Solar radiation prediction method, prediction device, electronic device and readable medium

Abstract

The present disclosure belongs to the field of applied meteorological technology, and specifically discloses a solar radiation prediction method, prediction device, electronic device and readable medium. The method includes: obtaining the first historical solar radiation data corresponding to the observation site of the target area within the first historical time and the first historical remote sensing data corresponding to the target area within the first historical time; inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain the historical solar radiation inversion data corresponding to the preset adjacent site of the observation site within the first historical time; constructing feature data according to the first historical solar radiation data and the historical solar radiation inversion data; inputting the feature data into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time. The method can predict solar radiation according to the temporal and spatial variation law of solar radiation data, thereby effectively improving the prediction accuracy of solar radiation.

Description

Technical Field

[0001] The present disclosure relates to the field of applied meteorological technology, and in particular to a solar radiation prediction method, a prediction device, an electronic device and a readable medium. Background Art

[0002] As a renewable and clean energy, solar energy plays an important role in electricity production. The efficient operation of solar power generation systems is heavily dependent on the prediction accuracy of solar radiation. In the field of photovoltaic power generation, the volatility and intermittency of solar radiation intensity directly affect the stability of power output and the dispatch efficiency of the power grid, making the new power system based on new energy sources face increasingly higher risks.

[0003] In traditional solar radiation prediction methods, predictions are usually made based on ground-based cloud image modeling. Specifically, it is based on cloud image observations from an all-sky imager, and by identifying and processing cloud images and predicting cloud movement, the relationship between clouds and solar radiation is established, thereby predicting solar radiation values. Although the temporal and spatial resolution of ground-based cloud images is very high, the observation range of the all-sky imager is limited. Therefore, for solar radiation predictions at different times, clouds may have moved out of the detection range of the all-sky imager, resulting in low accuracy in solar radiation predictions. Summary of the invention

[0004] To this end, the present disclosure provides a solar radiation prediction method, a prediction device, an electronic device and a readable medium to solve the problem of low solar radiation prediction accuracy in the prior art.

[0005] In order to achieve the above object, the present disclosure provides a method for predicting solar radiation in a first aspect, the method comprising:

[0006] Acquire first historical solar radiation data corresponding to an observation station in a target area within a first historical time and first historical remote sensing data corresponding to the target area within the first historical time;

[0007] Inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain historical solar radiation inversion data corresponding to a preset adjacent site of the observation site within the first historical time;

[0008] Constructing characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data;

[0009] The characteristic data is input into a solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0010] In an optional implementation, the number of the preset irradiance inversion models is multiple, and the multiple preset irradiance inversion models correspond one-to-one to multiple preset weather types, then the inputting of the first historical solar radiation data and the first historical remote sensing data into the preset irradiance inversion model includes:

[0011] Dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types;

[0012] From the first historical remote sensing data, respectively determine the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain a plurality of historical remote sensing sub-data corresponding to the plurality of solar radiation sub-data;

[0013] According to the first preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, select the historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data fragments corresponding to the multiple solar radiation sub-data;

[0014] Each solar radiation sub-data and the historical remote sensing sub-data fragment corresponding to each solar radiation sub-data are respectively input into a preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data.

[0015] In an optional implementation, selecting, from the historical remote sensing sub-data corresponding to each solar radiation sub-data according to the preset weather type corresponding to each solar radiation sub-data, a historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data comprises:

[0016] Determine, according to the association relationship between the preset weather type and the remote sensing attribute type, an associated remote sensing attribute type associated with the first preset weather type corresponding to each solar radiation sub-data;

[0017] From the historical remote sensing sub-data corresponding to each solar radiation sub-data, the historical remote sensing sub-data matching the associated remote sensing attribute type is selected to obtain the historical remote sensing sub-data fragment.

[0018] In an optional implementation, the historical remote sensing sub-data includes: a plurality of remote sensing attribute data corresponding to a plurality of remote sensing attribute types; the association relationship between the preset weather type and the remote sensing attribute type is determined by:

[0019] For any weather type among the plurality of preset weather types, calculating a correlation coefficient between the solar radiation sub-data corresponding to the any weather type and each remote sensing attribute data in the historical remote sensing sub-data corresponding to the any weather type, to obtain a plurality of correlation coefficients corresponding to the plurality of remote sensing attribute data;

[0020] For any weather type, the association relationship between any weather type and each remote sensing attribute type is determined according to a plurality of correlation coefficients corresponding to any weather type.

[0021] In an optional implementation, dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types includes:

[0022] For the solar radiation data at different times included in the first historical solar radiation data, calculating the clear sky index corresponding to the solar radiation data at each time;

[0023] Determining a preset weather type corresponding to the solar radiation data at each moment according to a clear sky index corresponding to the solar radiation data at each moment;

[0024] The solar radiation data at different times corresponding to the same preset weather type are divided into solar radiation sub-data corresponding to the same preset weather type.

[0025] In an optional implementation, the characteristic data includes characteristic data corresponding to the observation site and characteristic data corresponding to the preset adjacent site, and constructing the characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data includes:

[0026] Generate first weather attribute data corresponding to the first historical solar radiation data according to the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data; and generate second weather attribute data corresponding to the historical solar radiation inversion data according to the preset weather type corresponding to the solar radiation data at each moment in the historical solar radiation inversion data;

[0027] constructing characteristic data corresponding to the observation site according to the first historical solar radiation data and the first weather attribute data;

[0028] Based on the historical solar radiation inversion data and the second weather attribute data, characteristic data corresponding to the preset adjacent site are constructed.

[0029] In an optional implementation, after obtaining the predicted solar radiation data of the target area within the target time, the method further includes:

[0030] Acquire the solar radiation observation data corresponding to the observation site of the target area within the target time and the remote sensing data corresponding to the target area within the target time;

[0031] Calculating a data deviation value between the solar radiation observation data and the solar radiation prediction data;

[0032] In the case where the data deviation value meets the preset deviation condition, the solar radiation observation data and the remote sensing data corresponding to the target area within the target time are input into the preset irradiance inversion model to obtain the solar radiation inversion data of the preset adjacent station within the target time;

[0033] Incremental characteristic data are constructed according to the solar radiation observation data and the solar radiation inversion data of the preset adjacent stations within the target time, and model parameters of the solar radiation prediction model are optimized according to the incremental characteristic data.

[0034] A second aspect of the present disclosure provides a solar radiation prediction device, the device comprising:

[0035] An acquisition module, used to acquire first historical solar radiation data corresponding to an observation station of a target area within a first historical time and first historical remote sensing data corresponding to the target area within the first historical time;

[0036] A generation module, used for inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain historical solar radiation inversion data corresponding to a preset adjacent station of the observation station within the first historical time;

[0037] A construction module, used to construct characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data;

[0038] The prediction module is used to input the characteristic data into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0039] A third aspect of the present disclosure provides an electronic device, including:

[0040] one or more processors;

[0041] A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned solar radiation prediction method;

[0042] One or more I / O interfaces are connected between the processor and the memory and are configured to implement information interaction between the processor and the memory.

[0043] A fourth aspect of the present disclosure provides a computer-readable medium having a computer program stored thereon, wherein the program implements the above-mentioned solar radiation prediction method when executed by a processor.

[0044] The present disclosure has the following advantages:

[0045] In the solar radiation prediction method provided by the embodiment of the present disclosure, first, the first historical solar radiation data corresponding to the observation station in the target area within the first historical time and the first historical remote sensing data corresponding to the target area within the first historical time are obtained; secondly, by inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model, the model can generate the historical solar radiation inversion data corresponding to the preset adjacent station of the observation station within the first historical time according to the first historical solar radiation data and the first historical remote sensing data, that is, the historical solar radiation data of the preset adjacent station having spatial correlation with the observation station is generated; finally, feature data is constructed according to the first historical solar radiation data and the historical solar radiation inversion data, and the feature data is input into the solar radiation prediction model, so that the solar radiation prediction model can predict the solar radiation data of the target area within the target time according to the change law of the first historical solar radiation data and the historical solar radiation inversion data in the time dimension and the spatial change law between the first historical solar radiation data and the historical solar radiation inversion data, thereby effectively improving the prediction accuracy of the solar radiation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide further understanding of the present disclosure and constitute a part of the specification. They are used to explain the present disclosure together with the following detailed description, but do not constitute a limitation of the present disclosure.

[0047] Figure 1 A flow chart of a solar radiation prediction method provided in an embodiment of the present disclosure;

[0048] Figure 2 A schematic diagram of a solar radiation prediction method provided by an embodiment of the present disclosure;

[0049] Figure 3 Schematic diagram of solar radiation prediction data predicted by the solar radiation prediction model and actual solar radiation data;

[0050] Figure 4 A block diagram of a solar radiation prediction device provided by an embodiment of the present disclosure;

[0051] Figure 5 A block diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0052] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0053] As used in this disclosure, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0054] The terms used in the present disclosure are only used to describe specific embodiments and are not intended to limit the present disclosure.As used in the present disclosure, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0055] When the terms “comprising” and / or “made of…” are used in the present disclosure, it specifies the existence of the stated features, integers, steps, operations, elements and / or components, but does not exclude the existence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0056] The embodiments described in the present disclosure may be described with reference to plan views and / or cross-sectional views by way of ideal schematic diagrams of the present disclosure. Therefore, the exemplary illustrations may be modified according to manufacturing techniques and / or tolerances.

[0057] Unless otherwise defined, all terms (including technical and scientific terms) used in this disclosure have the same meaning as those commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined in this disclosure.

[0058] In view of this, the embodiments of the present disclosure provide a solar radiation prediction method, a prediction device, an electronic device, and a readable medium.

[0059] In the solar radiation prediction method provided by the embodiment of the present disclosure, first, the first historical solar radiation data corresponding to the observation station in the target area within the first historical time and the first historical remote sensing data corresponding to the target area within the first historical time are obtained; secondly, by inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model, the model can generate the historical solar radiation inversion data corresponding to the preset adjacent station of the observation station within the first historical time according to the first historical solar radiation data and the first historical remote sensing data, that is, the historical solar radiation data of the preset adjacent station having spatial correlation with the observation station is generated; finally, feature data is constructed according to the first historical solar radiation data and the historical solar radiation inversion data, and the feature data is input into the solar radiation prediction model, so that the solar radiation prediction model can predict the solar radiation data of the target area within the target time according to the change law of the first historical solar radiation data and the historical solar radiation inversion data in the time dimension and the spatial change law between the first historical solar radiation data and the historical solar radiation inversion data, thereby effectively improving the prediction accuracy of the solar radiation data.

[0060] In a first aspect, an embodiment of the present disclosure provides a method for predicting solar radiation.

[0061] Figure 1 A flowchart of a solar radiation prediction method provided in an embodiment of the present disclosure, referring to Figure 1 , the method comprising:

[0062] Step S110: Acquire first historical solar radiation data corresponding to the observation site of the target area within the first historical time and first historical remote sensing data corresponding to the target area within the first historical time.

[0063] The observation site in the target area refers to a solar radiation observation site that actually exists in the target area, such as a photovoltaic power station, which can observe the surface solar radiation value at a fixed time granularity to obtain the observed historical solar radiation data, for example, the surface solar radiation value is obtained by observing every 15 minutes. The number of observation sites can be one or more, and the embodiments of the present disclosure are not limited to this.

[0064] The first historical time is an observation time interval selected from the historical observation time of the observation site according to actual prediction requirements. For example, the observation site in the target area observed and obtained historical solar radiation data from January 2015 to September 2024. The first historical time can be September 2023 to September 2024, May to September 2024, etc., and the embodiments of the present disclosure are not limited to this.

[0065] The first historical remote sensing data is the remote sensing data of the target area observed by the geosynchronous satellite in the first historical time. The remote sensing data is multispectral image remote sensing data, which may include remote sensing attribute data of various bands including visible light, near infrared, and infrared, and is used to analyze cloud cover and aerosol distribution.

[0066] The first historical remote sensing data can be obtained by performing time-space matching on the observed remote sensing data. Specifically, first, the full amount of historical remote sensing data obtained by geosynchronous satellite observation is obtained. Then, the full amount of historical remote sensing data is screened according to the first historical time and the geographical location information corresponding to the target area to obtain the first historical remote sensing data corresponding to the target area in the first historical time.

[0067] It should be noted that there is a time correspondence between the first historical remote sensing data and the first historical solar radiation data. For example, the first historical solar radiation data includes solar radiation data corresponding to every 15 minutes from May to September 2024, and the first historical remote sensing data also includes remote sensing data corresponding to every 15 minutes from May to September 2024. In other words, at the same time, there is only one first historical remote sensing data and one first historical solar radiation data.

[0068] It should also be noted that after obtaining the first historical solar radiation data and the first historical remote sensing data, the first historical solar radiation data can be subjected to standardization processing such as quality control and outlier removal, and the first historical remote sensing data can be subjected to denoising, correction, and geo-registration processing to improve the data quality of the first historical solar radiation data and the first historical remote sensing data.

[0069] Step S120: Input the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain the historical solar radiation inversion data corresponding to the preset adjacent stations of the observation station within the first historical time.

[0070] Among them, the preset irradiance inversion model is used to predict the historical solar radiation inversion data corresponding to the preset adjacent stations of the observation station within the first historical time.

[0071] It should be noted that the preset irradiance inversion model is obtained by training the initial irradiance inversion model, wherein the initial irradiance inversion model is any machine learning prediction model, such as a gradient boosting regression model (GBR), and the embodiments of the present disclosure do not limit this.

[0072] Furthermore, when training the initial irradiance inversion model, the training can be performed based on historical solar radiation training data and historical remote sensing training data corresponding to multiple observation sites.

[0073] There needs to be an adjacent position relationship between the multiple observation sites. The multiple observation sites can be located in the target area or in the non-target area, which is not limited in the embodiment of the present disclosure.

[0074] The historical solar radiation training data is the solar radiation data observed by multiple observation sites within a preset historical time. The historical remote sensing training data is the remote sensing data corresponding to multiple observation sites within a preset historical time, which is obtained through geosynchronous satellite observations.

[0075] Specifically, during training, at least one observation site may be selected from a plurality of observation sites as an initial observation site, and the remaining observation sites may be selected as target observation sites for solar radiation data to be predicted.

[0076] Thus, the historical solar radiation training data corresponding to the initial observation site, the historical remote sensing training data corresponding to the initial observation site, and the historical remote sensing training data corresponding to the target observation site are input into the initial irradiance inversion model to obtain the historical solar radiation prediction data corresponding to the target observation site output by the initial irradiance inversion model; by comparing the historical solar radiation prediction data corresponding to the target observation site with the historical solar radiation training data corresponding to the target observation site, the model parameters of the initial irradiance inversion model are updated according to the comparison results to obtain the preset irradiance inversion model.

[0077] The preset adjacent site refers to a preset logical site that has an adjacent position relationship with the observation site in the target area. It should be noted that the preset adjacent site is a non-actual site, that is, a virtual site.

[0078] Correspondingly, the historical solar radiation inversion data is the solar radiation data corresponding to the preset adjacent stations in the first historical time predicted by the preset irradiance inversion model, that is, the historical solar radiation inversion data is not the actual solar radiation observation data, but the predicted data.

[0079] It should be noted that there is a time correspondence between the historical solar radiation inversion data and the first historical solar radiation data. For example, the first historical solar radiation data contains the solar radiation data corresponding to the observation site every 15 minutes from May to September 2024, and the historical solar radiation inversion data also contains the solar radiation inversion data corresponding to the preset adjacent site every 15 minutes from May to September 2024.

[0080] It should also be noted that the number of preset adjacent sites may be one or more, and the embodiment of the present disclosure does not limit this.

[0081] In addition, the geographical location of the preset adjacent site may be determined according to the geographical location of the target area and the adaptability of the geographical location of the observation site, which is not limited in this embodiment of the present disclosure.

[0082] For example, the preset direction and the preset distance can be set to determine the geographical location of the preset adjacent site according to the geographical location of the target area and the geographical location of the observation site. For example, in the geographical location of the target area, the geographical location of the observation site is used as the starting point, and the geographical location of the preset adjacent site is determined at each preset distance along the preset direction.

[0083] In addition, the geographical locations of the target area except for the observation sites may be directly divided into the geographical locations of the preset adjacent sites.

[0084] In an optional implementation, in order to enable the preset irradiance inversion model to effectively predict the historical solar radiation inversion data corresponding to the preset adjacent station within the first historical time, the site remote sensing data corresponding to the observation station and the adjacent remote sensing data corresponding to the preset adjacent station can be extracted from the first historical remote sensing data, so that the preset irradiance inversion model can predict the historical solar radiation inversion data corresponding to the preset adjacent station within the first historical time based on the site remote sensing data, the adjacent remote sensing data and the first historical solar radiation data.

[0085] Correspondingly, inputting the first historical solar radiation data and the first historical remote sensing data into the preset irradiance inversion model may include: determining the site remote sensing data corresponding to the observation site and the adjacent remote sensing data corresponding to the preset adjacent site from the first historical remote sensing data according to the geographical location of the observation site and the geographical location of the preset adjacent site; inputting the first historical solar radiation data, the site remote sensing data and the adjacent remote sensing data into the preset irradiance inversion model to obtain the historical solar radiation inversion data corresponding to the preset adjacent site of the observation site within the first historical time.

[0086] The site remote sensing data and the adjacent remote sensing data are both screened from the first historical remote sensing data.

[0087] There is a time correspondence between the site remote sensing data and the adjacent remote sensing data, that is, at the same moment, there are site remote sensing data and adjacent remote sensing data that are uniquely corresponding to each other at that moment.

[0088] Similarly, there is a time correspondence between the first historical solar radiation data and the site remote sensing data, that is, at the same moment, there are first historical solar radiation data and site remote sensing data that are uniquely corresponding to each other at that moment.

[0089] Therefore, by inputting the first historical solar radiation data, site remote sensing data and adjacent remote sensing data into the preset irradiance inversion model, on the one hand, the preset irradiance inversion model can learn the correlation between the site remote sensing data and the adjacent remote sensing data, for example, the change law between the site remote sensing data and the adjacent remote sensing data corresponding to the same time.

[0090] On the other hand, the preset irradiance inversion model can also learn the correlation between the first historical solar radiation data and the site remote sensing data, for example, the change law between the first historical solar radiation data and the site remote sensing data corresponding to the same time. Thus, the preset irradiance inversion model predicts the historical solar radiation inversion data corresponding to the preset adjacent site based on the correlation between the site remote sensing data and the adjacent remote sensing data and the correlation between the first historical solar radiation data and the site remote sensing data.

[0091] Step S130: construct characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data.

[0092] The characteristic data is the input data required by the solar radiation prediction model for solar radiation prediction. The characteristic data includes characteristic data corresponding to the observation site and characteristic data corresponding to the preset adjacent site.

[0093] It should be noted that the feature data may also include other attribute data, for example, location attribute data for describing the location relationship between the observation site and the preset adjacent sites, which is not limited in the embodiments of the present disclosure.

[0094] In an optional implementation, in order to enable the solar radiation prediction model to accurately predict solar radiation data under complex weather conditions, feature data can be constructed based on the first historical solar radiation data, historical solar radiation inversion data and weather attribute data, so that the solar radiation prediction model can effectively combine weather factors when making predictions.

[0095] Correspondingly, the characteristic data includes characteristic data corresponding to the observation site and characteristic data corresponding to the preset adjacent site. Constructing the characteristic data based on the first historical solar radiation data and the historical solar radiation inversion data may include: generating first weather attribute data corresponding to the first historical solar radiation data based on the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data; and generating second weather attribute data corresponding to the historical solar radiation inversion data based on the preset weather type corresponding to the solar radiation data at each moment in the historical solar radiation inversion data; constructing characteristic data corresponding to the observation site based on the first historical solar radiation data and the first weather attribute data; and constructing characteristic data corresponding to the preset adjacent site based on the historical solar radiation inversion data and the second weather attribute data.

[0096] Among them, the preset weather types may include sunny, cloudy, rainy, snowy and other weather types.

[0097] It should be noted that the clear sky index can be calculated based on the data value of the solar radiation data at each moment and the solar radiation baseline value, and then the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data and the historical solar radiation inversion data can be determined based on the clear sky index.

[0098] It should also be noted that the first weather attribute data is the weather attribute data corresponding to the first historical solar radiation data. The first weather attribute data includes the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data.

[0099] Similarly, the second weather attribute data is the weather attribute data corresponding to the historical solar radiation inversion data. The second weather attribute data includes the preset weather type corresponding to the solar radiation data at each moment in the historical solar radiation inversion data.

[0100] The characteristic data corresponding to the observation site is obtained by combining the first historical solar radiation data and the first weather attribute data. The characteristic data corresponding to the preset adjacent site is obtained by combining the historical solar radiation inversion data and the second weather attribute data.

[0101] Therefore, by inputting the characteristic data corresponding to the observation site and the characteristic data corresponding to the preset adjacent site into the solar radiation prediction model, the model can not only combine the solar radiation data of the observation site and the preset adjacent site when making solar radiation predictions, but also combine the weather attribute data, thereby improving the prediction accuracy under complex weather conditions.

[0102] Step S140: Input the characteristic data into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0103] The solar radiation prediction model is obtained by training the initial solar radiation prediction model. The initial solar radiation prediction model can be a machine learning model with spatiotemporal prediction characteristics, such as a graph convolutional neural network model, a spatiotemporal graph convolutional neural network model, etc., which is not limited in the embodiments of the present disclosure.

[0104] The target time is the time interval corresponding to the solar radiation data to be predicted, such as the next hour, four hours, or one day.

[0105] It should be noted that the solar radiation prediction model can only predict the solar radiation prediction data of the observation site within the target time, and can also predict the solar radiation prediction data of the preset adjacent sites within the target time. The embodiments of the present disclosure are not limited to this.

[0106] It should also be noted that when the solar radiation prediction model only predicts the solar radiation prediction data of the observation site within the target time, the solar radiation prediction data of the observation site within the target time can be directly used as the solar radiation prediction data of the target area within the target time.

[0107] When the solar radiation prediction model predicts the solar radiation prediction data of the observation site and the preset adjacent sites within the target time, the solar radiation prediction data of the target area within the target time can be obtained based on the solar radiation prediction data of the observation site and the preset adjacent sites within the target time.

[0108] In an optional implementation, the solar radiation prediction model can be trained in the following manner: obtain the second historical solar radiation data corresponding to the observation site in the target area within the second historical time and the second historical remote sensing data corresponding to the target area within the second historical time; input the second historical solar radiation data and the second historical remote sensing data into a preset irradiance inversion model to obtain the second historical solar radiation inversion data corresponding to the preset adjacent site of the observation site within the second historical time; construct training data based on the second historical solar radiation data and the second historical solar radiation inversion data; train the initial solar radiation prediction model based on the training data to obtain the trained solar radiation prediction model.

[0109] The second historical time is an observation time interval selected from the historical observation time of the observation site according to actual training requirements. The second historical time may have a time intersection with the first historical time, or may be different from the first historical time, which is not limited in the embodiment of the present disclosure.

[0110] The second historical solar radiation data is historical solar radiation data corresponding to the second historical time obtained by screening out historical solar radiation data observed at observation stations in the target area.

[0111] The second historical remote sensing data is the remote sensing data of the target area observed by the geosynchronous satellite within the second historical time.

[0112] By inputting the second historical solar radiation data and the second historical remote sensing data into the preset irradiance inversion model, the second historical solar radiation inversion data corresponding to the preset adjacent station within the second historical time predicted by the preset irradiance inversion model can be obtained.

[0113] It should be noted that when constructing training data based on the second historical solar radiation data and the second historical solar radiation inversion data, the second historical solar radiation data can be used as the training data corresponding to the observation site, and the second historical solar radiation inversion data can be used as the training data corresponding to the preset adjacent site. In addition, relevant attribute data such as weather attribute training data can also be added to the training data, and the embodiments of the present disclosure are not limited to this.

[0114] It should also be noted that when training the initial solar radiation prediction model based on the training data, the model parameters can be updated only based on the solar radiation prediction result data of the observation site within the target time output by the initial solar radiation prediction model, and the model parameters can also be updated based on the solar radiation prediction result data of the preset adjacent site within the target time output by the initial solar radiation prediction model. The embodiments of the present disclosure are not limited to this.

[0115] In the solar radiation prediction method provided by the embodiment of the present disclosure, first, the first historical solar radiation data corresponding to the observation station in the target area within the first historical time and the first historical remote sensing data corresponding to the target area within the first historical time are obtained; secondly, by inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model, the model can generate the historical solar radiation inversion data corresponding to the preset adjacent station of the observation station within the first historical time according to the first historical solar radiation data and the first historical remote sensing data, that is, the historical solar radiation data of the preset adjacent station having spatial correlation with the observation station is generated; finally, feature data is constructed according to the first historical solar radiation data and the historical solar radiation inversion data, and the feature data is input into the solar radiation prediction model, so that the solar radiation prediction model can predict the solar radiation data of the target area within the target time according to the change law of the first historical solar radiation data and the historical solar radiation inversion data in the time dimension and the spatial change law between the first historical solar radiation data and the historical solar radiation inversion data, thereby effectively improving the prediction accuracy of the solar radiation data.

[0116] In addition, those skilled in the art may also make various modifications and variations to the method disclosed herein:

[0117] In an optional implementation, in order to improve the accuracy of the historical solar radiation inversion data output by the preset irradiance inversion model under complex weather conditions, corresponding preset irradiance inversion models can be set for different preset weather types, thereby generating historical solar radiation inversion data by dividing the first historical solar radiation data into solar radiation sub-data corresponding to different preset weather types, and inputting the solar radiation sub-data and the corresponding historical remote sensing sub-data into the preset irradiance inversion model corresponding to the preset weather type.

[0118] Correspondingly, there are multiple preset irradiance inversion models, and the multiple preset irradiance inversion models correspond one-to-one to the multiple preset weather types. Then, the first historical solar radiation data and the first historical remote sensing data are input into the preset irradiance inversion model, including: dividing the first historical solar radiation data into multiple solar radiation sub-data corresponding to the multiple preset weather types; from the first historical remote sensing data, respectively determining the historical remote sensing sub-data corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data corresponding to the multiple solar radiation sub-data; according to the first preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, selecting the historical remote sensing sub-data slice associated with the first preset weather type corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data slices corresponding to the multiple solar radiation sub-data; respectively inputting each solar radiation sub-data and the historical remote sensing sub-data slice corresponding to each solar radiation sub-data into the preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data.

[0119] Among them, by determining the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data, the first historical solar radiation data is divided into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types.

[0120] The historical remote sensing sub-data corresponding to the solar radiation sub-data refers to the first historical remote sensing data corresponding to the same time as the solar radiation sub-data.

[0121] For example, the solar radiation sub-data contains the solar radiation data from 09:15 to 10:00 on January 8, 2024, then the corresponding historical remote sensing sub-data also contains the remote sensing data from 09:15 to 10:00 on January 8, 2024.

[0122] The historical remote sensing sub-data include a plurality of remote sensing attribute data corresponding to a plurality of remote sensing attribute types.

[0123] Among them, the remote sensing attribute type refers to the various channel types of geosynchronous satellite observations, such as channel BAND1 to channel BAND16, which correspond to related attributes such as cloud thickness, water vapor, and surface albedo.

[0124] The remote sensing attribute data corresponding to the remote sensing attribute type refers to the channel data of each channel obtained by geosynchronous satellite observation.

[0125] Correspondingly, the historical remote sensing sub-data fragments are composed of the remote sensing attribute data associated with the first preset weather type corresponding to the solar radiation sub-data contained in the historical remote sensing sub-data.

[0126] The remote sensing attribute data associated with the first preset weather type means that, under the first preset weather type, there is a correlation between the remote sensing attribute data and the solar radiation sub-data corresponding to the first preset weather type.

[0127] It should be noted that when the solar radiation sub-data and the historical remote sensing sub-data slices corresponding to the solar radiation sub-data are input into the preset irradiance inversion model corresponding to the first preset weather type corresponding to the solar radiation sub-data, the historical remote sensing sub-data slices can be further divided into historical remote sensing sub-data slices corresponding to the observation site and historical remote sensing sub-data slices corresponding to the preset adjacent site.

[0128] It should also be noted that after obtaining multiple historical solar radiation inversion sub-data output by the preset irradiance inversion model corresponding to different preset weather types, the different historical solar radiation inversion sub-data can be merged in chronological order into historical solar radiation inversion data that has a time correspondence with the first historical solar radiation data.

[0129] By selecting historical remote sensing sub-data fragments associated with the first preset weather type corresponding to the solar radiation sub-data from the historical remote sensing sub-data, and inputting the solar radiation sub-data and the historical remote sensing sub-data fragments into the preset irradiance inversion model corresponding to the first preset weather type, the preset irradiance inversion model can perform targeted historical solar radiation inversion data prediction according to the preset weather type, which not only effectively improves the prediction accuracy of the historical solar radiation inversion data, but also reduces the computational complexity of the prediction.

[0130] In an optional implementation, in order to improve the matching of the selected historical remote sensing sub-data fragments, historical remote sensing sub-data fragments can be selected from the historical remote sensing sub-data corresponding to the solar radiation sub-data according to the association between the preset weather type and the remote sensing attribute type.

[0131] Correspondingly, according to the preset weather type corresponding to each solar radiation sub-data, a historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data is selected from the historical remote sensing sub-data corresponding to each solar radiation sub-data, including: determining the associated remote sensing attribute type associated with the first preset weather type corresponding to each solar radiation sub-data according to the association relationship between the preset weather type and the remote sensing attribute type; and selecting the historical remote sensing sub-data matching the associated remote sensing attribute type from the historical remote sensing sub-data corresponding to each solar radiation sub-data to obtain the historical remote sensing sub-data fragment.

[0132] Among them, in the association relationship between the preset weather types and the remote sensing attribute types, a certain preset weather type may be associated with multiple remote sensing attribute types.

[0133] The association between the preset weather type and the remote sensing attribute type means that, under the preset weather type, the remote sensing attribute data corresponding to the remote sensing attribute type is correlated with the solar radiation data.

[0134] Thus, by calculating the correlation between the solar radiation sub-data corresponding to the preset weather type and the remote sensing attribute data corresponding to the remote sensing attribute type in the historical remote sensing sub-data, it can be determined whether there is a correlation between the preset weather type and the remote sensing attribute type.

[0135] Furthermore, according to the first preset weather type corresponding to the solar radiation sub-data, the associated remote sensing attribute type associated with the first preset weather type can be determined from the association relationship between the preset weather type and the remote sensing attribute type.

[0136] For example, when the first preset weather type is sunny, the associated remote sensing attribute types associated therewith are channel BAND1, channel BAND2, channel BAND3, channel BAND6, channel BAND4, and channel BAND12.

[0137] Correspondingly, by selecting historical remote sensing sub-data that matches the associated remote sensing attribute type from the historical remote sensing sub-data corresponding to the solar radiation sub-data, historical remote sensing sub-data fragments can be obtained.

[0138] For example, in the historical remote sensing sub-data, the historical remote sensing sub-data corresponding to channel BAND1, channel BAND2, channel BAND3, channel BAND6, channel BAND4, and channel BAND12 are selected to obtain the historical remote sensing sub-data fragments.

[0139] In an optional implementation, in order to improve the accuracy of determining the association between preset weather types and remote sensing attribute types, the correlation coefficient between the solar radiation sub-data corresponding to the preset weather type and the remote sensing attribute data corresponding to each remote sensing attribute type in the historical remote sensing sub-data can be calculated, thereby determining the association between the preset weather type and each remote sensing attribute type through the correlation coefficient.

[0140] Correspondingly, the historical remote sensing sub-data include: multiple remote sensing attribute data corresponding to multiple remote sensing attribute types; the association relationship between the preset weather type and the remote sensing attribute type is determined in the following manner: for any weather type among the multiple preset weather types, calculate the correlation coefficient between the solar radiation sub-data corresponding to any weather type and each remote sensing attribute data in the historical remote sensing sub-data corresponding to any weather type, and obtain multiple correlation coefficients corresponding to the multiple remote sensing attribute data; for any weather type, determine the association relationship between any weather type and each remote sensing attribute type based on the multiple correlation coefficients corresponding to any weather type.

[0141] The historical remote sensing sub-data corresponding to any weather type refers to the historical remote sensing sub-data that has a time correspondence relationship with the solar radiation sub-data corresponding to any weather type.

[0142] The correlation coefficient between the solar radiation sub-data and each remote sensing attribute data in the historical remote sensing sub-data can be calculated by a preset correlation coefficient algorithm. For example, the correlation coefficient between the solar radiation sub-data and each remote sensing attribute data in the historical remote sensing sub-data can be calculated by a Pearson correlation coefficient algorithm.

[0143] It should be noted that when determining the association between any weather type and each remote sensing attribute type based on multiple correlation coefficients corresponding to any weather type, a preset correlation coefficient threshold can be set so that when the correlation coefficient corresponding to the remote sensing attribute type reaches the preset correlation coefficient threshold, it is determined that it has an association relationship with any weather type.

[0144] In an optional implementation, in order to further improve the accuracy of the association between the preset weather type and each remote sensing attribute type, the characteristic importance of the remote sensing attribute data corresponding to each remote sensing attribute type in the historical remote sensing sub-data can be calculated, thereby determining the association between the preset weather type and each remote sensing attribute type through the characteristic importance.

[0145] Correspondingly, determining the association relationship between any weather type and each remote sensing attribute type based on multiple correlation coefficients corresponding to any weather type may include: determining the association relationship between any weather type and each remote sensing attribute type based on multiple correlation coefficients and multiple feature importances corresponding to any weather type; wherein the multiple feature importances corresponding to any weather type are obtained in the following manner: calculating the feature importance of the data features of each remote sensing attribute data in the historical remote sensing sub-data corresponding to any weather type to the solar radiation sub-data corresponding to any weather type, to obtain multiple feature importances corresponding to any weather type.

[0146] Among them, the feature importance of the data features of each remote sensing attribute data in the historical remote sensing sub-data to the solar radiation sub-data corresponding to any weather type refers to the contribution of the data features of the remote sensing attribute data to the prediction of the solar radiation sub-data corresponding to any weather type.

[0147] Among them, the feature importance can be calculated by a preset feature importance algorithm. For example, by using the machine learning algorithm XGBoost of the gradient boosting decision tree, the weight, gain, coverage and other index values ​​corresponding to the data features of each remote sensing attribute data are calculated when the model predicts the solar radiation sub-data, and then the feature importance corresponding to the data features of each remote sensing attribute data is obtained according to the above index values.

[0148] It should be noted that when determining the association between any weather type and each remote sensing attribute type based on multiple correlation coefficients and multiple feature importances corresponding to any weather type, corresponding weight values ​​can be set for the correlation coefficients and feature importances, so that the correlation coefficients and feature importances corresponding to each remote sensing attribute type are cumulatively calculated according to the weight values, and the association between any weather type and each remote sensing attribute type is determined based on the calculation results.

[0149] In addition, corresponding preset correlation coefficient thresholds and preset feature importance thresholds may be set for the correlation coefficient and feature importance, respectively, so that when the correlation coefficient corresponding to the remote sensing attribute type reaches the preset correlation coefficient threshold and the feature importance corresponding to the remote sensing attribute type reaches the preset feature importance threshold, it is determined that the remote sensing attribute type has an association with any weather type, and the embodiments of the present disclosure do not limit this.

[0150] In an optional implementation, in order to accurately divide the first historical solar radiation data into multiple solar radiation sub-data corresponding to multiple preset weather types, the clear sky index corresponding to the solar radiation data at each moment in the first historical solar radiation data can be calculated, and then the first historical solar radiation data can be divided into multiple solar radiation sub-data corresponding to multiple preset weather types according to the clear sky index corresponding to the solar radiation data at each moment.

[0151] Correspondingly, the first historical solar radiation data is divided into multiple solar radiation sub-data corresponding to multiple preset weather types, including: for the solar radiation data at different times contained in the first historical solar radiation data, calculating the clear sky index corresponding to the solar radiation data at each time; determining the preset weather type corresponding to the solar radiation data at each time according to the clear sky index corresponding to the solar radiation data at each time; and dividing the solar radiation data at different times corresponding to the same preset weather type into the solar radiation sub-data corresponding to the same preset weather type.

[0152] The clear sky index can be obtained by calculating the ratio between the solar radiation data value at each moment and the solar radiation reference value, wherein the solar radiation reference value can be the solar radiation data value of the outer horizontal plane.

[0153] It should be noted that when determining the preset weather type corresponding to the solar radiation data at each moment based on the clear sky index corresponding to the solar radiation data at each moment, the clear sky index corresponding to the solar radiation data at different moments can be clustered, so as to determine the preset weather type corresponding to the solar radiation data at each moment based on the cluster cluster to which the clear sky index corresponding to the solar radiation data at each moment belongs.

[0154] It should also be noted that when clustering the clear sky index corresponding to the solar radiation data at different times, any preset clustering algorithm can be used, such as the K-means clustering algorithm, the density clustering algorithm, etc., and the embodiments of the present disclosure do not limit this.

[0155] In the embodiment of the present disclosure, the clear sky index is calculated for the solar radiation data, and thus the preset weather type corresponding to the solar radiation data is determined according to the clear sky index, thereby improving the accuracy of the division of the solar radiation sub-data.

[0156] In an optional implementation, in order to improve the prediction accuracy of the solar radiation prediction model, incremental feature data can be constructed based on the corresponding solar radiation observation data of the observation stations in the target area within the target time, thereby optimizing the model parameters of the solar radiation prediction model based on the incremental feature data.

[0157] Correspondingly, after predicting the solar radiation prediction data of the target area within the target time, it also includes: obtaining the solar radiation observation data corresponding to the observation site in the target area within the target time and the remote sensing data corresponding to the target area within the target time; calculating the data deviation value between the solar radiation observation data and the solar radiation prediction data; when the data deviation value meets the preset deviation condition, the solar radiation observation data and the remote sensing data corresponding to the target area within the target time are input into the preset irradiance inversion model to obtain the solar radiation inversion data of the preset adjacent site within the target time; constructing incremental feature data according to the solar radiation observation data and the solar radiation inversion data of the preset adjacent site within the target time, and optimizing the model parameters of the solar radiation prediction model according to the incremental feature data.

[0158] Among them, the solar radiation observation data corresponding to the observation station in the target area within the target time is the solar radiation data actually observed by the observation station within the target time.

[0159] The remote sensing data corresponding to the target area within the target time is the remote sensing data of the target area observed by the geosynchronous satellite within the target time.

[0160] Among them, the data deviation value between the solar radiation observation data and the solar radiation prediction data reflects the data deviation of the solar radiation prediction data predicted by the model.

[0161] The preset deviation condition may be that the data deviation value reaches a preset data deviation threshold. Thus, when the data deviation value satisfies the preset deviation condition, it indicates that the data deviation of the solar radiation prediction data is large, and at this time, it is necessary to construct incremental feature data based on the solar radiation observation data, so as to optimize the model parameters of the solar radiation prediction model based on the incremental feature data.

[0162] It should be noted that, in addition to the solar radiation observation data and the solar radiation inversion data of the preset adjacent stations within the target time, the incremental feature data may also include other attribute data, such as weather attribute data, which is not limited in this embodiment of the present disclosure.

[0163] It should also be noted that when optimizing the model parameters of the solar radiation prediction model based on the incremental feature data, the incremental feature data can be added to the training data for incremental training, thereby updating the model parameters of the solar radiation prediction model.

[0164] In an optional implementation, when the data deviation value does not meet the preset deviation condition, the model parameters of the solar radiation prediction model are optimized according to the solar radiation observation data through an online gradient descent algorithm.

[0165] When the data deviation value does not meet the preset deviation condition, it indicates that the data deviation of the solar radiation prediction data is small at this time. Therefore, the prediction accuracy of the solar radiation prediction model can be improved by simply fine-tuning the model parameters of the solar radiation prediction model according to the solar radiation observation data.

[0166] For ease of understanding, the following uses an example to describe the technical implementation details in the above embodiment:

[0167] In the related art, there are two main methods for predicting solar radiation:

[0168] First, traditional physical models and statistical methods often find it difficult to capture subtle and rapid changes in clouds, and cannot meet the new energy industry's demand for high-precision and high-frequency forecasts.

[0169] The second is the prediction method based on ground-based cloud image modeling, which mostly uses machine learning methods such as optical flow or convolutional neural networks, bidirectional long short-term memory networks, etc., based on the cloud image observation of the all-sky imager, through the recognition and processing of the cloud image and the prediction of cloud movement, the relationship between clouds and radiation is established, and finally the horizontal solar radiation intensity is predicted. Although the temporal and spatial resolution of the ground-based cloud image is very high, the observation range of the all-sky imager is limited (for example, the effective detection range of TSI-880 is 5km). For a forecast 4 hours in advance, the clouds may have already moved out of the detection range of the imager. Therefore, the all-sky imager is only suitable for 0-15 minutes or 0-1 hour forecasts.

[0170] It can be seen that the existing solar radiation prediction method has the problem of low accuracy. In view of this, the embodiment of the present disclosure provides a solar radiation prediction method to solve the problem of low accuracy in the existing solar radiation prediction.

[0171] Figure 2A schematic diagram showing a method for predicting solar radiation provided by an embodiment of the present disclosure is shown, referring to Figure 2 , the method comprising:

[0172] Step S201: Acquire first historical solar radiation data corresponding to an observation station in a target area within a first historical time and first historical remote sensing data corresponding to the target area within a first historical time.

[0173] Among them, the first historical time is an observation time interval selected from the historical observation time of the observation site according to actual prediction needs. For example, the first historical time can be 2023.

[0174] The first historical solar radiation data is obtained by observing an observation site such as a photovoltaic power station. The first historical remote sensing data is collected by a geosynchronous satellite such as the Himawari-8 satellite, which may include multiple remote sensing attribute data, such as remote sensing attribute data corresponding to 16 channels.

[0175] Exemplarily, the first historical solar radiation data and the first historical remote sensing data can be obtained in the following manner:

[0176] Firstly, the remote sensing attribute data corresponding to the 16 channels of the Himawari-8 satellite in the first historical time and the solar radiation observation data observed by the photovoltaic power station in the first historical time were collected; then, the remote sensing attribute data were denoised, corrected and geo-referenced, and the solar radiation observation data were quality controlled and outliers were eliminated according to the standards such as Solar Energy Resource Assessment Method (GB / T 37526-2019) and Technical Specifications for Review and Correction of Meteorological Observation Data of Photovoltaic Power Stations (GB / T 42477-2023), so as to obtain the first historical remote sensing data and the first historical solar radiation data corresponding to the first historical time.

[0177] Step S202: Divide the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types.

[0178] In an optional implementation method, the clear sky index corresponding to the solar radiation data at each moment in the first historical solar radiation data can be calculated, so as to determine the preset weather type corresponding to the solar radiation data at each moment according to the clear sky index, and the solar radiation data at different moments corresponding to the same preset weather type can be divided into solar radiation sub-data corresponding to the same preset weather type.

[0179] Among them, the clear sky index can be obtained by calculating the ratio of the data value of the solar radiation data and the solar radiation baseline value.

[0180] Specifically, the solar radiation benchmark value may be the solar radiation value of an extraterrestrial horizontal surface, which may be calculated according to B.2.1 in the Solar Energy Resource Assessment Method (GB / T37526—2019).

[0181] Thus, the clear sky index is obtained by calculating the ratio of the solar radiation data observed by the photovoltaic power station to the solar radiation on the outer horizontal surface.

[0182] Furthermore, when determining the preset weather type corresponding to the solar radiation data at each moment based on the clear sky index, the K-means clustering method can be used to cluster the clear sky index, and then the clear sky index can be divided into corresponding preset weather types such as sunny, cloudy, and rainy according to the clustering results.

[0183] Step S203: from the first historical remote sensing data, respectively determine the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain a plurality of historical remote sensing sub-data corresponding to the plurality of solar radiation sub-data.

[0184] The historical remote sensing sub-data corresponding to the solar radiation sub-data refers to the historical remote sensing sub-data having a time corresponding relationship with the solar radiation sub-data.

[0185] Step S204: According to the first preset weather type corresponding to each solar radiation sub-data, select the historical remote sensing sub-data slice associated with the first preset weather type corresponding to each solar radiation sub-data from the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain multiple historical remote sensing sub-data slices corresponding to multiple solar radiation sub-data.

[0186] The historical remote sensing sub-data includes remote sensing attribute data corresponding to multiple remote sensing attribute types. The remote sensing attribute type may be a channel attribute type, such as channel BAND1 and channel BAND2. Correspondingly, the remote sensing attribute data is the remote sensing data corresponding to the channel attribute type, such as channel BAND1 remote sensing data.

[0187] The historical remote sensing sub-data fragments are historical remote sensing sub-data selected from the historical remote sensing sub-data and matching the associated remote sensing attribute types.

[0188] Specifically, the associated remote sensing attribute type refers to the remote sensing attribute type associated with the first preset weather type, which can be determined by the association relationship between the preset weather type and the remote sensing attribute type.

[0189] In an optional implementation, for any preset weather type, the correlation coefficient between the solar radiation sub-data corresponding to the weather type and each remote sensing attribute data in the historical remote sensing sub-data corresponding to the weather type can be calculated, thereby determining the association relationship between any weather type and each remote sensing attribute type based on multiple correlation coefficients corresponding to any weather type.

[0190] In an optional implementation, for any preset weather type, the feature importance of each remote sensing attribute data in the historical remote sensing sub-data corresponding to any weather type to the solar radiation sub-data corresponding to any weather type can be calculated, thereby determining the correlation between the preset weather type and each remote sensing attribute type through multiple feature importances corresponding to any weather type.

[0191] In addition, for any preset weather type, the correlation relationship between the weather type and each remote sensing attribute type can be determined based on multiple correlation coefficients and multiple feature importances corresponding to the weather type.

[0192] For example, according to any preset weather type, the correlation coefficient between the solar radiation sub-data corresponding to the preset weather type and each remote sensing attribute data in the historical remote sensing sub-data can be obtained through the Pearson correlation coefficient, and the feature importance of each remote sensing attribute data in the historical remote sensing sub-data corresponding to the preset weather type to the solar radiation sub-data can be obtained through XGBoost feature importance analysis. Thus, the preset weather type is sorted according to the correlation coefficient and feature importance corresponding to the preset weather type, and the association relationship between the preset weather type and each remote sensing attribute type is determined according to the sorting result.

[0193] For example, when the preset weather type is sunny, the remote sensing attribute types associated with it are channel BAND1, channel BAND2, channel BAND3, channel BAND6, channel BAND4, and channel BAND12; when the preset weather type is cloudy, the remote sensing attribute types associated with it are channel BAND2, channel BAND3, channel BAND1, and channel BAND12; when the preset weather type is rainy, the remote sensing attribute types associated with it are channel BAND1, channel BAND2, channel BAND3, channel BAND5, channel BAND6, channel BAND4, and channel BAND11.

[0194] Step S205: Input each solar radiation sub-data and the historical remote sensing sub-data fragment corresponding to each solar radiation sub-data into the preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data, and obtain the historical solar radiation inversion data corresponding to the preset adjacent stations of the observation station within the first historical time.

[0195] Among them, corresponding preset irradiance inversion models are established for different preset weather types such as sunny, cloudy, rainy, etc. The preset irradiance inversion model can be any trained machine learning prediction model, such as a gradient boosting regression model.

[0196] Therefore, according to the data screening results, by inputting the solar radiation sub-data and the historical remote sensing sub-data fragments obtained after screening into the preset irradiance inversion model of the corresponding weather type, the model can invert the historical solar radiation inversion data of the preset adjacent stations of the observation station based on the data characteristics of the historical remote sensing sub-data fragments and the data characteristics of the solar radiation sub-data.

[0197] Moreover, the historical solar radiation inversion data inverted by the preset irradiance inversion model corresponds to the same time resolution as the solar radiation sub-data, so that it can meet the needs of ultra-short-term solar radiation prediction.

[0198] Step S206: constructing characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data.

[0199] Step S207: input the characteristic data into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0200] The solar radiation prediction model is obtained by training the initial solar radiation prediction model. The initial solar radiation prediction model is a machine learning model with spatial correlation and time series adaptability, such as a convolutional neural network model.

[0201] Exemplarily, the remote sensing attribute data corresponding to the 16 channels of the geosynchronous satellite in the second historical time, such as from 2020 to 2022, and the solar radiation observation data observed by the observation station in the second historical time can be input into the preset irradiance inversion model to obtain the solar radiation inversion data corresponding to the preset adjacent stations of the observation station in the second historical time, and a historical data set is constructed based on the solar radiation observation data and the solar radiation inversion data, so that the initial solar radiation prediction model such as the convolutional neural network model in the figure can be trained based on the historical data set.

[0202] Specifically, the historical data set can be divided into a training set and a test set according to a preset ratio, such as 8:2, so that the initial solar radiation prediction model is trained and tested according to the training set and the test set to obtain a trained solar radiation prediction model.

[0203] Furthermore, during training, evaluation indicators can also be used to analyze the solar radiation data prediction effect of the solar radiation prediction model.

[0204] Exemplarily, Table 1 shows the evaluation results of the solar radiation prediction model based on graph convolutional neural network on the solar radiation prediction data of a photovoltaic power station in the next 4 hours using the correlation coefficient, mean absolute error and root mean square error, wherein the prediction time step interval of the model is 15 minutes.

[0205] Table 1 Mean values ​​of evaluation indexes for each step prediction result of solar radiation prediction model of a photovoltaic power station

[0206]

[0207] After obtaining the trained solar radiation prediction model, the characteristic data corresponding to the latest time, that is, the first historical time, can be input into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0208] Among them, the target time can be adaptively adjusted according to the forecast demand. For example, when the forecast demand is an ultra-short-term forecast, the target time can be 1 hour, 4 hours, etc. When the forecast demand is a short-term forecast, the target time can be 1 day, 3 days, etc.

[0209] Figure 3 Schematic diagram of solar radiation prediction data predicted by the solar radiation prediction model and actual solar radiation data.

[0210] Reference Figure 3 , which shows the actual solar radiation data from 9:15 to 13:00 on May 15, 2023 and the solar radiation forecast data predicted by the solar radiation forecast model. The root mean square error between the two is 27.2W / m 2 , the mean absolute error is 24.5W / m 2 , the maximum absolute error is 43.6W / m 2 .

[0211] It should be noted that in order to improve the prediction accuracy of the solar radiation prediction model, the model parameters of the solar radiation prediction model can be calibrated or dynamically adjusted based on the real-time observation results of multi-spectral image remote sensing data of geosynchronous satellites and surface solar radiation observation data of observation sites.

[0212] For example, after obtaining the latest solar radiation observation data, the model parameters of the solar radiation prediction model are fine-tuned using the online gradient descent method.

[0213] For another example, when the data deviation between the solar radiation observation data and the solar radiation prediction data is large, it indicates that there is a systematic deviation in the solar radiation prediction model. The incremental training method can be used to add the latest solar radiation observation data to the historical data set to form an extended data set, and the extended data set can be used to retrain the model to update the model parameters of the solar radiation prediction model.

[0214] In the solar radiation prediction method provided in the embodiment of the present disclosure, by integrating satellite remote sensing data and artificial intelligence technology, an accurate prediction of solar radiation intensity is achieved, which can not only capture the temporal and spatial variation of solar radiation, but also effectively handle the fusion problem of multi-source data, improve adaptability under complex weather conditions, and thus improve the prediction efficiency and accuracy of solar radiation. In addition, the solar radiation prediction method provided in the embodiment of the present disclosure comprehensively utilizes satellite remote sensing data and ground observation data of solar radiation, which can significantly improve the accuracy and timeliness of solar radiation prediction, provide a scientific basis for the efficient operation of photovoltaic power stations and grid load scheduling, and has important economic and social benefits.

[0215] The step division of the above methods is only for the purpose of clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0216] In a second aspect, an embodiment of the present disclosure provides a solar radiation prediction device.

[0217] Figure 4 A block diagram of a solar radiation prediction device 40 provided in an embodiment of the present disclosure is shown in FIG. Figure 4 , the device comprises:

[0218] An acquisition module 41 is used to acquire first historical solar radiation data corresponding to an observation station of a target area within a first historical time and first historical remote sensing data corresponding to the target area within the first historical time;

[0219] A generating module 42, configured to input the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model, and obtain historical solar radiation inversion data corresponding to a preset adjacent station of the observation station within the first historical time;

[0220] A construction module 43, configured to construct characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data;

[0221] The prediction module 44 is used to input the characteristic data into the solar radiation prediction model to predict the solar radiation prediction data of the target area within the target time.

[0222] In an optional implementation, the number of the preset irradiance inversion models is multiple, and the multiple preset irradiance inversion models correspond one-to-one to the multiple preset weather types. Then the generation module 42 inputs the first historical solar radiation data and the first historical remote sensing data into the preset irradiance inversion model, including:

[0223] Dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types;

[0224] From the first historical remote sensing data, respectively determine the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain a plurality of historical remote sensing sub-data corresponding to the plurality of solar radiation sub-data;

[0225] According to the first preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, select the historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data fragments corresponding to the multiple solar radiation sub-data;

[0226] Each solar radiation sub-data and the historical remote sensing sub-data fragment corresponding to each solar radiation sub-data are respectively input into a preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data.

[0227] In an optional implementation, selecting, from the historical remote sensing sub-data corresponding to each solar radiation sub-data according to the preset weather type corresponding to each solar radiation sub-data, a historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data comprises:

[0228] Determine, according to the association relationship between the preset weather type and the remote sensing attribute type, an associated remote sensing attribute type associated with the first preset weather type corresponding to each solar radiation sub-data;

[0229] From the historical remote sensing sub-data corresponding to each solar radiation sub-data, the historical remote sensing sub-data matching the associated remote sensing attribute type is selected to obtain the historical remote sensing sub-data fragment.

[0230] In an optional implementation, the historical remote sensing sub-data includes: a plurality of remote sensing attribute data corresponding to a plurality of remote sensing attribute types; the association relationship between the preset weather type and the remote sensing attribute type is determined by:

[0231] For any weather type among the plurality of preset weather types, calculating a correlation coefficient between the solar radiation sub-data corresponding to the any weather type and each remote sensing attribute data in the historical remote sensing sub-data corresponding to the any weather type, to obtain a plurality of correlation coefficients corresponding to the plurality of remote sensing attribute data;

[0232] For any weather type, the association relationship between any weather type and each remote sensing attribute type is determined according to a plurality of correlation coefficients corresponding to any weather type.

[0233] In an optional implementation, dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types includes:

[0234] For the solar radiation data at different times included in the first historical solar radiation data, calculating the clear sky index corresponding to the solar radiation data at each time;

[0235] Determining a preset weather type corresponding to the solar radiation data at each moment according to a clear sky index corresponding to the solar radiation data at each moment;

[0236] The solar radiation data at different times corresponding to the same preset weather type are divided into solar radiation sub-data corresponding to the same preset weather type.

[0237] In an optional implementation, the characteristic data includes characteristic data corresponding to the observation site and characteristic data corresponding to the preset adjacent site, and the construction module 43 constructs the characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data, including:

[0238] Generate first weather attribute data corresponding to the first historical solar radiation data according to the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data; and generate second weather attribute data corresponding to the historical solar radiation inversion data according to the preset weather type corresponding to the solar radiation data at each moment in the historical solar radiation inversion data;

[0239] constructing characteristic data corresponding to the observation site according to the first historical solar radiation data and the first weather attribute data;

[0240] Based on the historical solar radiation inversion data and the second weather attribute data, characteristic data corresponding to the preset adjacent site are constructed.

[0241] In an optional implementation, after the prediction data of solar radiation of the target area within the target time is obtained, the device is further used to:

[0242] Acquire the solar radiation observation data corresponding to the observation site of the target area within the target time and the remote sensing data corresponding to the target area within the target time;

[0243] Calculating a data deviation value between the solar radiation observation data and the solar radiation prediction data;

[0244] In the case where the data deviation value meets the preset deviation condition, the solar radiation observation data and the remote sensing data corresponding to the target area within the target time are input into the preset irradiance inversion model to obtain the solar radiation inversion data of the preset adjacent station within the target time;

[0245] Incremental characteristic data are constructed according to the solar radiation observation data and the solar radiation inversion data of the preset adjacent stations within the target time, and model parameters of the solar radiation prediction model are optimized according to the incremental characteristic data.

[0246] The functions or modules included in the device provided in the embodiments of the present disclosure can be used to execute the method described in the first aspect of the method embodiment above. Its specific implementation and technical effects can be referred to the description of the method embodiment above. For the sake of brevity, they will not be repeated here.

[0247] It should be noted that all modules involved in this embodiment are logic modules. In practical applications, a logic unit may be a physical unit, or a part of a physical unit, or may be implemented as a combination of multiple physical units. In addition, in order to highlight the innovative part of the present disclosure, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by the present disclosure, but this does not mean that there are no other units in this embodiment.

[0248] Reference Figure 5 , an embodiment of the present disclosure provides an electronic device, comprising:

[0249] One or more processors 501;

[0250] A memory 502 having one or more programs stored thereon, which, when executed by one or more processors, enables the one or more processors to implement any one of the above solar radiation prediction methods;

[0251] One or more I / O interfaces 503 are connected between the processor and the memory and are configured to implement information exchange between the processor and the memory.

[0252] Among them, the processor 501 is a device with data processing capabilities, including but not limited to a central processing unit (CPU) and the like; the memory 502 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); the I / O interface (read-write interface) 503 is connected between the processor 501 and the memory 502, and can realize information interaction between the processor 501 and the memory 502, including but not limited to a data bus (Bus) and the like.

[0253] In some embodiments, the processor 501 , the memory 502 , and the I / O interface 503 are connected to each other through a bus, and further connected to other components of the computing device.

[0254] This embodiment also provides a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the solar radiation prediction method provided in this embodiment is implemented. To avoid repeated description, the specific steps of the solar radiation prediction method are not repeated here.

[0255] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods invented above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is well known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0256] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0257] Those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present embodiment and to form different embodiments.

[0258] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present disclosure, but the present disclosure is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and substance of the present disclosure, and these modifications and improvements are also considered to be within the scope of protection of the present disclosure.

Claims

1. A method for predicting solar radiation, characterized in that: The method comprises: Acquire first historical solar radiation data corresponding to an observation station in a target area within a first historical time and first historical remote sensing data corresponding to the target area within the first historical time; Inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain historical solar radiation inversion data corresponding to a preset adjacent site of the observation site within the first historical time; Constructing characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data; Inputting the characteristic data into a solar radiation prediction model to predict solar radiation prediction data for the target area within a target time; Wherein, the first historical remote sensing data includes the site remote sensing data corresponding to the observation site and the adjacent remote sensing data corresponding to the preset adjacent site; The solar radiation prediction model is obtained by training a machine learning model with spatiotemporal prediction characteristics; Wherein, the number of the preset irradiance inversion models is multiple, and the multiple preset irradiance inversion models correspond one-to-one to the multiple preset weather types, and the inputting of the first historical solar radiation data and the first historical remote sensing data into the preset irradiance inversion model includes: Dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types; From the first historical remote sensing data, respectively determine the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain a plurality of historical remote sensing sub-data corresponding to the plurality of solar radiation sub-data; According to the first preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, select the historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data fragments corresponding to the multiple solar radiation sub-data; Each solar radiation sub-data and the historical remote sensing sub-data fragment corresponding to each solar radiation sub-data are respectively input into a preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data.

2. The method for predicting solar radiation according to claim 1, characterized in that: The selecting, according to the preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, a historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data, comprises: Determine, according to the association relationship between the preset weather type and the remote sensing attribute type, an associated remote sensing attribute type associated with the first preset weather type corresponding to each solar radiation sub-data; From the historical remote sensing sub-data corresponding to each solar radiation sub-data, the historical remote sensing sub-data matching the associated remote sensing attribute type is selected to obtain the historical remote sensing sub-data fragment.

3. The method for predicting solar radiation according to claim 2, characterized in that: The historical remote sensing sub-data includes: a plurality of remote sensing attribute data corresponding to a plurality of remote sensing attribute types; the association relationship between the preset weather type and the remote sensing attribute type is determined in the following manner: For any weather type among the plurality of preset weather types, calculating a correlation coefficient between the solar radiation sub-data corresponding to the any weather type and each remote sensing attribute data in the historical remote sensing sub-data corresponding to the any weather type, to obtain a plurality of correlation coefficients corresponding to the plurality of remote sensing attribute data; For any weather type, the association relationship between any weather type and each remote sensing attribute type is determined according to a plurality of correlation coefficients corresponding to any weather type.

4. The method for predicting solar radiation according to any one of claims 1 to 3, characterized in that: The dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types includes: For the solar radiation data at different times included in the first historical solar radiation data, calculating the clear sky index corresponding to the solar radiation data at each time; Determining a preset weather type corresponding to the solar radiation data at each moment according to a clear sky index corresponding to the solar radiation data at each moment; The solar radiation data at different times corresponding to the same preset weather type are divided into solar radiation sub-data corresponding to the same preset weather type.

5. The method for predicting solar radiation according to any one of claims 1 to 3, characterized in that: The characteristic data includes characteristic data corresponding to the observation site and characteristic data corresponding to the preset adjacent site, and constructing the characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data includes: Generate first weather attribute data corresponding to the first historical solar radiation data according to the preset weather type corresponding to the solar radiation data at each moment in the first historical solar radiation data; and generate second weather attribute data corresponding to the historical solar radiation inversion data according to the preset weather type corresponding to the solar radiation data at each moment in the historical solar radiation inversion data; constructing characteristic data corresponding to the observation site according to the first historical solar radiation data and the first weather attribute data; Based on the historical solar radiation inversion data and the second weather attribute data, characteristic data corresponding to the preset adjacent site are constructed.

6. The method for predicting solar radiation according to claim 1, characterized in that: After obtaining the predicted solar radiation data of the target area within the target time, the method further includes: Acquire the solar radiation observation data corresponding to the observation site of the target area within the target time and the remote sensing data corresponding to the target area within the target time; Calculating a data deviation value between the solar radiation observation data and the solar radiation prediction data; In the case where the data deviation value meets the preset deviation condition, the solar radiation observation data and the remote sensing data corresponding to the target area within the target time are input into the preset irradiance inversion model to obtain the solar radiation inversion data of the preset adjacent station within the target time; Incremental characteristic data are constructed according to the solar radiation observation data and the solar radiation inversion data of the preset adjacent stations within the target time, and model parameters of the solar radiation prediction model are optimized according to the incremental characteristic data.

7. A solar radiation prediction device, characterized in that: The device comprises: An acquisition module, used to acquire first historical solar radiation data corresponding to an observation station of a target area within a first historical time and first historical remote sensing data corresponding to the target area within the first historical time; A generation module, used for inputting the first historical solar radiation data and the first historical remote sensing data into a preset irradiance inversion model to obtain historical solar radiation inversion data corresponding to a preset adjacent station of the observation station within the first historical time; A construction module, used to construct characteristic data according to the first historical solar radiation data and the historical solar radiation inversion data; A prediction module, used for inputting the characteristic data into a solar radiation prediction model to predict the solar radiation prediction data of the target area within a target time; Wherein, the first historical remote sensing data includes the site remote sensing data corresponding to the observation site and the adjacent remote sensing data corresponding to the preset adjacent site; The solar radiation prediction model is obtained by training a machine learning model with spatiotemporal prediction characteristics; Wherein, the number of the preset irradiance inversion models is multiple, and the multiple preset irradiance inversion models correspond one-to-one to the multiple preset weather types, and the inputting of the first historical solar radiation data and the first historical remote sensing data into the preset irradiance inversion model includes: Dividing the first historical solar radiation data into a plurality of solar radiation sub-data corresponding to a plurality of preset weather types; From the first historical remote sensing data, respectively determine the historical remote sensing sub-data corresponding to each solar radiation sub-data, and obtain a plurality of historical remote sensing sub-data corresponding to the plurality of solar radiation sub-data; According to the first preset weather type corresponding to each solar radiation sub-data, from the historical remote sensing sub-data corresponding to each solar radiation sub-data, select the historical remote sensing sub-data fragment associated with the first preset weather type corresponding to each solar radiation sub-data, to obtain multiple historical remote sensing sub-data fragments corresponding to the multiple solar radiation sub-data; Each solar radiation sub-data and the historical remote sensing sub-data fragment corresponding to each solar radiation sub-data are respectively input into a preset irradiance inversion model corresponding to the first preset weather type corresponding to each solar radiation sub-data.

8. An electronic device, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6; One or more I / O interfaces are connected between the processor and the memory and are configured to implement information interaction between the processor and the memory.

9. A computer readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images