Power prediction method for distributed photovoltaic power stations

Abstract

Provided in the present invention is a power prediction method for distributed photovoltaic power stations, the method comprising: acquiring a satellite cloud image of distributed photovoltaic power stations at a historical moment and a satellite cloud image of the distributed photovoltaic power stations at a current moment to predict satellite cloud image features at a future moment; selecting several photovoltaic power stations to install solar irradiance collection apparatuses, and on the basis of the satellite cloud images, collected solar irradiance, and location information of the photovoltaic power stations, calculating the solar irradiance of all the photovoltaic power stations; on the basis of the satellite cloud image features at the future moment and the collected solar irradiance of the photovoltaic power stations, obtaining the solar irradiance of the distributed photovoltaic power stations at the future moment; and on the basis of the solar irradiance of the distributed photovoltaic power stations at the future moment and weather forecast information at the future moment, predicting the power of the distributed photovoltaic power stations at the future moment. In the solution of the present invention, solar irradiance data is expanded, and a mapping relationship between satellite cloud images and solar irradiance and a mapping relationship between solar irradiance and power are constructed by means of prediction, so that the power of distributed photovoltaic power stations is predicted, thereby improving the prediction precision.

Description

A method for predicting power of distributed photovoltaic power stations Technical Field

[0001] The present invention relates to the technical field of photovoltaic power station power prediction, and in particular to a distributed photovoltaic power station power prediction method. Background Art

[0002] With the rapid development of solar energy resources in the power industry, humans have realized that the development of only centralized photovoltaic power stations will result in most solar energy resources not being utilized. As a result, distributed photovoltaic power stations have attracted widespread attention. Due to the ever-changing cloud clusters, distributed photovoltaic power generation is highly volatile, and its installation locations are scattered, making power generation difficult to predict.

[0003] In existing studies, most of them use a large amount of historical power data of power stations to explore their periodic laws through statistical models; or use the power prediction of nearby centralized photovoltaic power stations to find the power relationship between them and the site to be predicted through spatiotemporal correlation to predict the power of distributed photovoltaic power stations. However, using a large amount of historical power data for prediction ignores the sudden changes in the photovoltaic environment of the site to be predicted in a short period of time, such as the formation and disappearance of clouds. At the same time, it ignores the problem that most distributed photovoltaic power stations are scattered and lack data collection devices. Some distributed power stations do not have nearby centralized photovoltaic power stations. Some studies also use cloud maps for power prediction. Due to their high price and maintenance costs, ground-based cloud maps cannot be applied to distributed photovoltaic power stations with small installation capacity and wide distribution.

[0004] Satellite cloud images have a large coverage area and are well matched with distributed photovoltaic power stations. Satellite cloud image data most intuitively reflects the changes in solar irradiance. The conversion from solar irradiance to power, the existing technology only considers satellite cloud image characteristics and meteorological characteristics as the input of the prediction model, and the power of distributed photovoltaic power stations as the output of the prediction model. However, clouds directly affect solar irradiance, which is then converted into electricity. Since distributed photovoltaic power stations lack data collection devices and solar irradiance collection data, the intermediate variable solar irradiance is not introduced to correct the model, resulting in large prediction errors.

[0005] Summary of the Invention

[0006] In order to solve the problems existing in the prior art, the present invention provides a distributed photovoltaic power station power prediction method, comprising:

[0007] Obtain historical and current satellite cloud images and weather forecast information for the area where the distributed photovoltaic power station is located;

[0008] Preprocessing the satellite cloud images at historical moments and the satellite cloud images at the current moment, and performing feature extraction on the preprocessed satellite cloud images to obtain features of the satellite cloud images at historical moments and features of the satellite cloud images at the current moment;

[0009] Predict the characteristics of satellite cloud images at future moments based on the characteristics of satellite cloud images at historical moments and the characteristics of satellite cloud images at current moments;

[0010] Select several photovoltaic power station locations in the area where the distributed photovoltaic power station is located to install solar irradiance collection devices, and calculate the solar irradiance of photovoltaic power stations that do not have solar irradiance collection devices installed based on satellite cloud images, the collected solar irradiance and photovoltaic power station location information;

[0011] According to the characteristics of satellite cloud images at historical moments and the solar irradiance at corresponding moments, the mapping relationship between satellite cloud image characteristics and solar irradiance is determined;

[0012] Based on the predicted satellite cloud image features at future times and the mapping relationship between satellite cloud image features and solar irradiance, the solar irradiance of the distributed photovoltaic power station at the corresponding time in the future is obtained;

[0013] According to the solar irradiance of distributed photovoltaic power stations and weather forecast information at future times, the power of distributed photovoltaic power stations at future times is predicted.

[0014] Optionally, the preprocessing of the historical satellite cloud images and the current satellite cloud images, and performing feature extraction on the preprocessed satellite cloud images, includes:

[0015] Grayscale processing is performed on the satellite cloud images at historical moments and the satellite cloud images at the current moment to obtain the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment;

[0016] Perform linear transformation and histogram equalization on the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment to obtain the preprocessed satellite cloud images;

[0017] Cloud pixel values ​​are extracted from the preprocessed satellite cloud images as cloud thickness features, and cloud texture features are extracted from the preprocessed satellite cloud images based on the gray-level co-occurrence matrix.

[0018] Optionally, the predicting of satellite cloud image features at future moments based on satellite cloud image features at historical moments and satellite cloud image features at current moments includes:

[0019] The satellite cloud image features at historical moments and the satellite cloud image features at current moment are time-sorted to obtain a time series of satellite cloud image feature quantities;

[0020] Draw a trend graph of the satellite cloud image feature sequence based on the satellite cloud image feature time series, and determine whether the satellite cloud image feature time series is a stationary time series;

[0021] If the satellite cloud image feature time series is a non-stationary time series, the d-order difference operation is performed on the satellite cloud image feature time series to obtain a stationary time series;

[0022] Draw the autocorrelation diagram and partial autocorrelation diagram of the satellite cloud image feature time series, and determine the p-value and q-value of the ARIMA model based on the tailing and truncation of the autocorrelation diagram and partial autocorrelation diagram;

[0023] The satellite cloud image features at historical moments and the satellite cloud image features at the current moment are input into the ARIMA model to obtain the satellite cloud image features at future moments.

[0024] Optionally, the step of selecting several photovoltaic power station locations in the area where the distributed photovoltaic power station is located to install solar irradiance collection devices, and estimating the solar irradiance of photovoltaic power stations that do not have solar irradiance collection devices installed based on satellite cloud images, the collected solar irradiance, and photovoltaic power station location information, includes:

[0025] Select n photovoltaic power stations in the area where the distributed photovoltaic power station is located, install solar irradiance collection devices, record them as data-complete power stations, and obtain the solar irradiance of the data-complete power stations;

[0026] Extract the pixel value of the distributed photovoltaic power station from the corresponding position of the satellite cloud image according to the location information of the distributed photovoltaic power station;

[0027] Obtain the longitude and latitude τ, tilt angle β, declination angle δ and hour angle ω of the data-complete power station, and calculate the unit irradiance cosz of the data-complete power station. The formula is as follows: cosz=sinδsin(τ-β)+cosδcos(τ-β)cosω

[0028] According to the solar irradiance Ex of the data-complete power station, the unit irradiance cosz of the data-complete power station and the atmospheric radiation intensity Calculate the data set λf(p x ), the formula is as follows:

[0029] The pixel value data set p of the data complete station x With the data set λf(p x ) performs function fitting to obtain function F(p);

[0030] Select a photovoltaic power station that is not equipped with a solar irradiance collection device and select a corresponding reference power station from the power stations with complete data, and obtain temperature data of each reference power station and the photovoltaic power station that is not equipped with a solar irradiance collection device;

[0031] The Pearson product-moment correlation coefficient r between each reference power station and the photovoltaic power station without solar irradiance collection device is calculated based on the temperature data of each reference power station and the photovoltaic power station without solar irradiance collection device. p , and the reference power station with the largest Pearson product-moment correlation coefficient is selected as the most relevant power station. The Pearson product-moment correlation coefficient r p The calculation formula is as follows:

[0032] Among them, x i is the temperature value of the reference power station, y i is the temperature value of the photovoltaic power station without solar irradiance collection device installed, x ave is the average temperature of the reference power station, y ave is the average temperature of the photovoltaic power station without solar irradiance collection device installed;

[0033] According to the solar irradiance E of the most relevant power station a , the pixel value p of the most relevant power station a And the pixel value p of the photovoltaic power station without solar irradiance collection device A Calculate the solar irradiance E of a photovoltaic power station without a solar irradiance collection device installed A , the formula is as follows: E A =F(p A )E a / F(p a )

[0034] Other photovoltaic power stations that are not equipped with solar irradiance collection devices are selected, and the solar irradiance of the photovoltaic power stations that are not equipped with solar irradiance collection devices is calculated according to the above method to obtain the solar irradiance of all photovoltaic power stations.

[0035] Optionally, determining a mapping relationship between satellite cloud image features and solar irradiance based on the satellite cloud image features and the solar irradiance at a corresponding moment includes:

[0036] The satellite cloud image features at historical moments are used as the input of the Elman neural network, and the solar irradiance of the photovoltaic power station at the corresponding moment is used as the output of the Elman neural network. The Elman neural network is trained to obtain the mapping relationship between satellite cloud image features and solar irradiance.

[0037] Optionally, predicting the power of the distributed photovoltaic power station at a future time based on the solar irradiance of the distributed photovoltaic power station at a future time and weather forecast information includes:

[0038] The solar irradiance of photovoltaic power stations at historical moments and the temperature at corresponding moments in weather forecast information are used as inputs to the LSTM neural network, and the power of distributed photovoltaic power stations at corresponding moments is used as the output of the LSTM neural network. The LSTM neural network is trained to obtain the mapping relationship between solar irradiance and distributed photovoltaic power station power.

[0039] The solar irradiance of the distributed photovoltaic power station at the future moment and the temperature in the weather forecast information are input into the mapping relationship between the solar irradiance and the power of the distributed photovoltaic power station to obtain the power of the photovoltaic power station at the future moment.

[0040] After adopting the above technical solution, the present invention has at least the following beneficial effects:

[0041] 1. Expand the solar irradiance data collected from some photovoltaic power stations in the region to the solar irradiance data of all distributed photovoltaic power stations in the region, reducing the number of solar irradiance collection devices installed, saving costs, and ensuring sufficient data;

[0042] 2. First, the mapping relationship between satellite cloud images and solar irradiance is obtained, and then the mapping relationship between solar irradiance and distributed photovoltaic power station power is obtained. This fully utilizes the satellite cloud image information and solar irradiance information to ensure the accuracy of distributed photovoltaic power station power prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] FIG1 is a flow chart of a method for predicting power of a distributed photovoltaic power station according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] FIG1 is a flow chart of a method for predicting power of a distributed photovoltaic power station according to an embodiment of the present disclosure. As shown in FIG1 , the method includes:

[0047] Step 1: Obtain historical satellite cloud images and weather forecast information for the area where the distributed photovoltaic power station is located, as well as current satellite cloud images and weather forecast information;

[0048] In this embodiment, the Fengyun-4 satellite can be used to obtain satellite cloud images at multiple historical moments and satellite cloud images at the current moment. The Fengyun-4 satellite conducts a full-disk observation every hour, from 15 minutes after the hour to 15 minutes after the hour. When there is no full-disk observation, a 5-minute Chinese regional planning is performed. The ultra-short-term prediction time scale is 15 minutes. The satellite cloud images obtained by the Fengyun-4 satellite meet the needs of ultra-short-term prediction.

[0049] Step 2: preprocessing the satellite cloud images at historical moments and the current moment, and performing feature extraction on the preprocessed satellite cloud images to obtain features of the satellite cloud images at historical moments and features of the satellite cloud images at the current moment;

[0050] In this embodiment, step 2 includes:

[0051] Grayscale processing is performed on the satellite cloud images at historical moments and the satellite cloud images at the current moment to obtain the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment;

[0052] Perform linear transformation and histogram equalization on the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment to obtain the preprocessed satellite cloud images;

[0053] Since the smaller the grayscale value, the lower the brightness of the grayscale image, which means the thinner the cloud, there is interference from the earth reflection in the satellite cloud image, and its grayscale value is not 0. During processing, part of the earth will be mistakenly classified as a cloud, causing interference to the satellite cloud image. The grayscale image is linearly changed to adjust the grayscale range of the grayscale image and improve the image contrast. The original image grayscale range is 0-255. The grayscale value less than the grayscale value 255*0.1 is set to 0 to reduce the error caused by the earth reflection on the satellite cloud image reflection. Then histogram equalization is performed to increase the global contrast of the image and enhance the image details and edges.

[0054] Cloud pixel values ​​are extracted from the preprocessed satellite cloud images as cloud thickness features, and cloud texture features are extracted from the preprocessed satellite cloud images based on the gray-level co-occurrence matrix, wherein the cloud thickness features are cloud pixel values, and the cloud texture features include angular second moment, entropy, contrast, and correlation.

[0055] Each element in the gray-level co-occurrence matrix represents the probability density of the simultaneous appearance of a pixel with gray-level value i and a pixel with gray-level value j at a certain distance and direction. Take any point (x, y) in the image (M×N) and another point (x+a, y+b) deviating from it, and let the gray-level value of the point pair be (i, j). If the point (x, y) moves across the entire image, various (i, j) values ​​will be obtained. Let the number of gray-level values ​​be k, then there are k combinations of (i, j). 2 For the entire picture, count the number of times each (i, j) value appears, then arrange them into a square matrix, and then use the total number of times (i, j) appears to normalize them into the probability of occurrence P(i, j). Such a square matrix is ​​called a gray-level co-occurrence matrix.

[0056] The texture eigenvalues ​​of each cloud cluster are calculated as follows:

[0057] The angular second moment is the sum of the squares of the gray-level co-occurrence matrix, which reflects the uniformity of the gray-level distribution and the coarseness of the texture of the image. The formula is as follows:

[0058] Entropy represents the complexity of the grayscale distribution of an image. The formula is as follows:

[0059] Contrast represents the local change of the gray-level co-occurrence matrix, reflecting the clarity of the image and the depth of the texture grooves. The formula is as follows

[0060] Correlation is used to measure the similarity of the grayscale of an image in the row or column direction, reflecting the local grayscale correlation. The formula is as follows:

[0061] Satellite cloud image features at all times;

[0062] The ARIMA model is used to predict the characteristics of satellite cloud images at future moments. The basic idea of ​​the ARIMA model is to regard the data sequence formed by the prediction object over time as a random sequence, and use a certain mathematical model to approximate this sequence. Once this model is identified, the future value can be predicted from the past and present values ​​of the time series.

[0063] In this embodiment, step 3 includes:

[0064] The satellite cloud image features at historical moments and the satellite cloud image features at current moment are time-sorted to obtain a time series of satellite cloud image feature quantities;

[0065] Draw a trend graph of the satellite cloud image feature sequence based on the satellite cloud image feature time series, and determine whether the satellite cloud image feature time series is a stationary time series;

[0066] If the satellite cloud image feature time series is a non-stationary time series, the d-order difference operation is performed on the satellite cloud image feature time series to obtain a stationary time series;

[0067] Draw the autocorrelation diagram and partial autocorrelation diagram of the satellite cloud image feature time series, and determine the p-value and q-value of the ARIMA model based on the tailing and truncation of the autocorrelation diagram and partial autocorrelation diagram;

[0068] The satellite cloud image features at historical moments and the satellite cloud image features at the current moment are input into the ARIMA model to obtain the satellite cloud image features at future moments.

[0069] Step 4: Select several photovoltaic power station locations in the area where the distributed photovoltaic power station is located to install solar irradiance collection devices, and calculate the solar irradiance of photovoltaic power stations that do not have solar irradiance collection devices installed based on satellite cloud images, the collected solar irradiance, and the location information of the photovoltaic power stations;

[0070] Since the meteorological factors in the same area are similar, the solar radiation intensity and solar altitude angle are the same, assuming there is no cloud obstruction, the solar irradiance E0 remains approximately unchanged after attenuation. After the influence of clouds, the solar irradiance is E1. E1 is positively correlated with E0 and is only affected by the thickness of the clouds. The thickness of the clouds is reflected in the size of the pixel value p in the satellite cloud image, and the formula E1 = f(p)E0 is obtained.

[0071] In this embodiment, step 4 includes:

[0072] Select n photovoltaic power stations in the area where the distributed photovoltaic power station is located, install solar irradiance collection devices, record them as data-complete power stations, and obtain the solar irradiance of the data-complete power stations;

[0073] According to the location information of the distributed photovoltaic power station, the pixel value p of the distributed photovoltaic power station is extracted from the corresponding position of the satellite cloud image;

[0074] Solar irradiance E0 without cloud attenuation and atmospheric external radiation intensity The relationship is as follows:

[0075] Where λ is the attenuation coefficient without considering cloud attenuation;

[0076] Obtain the longitude and latitude τ, tilt angle β, declination angle δ and hour angle ω of the data-complete power station, and calculate the unit irradiance cosz of the data-complete power station. The formula is as follows: cosz=sinδsin(τ-β)+cosδcos(τ-β)cosω

[0077] Complete the solar irradiance E of the power station based on the data x , complete data of the power station's unit irradiance cosz and atmospheric radiation intensity Calculate the data set λf(p x ), the formula is as follows:

[0078] The pixel value data set p of the data complete station x With the data set λf(p x ) performs function fitting to obtain function F(p);

[0079] Select a photovoltaic power station that is not equipped with a solar irradiance collection device and select a corresponding reference power station from the power stations with complete data, and obtain temperature data of each reference power station and the photovoltaic power station that is not equipped with a solar irradiance collection device;

[0080] The Pearson product-moment correlation coefficient r between each reference power station and the photovoltaic power station without solar irradiance collection device is calculated based on the temperature data of each reference power station and the photovoltaic power station without solar irradiance collection device. p , and the reference power station with the largest Pearson product-moment correlation coefficient is selected as the most relevant power station. The Pearson product-moment correlation coefficient r p The calculation formula is as follows:

[0081] Among them, x i is the temperature value of the reference power station, y i is the temperature value of the photovoltaic power station without solar irradiance collection device installed, x ave is the average temperature of the reference power station, y ave is the average temperature of the photovoltaic power station without solar irradiance collection device installed;

[0082] According to the solar irradiance E of the most relevant power station a , the pixel value p of the most relevant power station a And the pixel value p of the photovoltaic power station without solar irradiance collection device A Calculate the solar irradiance E of a photovoltaic power station without a solar irradiance collection device installed A , the formula is as follows: E A =F(p A )E a / F(p a )

[0083] Other photovoltaic power stations that are not equipped with solar irradiance collection devices are selected, and the solar irradiance of the photovoltaic power stations that are not equipped with solar irradiance collection devices is calculated according to the above method to obtain the solar irradiance of all photovoltaic power stations.

[0084] Step 5: Determine the mapping relationship between satellite cloud image characteristics and solar irradiance based on the characteristics of the satellite cloud image at the historical moment and the solar irradiance at the corresponding moment;

[0085] Based on the basic structure of BP network, Elman neural network adds a connection layer in the hidden layer as a one-step delay operator to achieve the purpose of memory, so that the system has the ability to adapt to time-varying characteristics. Satellite cloud image feature sequence is a typical time series, and Elman neural network has good adaptability to it.

[0086] In this embodiment, the satellite cloud image features at historical moments are used as the input of the Elman neural network, and the solar irradiance of the photovoltaic power station at the corresponding moment is used as the output of the Elman neural network. The Elman neural network is trained to obtain the mapping relationship between the satellite cloud image features and the solar irradiance.

[0087] Step 6: Based on the predicted satellite cloud image features at the future time and the mapping relationship between the satellite cloud image features and the solar irradiance, the solar irradiance of the distributed photovoltaic power station at the corresponding time in the future is obtained;

[0088] Step 7: Predict the power of the distributed photovoltaic power station at a future time based on the solar irradiance of the distributed photovoltaic power station and weather forecast information at a future time.

[0089] Since the power output of distributed photovoltaic power stations at future moments is affected by the power output of distributed photovoltaic power stations at previous moments, LSTM neural networks are selected for training. LSTM neural networks introduce memory units and gating units, which can selectively retain and forget information. The training takes solar irradiance information and the temperature at the corresponding moment in the weather forecast information as input, and the power of distributed photovoltaic power stations at the corresponding moment as output.

[0090] In this embodiment, step 7 includes:

[0091] The solar irradiance of photovoltaic power stations at historical moments and the temperature at corresponding moments in weather forecast information are used as inputs to the LSTM neural network, and the power of distributed photovoltaic power stations at corresponding moments is used as the output of the LSTM neural network. The LSTM neural network is trained to obtain the mapping relationship between solar irradiance and distributed photovoltaic power station power.

[0092] The solar irradiance of the distributed photovoltaic power station at the future moment and the temperature in the weather forecast information are input into the mapping relationship between the solar irradiance and the power of the distributed photovoltaic power station to obtain the power of the photovoltaic power station at the future moment.

[0093] Although the present invention has been disclosed above by way of embodiments, they are not intended to limit the present invention. Any person skilled in the art may make slight changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A distributed photovoltaic power station power prediction method, characterized in that: include: Obtain the satellite cloud images and weather forecast information at the historical moment and the current moment for the area where the distributed photovoltaic power station is located; Preprocessing the satellite cloud images at historical moments and the satellite cloud images at the current moment, and extracting features from the preprocessed satellite cloud images to obtain features of the satellite cloud images at historical moments and features of the satellite cloud images at the current moment; Predict the characteristics of satellite cloud images at future moments based on the characteristics of satellite cloud images at historical moments and the characteristics of satellite cloud images at current moments; Select several photovoltaic power station locations in the area where the distributed photovoltaic power station is located to install solar irradiance collection devices, and calculate the solar irradiance of photovoltaic power stations that do not have solar irradiance collection devices installed based on satellite cloud images, the collected solar irradiance and photovoltaic power station location information; According to the satellite cloud image characteristics at historical moments and the solar irradiance at corresponding moments, the mapping relationship between satellite cloud image characteristics and solar irradiance is determined; According to the predicted satellite cloud image features at future moments and the mapping relationship between satellite cloud image features and solar irradiance, the solar irradiance of the distributed photovoltaic power station at the corresponding moment in the future is obtained; According to the solar irradiance and weather forecast information of the distributed photovoltaic power station in the future, the power of the distributed photovoltaic power station in the future is predicted.

2. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The preprocessing of the satellite cloud images at the historical moment and the satellite cloud images at the current moment, and feature extraction of the preprocessed satellite cloud images, includes: Performing grayscale processing on the satellite cloud images at historical moments and the satellite cloud images at the current moment to obtain the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment; Perform linear change and histogram equalization on the satellite grayscale cloud images at historical moments and the satellite grayscale cloud images at the current moment to obtain the preprocessed satellite cloud images; Cloud pixel values ​​are extracted from the preprocessed satellite cloud images as cloud thickness features, and cloud texture features are extracted from the preprocessed satellite cloud images based on the gray-level co-occurrence matrix.

3. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The method of predicting the satellite cloud image features at future moments based on the satellite cloud image features at historical moments and the satellite cloud image features at current moments includes: The satellite cloud image features at historical moments and the satellite cloud image features at the current moment are sorted in time to obtain Satellite cloud image feature time series; Draw a trend graph of the satellite cloud image feature sequence based on the satellite cloud image feature time series, and determine whether the satellite cloud image feature time series is a stationary time series; If the satellite cloud image feature time series is a non-stationary time series, a d-order difference operation is performed on the satellite cloud image feature time series to obtain a stationary time series; Draw the autocorrelation diagram and partial autocorrelation diagram of the satellite cloud image feature time series, and determine the p-value and q-value of the ARIMA model based on the tailing and truncation of the autocorrelation diagram and partial autocorrelation diagram; The satellite cloud image features at historical moments and the satellite cloud image features at the current moment are input into the ARIMA model to obtain the satellite cloud image features at future moments.

4. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The method comprises: selecting several photovoltaic power station locations in the area where the distributed photovoltaic power station is located to install solar irradiance collection devices, and calculating the solar irradiance of the photovoltaic power station without the solar irradiance collection device installed based on the satellite cloud map, the collected solar irradiance and the photovoltaic power station location information, including: In the area where the distributed photovoltaic power station is located, n photovoltaic power stations are selected to install solar irradiance collection devices and recorded as data-complete power stations, and the solar irradiance of the data-complete power stations is obtained; Extract the pixel value of the distributed photovoltaic power station from the corresponding position of the satellite cloud image according to the location information of the distributed photovoltaic power station; Obtain the longitude and latitude τ, inclination angle β, declination angle δ and hour angle ω of the power station with complete data, and calculate the unit irradiance cosz of the power station with complete data. The formula is as follows: cosz=sinδsin(τ-β)+cosδcos(τ-β)cosω Complete the solar irradiance E of the power station based on the data x , complete data of the power station unit irradiance cosz and atmospheric radiation intensity Calculate the data set λf(p x ), the formula is as follows: The pixel value data set p of the data complete station x With the data set λf(p x ) to perform function fitting and obtain function F(p); Select a photovoltaic power station that is not equipped with a solar irradiance collection device and select a corresponding reference power station from the power stations with complete data, and obtain temperature data of each reference power station and the photovoltaic power station that is not equipped with a solar irradiance collection device; The Pearson product-moment relationship between each reference power station and the photovoltaic power station without the solar irradiance collection device is calculated based on the temperature data of each reference power station and the photovoltaic power station without the solar irradiance collection device. Number r p , and select the reference power station with the largest Pearson product-moment correlation coefficient as the most relevant power station. The Pearson product-moment correlation coefficient r p The calculation formula is as follows: Among them, x i is the temperature value of the reference power station, y i is the temperature value of the photovoltaic power station without solar irradiance collection device installed, x ave is the average temperature of the reference power station, y ave is the average temperature of the photovoltaic power station without solar irradiance collection device installed; According to the solar irradiance E of the most relevant power station a , the pixel value p of the most relevant power station a And the pixel value p of the photovoltaic power station without solar irradiance collection device installed A Calculate the solar irradiance E of a photovoltaic power station without a solar irradiance collection device installed A , the formula is as follows: E A =F(p A )E a / F(p a ) Other photovoltaic power stations that are not equipped with solar irradiance collection devices are selected, and the solar irradiance of the photovoltaic power stations that are not equipped with solar irradiance collection devices is calculated according to the above method to obtain the solar irradiance of all photovoltaic power stations.

5. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: Determining the mapping relationship between the satellite cloud image features and the solar irradiance at the corresponding time according to the satellite cloud image features includes: The satellite cloud image features at historical moments are used as the input of the Elman neural network, and the solar irradiance of the photovoltaic power station at the corresponding moment is used as the output of the Elman neural network. The Elman neural network is trained to obtain the mapping relationship between satellite cloud image features and solar irradiance.

6. The distributed photovoltaic power station power prediction method according to claim 1, characterized in that: The method of predicting the power of the distributed photovoltaic power station at a future time according to the solar irradiance of the distributed photovoltaic power station at a future time and the weather forecast information includes: The solar irradiance of photovoltaic power stations at historical moments and the temperature at corresponding moments in weather forecast information are used as the input of LSTM neural network, and the power of distributed photovoltaic power stations at corresponding moments is used as the output of LSTM neural network. The LSTM neural network is trained to obtain the mapping relationship between solar irradiance and distributed photovoltaic power station power. The solar irradiance of the distributed photovoltaic power station at the future moment and the temperature in the weather forecast information are input into the mapping relationship between the solar irradiance and the power of the distributed photovoltaic power station to obtain the power of the photovoltaic power station at the future moment.

Images