CN113504582A

CN113504582A - Photovoltaic power station solar radiation short-term forecasting method based on satellite radiation product

Info

Publication number: CN113504582A
Application number: CN202111050835.4A
Authority: CN
Inventors: 徐丽娜; 申彦波; 李忠; 姚锦烽; 叶冬; 胡玥明; 谷新波; 叶虎; 冯震; 李利秋
Original assignee: Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center
Current assignee: Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center
Priority date: 2021-06-24
Filing date: 2021-09-08
Publication date: 2021-10-15
Anticipated expiration: 2041-09-08
Also published as: CN113311512A; CN113504582B

Abstract

The present disclosure provides a photovoltaic power station solar radiation short-term forecasting method, device and computer readable medium based on satellite radiation products, the method comprises: determining a space window of the position to be forecasted according to the position to be forecasted; determining a total irradiance correction scheme of the position to be forecasted corresponding to the space window; determining the ground observation total irradiance and the satellite observation total irradiance of the adopted space window according to the total irradiance correction scheme of the position to be predicted, and establishing a total irradiance correction PDF model of the position to be predicted; correcting the satellite observation total irradiance of the position to be forecasted to obtain a corresponding satellite correction total irradiance data set; and correcting a total irradiance data set according to the satellite of the position to be forecasted, and forecasting the total irradiance of the position to be forecasted.

Description

Photovoltaic power station solar radiation short-term forecasting method based on satellite radiation product

Technical Field

The disclosure relates to the field of weather forecasting, and in particular relates to a photovoltaic power station solar radiation short-term forecasting method and device based on satellite radiation products and a computer readable medium.

Background

The installed capacity of the photovoltaic power generation system is continuously enlarged, so that the influence of the photovoltaic power generation system on the power system is more obvious. The photovoltaic array power generation forecasting plays an important role in reducing the influence of the uncertainty of the photovoltaic power generation on the safe and stable operation of the power grid. The fundamental reason for the uncertainty of the photovoltaic power generation power is the intermittency, randomness and fluctuation of solar radiation, so that the improvement of the accuracy of solar radiation prediction and photovoltaic power generation power prediction is one of the directions of research of scholars at home and abroad. Researches show that a solar radiation forecasting mode combining power and machine learning, namely numerical mode forecasting and a machine learning model, is the best way for improving the solar radiation forecasting accuracy at present. However, the machine learning model is established depending on the measured solar radiation data around the photovoltaic power station.

At present, the number of nationwide ground solar radiation observation stations is small, the distribution is uneven, the radiation observation instruments built by photovoltaic power stations are neglected for calibration, the data quality is relatively poor, the application effect is poor, and the shortage of measured data becomes a pain point in short-term photovoltaic power generation prediction. Therefore, under the condition that the existing ground solar radiation observation is difficult to meet the solar radiation forecast requirement, the application of the satellite remote sensing data becomes an effective way for making up the deficiency.

However, due to the influence of the sensitivity of the instrument and the change of response characteristics, the satellite inversion data not only has obvious non-independent system errors, but also has large error change along with time and space, and a common linear method is difficult to correct.

Disclosure of Invention

The disclosure provides a photovoltaic power station solar radiation short-term forecasting method, device and computer readable medium based on satellite radiation products.

As a first aspect of the present disclosure, there is provided a photovoltaic power station solar radiation short-term forecasting method based on satellite radiation products, including:

determining a spatial window of the position to be forecasted according to the position to be forecasted, wherein the spatial window comprises a position relevant to the position to be forecasted;

determining a total irradiance correction scheme of the position to be forecasted corresponding to the space window;

determining the ground observation total irradiance and the satellite observation total irradiance of the adopted space window according to the total irradiance correction scheme of the position to be predicted, and establishing a total irradiance correction Probability Density Function (PDF) model of the position to be predicted;

correcting a PDF model according to the total irradiance of the position to be forecasted, and correcting the satellite observation total irradiance of the position to be forecasted to obtain a corresponding satellite correction total irradiance data set;

correcting a total irradiance data set according to the satellite of the position to be forecasted, and forecasting the total irradiance of the position to be forecasted;

the correction scheme of the total irradiance of the position to be forecasted comprises the value ranges of the ground observed total irradiance and the satellite observed total irradiance of the space window in a preset time window, and an algorithm for establishing a PDF (Portable document Format) model for correcting the total irradiance of the position to be forecasted.

As an alternative embodiment, the determining the spatial window of the position to be forecasted includes:

when the data reliability of the ground observation total irradiance of the position to be forecasted meets a preset threshold value, taking the position to be forecasted as a space window;

and the data reliability of the ground observed total irradiance of the position to be forecasted represents the degree that the ground observed total irradiance of the position to be forecasted can reflect the actual total irradiance.

As an alternative embodiment, when the spatial window includes a position to be forecasted, the determining the total terrestrial observed irradiance and the total satellite observed irradiance of the spatial window to be used includes:

and determining the total terrestrial observed irradiance and the total satellite observed irradiance of the position to be forecasted in the preset time window.

As an optional implementation, the establishing a total irradiance correction PDF model of the location to be forecasted includes:

obtaining a Cumulative probability density Distribution Function (CDF) of the ground observed total irradiance of the position to be forecasted and a CDF of the satellite observed total irradiance according to the ground observed total irradiance and the satellite observed total irradiance of the position to be forecasted at multiple times;

calculating the difference value (Δ r (P) between the total observed terrestrial irradiance and the total observed terrestrial irradiance of the position to be forecasted according to the CDF of the total observed terrestrial irradiance and the CDF of the total observed satellite irradiance of the position to be forecasted and by calculating the corresponding total observed terrestrial irradiance and total observed satellite irradiance when the cumulative probability density value is P, wherein the cumulative probability density value P is a real number which is more than or equal to 0 and less than or equal to 1;

the total irradiance correction PDF model of the position to be forecasted is used for: and correcting the satellite observation total irradiance of the position to be forecasted under the corresponding accumulated probability density value P according to the difference value ar (P) to obtain the corresponding satellite correction total irradiance.

when the data reliability of the total ground observation irradiance of the position to be forecasted does not meet a preset threshold value, determining at least one matched ground radiation observation station according to the distance between the plurality of ground radiation observation stations and the position to be forecasted, determining the position comprising the matched ground radiation observation stations as an observation position matched with the position to be forecasted, and forming a space window by the matched observation position and the position to be forecasted;

As an optional implementation, when the spatial window includes the matched observed position and the position to be forecasted, the determining the total terrestrial observed irradiance and the total satellite observed irradiance of the spatial window to be used includes:

and determining the satellite observation total irradiance of the position to be forecasted in a preset time window, the ground observation total irradiance of the matched observation position and the satellite observation total irradiance of the matched observation position.

The ground observation total irradiance of the matched observation position is a set of ground observation total irradiance of each ground radiation observation station in the matched observation position;

and the satellite observation total irradiance of the matched observation position is a set of satellite observation total irradiance of each ground radiation observation station in the matched observation position.

respectively obtaining the CDF of the ground observed total irradiance distribution of the matched observed position, the CDF of the satellite observed total irradiance distribution of the matched observed position and the CDF of the satellite observed total irradiance distribution of the position to be forecasted according to the ground observed total irradiance of the matched observed position and the satellite observed total irradiance of the matched observed position at multiple times and the satellite observed total irradiance of the position to be forecasted;

according to the CDF of the ground observed total irradiance distribution of the matched observed position, the CDF of the satellite observed total irradiance distribution of the matched observed position and the CDF of the satellite observed total irradiance distribution of the position to be forecasted, calculating a difference value r1(P) between the total ground observed irradiance of the matched observed position and the total ground observed irradiance of the matched observed position, a difference value r2(P) between the total satellite observed irradiance of the position to be forecasted and the total satellite observed irradiance of the matched observed position by calculating the total ground observed irradiance of the matched observed position, the total satellite observed irradiance of the matched observed position and the total satellite observed irradiance of the position to be forecasted corresponding to an accumulated probability density value P, wherein the accumulated probability density value P is a real number which is more than or equal to 0 and less than or equal to 1;

the total irradiance correction PDF model of the position to be forecasted is used for: and correcting the satellite observation total irradiance of the position to be forecasted under the corresponding accumulated probability density value P according to the difference value ar 1(P) and the difference value r2(P), so as to obtain the corresponding satellite correction total irradiance.

As an alternative embodiment, the correcting a total irradiance data set according to the satellite of the position to be forecasted, and predicting a forecasted total irradiance of the position to be forecasted includes:

determining a current forecast of the position to be forecasted, and taking at least part of historical forecasts of the position to be forecasted before the current forecast as a historical forecast sample set;

according to the current forecast, calculating the similarity distance between each historical forecast in the historical forecast sample set and the current forecast, and finding out the similar historical forecast of the current forecast;

according to the similar historical forecast, calculating to obtain the satellite corrected total irradiance corresponding to the similar historical forecast through a total irradiance correction PDF model of the position to be forecasted, and using the satellite corrected total irradiance as an observed value of the similar historical forecast;

and taking the similarity distance between the similar historical forecast and the current forecast as a weight, and carrying out weighted average on the observed value of the similar historical forecast to obtain the forecast total irradiance of the position to be forecasted.

As an optional embodiment, the shorter the similarity distance between the similar historical forecast and the current forecast is, the more weight the observed value of the similar historical forecast occupies;

the sum of the weights of the observations of all the similar historical forecasts is 1.

Optionally, the predetermined time window is one or more times within a predetermined time period.

Optionally, each of the predetermined time periods is a season.

As a second aspect of the present disclosure, there is provided a photovoltaic power station solar radiation short-term forecasting device based on satellite radiation products, including:

one or more processors;

memory having one or more programs stored thereon that, when executed by the one or more processors, cause the one or more processors to implement the method of any of the first aspects;

one or more I/O interfaces connected between the processor and the memory and configured to enable information interaction between the processor and the memory.

As a third aspect of the present disclosure, there is provided a computer readable medium having stored thereon a computer program which, when executed by a processor, implements the method of any one of the first aspects.

According to the photovoltaic power station solar radiation short-term forecasting method based on the satellite radiation product, the system error of satellite data is effectively eliminated or reduced in time and space, the surface-reaching solar radiation data set based on the position to be forecasted is generated, the solar radiation numerical mode forecasting effect is improved, and the correction of the photovoltaic power station solar radiation short-term forecasting product is achieved.

Drawings

The accompanying drawings are included to provide a further understanding of the embodiments of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the principles of the disclosure and not to limit the disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing in detail exemplary embodiments thereof with reference to the attached drawings, in which:

FIG. 1 is a flow chart of a method for short-term forecasting of solar radiation of a photovoltaic power station based on a satellite radiation product provided by the present disclosure;

fig. 2 is a flowchart of a procedure for establishing a total irradiance correction PDF model according to the present disclosure;

fig. 3 is a flowchart of another step provided by the present disclosure for establishing a total irradiance correction PDF model;

FIG. 4 is a flow chart of the forecast total irradiance step for predicting a location to be forecasted provided by the present disclosure;

FIG. 5 is a CDF and PDF distribution plot of spring ground observations and FY-4A inverted total irradiance provided by the present disclosure;

FIG. 6 is a graph comparing correlation coefficients and mean absolute error before and after correction of a model validation period as provided by the present disclosure;

FIG. 7 is a plot of average deviation versus total irradiance level before and after correction during a model validation period as provided by the present disclosure;

FIG. 8 is a comparison graph of FY-4A observed total irradiance before and after correction provided by the present disclosure;

FIG. 9 is a schematic view of the spatial locations of selected photovoltaic power plants and ground radiation observation stations in a sparse area for ground radiation observation provided by the present disclosure;

FIG. 10 is a CDF distribution plot of ground observations and FY-4A observed total irradiance within a time-space window of each photovoltaic power plant provided by the present disclosure;

FIG. 11 is a graph comparing the mean deviation, mean absolute error, mean absolute percentage error before and after correction according to the present disclosure;

FIG. 12 is a graph comparing the average deviation, average absolute error, average absolute percent error versus total irradiance level before and after correction as provided by the present disclosure;

FIG. 13 is a comparison of FY-4A observed total irradiance versus live before and after correction as provided by the present disclosure;

FIG. 14 is a block diagram illustrating a short-term solar radiation forecasting device for a photovoltaic power plant according to the present disclosure;

FIG. 15 is a block diagram of a computer readable medium provided by the present disclosure.

Detailed Description

For those skilled in the art to better understand the technical solutions of the present disclosure, the following detailed description of the present disclosure is provided in conjunction with the accompanying drawings.

Example embodiments will be described more fully hereinafter with reference to the accompanying drawings, but which may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

Embodiments of the present disclosure and features of embodiments may be combined with each other without conflict.

As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

As a first aspect of the present disclosure, there is provided a photovoltaic power plant solar radiation short-term forecasting method based on satellite radiation products, as shown in fig. 1, the method including:

step S110: and determining a space window of the position to be forecasted according to the position to be forecasted.

It should be noted that the positions to be forecasted in the present disclosure are mainly directed to photovoltaic power stations, but not limited to photovoltaic power stations, and the short-term solar radiation forecasting described in the present disclosure is not limited to forecasting the positions of photovoltaic power stations, and even if no photovoltaic power station is built at the positions to be forecasted, forecasting of the positions is not affected. For example: the first scheme to be introduced in the disclosure is that, in the case that a ground radiation observation station is built at the position to be forecasted or a photovoltaic power station with ground radiation observation capability is built, the ground observation data and the satellite observation data of the position to be forecasted can be directly used for correcting, and forecasting is performed according to the correction result; the second scheme to be introduced in the present disclosure is to determine at least one matched ground radiation observation station according to the distance between the positions to be forecasted, for the situation that no ground radiation observation station is built at the position to be forecasted, or even only one arbitrarily specified position to be forecasted (for example, the position is used for feasibility analysis of photovoltaic station construction, but is still on the empty ground at present), analyze and correct the ground observation data and satellite observation data of the matched ground radiation observation stations, perform secondary correction on the satellite observation data of the position to be forecasted according to the correction result, and then forecast according to the result of the secondary correction. The specific embodiments are described in detail later.

Some photovoltaic power stations are built with radiation observation instruments, the total irradiance data observed by the radiation observation instruments are good in integrity and reliability, and when solar radiation short-term prediction is carried out, the total irradiance data can be directly used as the total irradiance observed on the ground, so that the position of the photovoltaic power station to be predicted can be directly used as a space window of the position to be predicted. Some photovoltaic power station self-built radiation observation instruments are easy to calibrate, and the integrity and the reliability of data are relatively poor. In addition, the number of ground solar radiation observation stations is small, the geographical distribution is uneven, and the position to be forecasted is likely to have no ground observation station, so that at least one matched observation position (namely, the position with the ground observation station, so that the total ground observation irradiance can be directly obtained) is required to be selected around the photovoltaic power station to be matched with the position to be forecasted, and the position to be forecasted is used as a space window of the position to be forecasted. And correcting the satellite observation total irradiance of the position to be forecasted by adopting the matched ground observation total irradiance and satellite observation total irradiance of the observation positions. In the following embodiments, the FY-4A satellite is taken as an example to perform satellite observation.

As an alternative implementation manner, if the data reliability of the total ground observed irradiance of the position to be forecasted reaches a predetermined threshold, the total ground observed irradiance of the position to be forecasted may be used, and the position to be forecasted serves as a space window. And the data reliability of the ground observed total irradiance of the position to be forecasted represents the degree that the ground observed total irradiance of the position to be forecasted can reflect the actual total irradiance.

As an optional implementation manner, if the data reliability of the total ground observed irradiance of the position to be forecasted does not reach a predetermined threshold, the total ground observed irradiance data of the position to be forecasted is unavailable, the position of at least one ground radiation observation station is matched according to the distance to the position to be forecasted, the position is used as a matched observed position to the position to be forecasted, and a space window is formed by the matched observed position and the position to be forecasted. And the data reliability of the ground observed total irradiance of the position to be forecasted represents the degree that the ground observed total irradiance of the position to be forecasted can reflect the actual total irradiance.

And taking the indexes such as the integrity, the reliability, the season and the like of the ground observation total irradiance data as a preset reliability threshold, and if the ground observation total irradiance data acquired by the photovoltaic power station position to be forecasted reaches a preset threshold, considering that the ground observation total irradiance is available, and performing modeling analysis by using the ground observation total irradiance and the satellite observation total irradiance of the position to be forecasted as a space window.

If the data of the total irradiance of the ground observation collected by the position to be forecasted does not reach the preset threshold value, or the position does not collect the total irradiance of the ground observation, the total irradiance of the ground observation is considered to be unavailable, a space window is required to be formed by the matched observation position and the position to be forecasted together, modeling analysis is carried out by adopting the matched total irradiance of the ground observation collected by the observation station and the total irradiance of the satellite observation, and then comprehensive analysis is carried out by combining the total irradiance of the satellite observation collected by the position to be forecasted.

Step S120: and determining a total irradiance correction scheme of the position to be forecasted corresponding to the space window.

Because the available ground observed irradiance conditions of the positions to be forecasted are different, and the observed positions selected by the space window are different, a total irradiance correction scheme needs to be determined for different scenes. Specifically, the correction scheme comprises a value range of the total terrestrial observed irradiance and the total satellite observed irradiance of the space window within a preset time window, and an algorithm for establishing a corrected PDF model of the total terrestrial observed irradiance of the position to be forecasted. According to different available situations of the ground observed irradiance of the position to be forecasted, the data of the ground observation or the satellite observation of which positions are selected for analysis are determined, and the mathematical formula is selected for modeling operation, so that a better correction effect is achieved.

Step S130: and according to the total irradiance correction scheme of the position to be predicted, determining the ground observed total irradiance and the satellite observed total irradiance of the adopted space window, and establishing a total irradiance correction PDF model of the position to be predicted.

According to the total irradiance correction scheme of the position to be forecasted determined in the step S120, when the total terrestrial observed irradiance of the position to be forecasted is available, the spatial window can only comprise the position to be forecasted, and analysis is carried out according to the total terrestrial observed irradiance and the total satellite observed irradiance of the position to be forecasted; when the non-collected ground observation total irradiance of the position to be forecasted is available or the collected ground observation total irradiance is unavailable, the space window further comprises a matched observation position, the ground observation total irradiance and the satellite observation total irradiance of the matched observation position are selected for analysis, and then the satellite observation total irradiance of the position to be forecasted is combined; and establishing a total irradiance correction PDF model of the position to be forecasted.

And correcting a PDF model through the total irradiance, correcting the satellite observation total irradiance of the position to be forecasted, and obtaining the satellite correction total irradiance of the position to be forecasted.

Because the non-independent system error shows larger change along with time and space, the satellite observation total irradiance is difficult to correct by adopting a conventional linear method, a PDF model is established, the cumulative probability density distribution function CDF of the ground observation total irradiance in the space window of the position to be forecasted and the cumulative probability density distribution function CDF of the satellite observation total irradiance are respectively calculated, the difference value of the satellite observation total irradiance and the ground observation total irradiance in the space window of the position to be forecasted under the same cumulative probability density is utilized to obtain a non-linear difference value rule, the corresponding satellite observation total irradiance rule is corrected by utilizing the difference value rule, and finally the satellite correction total irradiance of the position to be forecasted is obtained.

As an optional implementation manner, the satellite observation total irradiance is the satellite observation total irradiance collected by a pixel point closest to a corresponding observation position in the satellite pie chart. For example, the satellite observation total irradiance of the position to be forecasted is the satellite observation total irradiance collected by the pixel point closest to the position to be forecasted in the satellite pie chart; and the satellite observation total irradiance of the matched observation position is the satellite observation total irradiance collected by the pixel point which is closest to the matched observation position in the satellite disk image.

In an alternative embodiment, the predetermined time window is one or more times within a predetermined time period.

Optionally, each of the predetermined time periods is a season.

For example, several months in a season may be selected as a time window, a total irradiance correction PDF model for the season is established, and the selected month of the satellite observed total irradiance in the spatial window of the location to be forecasted may be the same as or different from the selected month of the ground observed total irradiance of the matched observed location, but all in the same season. Similarly, the upper, middle and lower ten days in a month can be selected as time windows, a PDF model for correcting the total irradiance of the month is established, and the selected ten days of the satellite observation total irradiance in the space window of the position to be forecasted can be the same as or different from the selected ten days of the ground observation total irradiance of the matched observation position, but are all in the same month. This is not to be taken as an example.

Step S140: and correcting a PDF model according to the total irradiance of the position to be forecasted, and correcting the satellite observation total irradiance of the position to be forecasted to obtain a corresponding satellite correction total irradiance data set.

And correcting the PDF model according to the total irradiance of the position to be forecasted, and correcting the historical satellite observation total irradiance of the position. For example, a PDF model is corrected through the total irradiance in spring at the position, the satellite observation total irradiance in spring of each historical year is corrected, and a revised data set of the corrected total irradiance of the historical satellite is obtained.

Step S150: and correcting a total irradiance data set according to the satellite of the position to be forecasted, and forecasting the forecast total irradiance of the position to be forecasted.

And acquiring a historical forecast of the position to be forecasted as a historical forecast sample. And acquiring satellite corrected total irradiance predicted by similar histories in a satellite corrected total irradiance data set obtained by calculating a total irradiance correction PDF model of the position to be predicted, taking the satellite corrected total irradiance as an observed value of the similar histories, taking the similarity distance as a weight, and carrying out weighted average on the observed values of the similar histories to finally obtain the predicted total irradiance of the position to be predicted.

According to the difference between the position of the selected space window and the correction scheme, the step S120 of establishing the total irradiance correction PDF model of the position to be forecasted has different implementation manners, which are described below.

As a case of the present disclosure, when the ground observation irradiance of the position to be forecasted is available, as shown in fig. 2, a scheme that the spatial window only includes the position to be forecasted is selected, and then the step S120 specifically includes:

step S121: and obtaining the cumulative probability density distribution function CDF of the ground observed total irradiance of the position to be forecasted and the CDF of the satellite observed total irradiance according to the ground observed total irradiance and the satellite observed total irradiance of the position to be forecasted at multiple times.

Step S122: and calculating the difference value (Δ r (P) between the total observed terrestrial irradiance and the total observed terrestrial irradiance of the position to be forecasted by calculating the corresponding total observed terrestrial irradiance and total observed satellite irradiance when the cumulative probability density value P is according to the CDF of the total observed terrestrial irradiance and the CDF of the total observed satellite irradiance of the position to be forecasted, wherein the cumulative probability density value P is a real number which is more than or equal to 0 and less than or equal to 1.

And if the total irradiance of the position to be forecasted is in the correction PDF model, correcting the satellite observation total irradiance of the position to be forecasted under the corresponding cumulative probability density value P according to the difference value Δ r (P) to obtain the corresponding satellite correction total irradiance.

As another aspect of the present disclosure, when the ground observed irradiance of the photovoltaic power station to be predicted is not available, as shown in fig. 3, a scheme that the spatial window further includes the position to be predicted and the matched observed position is selected, and step S120 specifically includes:

step S123: and respectively obtaining the CDF of the ground observed total irradiance distribution of the matched observed position, the CDF of the satellite observed total irradiance distribution of the matched observed position and the CDF of the satellite observed total irradiance distribution of the position to be forecasted according to the ground observed total irradiance of the matched observed position, the satellite observed total irradiance of the matched observed position and the satellite observed total irradiance of the position to be forecasted at multiple times.

Step S124: according to the CDF of the ground observed total irradiance distribution of the matched observed position, the CDF of the satellite observed total irradiance distribution of the matched observed position and the CDF of the satellite observed total irradiance distribution of the position to be forecasted, and calculating the difference value Δ r1(P) between the total ground observed irradiance of the matched observed position and the total ground observed irradiance of the matched observed position, the difference value Δ r2(P) between the total satellite observed irradiance of the position to be forecasted and the total satellite observed irradiance of the matched observed position by calculating the total ground observed irradiance of the matched observed position, the total satellite observed irradiance of the matched observed position and the total satellite observed irradiance of the position to be forecasted corresponding to the position to be forecasted when the cumulative probability density value P is P, wherein the cumulative probability density value P is a real number which is more than or equal to 0 and less than or equal to 1.

With the correction of the total irradiance of the position to be forecasted, the satellite observation total irradiance of the position to be forecasted under the corresponding accumulated probability density value P can be corrected according to the difference Δ r1(P) and Δ r2(P), and the corresponding satellite correction total irradiance can be obtained.

According to actual needs, the matched observation positions can be multiple, at this time, the ground observation total irradiance and the satellite observation total irradiance of one ground radiation observation station are not adopted for modeling, but the ground observation total irradiance and the satellite observation total irradiance of the matched ground radiation observation stations are respectively used as a set, so that the number of samples can be increased, the influence of individual abnormal data on a correction result is reduced, and the correction accuracy is improved. Therefore, in the above step, the total ground observed irradiance of the matched observed position is a set of total ground observed irradiance of each ground radiation observation station in the matched observed position; and the satellite observation total irradiance of the matched observation position is a set of satellite observation total irradiance of each ground radiation observation station in the matched observation position.

Accordingly, the step S140 of predicting the forecasted total irradiance of the location to be forecasted according to the satellite correction total irradiance data set of the location to be forecasted may include, as shown in fig. 4:

step S141: determining the current forecast of the position to be forecasted, and taking at least part of historical forecast of the position to be forecasted before the current forecast as a historical forecast sample set.

Step S142: and calculating the similarity distance between each historical forecast in the historical forecast sample set and the current forecast according to the current forecast, and finding out the similar historical forecast of the current forecast.

The key point of the similarity error correction method is to define a proper distance to measure the similarity degree of the historical forecast and the current forecast so as to describe the similarity degree of the weather process.

Step S143: and according to the similar historical forecast, calculating to obtain the satellite corrected total irradiance corresponding to the similar historical forecast through a total irradiance correction PDF model of the position to be forecasted, and using the satellite corrected total irradiance as an observed value of the similar historical forecast.

Step S144: and taking the similarity distance between the similar historical forecast and the current forecast as a weight, and carrying out weighted average on the observed value of the similar historical forecast to obtain the forecast total irradiance of the position to be forecasted.

As an alternative embodiment, the shorter the similarity distance between the similar history forecast and the current forecast, the greater the weight occupied by the observation values of the similar history forecast, and the sum of the weights of the observation values of all the similar history forecasts is 1.

And taking the similarity between the similar historical forecast and the current forecast as a distance, wherein the shorter the distance is, namely the higher the similarity is, the larger the weight occupied by the observed values of the similar historical forecast in the correction operation is, and the sum of the weights of the observed values of all the similar historical forecasts is 1.

As an alternative embodiment, in order to facilitate understanding of those skilled in the art, the short-term solar radiation forecasting method for the photovoltaic power station based on the satellite radiation product provided by the present disclosure is further explained by selecting total irradiance data observed by a ground radiation observation station and an FY-4A satellite in an inner Mongolia region.

Under the condition that the existing ground solar radiation observation is difficult to meet the solar radiation forecast requirement, the application of the satellite remote sensing data becomes an effective way for making up the deficiency. However, the satellite inversion data not only has obvious non-independent system errors, but also has large errors changing along with time and space. Therefore, in the photovoltaic power generation short-term forecasting application, how to effectively eliminate or reduce the system error of satellite data in time and space to obtain unbiased satellite inversion data and generate a surface-to-earth solar radiation data set based on the position of a photovoltaic power station so as to improve the solar radiation numerical mode forecasting effect is a key technical problem to be solved in the photovoltaic power generation short-term forecasting method based on satellite radiation products. In the disclosure, the main idea is that when the total irradiance data observed by the self-built radiation observation instrument of the photovoltaic power station is available, but the effective data time length cannot meet the solar radiation forecast requirement, the total irradiance data observed by the self-built radiation observation instrument of the photovoltaic power station is utilized to perform single-point correction on satellite observation data; when the total irradiance data observed by a radiation observation instrument built by the photovoltaic power station is unavailable, correcting a space region of satellite observation data by utilizing the peripheral ground radiation observation of the photovoltaic power station to generate a satellite inversion total irradiance data set based on the position of the photovoltaic power station; based on the satellite inversion total irradiance and the solar numerical prediction mode, the short-term prediction product correction of the photovoltaic power station is realized by quantitatively judging the similarity degree between the current prediction and the historical prediction.

As indicated above, the location to be forecasted according to the present disclosure is mainly directed to a photovoltaic power station, but not limited to a photovoltaic power station, and the short-term solar radiation forecast according to the present disclosure is not limited to a forecast of the location of the photovoltaic power station, and even if the photovoltaic power station is not built in the location to be forecasted, the forecast of the location is not affected. For example: the first technical scheme of the disclosure is that for the condition that a ground radiation observation station is built at the position to be forecasted or a photovoltaic power station with ground radiation observation capability is built, the ground observation data and the satellite observation data of the position to be forecasted can be directly used for correcting, and forecasting is carried out according to the correction result; the second technical scheme of the present disclosure is that, for the situation that the position to be forecasted is not provided with a ground radiation observation station, is not provided with a photovoltaic power station with ground radiation observation capability, or even is only provided with an arbitrarily specified position to be forecasted, at least one matched ground radiation observation station can be determined according to the distance of the position to be forecasted, the ground observation data and the satellite observation data of the matched ground radiation observation station are analyzed and corrected, then the satellite observation data of the position to be forecasted are secondarily corrected according to the correction result, and then forecasting is performed according to the secondary correction result.

As a first solution of the present disclosure, a ground radiation observation station is built at the position to be forecasted, or a photovoltaic power station with ground radiation observation capability is built at the position to be forecasted.

The method comprises the following steps of taking 8 ground radiation observation stations in an inner Mongolia region as positions to be forecasted to conduct research on a single-point correction method, namely correcting the satellite observation total irradiance at the position by calculating the difference between the ground radiation observation total irradiance and the satellite observation total irradiance at the position, wherein the probability density value is P. The latitude and longitude and altitude of 8 ground radiation observation stations are shown in table 1. Based on 8 ground radiation observation stations in inner Mongolia regions and the nearest pixel points in the FY-4A disk image from the 8 observation stations, the seasonal variation of solar radiation and the number of samples required by the establishment of a total irradiance correction PDF model are comprehensively considered, a mode of establishing the PDF model according to seasons is selected, the sample period is divided into 4 parts of spring samples, summer samples, autumn samples and winter samples, and each season contains 6 months samples. For example: the time range of the spring sample is 5/1/2018-5/31/2018, 3/1/2019-5/31/2019, 3/1/2020-4/30/2020; the summer sample time range is from 1 day at 6 months in 2018 to 31 days at 8 months in 2018, from 1 day at 6 months in 2019 to 31 days at 8 months in 2019; the time range of the autumn sample is from 1 day of 9 months in 2018 to 30 days of 11 months in 2018, from 1 day of 9 months in 2019 to 30 days of 11 months in 2019; the winter sample time ranges from 12 months 1 days in 2018 to 2 months 28 days in 2019, 12 months 1 days in 2019 to 2 months 28 days in 2020. And (4) adopting a random independent test mode, namely establishing a PDF model by using any 5-month sample in each season (model construction period), and using the rest 1-month sample for model effect cross test (model verification period).

TABLE 1

	Forehead Ji Na	Wulat middle flag	Dongsheng (Dongsheng)	Two link	Cylinder very high	Soren	Tongliao (traditional Chinese medicine)	Heilaer
									Longitude (G)	101.06	108.52	110.01	111.94	116.12	121.22	122.27	119.7
Latitude	41.96	41.57	39.82	43.63	43.95	46.6	43.6	49.25
									Altitude height	939	1288	1462.2	963.1	1003	499.7	178.7	649.6

The method of Probability Density Function (PDF) is suitable for correcting the error of the product with strong dependence on the numerical range of the data, and the main idea is to adjust the meteorological element value of the satellite data to be consistent with the Probability Density distribution of the ground observation meteorological element, thereby achieving the purpose of correcting the systematic error of the data.

The PDF represents the probability that the instantaneous amplitude falls within some specified range, i.e., the function f (x) of the amplitude x. The Cumulative probability density Distribution Function (CDF) f (X) of the random variable X over time t for any real number X is expressed as:

（1）

for the radiation data of ground observation and FY-4A observation, when the sample amount is large enough, the CDF of the total irradiance of the ground observation and the FY-4A observation can be obtained by the formula (1) respectively. When the probability density is P, the corresponding total irradiance observed for the ground radiation and the total irradiance observed for FY-4A are respectively recorded as obs (P) and FY (P), and the two have a stable difference value at this time:

（2）

therefore, by calculating the cumulative probability density of each time ground radiation observation, the corresponding FY-4A and total ground observation irradiance value of the probability density, the Δ r of the time can be obtained, and the total irradiance FY of each time FY-4A can be corrected to obtain the correction value FY_C：

（3）

The non-independent system error appears to be large in time and space variation, and the conventional linear method is difficult to correct. However, the probability density distribution of solar radiation is relatively stable in a certain space-time range, and it can be known from the above analysis that the magnitude of the average deviation of the total irradiance observed by FY-4A has strong dependency on the total irradiance level. Considering the time scale change of solar radiation, the east-west span of inner Mongolia is large, the difference of the solar radiation of the east and the west is obvious, and in order to ensure that the selected samples are enough, the stable cumulative probability density can be obtained statistically, and a PDF model is established for each station by season.

CDF curves of the total irradiance observed by the ground and the FY-4A in 4 seasons are always intersected at a certain point, and when the CDF value is less than the CDF value of the point, the total irradiance observed by the FY-4A is higher; above this point CDF value, FY-4A observes a lower total irradiance, which correlates to the average deviation characteristic of FY-4A "overestimate low radiation, underestimate high radiation".

With WLTZQ (Observation station name)Abbreviated) ground radiation observation station for illustration. In fig. 5 (a), the ordinate represents CDF, and the probability density value P is a real number of 0 to 1; the abscissa represents irradiance distribution, and the value is from 0 to 1200 W.m according to the highest irradiance in spring of the total irradiance observed on the ground of the WLTZQ radiation observation station^-2(ii) a In a rectangle enclosed by coordinate axes, a solid line curve is a curve obtained in the process that the ground observation total irradiance of the WLTZQ radiation observation station changes from 0 to 1 along with the probability density value P, and a dashed line curve is a curve obtained in the process that the FY-4A observation total irradiance of the WLTZQ changes from 0 to 1 along with the probability density value P.

For example, according to the above equation (2), when the vertical axis probability density value P is 0.2, the OBS (0.2) is a numerical value corresponding to the point on the solid line curve on the lower horizontal axis, and the FY (0.2) is a numerical value corresponding to the point on the dashed line curve on the lower horizontal axis. As can be seen from FIG. 5, when P is 0.2, OBS (0.2) is smaller than FY (0.2), and Δ r (0.2) at this time is FY (0.2) -OBS (0.2) according to formula (2).

As can be seen from (a) in FIG. 5, on the CDF and PDF distribution diagrams of the total irradiance observed in spring ground and FY-4A, when the CDF is less than or equal to 0.5, the total irradiance observed in ground corresponding to the same CDF is lower than the total irradiance observed in FY-4A; when the CDF is more than or equal to 0.5, the ground observation corresponding to the same CDF is higher than the total irradiance observed by FY-4A.

According to the correction idea of the PDF method: when the same accumulated probability density value corresponds to different ground observation and FY-4A observation total irradiance, the deviation of the ground observation and the FY-4A observation total irradiance is used for correcting the FY-4A observation total irradiance.

In FIG. 5 (b) the ordinate represents the random probability density and the abscissa represents the piecewise distribution of irradiance, with the total observed terrestrial irradiance from 0 to a maximum of 1200 W.m. according to the WLTZQ radiometric observation station^-2Dividing the random probability density into 30 sections, and respectively counting the random probability density of each section; the histogram shows the change of random probability density of the satellite observation total irradiance of the WLTZQ corresponding to each section; the solid curve represents the variation curve of the random probability density corresponding to each section of the total ground observation irradiance of the WLTZQ radiation observation station; the dotted curve represents the satellite view of the WLTZQ radiation observation station corrected by the PDF methodAnd measuring the change curve of the random probability density corresponding to each section of the total irradiance.

Therefore, the PDF of the total observed irradiance of the FY-4A is more concentrated than that of the ground observation before correction, namely, the distribution interval of the total observed irradiance of the FY-4A is obviously smaller than that of the ground radiation observation, and after correction, the PDF of the FY-4A is obviously close to that of the actual observation.

FIG. 6 is a correlation coefficient and mean absolute error comparison of FY-4A observed total irradiance with ground radiation observed before and after model validation period correction. Wherein, the (a) to (d) are respectively a comparison graph of the relevant coefficients and the average absolute error of four seasons of spring, summer, autumn and winter, the left ordinate is the relevant coefficient, the value is a real number from 0 to 1, the right ordinate is the average absolute error of the season, and the unit is W.m^-2The horizontal axis represents each ground radiation observation station, the histogram corrects the comparison of the mean absolute error before and after each ground radiation observation station, and the curve corrects the comparison of the correlation coefficient before and after each ground radiation observation station. As can be seen from FIG. 6, the number of the quaternary phase relations and the average absolute error are greatly improved compared with those before correction, the correlation coefficient is improved from 0.37 to 0.91 to 0.83 to 0.96, and the average improvement is 25.7% through the confidence test of alpha = 0.01. The average absolute error is from 5.2 to 404.9 W.m^-2Reduce the temperature to 5.3-139.1 W.m^-2Average reduction of 67 W.m^-2. After correction, the correlation coefficient curve is roughly consistent with that before correction, and the region and seasonal distribution characteristics of the correlation coefficient after correction are basically consistent with those of the original FY-4A. If the average absolute error before correction is large, the average absolute error after correction is reduced obviously, and if the average absolute error before correction is small, the average absolute error may be increased, even so, the average absolute error is greatly improved in the whole view.

FIG. 7 is the average deviation of FY-4A observed total irradiance at each season before and after correction during the verification period as a function of total irradiance level. Wherein, the (a) to (d) are respectively a comparison graph of the relevant coefficients and the average absolute errors of spring seasons, summer seasons, autumn seasons and winter seasons, the left side ordinate is the average deviation obtained by subtracting the total irradiance observed on the ground from the total irradiance observed on the FY-4A, and the unit is W.m^-2With the lower abscissa being a spokeIlluminance distribution in W.m^-2The histogram is the comparison of the corresponding average deviation under different irradiance distributions before and after correction, and the curve is the comparison of the correlation coefficients before and after correction of each ground radiation observation station. As can be seen from FIG. 7, the average deviation is from-150.7 to 305.2 W.m^-2The temperature is reduced to-98.2 to 78.9 W.m^-2After correction, the distribution of positive and negative deviation is basically consistent with that before correction and is lower than 400 W.m^-2Mainly positive deviation, higher than 600 W.m^-2In the method, the negative deviation is taken as the main, because the value of FY-4A is adjusted by the PDF method according to the ground radiation observation CDF, and when the same CDF is obtained, r is the same, the PDF method effectively reduces the system error, so that the total observed irradiance of the FY-4A is closer to the true value, and the error distribution characteristic of the satellite data can be better kept. When total irradiance is<400 W•m^-2The correction effect is particularly remarkable. In a word, the PDF method has a good correction effect on the total observed irradiance of the FY-4A, can well improve the total irradiance in cloud days, and has an obvious effect on improving the quality of an FY-4A surface incident solar radiation inversion product.

From the change curve of the total irradiance with time before and after correction, the corrected total irradiance curve is more consistent with the ground observation. Fig. 8 is a comparison of FY-4A observed total irradiance variation curves before and after correction of XLHT (abbreviation for name of observation station) observation station during autumn verification period (10/1/2018-10/31/2018). Wherein the left ordinate is irradiance distribution, and the unit is W.m^-2The lower abscissa corresponds to each day of month 10, in the format YYMMDD, for example 181001 for month 10 and day 1 of 2018, and the three curves are the comparison of the total irradiance observed on the ground, the total irradiance observed at FY-4A, and the total irradiance observed at FY-4A after correction, respectively, each day. As can be seen, the PDF method makes the corrected FY-4A observed total irradiance closer to the ground observed value, and the average deviation is 47.20 W.m^-2Reduced to-5.2 W.m^-2The improvement effect on low-value total irradiance is particularly obvious, such as 7 days at 10 months to 10 months and 10 days at 10 months, the improvement effect is influenced by snowfall weather in eastern areas of inner Mongolia, continuous solar radiation observation of an XLH station for a plurality of days goes low, the total irradiance value observed by FY-4A is higher and obvious, and after the improvement effect is corrected by a PDF method, the improvement effect of FY-4A on cloud is effectively improvedThe defect of overestimating the total irradiance, but the improvement effect of underestimating the total irradiance with high value needs to be further improved. In general, by using historical data of two years, namely 5/1/2018-2020/4/30, a PDF model is built by seasons, so that the distribution characteristics of ground observation and FY-4A observation total irradiance in inner Mongolia regions can be reflected, and the method has an obvious effect on improving the applicability of FY-4A observation total irradiance in inner Mongolia regions.

As a second solution of the present disclosure, a ground radiation observation station is not established for the location to be forecasted, and the photovoltaic power station at the location does not have ground radiation observation capability, even if only one location to be forecasted is arbitrarily specified.

When total irradiance data observed by a radiation observation instrument built by the photovoltaic power station is unavailable, a space region correction mode is adopted, three photovoltaic power stations (named as FWMSG, GDBRB and GDWET respectively) in a ground radiation observation sparse area in the midwest inner Mongolia are taken as a research area, and as shown in fig. 9, the distance between the photovoltaic power station and a nearest national-level ground radiation observation station (ground radiation observation station for short, the same below) is more than 200 km. The data information comprises the time-by-time surface radiation observation of a ground radiation observation station and a photovoltaic power station and an FY-4A SSI product, wherein the ground radiation observation station information is used for model construction, and the photovoltaic power station radiation observation information is used for model effect inspection.

And selecting a proper time-space window for each photovoltaic power station, collecting time-by-time data of all ground radiation observation stations in the time-space window and nearest pixel points in the FY-4A disk image to the observation stations and the photovoltaic power stations, and calculating to obtain the ground observation total irradiance CDF of the observation position matched with the photovoltaic power station to be forecasted, the CDF of the satellite observation total irradiance and the CDF of the satellite observation total irradiance of the position to be forecasted. A probability density of

Then, the total ground observed irradiance and the total FY-4A satellite observed irradiance of the matched observed position are recorded as

And

when the two have a stable difference:

（4）

meanwhile, the total FY-4A irradiance of the position to be forecasted is recorded as

And FY-4A total irradiance in the space-time window of the position to be forecasted

With a stable difference between:

（5）

retrieving each time on the FY-4A total irradiance CDF curve of the position to be forecasted

The corresponding CDF and the ground observation total irradiance value and FY-4A total irradiance value of the matched observation position corresponding to the CDF are obtained to obtain the current time

And

the correction value of the position to be forecasted is calculated by correcting the total irradiance FY observed by each time FY-4A

：

（6）

Solar radiation has obvious seasonal variation rules. For the inner Mongolia region, the seasonal variation of solar radiation appears to be monomodal, i.e., maximum in summer, second in spring and autumn, and minimum in winter. From the spatial distribution of the average total solar radiation in the four seasons of 1961-2017, the solar radiation in each season conforms to the increasing trend from the northeast to the southwest.

The selection of the space-time window needs to consider the space-time variation characteristics of the solar radiation. In the disclosure, seasons are taken as time windows, a space window adopts a high correlation determination principle, and correlation coefficients of total irradiance of radiation observation and peripheral ground radiation observation in three photovoltaic power station sample periods are calculated respectively, and are shown in table 2. The correlation coefficients all pass

The degree of correlation depends on the distance between the photovoltaic power station and the ground radiation observation station, and the closer the distance is, the larger the correlation coefficient is, and vice versa. And selecting an observation station with a higher radiation observation correlation coefficient with the photovoltaic power station as a ground radiation observation station matched with a space window of the photovoltaic power station, wherein the result of the ground radiation observation station matched with each photovoltaic power station is shown in a table 3.

TABLE 2

TABLE 3

Photovoltaic power station	Ground radiation observation station
		FWMSG	Wulat middle flag
GDBRB	Dongsheng (Dongsheng)
		GDWET	Wulat Zhongqi and Dongsheng

And sequentially utilizing the non-zero effective samples in the time-space window of each photovoltaic power station to calculate the stable CDF distribution of ground observation and FY-4A in each season of the matched observation position. As shown in fig. 10, in (a) - (d), CDF distributions of ground observations and total FY-4A irradiance in spring, summer, autumn, and winter of GDWET are shown, respectively, and the abscissa is irradiance and the ordinate is cumulative probability density. The ground observation and the total FY-4A irradiance are intersected at a certain point, and when the CDF value is below the point, the total FY-4A irradiance is higher than the ground observation; when the CDF value is above the point, the total irradiance of FY-4A is lower than that of ground observation, which shows that the FY-4A has the deviation distribution characteristics of overestimation and underestimation of low-value radiation. That is, the magnitude of the deviation of the total irradiance at a certain time FY-4A has a certain dependency on the numerical range of the data itself at that time. In addition, the size of the CDF value at the intersection point is dependent on the season, which is the smallest in winter, the next in spring and autumn, and the largest in summer. The FY-4A is more prone to negative deviation in winter and positive deviation in summer. Calculating the CDF value corresponding to each time by using the total irradiance value of each time photovoltaic power station position FY-4A, and calculating the corresponding CDF value

、

And finally obtaining the correction value of the FY-4A total irradiance.

Fig. 11 is a comparison of correlation coefficients, average error, average absolute error, and average absolute percentage error before and after correction of three photovoltaic power stations. As can be seen from the graph, after correction, the correlation coefficient, the average error, the average absolute error and the average absolute percentage error of the three photovoltaic power stations have different degreesThe improvement of the average error and the average absolute percentage error has obvious improvement effect. Wherein, the FEMSG correlation coefficient is improved from 0.83 to 0.84, and the average error is improved from 113.5 W.m^-2Reduced to 68.4 W.m^-2Mean absolute error of 174.0 W.m^-2Reducing to 135.3 W.m^-2The average absolute percentage error is reduced from 92.6 percent to 57.4 percent; the GDBRB correlation coefficient is improved to 0.85 from 0.84, and the average error is improved to 50.8 W.m^-2Reduced to 12.6 W.m^-2The mean absolute error is 107.8 W.m^-2Reduced to 92.8 W.m^-2The average absolute percentage error is reduced from 44.4 percent to 27.0 percent; the GDWET correlation coefficient is improved to 0.85 from 0.83, and the average error is improved to 43.0 W.m^-2Reduced to 24.6 W.m^-2The mean absolute error is 113.0 W.m^-2Reduced to 101.2 W.m^-2The mean absolute percentage error decreased from 54.8% to 39.1%. On the whole, the position of the FY-4A in three photovoltaic power stations is larger than the actual situation and is mainly based on positive deviation, the problem of over-high estimation of the FY-4A can be effectively solved by the PDF method, and the larger the original deviation is, the more obvious the reduction after correction is.

The average error, the average absolute error and the average absolute percentage error before and after correction are distributed along with the irradiance level, as shown in fig. 12, wherein (a) - (c) are comparison graphs of the corresponding average errors of the FWMSG, the GDBRB and the GDWET under different irradiance distributions before and after correction, the abscissa is the irradiance distribution, and the ordinate is the average error; (d) the (f) is a comparison graph of corresponding average absolute errors of the FWMSG, the GDBRB and the GDWET under different irradiation degree distributions before and after correction, the abscissa is the same as the above, and the ordinate is the average absolute error; (g) and (i) is a comparison graph of corresponding average absolute percentage errors of the FWMSG, the GDBRB and the GDWET under different irradiation degree distributions before and after correction, the abscissa is the same as the above, and the ordinate is the average absolute percentage error.

Before correction, as the irradiance level increases, the average error and the average absolute percentage error of FY-4A gradually decrease, and the average absolute error shows the trend of decreasing firstly and then increasing: total irradiance<400 W·m^-2FY-4A shows a significant positive deviation, and the average error and the average absolute error have approximately the same value, 400 W.m^-2Total irradiance less than or equal to<800 W·m^-2The uncertainty of the average error is large, and the total irradiance is more than or equal to 800 W.m^-2FY-4A exhibits a significant negative bias. Total irradiance<400 W·m^-2The average deviation after correction is from 129.8 to 312.2 W.m^-2Reducing the temperature to-1.9 to 183.4 W.m^-2The mean absolute error is from 138.4 to 312.2 Wm^-2Reducing the temperature to 87.4-197.5 Wm^-2The average absolute percentage error is reduced from 40.3-605.5% to 39.2-307.5%, the improvement effect of the PDF method is obvious, and 400 W.m^-2Total irradiance less than or equal to<800 W·m^-2From the calibration results, the FY-4A error is relatively small in the irradiance level range, and the PDF method has negative return, which is related to the uncertainty of positive and negative deviation of the FY-4A SSI product in the irradiance level range. When the total irradiance is more than or equal to 800 W.m^-2The average error, average absolute error and average absolute percentage error after correction are all improved compared with those before correction. As can be seen, the PDF method is less than 400 W.m^-2Low value radiation sum of more than 800 W.m^-2The high-value radiation has better improvement effect, and can effectively make up the problems of low-value radiation overestimation and high-value radiation underestimation of FY-4A.

Compared with the correction results of the three photovoltaic power stations, the PDF method has the same correction effect on the three photovoltaic power stations. Therefore, a proper space-time window is selected by combining the space-time change characteristics of solar radiation, and the FY-4A SSI product in the space window is observed and corrected based on the ground radiation in the space-time window, so that the method is feasible.

As shown in fig. 13, the four seasons of spring, summer, autumn and winter are represented by 3 months, 6 months, 9 months and 12 months respectively, the time sequence change of the total irradiance observed by the photovoltaic power station is compared with the FY-4A before and after correction, wherein the ordinate is the irradiance, and the unit is W.m^-2The abscissa is the corresponding date in the format YYMMDD, e.g. 190301 for 3 months and 1 day of 2019. As can be seen from fig. 13, the PDF method improves the effect in spring, summer and autumn better than in winter. Before correction, the total irradiance value of the position FY-4A of the three photovoltaic power stations is basically between 400 W.m^-2~1000 W·m^-2Meanwhile, the daily variation amplitude is obviously smaller than that of the live situation. After being corrected by the PDF method, the peak value or the valley value tends to be moreClose to live. However, due to the limitation of the FY-4A surface solar incident radiation inversion algorithm, the radiation value is only around noon in winter, and the inversion value is basically 400-500 W.m^-2When modeling is carried out by PDF method, the model is less than 400 W.m^-2And is greater than 500 W.m^-2The number of FY-4A samples is limited, and a certain deviation exists between the simulated CDF curve and the reality, so that the correction effect of the PDF method in winter is better than that of the PDF method in other three seasons.

Further, by taking 6 photovoltaic power stations (the names are abbreviated as FWMSG, GDBRB, GDWET, GDBLN, THTGF and GHDL respectively) in the midwest of inner Mongolia as a research area, photovoltaic power generation short-term prediction is carried out in a model verification period based on a satellite inversion total irradiance correction product and a numerical mode prediction WRF total irradiance prediction product.

The main idea of the similarity error correction method is to transform the predictions arranged according to the time sequence onto the similarity space to find the historical prediction most similar to the current prediction (defined by specific distance), and at the same time, because the true simulation error of the historical prediction can be obtained, the simulation error of the current prediction can be estimated. And grading the historical forecasts according to the similarity with the current forecast in the similarity space, wherein the correction aiming at the current forecast provides more weight for the similar forecast which is closest to the current forecast from the similar forecast which is farthest to the current forecast and is the worst similar forecast and the best similar forecast respectively. If appropriate similar forecast selection criteria are found and defined, they can also be corrected in the case of a drastic change in forecast error. Therefore, the key point of the similarity error correction method is to define an appropriate distance to measure the similarity between the historical forecast and the current forecast, so as to describe the similarity of the weather process.

The distance between the forecast for a particular time and location and the historical forecasts for all times before the same location is defined as:

（7）

wherein the content of the first and second substances,

is the current forecast that needs to be corrected at a given time t, is

Historical forecasts at time t' before the start of the forecast,

(v represents a variable) and

respectively the vectors of the relevant physical variables and their weights,

(f denotes forecast) is the standard deviation of the time series of past forecasts for a certain variable,

is half the length of the time window (range of effective influence) over which the distance is calculated,

and

is the specific value of the current forecast and the historical forecast for a given variable within a time window.

The corrected forecast is defined as the weighted average of the observations of similar historical forecasts:

（8）

wherein the content of the first and second substances,

is the predicted order value at time t,

(a represents similarity) is the number of similar historical forecasts,

is the best

The observed values of the similarity predictions are used,

is the time of the similar forecast. Weight of each affinity forecast

Comprises the following steps:

（9）

the shorter the distance between the similar forecast and the current forecast, i.e., the more similar, the greater the weight occupied by its observation, and the sum of all weights is 1.

The similar forecasting method has the advantages that the error correction is carried out by searching the historical forecasts similar to the current forecast, the forecasts arranged according to the time sequence are transformed to the similar space, the defect of the common error correction method based on the time sequence is overcome, and the rapid change of the forecast error caused by the severe change of a weather system can be processed.

Selecting a proper correction factor, determining model parameters such as a time window, correction factor weight and the like through a sensitivity test, taking a historical WRF forecast product as a historical forecast sample, taking FY-4A historical radiation data based on the position of a photovoltaic power station as a historical live sample, and establishing a total irradiance short-term forecast model so as to effectively improve WRF cloud-day radiation forecast, serve the photovoltaic power generation industry and improve the solar resource short-term forecast service capability.

The correlation coefficient, average deviation and average absolute error ratio of 6 photovoltaic power stations before and after correction are improved to different degrees as shown in table 4.

TABLE 4

In a second aspect, referring to fig. 14, an embodiment of the present disclosure provides a short-term solar radiation forecasting device for a photovoltaic power station based on a satellite radiation product, including:

one or more processors 501;

a memory 502 having one or more programs stored thereon which, when executed by the one or more processors, cause the one or more processors to implement the satellite radiation product-based photovoltaic plant solar radiation short-term forecasting method of the first aspect of the present disclosure;

one or more I/O interfaces 503 coupled between the processor and the memory and configured to enable information interaction between the processor and the memory.

The processor 501 is a device with data processing capability, and includes but is not limited to a Central Processing Unit (CPU) and the like; memory 502 is a device having 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); an 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, which includes but is not limited to a data Bus (Bus) and the like.

In some embodiments, the processor 501, memory 502, and I/O interface 503 are interconnected by a bus 504, which in turn connects with other components of the computing device.

In a third aspect, referring to fig. 15, an embodiment of the present disclosure provides a computer readable medium, on which a computer program is stored, which when executed by a processor, implements the short-term solar radiation forecasting method for a photovoltaic power station based on a satellite radiation product according to the first aspect of the present disclosure.

It will be understood by those of ordinary skill in the art that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed by several physical components in cooperation. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is 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 tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. In addition, communication media typically embodies 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 includes any information delivery media as known to those skilled in the art.

Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted in a generic and descriptive sense only and not for purposes of limitation. In some instances, features, characteristics and/or elements described in connection with a particular embodiment may be used alone or in combination with features, characteristics and/or elements described in connection with other embodiments, unless expressly stated otherwise, as would be apparent to one skilled in the art. Accordingly, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the disclosure as set forth in the appended claims.

Claims

1. A photovoltaic power station solar radiation short-term forecasting method based on satellite radiation products comprises the following steps:

according to the total irradiance correction scheme of the position to be predicted, determining the ground observed total irradiance and the satellite observed total irradiance of the adopted space window, and establishing a PDF model for correcting the total irradiance of the position to be predicted according to the probability density function;

the correction scheme of the total irradiance of the position to be forecasted comprises the value ranges of the ground observed total irradiance and the satellite observed total irradiance of the space window in a preset time window, and an algorithm for establishing a correction PDF model of the total irradiance of the position to be forecasted.

2. The method of claim 1, wherein the determining a spatial window of locations to be forecasted comprises:

the data reliability of the ground observed total irradiance of the position to be forecasted characterizes the degree that the ground observed total irradiance of the position to be forecasted can reflect the actual total irradiance.

3. The method of claim 2, wherein when the spatial window includes a location to be forecasted, the determining the total terrestrial observed irradiance and the total satellite observed irradiance for the spatial window employed comprises:

determining the total ground observed irradiance and the total satellite observed irradiance of the position to be forecasted in the preset time window;

the establishing of the total irradiance correction PDF model of the position to be forecasted comprises the following steps:

4. The method of claim 1, wherein the determining a spatial window of locations to be forecasted comprises:

5. The method of claim 4, wherein when the spatial window comprises the matched observed position and the position to be forecasted, the determining the total terrestrial observed irradiance and the total satellite observed irradiance of the employed spatial window comprises:

determining the satellite observation total irradiance of the position to be forecasted in a preset time window, the ground observation total irradiance of the matched observation position and the satellite observation total irradiance of the matched observation position;

the ground observation total irradiance of the matched observation position is a set of ground observation total irradiance of each ground radiation observation station in the matched observation position; the satellite observation total irradiance of the matched observation position is a set of satellite observation total irradiance of each ground radiation observation station in the matched observation position;

6. The method of any one of claims 1 to 5, wherein the correcting a total irradiance data set from the satellite for the location to be forecasted to predict a forecasted total irradiance for the location to be forecasted comprises:

7. The method of claim 6, wherein,

the shorter the similarity distance between the similar historical forecast and the current forecast is, the larger the weight occupied by the observation value of the similar historical forecast is;

8. The method of any one of claims 1 to 5,

the preset time window is one or more times in a preset time period;

each of the predetermined time periods is a season.

9. A photovoltaic power station solar radiation short-term forecasting device based on satellite radiation products comprises:

one or more processors;

memory having one or more programs stored thereon that, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-8;

10. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 8.