and temporal resolution, they can clearly capture the cloud distribution around photovoltaic sites. By processing and analyzing ground cloud images, cloud features such as shape, position, and motion are extracted and combined with specific algorithmic models to predict solar irradiance changes.

[0008] However, existing solar irradiance prediction methods have numerous shortcomings. Statistical methods rely on the stability and regularity of historical data and are less adaptable to environmental changes. In reality, solar irradiance is affected by a variety of complex factors, such as sudden weather changes and rapid changes in cloud cover. These factors make it difficult for historical data to accurately reflect future irradiance trends. Furthermore, statistical models are often constructed based on assumptions. When the actual situation does not meet these assumptions, the accuracy of the prediction results will drop significantly. For example, in areas with unpredictable weather, the ARMA model struggles to capture the impact of rapid cloud changes on irradiance, resulting in large prediction errors.

[0009] While traditional AI methods possess certain learning and prediction capabilities, they still suffer from several drawbacks when it comes to solar irradiance prediction. For one thing, these methods require high data quality and quantity. Noisy or missing data can severely impact model performance. Furthermore, traditional AI models lack interpretability, making it difficult to intuitively understand the basis for their predictions. For example, the complex structure of BP neural networks makes it difficult to analyze the specific impact of input variables on output, hindering troubleshooting and optimization in practical applications.

[0010] While ground-based imagery-based forecasting methods can capture localized cloud information around a site, their limited monitoring range prevents them from fully capturing cloud conditions and changing trends over a larger area. When clouds move rapidly from a distance, ground-based imagery may not capture relevant information in advance, resulting in forecast delays. Furthermore, ground-based imagery acquisition is susceptible to weather conditions. For example, in severe weather conditions (such as heavy rain or fog), image quality can severely degrade, or even render it impossible to obtain valid images, compromising forecast accuracy and reliability.

[0011] Therefore, a new method is needed to improve the prediction accuracy of solar irradiance. Summary of the Invention

[0012] In order to overcome the above-mentioned defects, the present invention proposes a method and device for predicting ultra-short-term solar irradiance based on satellite cloud images.

[0013] In a first aspect, a method for predicting ultra-short-term solar irradiance based on satellite cloud images is provided, wherein the method comprises:

[0014] Obtain the error evaluation indicators of the prediction period and parameter type of the prediction results relative to each pre-trained multilayer perceptron model;

[0015] Selecting a pre-trained multi-layer perceptron model with the smallest error evaluation index to predict the prediction result;

[0016] The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

[0017] Preferably, the error evaluation indicators include: mean absolute error, root mean square error, normalized root mean square error, correlation coefficient, determination coefficient, and mean absolute percentage error.

[0018] Furthermore, the training process of the pre-trained multi-layer perceptron model includes:

[0019] The training data is constructed using historical meteorological data and its corresponding solar irradiance data of different parameter types;

[0020] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0021] Furthermore, the meteorological data includes: upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and wind direction.

[0022] Preferably, the training process of the pre-trained multilayer perceptron model includes:

[0023] Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image;

[0024] The training data is constructed using cloud feature data from historical satellite cloud images and their corresponding solar irradiance data of different parameter types;

[0025] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0026] Furthermore, the cloud features include: cloud area, cloud average grayscale value, and cloud texture features.

[0027] Preferably, the training process of the pre-trained multilayer perceptron model includes:

[0028] Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image;

[0029] The training data is constructed using historical meteorological data, cloud feature data from historical satellite cloud images, and their corresponding solar irradiance data of different parameter types;

[0030] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0031] Furthermore, the acquisition of historical satellite cloud image data includes: preprocessing the historical satellite cloud image data, and the preprocessing includes: image enhancement, grayscale processing and research area determination.

[0032] In a second aspect, a device for predicting ultra-short-term solar irradiance based on a satellite cloud image is provided, wherein the device comprises:

[0033] An acquisition module is used to obtain the error evaluation index of the prediction period and parameter type of the prediction result relative to each pre-trained multi-layer perceptron model;

[0034] A prediction module, configured to select a pre-trained multi-layer perceptron model with a minimum error evaluation index to predict the prediction result;

[0035] The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

[0036] In a third aspect, a computer device is provided, comprising: one or more processors;

[0037] The processor is configured to execute one or more programs;

[0038] When the one or more programs are executed by the one or more processors, the ultra-short-term solar irradiance prediction method based on satellite cloud images is implemented.

[0039] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed, the ultra-short-term solar irradiance prediction method based on satellite cloud images is implemented.

[0040] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0041] The present invention provides a method and device for predicting ultra-short-term solar irradiance based on satellite cloud images, comprising: obtaining an error evaluation index for the prediction period and parameter type of the prediction result relative to each pre-trained multi-layer perceptron model; selecting the pre-trained multi-layer perceptron model with the smallest error evaluation index to predict the prediction result; wherein the parameter types include: global horizontal irradiance, direct normal irradiance, and diffuse horizontal irradiance. The technical solution provided by the present invention optimizes and selects models for different irradiance types. The system can acquire and process data in real time and output prediction results promptly, ensuring the real-time nature of the prediction. Furthermore, the multi-model comparison and evaluation mechanism improves the reliability and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1The present invention is a schematic flow chart of the main steps of the ultra-short-term solar irradiance prediction method based on satellite cloud images. DETAILED DESCRIPTION

[0043] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0045] Example 1

[0046] See attached Figure 1 , Figure 1 FIG1 is a flow chart of the main steps of a method for predicting ultra-short-term solar irradiance based on satellite cloud images according to an embodiment of the present invention. Figure 1 As shown, the ultra-short-term solar irradiance prediction method based on satellite cloud images in the embodiment of the present invention mainly includes the following steps:

[0047] Step S101: Obtaining the error evaluation index of the prediction period and parameter type of the prediction result relative to each pre-trained multi-layer perceptron model;

[0048] Step S102: selecting a pre-trained multi-layer perceptron model with the smallest error evaluation index to predict the prediction result;

[0049] The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

[0050] In one embodiment, data acquisition and preprocessing are first performed, including satellite cloud image acquisition, meteorological data collection, data cleaning and preprocessing.

[0051] Satellite cloud image acquisition: Utilize satellite remote sensing technology to regularly acquire satellite cloud image data for the target area. Select an appropriate satellite data source to ensure that the image resolution, coverage, and update frequency meet forecasting requirements. For example, choose meteorological satellite cloud imagery with high temporal and spatial resolution, with an update frequency set to every 15 minutes.

[0052] Meteorological data collection: Meteorological data related to solar irradiance, including upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and wind direction, are collected from meteorological monitoring stations, meteorological databases, and other channels. To ensure the accuracy and timeliness of meteorological data, the data collection frequency is consistent with the update frequency of satellite cloud images.

[0053] Data cleaning and preprocessing: Acquired satellite cloud images and meteorological data are cleaned to remove noise, missing values, and outliers. For satellite cloud images, preprocessing operations such as geometric correction and radiometric calibration are performed to improve image quality. For meteorological data, missing values ​​and outliers are addressed using methods such as interpolation and smoothing.

[0054] Then the satellite cloud image processing is carried out, including image enhancement, grayscale processing, study area determination and cloud feature extraction.

[0055] Image enhancement: Adjust the contrast and brightness of pre-processed satellite cloud images to enhance cloud features. Use algorithms such as histogram equalization and adaptive enhancement to improve the visual quality and analyzability of cloud images.

[0056] Grayscale processing: The enhanced color satellite cloud image is converted into a grayscale matrix to facilitate subsequent feature extraction and analysis. Grayscale processing uses a weighted average method to assign weights based on the importance of different color channels and calculate the grayscale value.

[0057] Determine the study area: Based on the geographic location of the PV plant, determine the study area in the grayscale matrix. A circular or rectangular area with a certain radius, centered on the PV plant, is defined as the study area.

[0058] Cloud feature extraction: Cloud features are extracted from the grayscale matrix of the study area, obtaining at least three cloud feature values. For example, cloud area, average cloud grayscale value, and cloud texture characteristics are extracted. Cloud area is calculated through binarization and connected region analysis; average cloud grayscale value is obtained by averaging the grayscale values ​​within the study area; and cloud texture characteristics are extracted using methods such as the grayscale co-occurrence matrix.

[0059] Then the model is constructed, including the construction of Model A, Model B and Model C.

[0060] Multilayer Perceptron (MLP) model selection: MLP is a feedforward artificial neural network with strong nonlinear mapping capabilities, making it suitable for solar irradiance prediction. Select an appropriate MLP model structure, including the number of neurons in the input layer, hidden layer, and output layer.

[0061] Model A Construction: Model A's inputs include meteorological elements such as upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and wind direction. These elements serve as neurons in the input layer, which, after undergoing nonlinear transformations in the hidden layer, output the predicted solar irradiance value. The number of neurons and layers in the hidden layer were determined through experimentation and optimization to improve the model's prediction accuracy.

[0062] Model B Construction: The input to Model B is the extracted cloud feature values. These cloud feature values ​​are used as input to neurons in the hidden layer, which also undergo nonlinear transformations to output predicted solar irradiance values. The structure and parameters of the hidden layer are adjusted based on the characteristics of the cloud features.

[0063] Model C Construction: The input to Model C is a combination of meteorological factors and cloud characteristics. These factors and cloud characteristics serve as neurons in the input layer. After processing in the hidden layer, the predicted solar irradiance value is output. The design of the hidden layer requires comprehensive consideration of the interaction between meteorological factors and cloud characteristics.

[0064] Then the model training is carried out, including data set division, model initialization and model training.

[0065] Dataset division: The collected historical solar irradiance data, corresponding meteorological data, and satellite cloud image data are divided into training, validation, and test sets according to a certain ratio of 70%, 15%, and 15%.

[0066] Model initialization: Randomly initialize the weights and biases of Model A, Model B, and Model C to prepare for the training process.

[0067] Training Process: The model is trained using the training set, using the backpropagation algorithm to adjust the model's weights and biases to minimize the error between the model's predictions and the actual values. During training, the model's performance is evaluated using the validation set to prevent overfitting. Based on the validation set's evaluation results, model hyperparameters such as the learning rate and the number of hidden layer neurons are adjusted.

[0068] Training termination condition: When the performance of the model on the validation set no longer improves or reaches the preset number of training rounds, the training process is stopped.

[0069] Then the irradiance prediction is carried out, including real-time data input, prediction process and prediction result output.

[0070] Real-time data input: After preprocessing and cloud feature extraction, the real-time meteorological data and satellite cloud image data are input into the trained models A, B, and C.

[0071] Prediction process: The model calculates the global horizontal irradiance (GHI), direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI) based on the input data and outputs the predicted values ​​of global horizontal irradiance (GHI), direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI) in the next two hours with a time resolution of 15 minutes.

[0072] Prediction result output: The model's prediction results are organized and formatted, and output in a visual manner, such as drawing line graphs, bar charts, etc., to facilitate user viewing and analysis.

[0073] Finally, model evaluation and model selection are carried out, including error evaluation indicator selection, model evaluation and optimal model selection.

[0074] In this embodiment, the error evaluation indicators include: mean absolute error, root mean square error, normalized root mean square error, correlation coefficient, determination coefficient, and mean absolute percentage error.

[0075] Evaluation process: Compare the model's predicted values ​​with the actual solar irradiance data and calculate the values ​​of various error evaluation indicators. Based on the values ​​of these indicators, the model's prediction accuracy and reliability are evaluated.

[0076] Optimal model selection: Based on the evaluation results, the model that performs best in the corresponding prediction task is selected for actual solar irradiance prediction. For example, when predicting GHI and DNI, the model with the smallest error index is selected; when predicting DHI, the optimal model is also selected based on the error evaluation results.

[0077] In one specific implementation, satellite cloud images of the area where a photovoltaic power station is located are first obtained, using specialized equipment to ensure stable data reception. Meteorological data covering upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and direction are also collected from surrounding meteorological monitoring stations. The satellite cloud images are processed using techniques such as image filtering, grayscale conversion, and normalization to remove noise and standardize the data scale. Meteorological data is cleaned to fill missing values ​​and correct outliers to ensure data accuracy. The cloud image study area is determined based on the location of the photovoltaic power station. A specific algorithm is used to extract cloud features from the regional grayscale matrix. For example, by calculating the statistical characteristics and spatial relationships of the pixel grayscale values ​​within the area, three cloud feature values ​​are obtained, reflecting cloud thickness, distribution, and texture information, respectively. These values ​​provide key information for subsequent model predictions. Three MLP models are constructed. Model A inputs meteorological elements, Model B inputs cloud features, and Model C inputs a combination of meteorological elements and cloud features. The models are trained using historical data, and the backpropagation algorithm is used to optimize model weights and biases. The number of hidden layers and neurons is determined based on the data size and problem complexity. Parameters are adjusted through multiple experiments to improve model prediction accuracy. Real-time satellite cloud images and meteorological data are processed and fed into a trained model to predict global horizontal irradiance (GHI), direct normal irradiance (DNI), and diffuse horizontal irradiance (DHI) at different time points within the next two hours. The prediction results are evaluated using six error metrics, including mean absolute error (MAE) and root mean square error (RMSE), and compared with actual irradiance data to determine model accuracy.

[0078] When predicting GHI and DNI, Model B performs better in terms of error metrics. For example, using a 30-minute forecast, Model B achieves lower NRMSE values ​​than Models A and C, as shown in Table 1, demonstrating its more accurate GHI and DNI forecasts. When predicting DHI, Model C achieves the highest accuracy within 30 minutes, better capturing variations in diffuse solar irradiance.

[0079] Table 1 shows the model evaluation results. As can be seen from Table 1, Global Horizontal Irradiance (GHI) predictions: Model B performs relatively well overall, with its NRMSE values ​​lower than those of Models A and C at most time points. For example, when forecasting 15 minutes ahead, Model B's NRMSE value is 0.10569, significantly lower than Model A's 0.12604 and Model C's 0.13676. However, when forecasting 30 minutes ahead, Model C's NRMSE value is 0.13500, lower than Model B's 0.14572, indicating that Model C has higher prediction accuracy. As the forecast interval increases, the NRMSE values ​​of all three models increase, indicating increased prediction difficulty.

[0080] Table 1

[0081]

[0082] Direct Normal Irradiance (DNI) Prediction: Model B has a significant advantage in DNI prediction, with its NRMSE values ​​being lower than those of Models A and C at all prediction time points. For example, when predicting 15 minutes in advance, Model B's NRMSE value is 0.12219, while Model A's is 0.17898 and Model C's is 0.18779. Similarly, the NRMSE values ​​of each model for DNI prediction gradually increase with increasing prediction time.

[0083] Diffuse Horizontal Irradiance (DHI) Forecasting: Each model has its own strengths and weaknesses in DHI prediction at different time points. When forecasting 15 minutes ahead, Model C achieved an NRMSE value of 0.13510, the lowest of the three models. Within 30 minutes, Model C's prediction accuracy was relatively higher, indicating that Model C, which combines meteorological factors and satellite cloud imagery, performs best in ultra-short-term DHI forecasting. However, as the forecast time exceeds 30 minutes, the NRMSE values ​​of each model show increasing difficulty, and the gap between the models gradually narrows.

[0084] Example 2

[0085] Based on the same inventive concept, the present invention also provides an ultra-short-term solar irradiance prediction device based on satellite cloud images, the ultra-short-term solar irradiance prediction device based on satellite cloud images comprising:

[0086] An acquisition module is used to obtain the error evaluation index of the prediction period and parameter type of the prediction result relative to each pre-trained multi-layer perceptron model;

[0087] A prediction module, configured to select a pre-trained multi-layer perceptron model with a minimum error evaluation index to predict the prediction result;

[0088] The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

[0089] Preferably, the error evaluation indicators include: mean absolute error, root mean square error, normalized root mean square error, correlation coefficient, determination coefficient, and mean absolute percentage error.

[0090] Furthermore, the training process of the pre-trained multi-layer perceptron model includes:

[0091] The training data is constructed using historical meteorological data and its corresponding solar irradiance data of different parameter types;

[0092] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0093] Furthermore, the meteorological data includes: upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and wind direction.

[0094] Preferably, the training process of the pre-trained multilayer perceptron model includes:

[0095] Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image;

[0096] The training data is constructed using cloud feature data from historical satellite cloud images and their corresponding solar irradiance data of different parameter types;

[0097] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0098] Furthermore, the cloud features include: cloud area, cloud average grayscale value, and cloud texture features.

[0099] Preferably, the training process of the pre-trained multilayer perceptron model includes:

[0100] Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image;

[0101] The training data is constructed using historical meteorological data, cloud feature data from historical satellite cloud images, and their corresponding solar irradiance data of different parameter types;

[0102] The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

[0103] Furthermore, the acquisition of historical satellite cloud image data includes: preprocessing the historical satellite cloud image data, and the preprocessing includes: image enhancement, grayscale processing and research area determination.

[0104] Example 3

[0105] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, the memory being used to store a computer program, the computer program including program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of an ultra-short-term solar irradiance prediction method based on satellite cloud images in the above embodiment.

[0106] Example 4

[0107] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space that stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a method for ultra-short-term solar irradiance prediction based on satellite cloud images in the above embodiment.

[0108] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for predicting ultra-short-term solar irradiance based on satellite cloud images, characterized in that: The method comprises: Obtain the error evaluation indicators of the prediction period and parameter type of the prediction results relative to each pre-trained multilayer perceptron model; Selecting a pre-trained multi-layer perceptron model with the smallest error evaluation index to predict the prediction result; The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

2. The method according to claim 1, wherein The error evaluation indicators include: mean absolute error, root mean square error, normalized root mean square error, correlation coefficient, determination coefficient, and mean absolute percentage error.

3. The method according to claim 1, wherein The training process of the pre-trained multilayer perceptron model includes: The training data is constructed using historical meteorological data and its corresponding solar irradiance data of different parameter types; The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

4. The method according to claim 3, wherein The meteorological data include: upper atmospheric irradiance, relative humidity, temperature, pressure, wind speed, and wind direction.

5. The method according to claim 1, wherein The training process of the pre-trained multilayer perceptron model includes: Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image; The training data is constructed using cloud feature data from historical satellite cloud images and their corresponding solar irradiance data of different parameter types; The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

6. The method according to claim 5, wherein The cloud features include: cloud area, cloud average gray value, and cloud texture features.

7. The method according to claim 1, wherein The training process of the pre-trained multilayer perceptron model includes: Acquire historical satellite cloud image data, and perform cloud feature extraction on the historical satellite cloud image data to obtain cloud feature data of the historical satellite cloud image; The training data is constructed using historical meteorological data, cloud feature data from historical satellite cloud images, and their corresponding solar irradiance data of different parameter types; The initial multilayer perceptron model is trained using the training data to obtain the pre-trained multilayer perceptron model.

8. The method according to claim 7, wherein The method of obtaining the historical satellite cloud image data includes: pre-processing the historical satellite cloud image data, wherein the pre-processing includes: image enhancement, grayscale processing and research area determination.

9. A device based on the ultra-short-term solar irradiance prediction method based on satellite cloud images according to any one of claims 1 to 8, characterized in that: The device comprises: An acquisition module is used to obtain the error evaluation index of the prediction period and parameter type of the prediction result relative to each pre-trained multi-layer perceptron model; A prediction module, configured to select a pre-trained multi-layer perceptron model with a minimum error evaluation index to predict the prediction result; The parameter types include: global horizontal irradiance, direct normal irradiance and diffuse horizontal irradiance.

10. A computer device, characterized in that: include: one or more processors; The processor is configured to execute one or more programs; When the one or more programs are executed by the one or more processors, the ultra-short-term solar irradiance prediction method based on satellite cloud images as described in any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, the ultra-short-term solar irradiance prediction method based on satellite cloud images as described in any one of claims 1 to 8 is implemented.

Images