CN114004405B - Photovoltaic power prediction method and system based on Elman neural network and satellite cloud image - Google Patents

Abstract

The invention belongs to the technical field of photovoltaic power generation, and provides a photovoltaic power prediction method and a system based on an Elman neural network and a satellite cloud picture, wherein the method comprises the following steps: acquiring historical electricity utilization data and satellite images of an electricity utilization system to be predicted and preprocessing the historical electricity utilization data and satellite images; building an Elman dynamic recurrent neural network model, and inputting preprocessed data for training; inputting the preprocessed data into a trained Elman model, and outputting a prediction result; extracting the gray value of the satellite cloud picture with high precision, and preparing for modeling an Elman neural network prediction model; simultaneously considering historical operation data and an Elman photovoltaic power prediction model and algorithm of a satellite cloud image gray value; and obtaining model error evaluation by an error calculation mode of the measured value and the true value, thereby improving the accuracy of power prediction.

Description

Photovoltaic power prediction method and system based on Elman neural network and satellite cloud image

Technical Field

The present invention belongs to the technical field related to photovoltaic power generation, and in particular relates to a photovoltaic power prediction method and system based on Elman neural network and satellite cloud image.

Background Art

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

With the acceleration of industrialization and electrification, human demand for energy has soared, especially the demand for electricity, which has shown an upward trend year by year. For power generation, the consumption burden of disposable energy such as coal, oil, and natural gas has increased, and the problems of cost and pollution are particularly prominent. New energy power generation dominated by photovoltaics has been vigorously developed, but due to the instability of weather conditions, photovoltaic power generation has a strong intermittent and random nature, which poses a challenge to the planning and operation of the existing power system.

At present, the photovoltaic power prediction tools that are used more are mainly artificial neural networks. Among them, Elman and BP neural networks, as recursive neural networks, both belong to artificial neural networks. However, due to the memory function of its receiving layer, the global stability of the Elman neural network is higher than that of the BP neural network. In addition, considering the weather factors, on rainy days, the movement and dissipation of clouds present inertia-free mutations, which brings difficulties to the prediction of photovoltaic power. The use of cloud maps to predict photovoltaic power has begun to be explored. The cloud maps used for photovoltaic power prediction in existing studies are mainly divided into two categories: ground-based cloud maps and satellite cloud maps. Among them, there are many studies on photovoltaic power prediction methods based on ground-based cloud maps. However, the observation range of ground-based cloud maps is limited, the cost of installation and maintenance is high, and it is difficult to apply them widely. Satellite cloud maps are relatively easy to obtain on domestic meteorological data websites and have a wide range, so the research on photovoltaic power prediction based on satellite cloud maps has gradually attracted attention. The present invention mainly solves the accurate prediction of photovoltaic power to improve the reliability of power system operation, which can reduce electricity costs, reduce energy consumption, save energy and reduce emissions, and improve economic benefits.

The inventors found that in the photovoltaic power prediction model, the historical power data contains limited features, and the traditional recursive network BP neural network has no dynamic characteristics, slow convergence speed, and poor global stability. This results in large errors in photovoltaic power prediction and cannot improve the prediction accuracy.

Summary of the invention

In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a photovoltaic power prediction method and system based on Elman neural network and satellite cloud map, which solves the problems of limited data characteristics and global stability, effectively utilizes meteorological laws, and thus improves the accuracy of power prediction.

In order to achieve the above object, the present invention adopts the following technical solution:

A first aspect of the present invention provides a photovoltaic power prediction method based on Elman neural network and satellite cloud images, comprising:

Obtain historical power consumption data and satellite images of the power consumption system to be predicted and perform preprocessing;

Build the Elman dynamic recurrent neural network model and input the preprocessed data for training;

Input the preprocessed data into the trained Elman model and output the prediction results.

A second aspect of the present invention provides a photovoltaic power prediction system based on Elman neural network and satellite cloud images, comprising:

A data acquisition module is configured to acquire historical power consumption data and satellite images of the power consumption system to be predicted and perform preprocessing;

The model building module is configured to build an Elman dynamic recurrent neural network model and input preprocessed data for training;

The photovoltaic power prediction module is configured to input the preprocessed data into the trained Elman model and output the prediction result.

A third aspect of the present invention provides a computer-readable storage medium.

A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the photovoltaic power prediction method based on Elman neural network and satellite cloud map as described in the first aspect above.

A fourth aspect of the present invention provides a computer device.

A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the photovoltaic power prediction method based on the Elman neural network and satellite cloud image as described in the first aspect above are implemented.

Compared with the prior art, the present invention has the following beneficial effects:

1. The present invention extracts the gray value of satellite cloud images with high precision to prepare for the modeling of Elman neural network prediction model; the Elman photovoltaic power prediction model and algorithm that considers historical operation data and satellite cloud image gray value at the same time; the error calculation method between the measured value and the true value obtains the model error evaluation, thereby improving the accuracy of power prediction;

2. The present invention uses satellite cloud images and historical power generation data as feature inputs, further enriching the data set and improving the prediction accuracy to a certain extent; the Elman model is selected, which is superior to the BP model and has the advantages of fast convergence speed and good global stability. At the same time, it also makes up for the shortcomings of the BP model of no memory and static nature.

3. The Elman model innovatively combined with satellite data is applied to the field of photovoltaic power prediction. ;

Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

FIG1 is a photovoltaic power prediction process of an Elman neural network according to a first embodiment of the present invention;

FIG. 2( a ) is an example of a satellite cloud image at 9:00 on March 21 according to the first embodiment of the present invention;

FIG2( b ) is an example of a satellite cloud image at 8:00 on April 13 according to the first embodiment of the present invention;

FIG3( a ) is a grayscale image of a satellite cloud image at 8 o'clock on March 4 according to the first embodiment of the present invention;

FIG3( b ) is a grayscale image of a satellite cloud image at 10:15 on March 9 according to the first embodiment of the present invention;

FIG3( c ) is a grayscale image of a satellite cloud image at 10:15 on March 22 according to the first embodiment of the present invention;

FIG3( d ) is a grayscale image of the satellite cloud image at 8 o'clock on April 13 according to the first embodiment of the present invention;

FIG4 is a statistical diagram of the number of clusters in each gray value interval according to the first embodiment of the present invention;

FIG5 is a diagram of the structure of an Elman neural network according to the first embodiment of the present invention;

FIG6 is a test error evaluation diagram of Model 1 according to Embodiment 1 of the present invention;

FIG7 is a test error evaluation diagram of Model 2 of Embodiment 1 of the present invention;

FIG8 is a bar graph comparing power prediction values and actual values according to the first embodiment of the present invention.

DETAILED DESCRIPTION

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and/or combinations thereof.

It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and/or block diagram, and the combination of boxes in the flowchart and/or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

Embodiment 1

As shown in Figures 1-8, this embodiment provides a photovoltaic power prediction method based on Elman neural network and satellite cloud map. This embodiment uses the method applied to the server as an example for illustration. It can be understood that the method can also be applied to terminals, and can also be applied to terminals, servers and systems, and implemented through the interaction between terminals and servers. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application. In this embodiment, the method includes the following steps:

Step S100: Acquire historical power consumption data and satellite images of the power consumption system to be predicted and perform preprocessing;

Data acquisition and satellite cloud image processing, data normalization, and division of data sets into training data and test data.

Step S200: Building an Elman dynamic recursive neural network model and inputting preprocessed data for training;

The training data is input into the Elman network model for model training.

Step S300: input the preprocessed data into the trained Elman model and output the prediction result;

The test data is input into the trained model to obtain the predicted value, and then denormalized to obtain the final predicted value;

The error between the final predicted value and the true value is calculated to obtain the model error evaluation.

Specifically, step S100 includes satellite image acquisition, image grayscale processing, and grayscale value interval division.

Step S101: First, the photovoltaic power data comes from a photovoltaic power plant in Zhangqiu District, Shandong Province. The photovoltaic power plant has an installed capacity of 10MW and covers an area of 405 acres. The annual average grid-connected power generation is expected to reach 12 million kWh, equivalent to saving 3,670 tons of coal. The historical photovoltaic power generation data of the power plant from March to April 2021 is mainly used, with an interval of 15 minutes.

The latitude of the selected area of the satellite cloud image is [117.00-118.00] and the longitude is [36-37], which contains 512*450 pixels and has a resolution of 1000m. The satellite cloud image from 8:00-10:30 in the morning was selected, and the power data was kept corresponding to the satellite cloud image data, and 103 data points were finally determined. When the cloud layer is thicker, the reflection of sunlight is stronger, and the pixel value of the corresponding area is larger. On the contrary, when the sky is clear, the pixel value of the corresponding area is smaller. As shown in Figure 2, examples of visible light images of Zhangqiu District, Shandong Province on March 21 and April 13, 2021 are shown.

It should be noted that the historical data is the power generation data of the power plant, with a resolution of 15 minutes. Combined with the cloud map data, the data from 8:00 to 10:30 is taken. The cloud map data is the image data clipped by the longitude and latitude of the power plant area, with a resolution of 1000m. Considering the characteristics of visible light images, the data from 8:00 to 10:30 is taken.

Step S102: Then grayscale the image. According to the precise longitude and latitude of the known power plant, the picture is cropped to obtain a satellite cloud image of 50*50 pixels. Then the cloud image is grayscaled, that is, each pixel in the pixel matrix satisfies R=G=B, and the grayscale value of each pixel is extracted. The present invention selects satellite cloud images at four different time points for grayscale, and obtains a grayscale image as shown in Figure 3. It is known that the power generation at four moments is Pa=1840kw, Pb=5902kw, Pc=7169kw, and Pd=2869kw. It can be obtained that: the smaller the grayscale value of the satellite cloud image, the sparser the cloud layer, and the greater the power generation; the larger the grayscale value, the thicker the cloud layer, and the smaller the power generation.

Step S103: Finally, grayscale value interval division. The grayscale value is divided into interval clusters. Taking the four satellite cloud images mentioned above as an example, the grayscale value range is 0-255, and 10 intervals are clustered. The statistical results are shown in Table 1. The bar chart comparison is shown in Figure 4. The results show that the number of pixels obtained by interval clustering still follows the relationship between the grayscale value and the power generation. Therefore, a method of using the interval clustering statistical results as a model input to predict the power generation is proposed. The following is an explanation with a specific example.

Table 1 is a statistical table of the number of pixels in each gray value interval.

Table 1

Step S104: Data normalization. Before the network training of the historical power generation and satellite cloud image gray value information, the gray value interval clustering statistics results and the historical power generation are normalized to make their value range [0,1].

Step S105: Divide the data set into training data and test data. 90% of the data points are used for training, and 10% of the data points are used for testing the model and performing error evaluation.

Step S200: including model building and basic parameter selection

Step S201: Elman neural network is a typical dynamic recursive neural network. Compared with BP neural network, it adds a successor layer in the hidden layer. Its network structure is generally divided into four layers: input layer, hidden layer, successor layer and output layer. Its structure diagram is shown in Figure 5. Among them, u is the input vector, y is the output vector, x is the n-dimensional hidden layer unit vector, c is the n-dimensional feedback vector of the successor layer, _w1 , _w2 , _w3 are the connection weights from the successor layer to the hidden layer, the input layer to the hidden layer, and the hidden layer to the output layer respectively. The calculation formula of the network is as follows:

y(t)＝g(w ₃ x(t)) (1)

x(t)＝f(w ₁ c(t)+w ₂ (u(t-1))) (2)

c(t)＝x(t-1) (3)

The input layer units play a role in signal transmission, and the output layer units play a role in weighting. The receiving layer is used to memorize the output value of the hidden layer unit at the previous moment, and can be considered as a delay operator with a one-step delay. The hidden layer unit usually adopts the Signmoid nonlinear activation function. It connects its output to its input through the delay and storage of the receiving layer. This self-connection method makes it sensitive to historical data. The addition of the internal feedback network increases the ability of the network itself to process dynamic information, thereby achieving the purpose of dynamic modeling. At the same time, it enables the system to adapt to time-varying characteristics and enhances the global stability of the network.

Step S202: The basic parameters of the Elman neural network are set as follows:

① Determination of input and output layer nodes. Since the present invention mainly studies the influence of satellite cloud image information on the accuracy of the prediction model, the input nodes of different models are based on the different data volumes. The input variable of model 1 is the historical power generation, so the number of input layer nodes is 1; model 2 adds the number of interval clusters of satellite cloud image gray values, and the number of input layer nodes is 11. The output is the actual power at the current moment.

② Selection of the number of hidden layer nodes. The number of hidden layer neurons has a great influence on the performance of the neural network and will affect the accuracy of the prediction. When the number of hidden layer neurons is too small, the network cannot conduct comprehensive learning and the prediction results are not accurate; when the number is too large, the learning speed is reduced and the "overfitting" phenomenon may occur. The present invention determines the number of hidden layer nodes based on the empirical formula:

Wherein: M _h , M _i , M _o are the number of nodes in the hidden layer, input layer and output layer respectively; a is usually 1 to 8. According to the empirical formula, different hidden layer nodes are tried to be input, and the error value of each training is compared. The present invention finally selects 3 hidden layer nodes.

In step S300, the following steps are specifically included:

Step S301: input the test data into the trained model to obtain the predicted value and perform denormalization. Denormalization is to convert the dimensionless pre-test output by the model into the predicted value in actual power units as the final predicted value.

Step S302: Calculate the error between the final predicted value and the true value to obtain a model error evaluation.

In order to compare the impact of adding satellite cloud images on the prediction accuracy, two sets of prediction values are output and measured by the error with the true value. The errors used are the root mean square error (RMSE) and the relative coefficient of variation. The root mean square error is a measure of the deviation between the predicted value and the true value, and has the same dimension as the power generation. In order to eliminate the dimension and more intuitively analyze the prediction accuracy of the model, the present invention introduces the relative coefficient of variation, that is, the ratio of the root mean square error to the average power generation. The corresponding calculation formula is as follows:

Where: _Pi is the actual value of photovoltaic output power; _Pf is the predicted value of photovoltaic power; N is the total number of data; is the average value of photovoltaic power generation, and RCV is the relative coefficient of variation.

Model 1: The power generation at the previous moment is used to predict the power generation at the next moment. The model test results are shown in Figure 6.

Model 2: The power generation at the previous moment and the number of grayscale interval clusters extracted from the satellite cloud image are used as input to predict the power generation at the next moment. The model test results are shown in Figure 7. The power prediction values and actual values of the two models are not much different.

In order to more intuitively compare the prediction accuracy of the two models, the prediction results are shown in Figure 8. The prediction accuracy of different models is listed in Table 2. Combined with the statistical results in Table 2, after adding the pixel information of the satellite cloud image, the accuracy of Model 2 is improved overall, among which the root mean square error is reduced from 191.3141 to 154.7513, and the RMSE-C is reduced from 0.05415 to 0.04010, indicating that the addition of satellite cloud image pixel information is feasible and reliable for improving accuracy. The reason is that the addition of satellite cloud image pixel information can reflect the solar irradiation information at the time of prediction. The more useful information is input, the more accurate the prediction will be.

Table 2 Prediction accuracy of different models

Table 2

RMSE RMSE-C Model 1 191.3141 0.05415 Model 2 142.0417 0.04010

Embodiment 2

This embodiment provides a photovoltaic power prediction system based on Elman neural network and satellite cloud images, including:

A data acquisition module is configured to acquire historical power consumption data and satellite images of the power consumption system to be predicted and perform preprocessing;

The model building module is configured to build an Elman dynamic recurrent neural network model and input preprocessed data for training;

The photovoltaic power prediction module is configured to input the preprocessed data into the trained Elman model and output the prediction result.

It should be noted that the above data acquisition module, model building module and photovoltaic power prediction module correspond to steps S100 to S300 in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Embodiment 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

Embodiment 3

This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the photovoltaic power prediction method based on Elman neural network and satellite cloud map as described in the first embodiment above are implemented.

Embodiment 4

This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the photovoltaic power prediction method based on the Elman neural network and satellite cloud map as described in the first embodiment are implemented.

Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program codes.

The present invention is described with reference to the flowchart and/or block diagram of the method, device (system), and computer program product according to the embodiment of the present invention. It should be understood that each process and/or box in the flowchart and/or block diagram, as well as the combination of the process and/or box in the flowchart and/or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims (7)

1. The photovoltaic power prediction method based on the Elman neural network and the satellite cloud picture is characterized by comprising the following steps of:

acquiring historical electricity utilization data and satellite images of an electricity utilization system to be predicted and preprocessing the historical electricity utilization data and satellite images;

building an Elman dynamic recurrent neural network model, and inputting preprocessed data for training;

inputting the preprocessed data into a trained Elman model, and outputting a prediction result;

the method comprises the steps of obtaining historical electricity consumption data and satellite images of an electricity consumption system to be predicted and preprocessing, and specifically comprises the following steps:

Acquiring historical photovoltaic power generation power data and a position satellite cloud picture of a power plant;

carrying out image graying treatment on the satellite cloud image, and carrying out interval clustering division on gray values;

Carrying out normalization processing on the gray value interval clustering division result and the historical power generation power to form a data set;

the clustering and dividing result of the gray value intervals is the number of pixel points of each gray value interval;

Dividing the dataset into training data and test data;

the Elman dynamic recurrent neural network model is built, and the preprocessed data is input for training, specifically:

building an Elman model, wherein the Elman model comprises an input layer, a hidden layer, a bearing layer and an output layer;

Setting parameters of an Elman model;

Inputting training data in the preprocessed data into an Elman model;

Obtaining a trained Elman model;

The setting of the parameters of the Elman model is specifically as follows:

the satellite cloud image and the historical power generation data are taken as characteristic input together, the output is the actual power at the current moment, and the nodes of an input layer and an output layer are determined;

the node number of the hidden layer is selected, and the specific formula is as follows:

；

Wherein: m _h、M_i、M_o is the number of nodes of the hidden layer, the input layer and the output layer in sequence; a is 1-8.

2. The photovoltaic power prediction method based on Elman neural network and satellite cloud image as claimed in claim 1, wherein the preprocessing data is input into a trained Elman model, and the prediction result is output, specifically:

inputting test data in the preprocessed data into a trained Elman model to obtain a predicted value;

Inversely normalizing the predicted value to obtain a predicted value of the actual power;

And carrying out error calculation on the predicted value of the actual power and the actual photovoltaic power generation power to obtain model error evaluation.

3. The photovoltaic power prediction method based on Elman neural network and satellite cloud image as claimed in claim 2, wherein the error calculation is performed on the predicted value of the actual power and the actual photovoltaic power generation power to obtain a model error evaluation, specifically:

；

；

；

Wherein: p _i is the actual value of the output power of the photovoltaic; p _f is the photovoltaic power prediction value; n is the total number of data; is the average value of the photovoltaic power generation power; RCV is the relative coefficient of variation.

4. A photovoltaic power prediction method based on Elman neural network and satellite cloud as claimed in claim 1, wherein the dataset is divided into training data and test data, 90% of the data points are used for training, 10% of the data points are used for testing the model and evaluating the error.

5. A photovoltaic power prediction system based on Elman neural network and satellite cloud image, comprising:

the data acquisition module is configured to acquire historical electricity utilization data and satellite images of the electricity utilization system to be predicted and perform preprocessing;

the model building module is configured to build an Elman dynamic recurrent neural network model, and inputs preprocessed data for training;

The photovoltaic power prediction module is configured to input the preprocessed data into a trained Elman model and output a prediction result;

the method comprises the steps of obtaining historical electricity consumption data and satellite images of an electricity consumption system to be predicted and preprocessing, and specifically comprises the following steps:

Acquiring historical photovoltaic power generation power data and a position satellite cloud picture of a power plant;

carrying out image graying treatment on the satellite cloud image, and carrying out interval clustering division on gray values;

Carrying out normalization processing on the gray value interval clustering division result and the historical power generation power to form a data set;

the clustering and dividing result of the gray value intervals is the number of pixel points of each gray value interval;

Dividing the dataset into training data and test data;

the Elman dynamic recurrent neural network model is built, and the preprocessed data is input for training, specifically:

building an Elman model, wherein the Elman model comprises an input layer, a hidden layer, a bearing layer and an output layer;

Setting parameters of an Elman model;

Inputting training data in the preprocessed data into an Elman model;

Obtaining a trained Elman model;

The setting of the parameters of the Elman model is specifically as follows:

the satellite cloud image and the historical power generation data are taken as characteristic input together, the output is the actual power at the current moment, and the nodes of an input layer and an output layer are determined;

the node number of the hidden layer is selected, and the specific formula is as follows:

；

Wherein: m _h、M_i、M_o is the number of nodes of the hidden layer, the input layer and the output layer in sequence; a is 1-8.

6. A computer readable storage medium, on which a computer program is stored, which program, when being executed by a processor, implements the steps of a photovoltaic power prediction method based on Elman neural network and satellite cloud according to any of claims 1-4.

7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps of a photovoltaic power prediction method based on Elman neural network and satellite cloud according to any of claims 1-4.