Photovoltaic prediction method and device based on sky image enhancement, equipment and medium
The acquisition of real sky images through all-sky imaging equipment and the generation model of the variational autoencoder is used to expand the enhanced image data set, which solves the challenge of obtaining diversified data sets in photovoltaic prediction and improves the accuracy and adaptability of photovoltaic prediction.
Patent Information
- Application Number
- CN202510331893.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-07-04
AI Technical Summary
Acquisition of diverse sky image datasets for photovoltaic predictions faces challenges, affecting the accurate prediction and energy management efficiency of photovoltaic systems.
Real sky images are collected through all-sky imaging equipment, and the image generation model is generated using a variational autoencoder, which expands and enhances the sky image dataset, builds training samples of the photovoltaic prediction model, and trains the photovoltaic prediction model.
Improve the accuracy and adaptability of the photovoltaic prediction model, ensuring effective prediction of unseen sky images.
Smart Images
Figure CN120259107A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of photovoltaic power generation, and particularly to a photovoltaic prediction method, device, electronic device, and storage medium based on sky image enhancement. Background Art
[0002] In recent years, with the global transition to renewable energy, photovoltaic (PV) systems have received extensive attention as a sustainable energy solution. To achieve the effective integration of PV systems with the power grid, accurate solar energy prediction has become crucial. Such prediction can improve the efficiency of energy management and optimize the operation of energy storage systems. Sky images have become a key data source in solar energy prediction because they can capture cloud dynamics (which directly affect the solar irradiance level). However, there are significant challenges in obtaining a comprehensive and diverse sky image dataset. Therefore, how to obtain diverse sky images for photovoltaic prediction has become a technical problem that urgently needs to be solved. Summary of the Invention
[0003] The present invention provides a photovoltaic prediction method, device, electronic device, and storage medium based on sky image enhancement.
[0004] According to an aspect of the present invention, there is provided a photovoltaic prediction method based on sky image enhancement, including:
[0005] Obtaining a real sky image collected by an all-sky imaging device installed near a photovoltaic power station;
[0006] Based on the real sky image, training a pre-constructed variational autoencoder generation model, and performing an image generation operation using the trained variational autoencoder generation model to obtain a generated new sky image;
[0007] According to the new sky image and the real sky image, constructing a training sample for the photovoltaic prediction model to be trained;
[0008] Based on the training sample, training the photovoltaic prediction model to be trained, and performing photovoltaic prediction using the trained photovoltaic prediction model.
[0009] In some embodiments, before training the pre-constructed variational autoencoder generation model based on the real sky image, it further includes:
[0010] Preprocessing the real sky image; wherein the preprocessing includes normalization processing and / or size adjustment processing.
[0011] In some embodiments, the pre-constructed variational autoencoder generation model includes an encoder and a decoder; the loss function of the pre-constructed autoencoder generation model is composed of a reconstruction loss and a KL divergence loss;
[0012] Training the pre-constructed variational autoencoder generation model based on the real sky image includes:
[0013] Embedding the input real sky image in the latent space through the encoder of the variational autoencoder generation model, and generating a latent vector representation;
[0014] Performing image generation processing on the basis of the latent vector representation through the decoder of the variational autoencoder generation model to obtain a new sky image;
[0015] Determining a reconstruction loss according to the difference between the new sky image and the real sky image; and determining a KL divergence loss according to the difference between the distribution of the real sky image in the latent space and the prior distribution;
[0016] Adjusting the parameters of the variational autoencoder generation model according to the reconstruction loss and the KL divergence loss.
[0017] In some embodiments, embedding the input real sky image in the latent space through the encoder of the variational autoencoder generation model and generating a latent vector representation includes:
[0018] Inputting the real sky image into the encoder of the variational autoencoder generation model for encoding processing, so that the image features of the real sky image are mapped to the probability distribution parameters in the latent space; generating a latent vector representation based on the probability distribution parameters; wherein the probability distribution parameters include a mean and a standard deviation.
[0019] In some embodiments, performing an image generation operation using the trained variational autoencoder generation model to obtain a generated new sky image includes:
[0020] Randomly sampling from the latent space to obtain a latent vector representation;
[0021] Performing image generation processing on the latent vector representation through the decoder of the trained variational autoencoder generation model to obtain a new sky image.
[0022] In some embodiments, the photovoltaic prediction model is a convolutional neural network model for predicting photovoltaic power or solar irradiance.
[0023] According to another aspect of the present invention, there is provided a photovoltaic prediction device based on sky image enhancement, including:
[0024] A data acquisition module for acquiring a real sky image collected by an all-sky imaging device installed near a photovoltaic power station;
[0025] An image reconstruction module, configured to train a pre - constructed variational auto - encoder generation model based on the real sky image, and perform an image generation operation using the trained variational auto - encoder generation model to obtain a generated new sky image;
[0026] A sample construction module, configured to construct training samples for the photovoltaic prediction model to be trained according to the new sky image and the real sky image;
[0027] A photovoltaic prediction module, configured to train the photovoltaic prediction model to be trained based on the training samples, and perform photovoltaic prediction using the trained photovoltaic prediction model.
[0028] In some embodiments, it further includes:
[0029] A pre - processing module, configured to pre - process the real sky image; wherein, the pre - processing includes normalization processing and / or size adjustment processing.
[0030] In some embodiments, the pre - constructed variational auto - encoder generation model includes an encoder and a decoder; the loss function of the pre - constructed auto - encoder generation model is composed of a reconstruction loss and a KL - divergence loss;
[0031] In terms of training the pre - constructed variational auto - encoder generation model based on the real sky image, the data reconstruction module is specifically configured to:
[0032] Embed the input real sky image into the latent space through the encoder of the variational auto - encoder generation model, and generate a latent vector representation;
[0033] Perform image generation processing according to the latent vector representation through the decoder of the variational auto - encoder generation model to obtain a new sky image;
[0034] Determine the reconstruction loss according to the difference between the new sky image and the real sky image; and determine the KL - divergence loss according to the difference between the distribution of the real sky image in the latent space and the prior distribution;
[0035] Adjust the parameters of the variational auto - encoder generation model according to the reconstruction loss and the KL - divergence loss.
[0036] In some embodiments, the step of embedding the input real sky image into the latent space through the encoder of the variational auto - encoder generation model and generating a latent vector representation includes:
[0037] Input the real sky image into the encoder of the variational autoencoder generation model for encoding, so that the image features of the real sky image are mapped to the probability distribution parameters in the latent space; generate a latent vector representation based on the probability distribution parameters; wherein, the probability distribution parameters include the mean and the standard deviation.
[0038] In some embodiments, the data reconstruction module is further configured to:
[0039] Randomly sample from the latent space to obtain a latent vector representation;
[0040] Perform image generation processing on the latent vector representation through the decoder of the trained variational autoencoder generation model to obtain a new sky image.
[0041] In some embodiments, the photovoltaic prediction model is a convolutional neural network model for predicting photovoltaic power or solar irradiance.
[0042] According to another aspect of the present invention, there is provided an electronic device, which includes:
[0043] At least one processor; and
[0044] A memory communicatively connected to the at least one processor; wherein,
[0045] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the photovoltaic prediction method based on sky image enhancement according to the embodiments of the present invention.
[0046] According to another aspect of the present invention, there is provided a computer-readable storage medium, which stores computer instructions for causing a processor to implement the photovoltaic prediction method based on sky image enhancement according to the embodiments of the present invention when executed.
[0047] The technical solution of the embodiments of the present invention can utilize the trained variational autoencoder generation model to expand and enhance the real sky image to obtain diversified sky images; and then use the enhanced sky images for training the photovoltaic prediction model, which can ensure the accuracy of the trained photovoltaic prediction model, and further ensure the accuracy of photovoltaic prediction based on the photovoltaic prediction model.
[0048] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.
[0050] Figure 1 is a schematic flowchart of a photovoltaic prediction method based on sky image enhancement provided by an embodiment of the present invention;
[0051] Figure 2 is a timing diagram of the photovoltaic prediction method based on sky image enhancement provided by an embodiment of the present invention;
[0052] Figure 3 is a schematic structural diagram of a photovoltaic prediction device based on sky image enhancement provided by an embodiment of the present invention;
[0053] Figure 4 is a schematic structural diagram of an electronic device for implementing the photovoltaic prediction method based on sky image enhancement in an embodiment of the present invention. Detailed implementation manners
[0054] In order to enable those skilled in the art of the present technology to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0055] Embodiment 1
[0056] Figure 1 is a flowchart of a photovoltaic prediction method based on sky image enhancement provided by an embodiment of the present invention. This embodiment is applicable to the scenario of expanding and enhancing the sky image for photovoltaic prediction. This method can be executed by a photovoltaic prediction device based on sky image enhancement. The photovoltaic prediction device based on sky image enhancement can be implemented in the form of hardware and / or software, and the photovoltaic prediction device based on sky image enhancement can be configured in an electronic device.
[0057] As Figure 1 shown, the photovoltaic prediction method based on sky image enhancement includes:
[0058] S101. Obtain a real sky image collected by an all-sky imaging device installed near a photovoltaic power station.
[0059] In the embodiments of the present invention, the all-sky imaging device may optionally be an all-sky camera, which can clearly capture cloud details under both strong light and shadows; the all-sky imaging device can be installed in an open area near the photovoltaic power station, for example, installed at the highest point of the photovoltaic power station. The real sky images collected by the all-sky imaging device can be stored locally in the device or transmitted to the cloud server for storage through the network. Therefore, the real sky images collected by the all-sky imaging device can be obtained locally from the all-sky imaging device, or the real sky images collected by the all-sky imaging device can be obtained from the cloud server, and no specific limitation is made here.
[0060] S102. Based on the real sky image, train a pre-constructed variational autoencoder generation model, and use the trained variational autoencoder generation model to perform an image generation operation to obtain a generated new sky image.
[0061] In this embodiment, the structure of the pre-constructed variational autoencoder generation model includes an encoder and a decoder. Among them, the encoder may include an input layer (for receiving real sky images), multiple convolutional layers (for extracting image features, the convolutional kernel size can be 3×3, and the number of feature channels is 32, 64, 128 respectively), and two fully connected layers (for outputting the distribution parameters of the real sky image in the latent space). The decoder may include: a fully connected layer (for receiving and converting the latent vector representation), a transposed convolutional layer (for gradually restoring the image size), and an output layer (for generating a new sky image). The loss function of the pre-constructed autoencoder generation model is composed of a reconstruction loss and a KL divergence loss; among them, the reconstruction loss is used to measure the difference between the generated new sky image and the real sky image used as the model input; the KL divergence loss is used to measure the difference between the distribution of the real sky image in the latent space and the prior distribution (such as a normal distribution).
[0062] On the above basis, training the pre-constructed variational autoencoder generation model based on the real sky image includes steps A - D:
[0063] A. Through the encoder of the constructed variational autoencoder generation model, embed the input real sky image in the latent space and generate a latent vector representation.
[0064] Specifically, input the real sky image into the encoder of the variational autoencoder generation model for encoding processing, so that the convolutional layers in the encoder extract image features from the real sky image and map the image features of the real sky image to the latent space. Each of the two fully connected layers of the encoder can output a probability distribution parameter; among them, the probability distribution parameter includes a mean and a standard deviation; furthermore, generate a latent vector representation based on the probability distribution parameter; among them, the latent vector representation is an abstract and low-dimensional expression of the image features of the real sky image, and it includes the key information required for generating the image.
[0065] It should be noted that before encoding the real sky image into the encoder of the variational autoencoder generation model, the real sky image can be normalized and resized, so as to convert the real sky image into input data suitable for the variational autoencoder generation model.
[0066] B. Through the decoder of the variational autoencoder generation model, image generation processing is performed according to the latent vector representation to obtain a new sky image.
[0067] In specific implementation, first, the input latent vector representation vector is input into the fully connected layer of the decoder. The role of the fully connected layer is to perform a linear transformation on the latent vector representation, mapping the low-dimensional latent vector to a higher-dimensional feature space. This higher-dimensional feature space contains more information, which helps to reconstruct a more complex image structure subsequently. For example, a latent vector representation with a dimension of 128 may be mapped to a feature vector with a dimension of 4096. The output of the fully connected layer usually undergoes a non-linear transformation through an activation function (such as ReLU) to introduce non-linear features and enhance the expressive power of the model. Then, the feature vector processed by the fully connected layer is reshaped into a multi-dimensional feature map, and the shape of this feature map is similar to the feature representation of an image. Next, the decoder uses a series of transposed convolutional layers (also known as deconvolutional layers) to process the feature map. The role of the transposed convolutional layer is to gradually enlarge the size of the feature map to make it close to the size of the target image. Each transposed convolutional layer performs a deconvolution operation on the feature map through the setting of the convolutional kernel and stride, gradually enlarging the size of the feature map. After being processed by multiple transposed convolutional layers, the decoder finally outputs a feature map with the same size as the real sky image. Subsequently, post-processing can also be performed on the generated feature map, such as performing denormalization processing to restore the pixels to the original range.
[0068] C. Determine the reconstruction loss according to the difference between the new sky image and the real sky image; and determine the KL divergence loss according to the difference between the distribution of the real sky image in the latent space and the prior distribution.
[0069] In specific implementation, first determine the first loss function (such as the mean squared error loss function or the binary cross-entropy loss function), input the new sky image and the real sky image into the first loss function for calculation to obtain the reconstruction loss. When determining the KL divergence loss, a normal distribution can be used as the prior distribution of the latent space, and then calculate the difference between the distribution of the real sky image in the latent space and the prior distribution to obtain the KL divergence loss.
[0070] D. Adjust the parameters of the variational autoencoder generation model according to the reconstruction loss and the KL divergence loss.
[0071] In specific implementation, the reconstruction loss and the KL divergence loss can be weighted and summed, and the sum value is used as the total loss of the variational autoencoder generation model; then, derivatives are taken with respect to all trainable parameters of the model according to the total loss to obtain the gradient of each parameter. These gradients represent the rate of change of the total loss with respect to each parameter and reflect the influence degree of each parameter on the total loss; based on the gradient of each parameter, a preset optimization method (such as the gradient descent method) is used to update the model parameters.
[0072] Through continuous training of the above process, the final variational autoencoder generation model can be obtained. Furthermore, the variational autoencoder generation model can be utilized to generate multiple new sky images by randomly sampling in the latent space. It can be understood that the new sky images simulate various weather conditions (such as sunny days, cloudy days, cloud changes, etc.), so that the generated new sky images and the real sky images together can form a diverse sky image dataset. In this way, the real sky images are expanded and enhanced.
[0073] It should be noted here that compared with traditional data augmentation methods (such as flipping, cropping, etc.), the present invention generates new sky images with higher diversity by learning the latent distribution of the data and can simulate rare scenarios and extreme conditions.
[0074] S103. Construct a training sample of the photovoltaic prediction model to be trained according to the new sky image and the real sky image.
[0075] In the embodiment of the present invention, the training sample of the photovoltaic prediction model to be trained can be constructed manually or by machine according to the new sky image and the real sky image, and no specific limitation is made here. Among them, the true value or label of the training sample can be irradiance or photovoltaic power, and the photovoltaic prediction model can be a convolutional neural network model for predicting photovoltaic power or solar irradiance.
[0076] S104. Train the photovoltaic prediction model to be trained based on the training sample, and perform photovoltaic prediction using the trained photovoltaic prediction model.
[0077] Using the expanded and enhanced sky images to construct the training sample can ensure the diversity of the samples, so that the photovoltaic prediction model can learn the essential features of the sky images during the training process, thereby improving the adaptability of the photovoltaic prediction model to unseen sky images.
[0078] In the embodiments of the present invention, a trained variational autoencoder generation model can be used to augment and enhance real sky images to obtain diverse sky images. Furthermore, the enhanced sky images can be used to train a photovoltaic prediction model, which can ensure the accuracy of the trained photovoltaic prediction model and thus ensure the accuracy of photovoltaic prediction based on the photovoltaic prediction model.
[0079] Embodiment 2
[0080] Figure 2 FIG. is a flowchart of a photovoltaic prediction method based on sky image enhancement provided by an embodiment of the present invention. Refer to Figure 2 , and the specific method includes the following steps:
[0081] S201. Obtain real sky images collected by an all-sky imaging device installed near a photovoltaic power station.
[0082] S202. Preprocess the real sky images.
[0083] Among them, the preprocessing includes normalization processing and / or size adjustment processing. It can be understood that through the preprocessing, the real sky images can be made suitable for the variational autoencoder generation model, ensuring the training efficiency of the variational autoencoder generation model.
[0084] S203. Based on the real sky images, train a pre-constructed variational autoencoder generation model, and use the trained variational autoencoder generation model to perform an image generation operation to obtain generated new sky images.
[0085] Optionally, use the preprocessed real sky images as the input of the pre-constructed variational autoencoder generation model, and according to the model output results, reversely adjust the model parameters to complete the model training. Using the trained variational autoencoder generation model to perform an image generation operation to obtain generated new sky images includes: randomly sampling from the latent space to obtain a latent vector representation; performing image generation processing on the latent vector representation through the decoder of the trained variational autoencoder generation model to obtain new sky images.
[0086] S204. According to the new sky images and the real sky images, construct training samples for the photovoltaic prediction model to be trained.
[0087] S205. Based on the training samples, train the photovoltaic prediction model to be trained, and use the trained photovoltaic prediction model to perform photovoltaic prediction.
[0088] In the embodiments of the present invention, by preprocessing the real sky images, it can be ensured that the processed images are suitable for the variational autoencoder generation model, effectively improving the model training efficiency.
[0089] Embodiment 3
[0090] Figure 3 The following is a schematic structural diagram of a photovoltaic prediction device based on sky image enhancement provided by an embodiment of the present invention. This embodiment is applicable to scenarios where sky images for photovoltaic prediction are augmented and enhanced. As Figure 3 shown, the device includes:
[0091] A data acquisition module 301, configured to acquire real sky images collected by an all-sky imaging device installed near a photovoltaic power station;
[0092] An image reconstruction module 302, configured to train a pre-constructed variational autoencoder generation model based on the real sky images, and perform image generation operations using the trained variational autoencoder generation model to obtain generated new sky images;
[0093] A sample construction module 303, configured to construct training samples for a photovoltaic prediction model to be trained according to the new sky images and the real sky images;
[0094] A photovoltaic prediction module 304, configured to train the photovoltaic prediction model to be trained based on the training samples, and perform photovoltaic prediction using the trained photovoltaic prediction model.
[0095] In some embodiments, it further includes:
[0096] A preprocessing module, configured to preprocess the real sky images; wherein, the preprocessing includes normalization processing and / or size adjustment processing.
[0097] In some embodiments, the pre-constructed variational autoencoder generation model includes an encoder and a decoder; the loss function of the pre-constructed autoencoder generation model is composed of a reconstruction loss and a KL divergence loss;
[0098] In terms of training the pre-constructed variational autoencoder generation model based on the real sky images, the data reconstruction module is specifically configured to:
[0099] Embed the input real sky images into the latent space through the encoder of the variational autoencoder generation model, and generate a latent vector representation;
[0100] Perform image generation processing according to the latent vector representation through the decoder of the variational autoencoder generation model to obtain new sky images;
[0101] Determine the reconstruction loss according to the difference between the new sky images and the real sky images; and determine the KL divergence loss according to the difference between the distribution of the real sky images in the latent space and the prior distribution;
[0102] Adjust the parameters of the variational autoencoder generation model according to the reconstruction loss and the KL divergence loss.
[0103] In some embodiments, the encoder of the variational autoencoder generation model embeds the input real sky image in the latent space and generates a latent vector representation, including:
[0104] Input the real sky image into the encoder of the variational autoencoder generation model for encoding, so that the image features of the real sky image are mapped to the probability distribution parameters in the latent space; generate a latent vector representation based on the probability distribution parameters; wherein, the probability distribution parameters include a mean and a standard deviation.
[0105] In some embodiments, the data reconstruction module is further configured to:
[0106] Randomly sample from the latent space to obtain a latent vector representation;
[0107] Perform image generation processing on the latent vector representation through the decoder of the trained variational autoencoder generation model to obtain a new sky image.
[0108] In some embodiments, the photovoltaic prediction model is a convolutional neural network model for predicting photovoltaic power or solar irradiance.
[0109] The photovoltaic prediction device based on sky image enhancement provided by the embodiments of the present invention can execute the photovoltaic prediction method based on sky image enhancement provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.
[0110] Embodiment IV
[0111] Figure 4 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement the embodiments of the present invention. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present invention described and / or claimed herein.
[0112] As Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory communicatively connected to the at least one processor 11, such as read-only memory (ROM) 12, random access memory (RAM) 13, etc. Among them, the memory stores a computer program executable by the at least one processor. The processor 11 can execute various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.
[0113] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0114] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as executing a photovoltaic prediction method based on sky image enhancement.
[0115] In some embodiments, the photovoltaic prediction method based on sky image enhancement can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the photovoltaic prediction method based on sky image enhancement described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the photovoltaic prediction method based on sky image enhancement by any other appropriate means (such as, by means of firmware).
[0116] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems-on-chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.
[0117] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable apparatus for sky image enhancement-based photovoltaic prediction, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart(s) and / or block diagram(s) to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on the remote machine or server.
[0118] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0119] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and the input received from the user can be in any form (including acoustic input, voice input, or tactile input).
[0120] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.
[0121] A computing system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.
[0122] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.
[0123] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.
Claims
1. A photovoltaic prediction method based on sky image enhancement, characterized in that, Including: Obtain a real sky image collected by an all-sky imaging device installed near a photovoltaic power station; Based on the real sky image, train a pre-constructed variational autoencoder generation model, and use the trained variational autoencoder generation model to perform an image generation operation to obtain a generated new sky image; Construct a training sample for the photovoltaic prediction model to be trained according to the new sky image and the real sky image; Based on the training sample, train the photovoltaic prediction model to be trained, and use the trained photovoltaic prediction model to perform photovoltaic prediction.
2. The method according to claim 1, wherein Before training the pre-constructed variational autoencoder generation model based on the real sky image, it further includes: Preprocess the real sky image; wherein, the preprocessing includes normalization processing and / or size adjustment processing.
3. The method according to claim 1, wherein The pre-constructed variational autoencoder generation model includes an encoder and a decoder; the loss function of the pre-constructed autoencoder generation model is composed of a reconstruction loss and a KL divergence loss; The training of the pre-constructed variational autoencoder generation model based on the real sky image includes: Embed the input real sky image into the latent space through the encoder of the variational autoencoder generation model, and generate a latent vector representation; Through the decoder of the variational autoencoder generation model, perform image generation processing according to the latent vector representation to obtain a new sky image; Determine the reconstruction loss according to the difference between the new sky image and the real sky image; and determine the KL divergence loss according to the difference between the distribution of the real sky image in the latent space and the prior distribution; Adjust the parameters of the variational autoencoder generation model according to the reconstruction loss and the KL divergence loss.
4. The method according to claim 3, characterized in that, The embedding of the input real sky image into the latent space through the encoder of the variational autoencoder generation model and generating a latent vector representation includes: Input the real sky image into the encoder of the variational autoencoder generation model for encoding processing, so that the image features of the real sky image are mapped to the probability distribution parameters in the latent space; generate a latent vector representation based on the probability distribution parameters; wherein, the probability distribution parameters include a mean and a standard deviation.
5. The method according to claim 3, characterized in that, The use of the trained variational autoencoder generation model to perform an image generation operation to obtain a generated new sky image includes: Randomly sample from the latent space to obtain a latent vector representation; Through the decoder of the trained variational autoencoder generation model, perform image generation processing on the latent vector representation to obtain a new sky image.
6. The method according to claim 1, characterized in that The photovoltaic prediction model is a convolutional neural network model for predicting photovoltaic power or solar irradiance.
7. A photovoltaic prediction device based on sky image enhancement, characterized in that, Including: A data acquisition module for obtaining a real sky image collected by an all-sky imaging device installed near a photovoltaic power station; An image reconstruction module for training a pre-constructed variational autoencoder generation model based on the real sky image, and using the trained variational autoencoder generation model to perform an image generation operation to obtain a generated new sky image; A sample construction module, configured to construct training samples for a photovoltaic prediction model to be trained according to the new sky image and the real sky image; A photovoltaic prediction module, configured to train the photovoltaic prediction model to be trained based on the training samples, and perform photovoltaic prediction using the trained photovoltaic prediction model.
8. The device according to claim 7, characterized in that, It further includes: A preprocessing module, configured to preprocess the real sky image; wherein, the preprocessing includes normalization processing and / or size adjustment processing.
9. An electronic device, characterized in that, It includes: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the method according to any one of claims 1-6.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1-6 is implemented.