Photovoltaic output prediction method, device, electronic device and storage medium

Abstract

The present application discloses a photovoltaic output prediction method, device, electronic device and storage medium. The photovoltaic output prediction method includes: obtaining multiple first historical satellite cloud images within a first preset time period in a first target area; extracting the first historical cloud occlusion feature in the first historical satellite cloud image; predicting the cloud occlusion feature within a second preset time period based on the first historical cloud occlusion feature; obtaining the first weather information within the second preset time period; inputting the cloud occlusion feature and the first weather information into a preset model to obtain a photovoltaic output prediction value, wherein the preset model is obtained based on ConvTransformer model training. Thus, the prediction accuracy of photovoltaic output is improved.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic output prediction, and specifically to a photovoltaic output prediction method, device, electronic device and storage medium. Background Art

[0002] Photovoltaic has the advantages of small scale, high practicality, and easy integration with buildings. It has been widely promoted in my country in recent years. However, large-scale photovoltaic multi-point and disorderly access to the distribution network has caused the distribution network to face a series of problems such as power quality, harmonics, and economic operation, which seriously affects the safe and stable operation of the power grid. Therefore, scientifically predicting photovoltaic output is an important guarantee for achieving reliable, efficient and economical photovoltaic grid-connected operation.

[0003] Affected by the weather, photovoltaic power generation is characterized by randomness and volatility. For example, in cloudy and rainy weather, the photovoltaic output fluctuates greatly. Clouds are one of the main factors affecting the solar irradiance. They block the solar radiation, thus affecting the photovoltaic output. How to solve the problem of unpredictable photovoltaic output caused by cloud blocking is a hot issue in current research.

[0004] It can be seen that there are technical problems in the existing photovoltaic output that are difficult to accurately predict. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a photovoltaic output prediction method, device, electronic device and storage medium to solve the technical problem in the prior art that photovoltaic output is difficult to accurately predict.

[0006] In order to achieve the above-mentioned purpose, the first aspect of the present application provides a photovoltaic output prediction method, comprising:

[0007] Acquire a plurality of first historical satellite cloud images within a first preset time period in a first target area;

[0008] Extract the first historical cloud cluster occlusion features in the first historical satellite cloud image;

[0009] Predicting cloud occlusion characteristics within a second preset time period based on the first historical cloud occlusion characteristics;

[0010] Obtaining first weather information within a second preset time period;

[0011] The cloud occlusion characteristics and the first weather information are input into the preset model to obtain the photovoltaic output prediction value, wherein the preset model is trained based on the ConvTransformer model.

[0012] In an embodiment of the present application, extracting the first historical cloud cluster occlusion feature in the first historical satellite cloud image includes:

[0013] The first historical satellite cloud image is segmented by using a gradient vector flow model to obtain a plurality of historical cloud cluster images in the first historical satellite cloud image;

[0014] The gray-level co-occurrence matrix is ​​used to extract features from multiple historical cloud images, and the first historical cloud occlusion features in the first historical satellite cloud image are obtained.

[0015] In the embodiment of the present application, before extracting the first historical cloud cluster occlusion feature in the first historical satellite cloud image, the method further includes:

[0016] Preprocess the first historical satellite cloud image.

[0017] In the embodiment of the present application, the first historical satellite cloud image is preprocessed, including:

[0018] Performing grayscale histogram equalization on the first historical satellite cloud image to obtain the equalized first historical satellite cloud image;

[0019] The equalized first historical satellite cloud image is smoothed and filtered.

[0020] In the embodiment of the present application, smoothing filtering is performed on the equalized first historical satellite cloud image, including:

[0021] The Gaussian noise in the equalized first historical satellite cloud image is removed by using mean filtering;

[0022] The median filter is used to remove the salt and pepper noise in the equalized first historical satellite cloud image.

[0023] In an embodiment of the present application, the first historical cloud occlusion feature includes historical cloud thickness and historical cloud texture.

[0024] In the embodiment of the present application, the training steps of the preset model include:

[0025] Acquire a plurality of second historical satellite cloud images, second weather information and corresponding photovoltaic output historical values ​​in a third preset time period in a second target area;

[0026] Extract the second historical cloud cluster occlusion features in the second historical satellite cloud image;

[0027] The preset model based on the ConvTransformer model is trained according to the second weather information, the second historical cloud occlusion characteristics and the historical values ​​of photovoltaic output.

[0028] A second aspect of the present application provides a photovoltaic output prediction device, characterized by comprising:

[0029] a memory configured to store instructions; and

[0030] A processor is configured to call instructions from a memory and implement the photovoltaic output prediction method according to any one of the first aspects when executing the instructions.

[0031] A third aspect of the present application provides an electronic device, characterized in that it includes:

[0032] According to the photovoltaic output prediction device described in the second aspect.

[0033] A fourth aspect of the present application provides a machine-readable storage medium, characterized in that instructions are stored on the machine-readable storage medium, and the instructions are used to enable a machine to execute the photovoltaic output prediction method according to any one of the first aspects.

[0034] Through the above technical scheme, the first historical cloud obstruction characteristics are obtained according to the first historical satellite cloud image, and combined with the first weather information, the photovoltaic output is predicted using a preset model based on the ConvTransformer model, which can reduce the impact of high temperature, rain, cloud obstruction or other severe weather on the photovoltaic output prediction, thereby effectively improving the prediction accuracy of photovoltaic output.

[0035] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0037] Figure 1 A schematic diagram of a photovoltaic output prediction method according to an embodiment of the present application is shown;

[0038] Figure 2 The schematic diagram shows the structure of a photovoltaic output prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0040] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0041] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0042] Figure 1 The following schematically shows a flow chart of a photovoltaic output prediction method according to an embodiment of the present application. Figure 1 As shown, an embodiment of the present application provides a photovoltaic output prediction method, which may include the following steps.

[0043] Step 110: Acquire a plurality of first historical satellite cloud images within a first preset time period in a first target area;

[0044] It is understandable that since the photovoltaic output prediction method provided in the embodiment of the present application needs to be predicted in combination with the cloud occlusion characteristics, it is necessary to first obtain a cloud map, and then obtain the cloud occlusion characteristics from the cloud map. Common cloud maps may include ground-based cloud maps and satellite cloud maps. Ground-based cloud maps are images of clouds above the test station collected in real time by a ground-based all-sky imager, and satellite cloud maps are images obtained by observing the earth as a whole through satellites. Compared with satellite cloud maps, ground-based cloud maps have higher temporal and spatial resolutions, but the price of all-sky imagers is high and is not suitable for photovoltaic power stations with low investment costs. Therefore, the embodiment of the present application uses satellite cloud maps. First, a plurality of first historical satellite cloud maps within a first preset time period in a first target area are obtained by using satellites to provide an image source for the subsequent extraction of the first historical cloud occlusion characteristics from the first historical satellite cloud maps.

[0045] It should be noted that the first target area and the first preset time period can be set according to actual needs, and the embodiment of the present application does not limit this.

[0046] Step 120, extracting the first historical cloud cluster occlusion feature in the first historical satellite cloud image;

[0047] It is understandable that photovoltaic output refers to the product of the current and voltage generated by a photovoltaic cell or photovoltaic module under specific conditions, and is used to represent the power output capacity of a photovoltaic site under given conditions. Affected by the weather, photovoltaic power generation is characterized by randomness and strong volatility. For example, in cloudy and rainy weather, the photovoltaic output fluctuates greatly, and clouds, as one of the main factors affecting the solar irradiance, will block solar radiation, thereby affecting photovoltaic output. Therefore, the embodiment of the present application extracts the first historical cloud occlusion feature from the first historical satellite cloud image obtained. There are many ways to extract the first historical cloud occlusion feature, such as deep learning methods, image segmentation methods or other texture feature extraction methods. By combining the first historical cloud occlusion feature to predict the photovoltaic output, the accuracy of the photovoltaic output prediction can be improved.

[0048] In the embodiment of the present application, step 120 includes:

[0049] Step 121: Use a gradient vector flow model to perform image segmentation on the first historical satellite cloud image to obtain a plurality of historical cloud cluster images in the first historical satellite cloud image;

[0050] Step 122: Use the gray level co-occurrence matrix to extract features from the multiple historical cloud images to obtain the first historical cloud occlusion features in the first historical satellite cloud image.

[0051] It can be understood that the Gradient Vector Flow (GVF) model is a computer vision model for edge detection and image segmentation. It is an improvement based on the active contour model, aiming to overcome some limitations of sensitivity and convergence speed in the traditional active contour model. The active contour model defines an initial contour in the image and uses energy minimization to make the contour converge to the edge of the object of interest. However, the traditional active contour model may converge to the wrong edge or stagnate at a local minimum when the edge is unclear or there is noise, while the gradient vector flow model introduces the gradient vector flow as a force field to help the active contour model better adapt to the complex edges in the image. It calculates the gradient field of the image to generate a smooth force field that attracts the contour to the target edge, thereby improving the perception and attraction ability of the active contour model to the edge.

[0052] Specifically, firstly, the area to be segmented is determined in the first historical satellite cloud image, and the gradient vector flow model is used to extract the edge of each area to be segmented. The extracted result will contain real edges and pseudo edges, wherein the real edge refers to the edge that can be used to represent the real cloud cluster, and the pseudo edge refers to the edge that cannot be used to represent the real cloud cluster. Therefore, it is necessary to remove the pseudo edge, connect the real edge end to end, and use the gradient vector flow model to extract the real edge after connecting the end to end again, and remove the pseudo edge from the extracted result again, and connect the real edge end to end, and repeat the above steps until the area to be segmented no longer contains pseudo edges and the real edges are all connected into multiple complete historical cloud cluster images.

[0053] By using the gradient vector flow model to perform image segmentation on the first historical satellite cloud image, multiple historical cloud images in the first historical satellite cloud image are obtained, which is beneficial to reducing the background and other interference information in the first historical satellite cloud image, and the gradient vector flow model can accurately capture the edge information of the first historical satellite cloud image. Using these clear edge features, the area of ​​the historical cloud image can be more accurately defined, so that when subsequently extracting the first historical cloud cluster occlusion features, more focus will be placed on the structure and texture features of the cloud cluster rather than the surrounding environment, thereby improving the accuracy of extracting the first historical cloud cluster occlusion features.

[0054] It can be understood that the Gray-Level Co-occurrence Matrix (GLCM) is a statistical method for describing image texture features. It is often used in image classification, texture recognition, and feature extraction. It can capture the spatial relationship and distribution characteristics between the gray levels of pixels in the image. The gray-level co-occurrence matrix describes the texture features based on the relative position relationship between the gray levels of pixels in the image. It constructs a matrix by counting the frequency and spatial distribution of specific pixel pairs in the image. The elements in this matrix represent the probability distribution of specific gray-level pairs in the image at a given direction and distance. It can be used to calculate a series of statistical features. By analyzing the statistical characteristics of the gray-level co-occurrence matrix, the texture features of the image can be quantitatively described, which can be used for image classification, recognition, or feature extraction.

[0055] Specifically, take any point (x, y) in the historical cloud image and another point (x+a, y+b) that deviates from this point by a distance a, and let the gray value corresponding to this point be (g1, g2). Let the point (x, y) move on the entire historical cloud image, and then various (g1, g2) values ​​can be obtained. Let the number of gray values ​​be L, then there are L combinations of (g1, g2) in total. 2For the entire image, the number of occurrences of each value (g1, g2) is counted, and then arranged into a square matrix, which is the gray-level co-occurrence matrix. The gray-level co-occurrence matrix can be used to calculate the pixel mean, entropy and correlation in the historical cloud image. By analyzing the pixel mean, entropy and correlation in the historical cloud image, the first historical cloud occlusion feature in the first historical satellite cloud image can be obtained.

[0056] By using the gray-level co-occurrence matrix, the first historical cloud occlusion features in the first historical satellite cloud image can be accurately extracted and quantified, which is beneficial for the subsequent prediction of the cloud occlusion features within the second preset time period.

[0057] In the embodiment of the present application, before step 120, the method further includes:

[0058] Preprocess the first historical satellite cloud image.

[0059] It is understandable that due to the large spatial resolution of the first historical satellite cloud image initially acquired, the detailed information on the image is insufficient, which will cause the extracted features to be inaccurate and inadequately detailed, and the low-resolution image will have more noise or pixel-level errors, which will affect the accuracy of image segmentation, causing the extracted features to be disturbed or distorted, thereby affecting the accuracy of subsequent photovoltaic output predictions. Therefore, before extracting the first historical cloud cluster occlusion features in the first historical satellite cloud image, the first historical satellite cloud image needs to be preprocessed, such as using image enhancement, high resolution, filtering and smoothing operations, which can help enhance the image features of the first historical satellite cloud image, reduce noise and blur, and provide a more reliable data basis for subsequent photovoltaic output predictions.

[0060] In the embodiment of the present application, the first historical satellite cloud image is preprocessed, including:

[0061] Performing grayscale histogram equalization on the first historical satellite cloud image to obtain the equalized first historical satellite cloud image;

[0062] The equalized first historical satellite cloud image is smoothed and filtered.

[0063] It is understandable that grayscale histogram equalization is an image processing technology used to enhance image contrast. In the process of image recognition, increasing grayscale contrast can highlight important features of the image. Grayscale histogram equalization is to change the distribution of pixels on each grayscale level so that they all have the same number of pixels. The purpose is to make the image evenly distributed within the entire dynamic range of grayscale values, improve the brightness distribution of the image, and thus enhance the visual effect of the image.

[0064] In an embodiment of the present application, the first historical satellite cloud image may not have obvious contrast of clouds or cloud layers due to lighting conditions and other reasons. Grayscale histogram equalization can enhance the contrast of the image, making the boundaries between clouds and other backgrounds or objects clearer; and the texture information of clouds in the satellite cloud image is very important for the subsequent extraction of occlusion features of the first historical clouds, while sometimes images with low contrast or narrow grayscale range may cause unclear texture information. Grayscale histogram equalization can make the texture information more prominent, which is convenient for the subsequent extraction of occlusion features of the first historical clouds.

[0065] It is understandable that since the first historical satellite cloud image is inevitably interfered by noise during transmission and reception, and the noise advantage will affect the quality of the first historical satellite cloud image, it is necessary to denoise the image after using grayscale histogram equalization to highlight the important features of the image. In the embodiment of the present application, the equalized first historical satellite cloud image is smoothed and filtered. Smoothing filtering is an image processing technology that aims to reduce noise in the image and blur the details in the image. It is often used to remove noise, reduce changes in the image, and smooth the brightness changes of the image. The core idea is to process the pixels in the image to reduce the drastic changes between pixel values. Smoothing filtering usually operates based on the neighborhood or window of the pixels in the image, and the pixel values ​​are counted or calculated within the window, and then these statistical values ​​are used to replace the value of the central pixel to reduce the impact of noise, blur details, and reduce image changes. Common smoothing filtering methods include mean filtering, Gaussian filtering, median filtering, and bilateral filtering. Different filtering methods are suitable for different types of noise. The selection of a suitable filtering method needs to be determined according to specific image features and processing requirements.

[0066] In the embodiment of the present application, smoothing filtering is performed on the equalized first historical satellite cloud image, including:

[0067] The Gaussian noise in the equalized first historical satellite cloud image is removed by using mean filtering;

[0068] The median filter is used to remove the salt and pepper noise in the equalized first historical satellite cloud image.

[0069] It can be understood that mean filtering and median filtering are two common smoothing filtering techniques. Mean filtering replaces the value of the central pixel by taking the average value of the pixel values ​​in the neighborhood around the pixel, so that each pixel is replaced by the average value of the surrounding pixels. Mean filtering is simple and easy to implement, and can effectively reduce continuous noise such as Gaussian noise in the image; median filtering replaces the central pixel by selecting the median value of the pixel value in the pixel neighborhood, which is suitable for removing discrete noise such as salt and pepper noise, can retain the edges and details of the image, and has a good effect on retaining image features.

[0070] In the first historical satellite cloud image, common noise is Gaussian noise or salt and pepper noise. Mean filtering and median filtering have good adaptability to these two types of noise, and are respectively suitable for removing Gaussian noise and salt and pepper noise. Compared with mean filtering and median filtering, Gaussian filtering and bilateral filtering are more complex and time-consuming in calculation. Therefore, when actually processing large-scale satellite cloud images, in order to improve processing efficiency and speed, the embodiment of the present application uses mean filtering to remove Gaussian noise in the equalized first historical satellite cloud image, and uses median filtering to remove salt and pepper noise in the equalized first historical satellite cloud image, which helps to improve the quality of the image, reduce the impact of noise, and improve the overall characteristics of the image, thereby facilitating the subsequent extraction of the first historical cloud cluster occlusion features.

[0071] In an embodiment of the present application, the first historical cloud occlusion feature includes historical cloud thickness and historical cloud texture.

[0072] Specifically, the first historical cloud occlusion feature includes the historical cloud thickness and the historical cloud texture. When the gray level co-occurrence matrix is ​​used to extract the first historical cloud occlusion feature, the gray level co-occurrence matrix can calculate the pixel mean, entropy and correlation in the historical cloud image. The historical cloud thickness can be described by the pixel mean of the historical cloud image. The pixel mean represents the overall brightness or gray level of the historical cloud image. Generally, thicker historical cloud images have higher pixel means, while thinner historical cloud images may have lower pixel means. The historical cloud texture can be described by the entropy and correlation in the historical cloud image. Entropy refers to the degree of information confusion in the image area. For the historical cloud image area with complex texture, the entropy value will be higher. The correlation reflects the statistical relationship between pixels in the historical cloud image. For the image area with more regular texture, the correlation will be higher.

[0073] Historical cloud thickness provides information about cloud density and occlusion, while historical cloud texture provides visual features about the internal structure and distribution of clouds. Combining historical cloud thickness with historical cloud texture helps to better describe the first historical cloud occlusion characteristics.

[0074] Step 130: predicting cloud occlusion characteristics within a second preset time period based on the first historical cloud occlusion characteristics;

[0075] It can be understood that since the first historical cloud blockage feature obtained in the above embodiment is a historical value, it cannot be directly used to predict the photovoltaic output in the future. Therefore, it is also necessary to predict the cloud blockage feature in the second preset time period based on the first historical cloud blockage feature, and then use the cloud blockage feature in the second preset time period to predict the corresponding photovoltaic output prediction value.

[0076] Specifically, the difference integrated moving average autoregression (ARIMA) model can be used to predict the cloud occlusion characteristics within the second preset time period. The difference integrated moving average autoregression model is a method used to process non-stationary time series data with trends and seasonality. It combines the autoregressive (AR) model, the integrated (I) part and the moving average (MA) model, and can capture trends, seasonality and other cyclical changes in the data. The autoregressive model uses past observations to predict current values, thereby capturing the autocorrelation in the time series; the integration part represents the difference operation on the original data. Through the difference, the non-stationary time series can be converted into a stationary series, which helps to eliminate the trend or seasonal influence; the moving average model uses a linear combination of past error terms to predict the current value, indicating the relationship between the current value and the error. This part helps to capture the irregularity or randomness in the time series. By inputting the first historical cloud occlusion feature into the pre-trained differential integrated moving average autoregressive model, the cloud occlusion feature within the second preset time period with higher prediction accuracy can be obtained, which is conducive to the subsequent prediction of the corresponding photovoltaic output prediction value based on the cloud occlusion feature.

[0077] Step 140: Obtain first weather information within a second preset time period;

[0078] It is understandable that photovoltaic output will not only be affected by the characteristics of cloud shading, but also by various weather information, such as solar irradiance, temperature, humidity, wind speed, air pressure, etc. When the solar irradiance is higher, the stronger the light received by the photovoltaic panels on the photovoltaic site, the greater the photovoltaic output will be, that is, the more electrical energy will be generated; and high temperature will also reduce the conversion efficiency of electrical energy, resulting in a decrease in photovoltaic output; at the same time, humidity, wind speed and air pressure will also have a certain impact on the power generation performance of the photovoltaic site. High humidity will affect the working efficiency of the electronic components inside the photovoltaic panel, and the wind speed will affect the heat dissipation effect of the photovoltaic panel. Changes in air pressure will affect the water vapor content in the atmosphere and the propagation of solar radiation. Therefore, the embodiment of the present application obtains the first weather information within the second preset time period. The first weather information can be obtained from the numerical weather prediction result (NWP). The first weather information includes solar irradiance, temperature, humidity, wind speed, air pressure, etc. By combining the first day's information to predict the photovoltaic output, it is possible to reduce the impact of high temperature, rain or other severe weather on the photovoltaic output prediction, thereby effectively improving the prediction accuracy.

[0079] Step 150: Input the cloud shading characteristics and the first weather information into a preset model to obtain a photovoltaic output prediction value, wherein the preset model is obtained based on ConvTransformer model training.

[0080] It can be understood that the ConvTransformer model is a hybrid architecture that combines the Convolutional Neural Network (CNN) and the Transformer model. It combines the advantages of the two models and is suitable for various types of tasks such as sequence data and image data. The ConvTransformer model can capture more global information through the combination of the Convolutional Neural Network and the Transformer model. Photovoltaic output data usually has a strong time series. Traditional convolutional neural networks often cannot capture global information well when processing time series data. By combining the Transformer model, the ConvTransformer model can better capture the long-term dependencies and global patterns in the time series data, so that the model can better capture the complex patterns and trends in the photovoltaic output data; and the ConvTransformer model adopts a parallel computing structure, which can be efficiently trained on large-scale distributed clusters, so that the model can quickly and efficiently process large-scale photovoltaic output data.

[0081] Specifically, the cloud shading characteristics and the first weather information obtained in the above embodiment are input into a preset model trained based on the ConvTransformer model, so that a photovoltaic output prediction value with high accuracy can be predicted.

[0082] In the embodiment of the present application, the training steps of the preset model include:

[0083] Acquire a plurality of second historical satellite cloud images, second weather information and corresponding photovoltaic output historical values ​​in a third preset time period in a second target area;

[0084] Extract the second historical cloud cluster occlusion features in the second historical satellite cloud image;

[0085] The preset model based on the ConvTransformer model is trained according to the second weather information, the second historical cloud occlusion characteristics and the historical values ​​of photovoltaic output.

[0086] Specifically, firstly, a plurality of second historical satellite cloud images, second weather information and corresponding photovoltaic output historical values ​​in the third preset time period in the second target area are obtained, wherein the second target area can be the same as or different from the first target area, and is set according to actual needs; the satellite cloud image can be obtained from the satellite, the second weather information can be obtained from the numerical forecast results, and the photovoltaic output historical value can be obtained from the historical monitoring data of the meteorological station near the second target area. The second historical satellite cloud image is preprocessed by using radian histogram equalization, the Gaussian noise in the second historical satellite cloud image after equalization is removed by using mean filtering, the salt and pepper noise in the second historical satellite cloud image after equalization is removed by using median filtering, the image segmentation of the second historical satellite cloud image after equalization and filtering is performed by using the gradient vector flow model, and the second historical cloud group occlusion feature is extracted from the segmented second historical satellite cloud image by using the gray level co-occurrence matrix, the second weather information, the second historical cloud group occlusion feature and the photovoltaic output historical value are input into the initial model based on the ConvTransformer model, and the initial model is trained to obtain the preset model based on the ConvTransformer model.

[0087] By training according to the second weather information, the second historical cloud occlusion characteristics and the historical values ​​of photovoltaic output, a preset model based on the ConvTransformer model is obtained, which can ensure that the model has a strong light output prediction ability.

[0088] The photovoltaic output prediction method provided in the embodiment of the present application obtains the first historical cloud occlusion feature through the first historical satellite cloud image, and combines it with the first weather information, and uses a preset model based on the ConvTransformer model to predict the photovoltaic output. It can reduce the impact of high temperature, rain, cloud occlusion or other severe weather on the photovoltaic output prediction, thereby effectively improving the prediction accuracy of photovoltaic output.

[0089] Figure 2 The schematic diagram of the structure of a photovoltaic output prediction device according to an embodiment of the present application is shown. Figure 2 As shown, in an embodiment of the present application, a photovoltaic output prediction device is provided, the device comprising:

[0090] Memory 210, configured to store instructions; and

[0091] The processor 220 is configured to call instructions from the memory 210 and implement the photovoltaic output prediction method according to any one of the first aspects when executing the instructions.

[0092] Specifically, in the embodiment of the present application, the processor 220 may be configured to:

[0093] Acquire a plurality of first historical satellite cloud images within a first preset time period in a first target area;

[0094] Extract the first historical cloud cluster occlusion features in the first historical satellite cloud image;

[0095] Predicting cloud occlusion characteristics within a second preset time period based on the first historical cloud occlusion characteristics;

[0096] Obtaining first weather information within a second preset time period;

[0097] The cloud occlusion characteristics and the first weather information are input into the preset model to obtain the photovoltaic output prediction value, wherein the preset model is trained based on the ConvTransformer model.

[0098] It can be understood that the photovoltaic output prediction device provided in the embodiment of the present application can implement each process of the photovoltaic output prediction method in the above embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0099] The present application also provides an electronic device, which may include:

[0100] According to the photovoltaic output prediction device described in the above embodiment.

[0101] It can be understood that the electronic device provided in the embodiment of the present application includes the photovoltaic output prediction device described in the above embodiment and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0102] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored. The instructions are used to enable a machine to execute the photovoltaic output prediction method described in the above embodiment and achieve the same technical effect. To avoid repetition, they are not repeated here.

[0103] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0105] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0107] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0108] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0109] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0110] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0111] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A photovoltaic output prediction method, characterized in that: The method comprises: Acquire a plurality of first historical satellite cloud images within a first preset time period in a first target area; Extracting the first historical cloud cluster occlusion feature in the first historical satellite cloud image; Predicting cloud occlusion characteristics within a second preset time period based on the first historical cloud occlusion characteristics; Acquire first weather information within the second preset time period; Inputting the cloud shading characteristics within the second preset time period and the first weather information into a preset model to obtain the photovoltaic output prediction value, wherein the preset model is obtained based on ConvTransformer model training; The extracting of the first historical cloud cluster occlusion feature in the first historical satellite cloud image comprises: Performing image segmentation on the first historical satellite cloud image using a gradient vector flow model to obtain a plurality of historical cloud cluster images in the first historical satellite cloud image; The gray level co-occurrence matrix is ​​used to perform feature extraction on the multiple historical cloud images to obtain the first historical cloud occlusion feature in the first historical satellite cloud image.

2. The photovoltaic output prediction method according to claim 1, characterized in that: Before extracting the first historical cloud cluster occlusion feature in the first historical satellite cloud image, the method further includes: The first historical satellite cloud image is preprocessed.

3. The photovoltaic output prediction method according to claim 2, characterized in that: The preprocessing of the first historical satellite cloud image comprises: Performing grayscale histogram equalization on the first historical satellite cloud image to obtain a first historical satellite cloud image after equalization; The equalized first historical satellite cloud image is subjected to smoothing filtering.

4. The photovoltaic output prediction method according to claim 3, characterized in that: The step of performing smoothing filtering on the equalized first historical satellite cloud image comprises: Using mean filtering to remove Gaussian noise in the equalized first historical satellite cloud image; The salt and pepper noise in the equalized first historical satellite cloud image is removed by using a median filter.

5. The photovoltaic output prediction method according to claim 1, characterized in that: The first historical cloud occlusion feature includes historical cloud thickness and historical cloud texture.

6. The photovoltaic output prediction method according to claim 1, characterized in that: The training steps of the preset model include: Acquire a plurality of second historical satellite cloud images, second weather information and corresponding photovoltaic output historical values ​​in a third preset time period in a second target area; Extracting second historical cloud cluster occlusion features from the second historical satellite cloud image; The preset model based on the ConvTransformer model is obtained by training according to the second weather information, the second historical cloud shading characteristics and the photovoltaic output historical values.

7. A photovoltaic output prediction device, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instruction from the memory and implement the photovoltaic output prediction method according to any one of claims 1 to 6 when executing the instruction.

8. An electronic device, characterized in that: include: The photovoltaic output prediction device according to claim 7.

9. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions, which are used to enable a machine to execute the photovoltaic output prediction method according to any one of claims 1 to 6.

Images