Ultra-short-term power prediction method and system for photovoltaic stations based on multi-modal large model

Abstract

The invention discloses a method and system for predicting ultra-short-term power of a photovoltaic station based on a multimodal large model, and relates to the technical field of power prediction of a photovoltaic new energy station. The method and system receive short-term meteorological data and satellite cloud images, input the short-term meteorological data and satellite cloud images into a pre-established first multimodal large model, and output corrected ultra-short-term meteorological prediction data; receive base cloud images and meteorological detection station detection information, input the base cloud images, satellite cloud images, meteorological detection station detection information and corrected ultra-short-term meteorological prediction data into a pre-established second multimodal large model, and output ultra-short-term meteorological prediction data; obtain timestamps, temperature values ​​and station power, input the timestamps, temperature values, station power and ultra-short-term meteorological prediction data into a pre-established time series prediction large model, and output ultra-short-term power prediction data.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic new energy station power prediction, and specifically to an ultra-short-term power prediction method and system for photovoltaic stations based on a multi-modal large model. Background Art

[0002] As the electricity seller in the market, new energy stations participate in the electricity trading market. Electricity trading is mainly for future power generation. Photovoltaic stations rely on sunlight for power generation, which has certain volatility. Therefore, improving the accuracy of ultra-short-term power prediction can effectively improve the trading ability of photovoltaic stations in power trading, reduce disturbances to the power grid, and enhance corporate profitability. The current mainstream algorithm is based on various simple ultra-short-term algorithms such as statistical learning, image learning, and physical models. It extracts features from satellite cloud images and base cloud images and predicts ultra-short-term power. However, due to the relatively low complexity of the model, manual participation is required to design feature extraction methods. Manual participation methods are difficult to effectively utilize all features, and the algorithm has poor versatility. Summary of the invention

[0003] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a method and system for ultra-short-term power prediction of photovoltaic stations based on a multi-modal large model, which can improve the accuracy of ultra-short-term prediction.

[0004] In a first aspect, the purpose of the present invention can be achieved by the following technical solution: a method for ultra-short-term power prediction of a photovoltaic station based on a multi-modal large model, the method comprising the following steps:

[0005] Receiving short-term meteorological data and satellite cloud images, inputting the short-term meteorological data and satellite cloud images into a pre-established first multimodal large model, and outputting corrected ultra-short-term meteorological forecast data;

[0006] Receive base cloud images and meteorological detection station detection information, input the base cloud images, satellite cloud images, meteorological detection station detection information and revised ultra-short-term meteorological forecast data into a pre-established second multimodal large model, and output the ultra-short-term meteorological forecast data;

[0007] The timestamp, temperature value and station power are obtained, and the timestamp, temperature value and station power as well as the ultra-short-term meteorological forecast data are input into a pre-established time series forecasting large model, and the ultra-short-term power forecast data is outputted.

[0008] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the calculation process of inputting the short-term meteorological data and the satellite cloud image into a pre-established first multimodal large model and outputting the corrected ultra-short-term meteorological forecast data is as follows:

[0009] The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows;

[0010] Loss f 1=mse(x_f1,x_c1)

[0011] Where, x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error.

[0012] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the calculation process of inputting the base cloud map, the satellite cloud map, the meteorological detection station detection information and the corrected ultra-short-term meteorological forecast data into a pre-established second multimodal large model, and outputting the ultra-short-term meteorological forecast data is as follows:

[0013] The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows:

[0014] Loss f2=mse(x_f2,x_c2)

[0015] Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error.

[0016] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a calculation process of inputting the timestamp, temperature value, station power, and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and outputting the ultra-short-term power forecast data:

[0017] The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction.

[0018] Loss f3=mse(p_f,p_c)

[0019] Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the loss function of the third model, and mse represents the mean square error.

[0020] In the second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a photovoltaic station ultra-short-term power prediction system based on a multi-modal large model, comprising:

[0021] A primary prediction module is used to receive short-term meteorological data and satellite cloud images, input the short-term meteorological data and satellite cloud images into a pre-established first multi-modal large model, and output corrected ultra-short-term meteorological forecast data;

[0022] The secondary prediction module is used to receive the base cloud map and the detection information of the meteorological detection station, input the base cloud map, the satellite cloud map, the detection information of the meteorological detection station and the revised ultra-short-term meteorological forecast data into the pre-established second multimodal large model, and output the ultra-short-term meteorological forecast data;

[0023] The power prediction module is used to obtain timestamps, temperature values ​​and site power, input the timestamps, temperature values, site power and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and output ultra-short-term power forecast data.

[0024] Alternatively, in a single forecast module, short-term meteorological data and satellite cloud images are input into a pre-established first multimodal large model, and the calculation process of outputting the corrected ultra-short-term meteorological forecast data is as follows:

[0025] The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows;

[0026] Loss f 1=mse(x_f1,x_c1)

[0027] Where x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error;

[0028] Alternatively, in the secondary prediction module, the base cloud map, satellite cloud map, meteorological detection station detection information and the revised ultra-short-term meteorological forecast data are input into the pre-established second multimodal large model, and the calculation process of outputting the ultra-short-term meteorological forecast data is as follows:

[0029] The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows:

[0030] Loss f2=mse(x_f2,x_c2)

[0031] Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error;

[0032] Or in the power prediction module, the timestamp, temperature value, station power and ultra-short-term meteorological forecast data are input into the pre-established time series prediction large model, and the calculation process of outputting the ultra-short-term power prediction data is obtained:

[0033] The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction.

[0034] Loss f3=mse(p_f,p_c)

[0035] Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the model loss function, and mse represents the mean square error.

[0036] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, the ultra-short-term power prediction method for a photovoltaic station based on a multimodal large model as described above is adopted.

[0037] Beneficial effects of the present invention:

[0038] The present invention can effectively understand the meteorological operation mechanism and can capture the algorithm of satellite cloud images and base cloud images over a period of time, thereby improving the accuracy of ultra-short-term predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 It is a schematic flow chart of the method of the present invention;

[0041] Figure 2 It is a schematic diagram of the ultra-short-term power prediction algorithm framework and reasoning flow of the present invention in a multi-modal large model;

[0042] Figure 3 It is a schematic diagram of the algorithm training process of the present invention;

[0043] Figure 4 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Embodiment 1:

[0046] The following is an introduction to the relevant terms involved in the embodiments of the present application:

[0047] Satellite cloud imagery: Satellite cloud imagery (scientific name: satellite cloud imagery) is an image of the cloud cover and surface features on the earth observed from top to bottom by meteorological satellites. The cloud images received mainly include infrared cloud images, visible light cloud images and water vapor images. Satellite cloud images can be used to identify different weather systems, determine their locations, estimate their intensity and development trends, and provide a basis for weather analysis and weather forecasts. In areas such as oceans, deserts, and plateaus that lack meteorological observation stations, the data provided by satellite cloud images make up for the lack of conventional detection data and play an important role in improving the accuracy of forecasts.

[0048] Multimodal big model: Multimodal big model is a machine learning technology based on deep learning. Its core idea is to fuse different media data (such as text, images, audio and video, etc.) and achieve more intelligent information processing by learning the associations between different modalities. In the multimodal big model, data of different modalities are pre-processed and input into a deep neural network. After multiple layers of feature extraction and fusion, the corresponding results are finally output.

[0049] The advantage of a large multimodal model is that it can fully utilize the information of different media data to extract richer and more comprehensive features, thereby improving the performance and generalization ability of the model. In addition, a large multimodal model can further enhance the semantic understanding and expression ability of the model by learning the associations between different modalities.

[0050] like Figure 1 As shown, a method for ultra-short-term power prediction of a photovoltaic station based on a multi-modal large model includes the following steps:

[0051] S101: receiving short-term meteorological data and satellite cloud images, inputting the short-term meteorological data and satellite cloud images into a pre-established first multimodal large model, and outputting corrected ultra-short-term meteorological forecast data;

[0052] Among them, the collection process of short-term meteorological data and satellite cloud images: short-term meteorological data and satellite cloud images are purchased from third parties.

[0053] The short-term meteorological data and satellite cloud images are input into the pre-established first multi-modal large model, and the calculation process of outputting the revised ultra-short-term meteorological forecast data is as follows:

[0054] The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows;

[0055] Loss f 1=mse(x_f1,x_c1)

[0056] Where x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error, which is a commonly used indicator to measure the difference between the model prediction value and the actual observation value.

[0057] S102: receiving base cloud images and meteorological detection station detection information, inputting the base cloud images, satellite cloud images, meteorological detection station detection information and revised ultra-short-term meteorological forecast data into a pre-established second multimodal large model, and outputting ultra-short-term meteorological forecast data;

[0058] Among them, the collection process of base cloud map and meteorological detection station detection information:

[0059] The base cloud map uses a full-sky imager, and the meteorological detection station detection information is the standard meteorological detection equipment for photovoltaic sites.

[0060] The base cloud map, satellite cloud map, meteorological station detection information and revised ultra-short-term meteorological forecast data are input into the pre-established second multimodal large model, and the calculation process of outputting the ultra-short-term meteorological forecast data is as follows:

[0061] The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows:

[0062] Loss f2=mse(x_f2,x_c2)

[0063] Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error.

[0064] S103: Obtain timestamp, temperature value and station power, input the timestamp, temperature value and station power and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and output ultra-short-term power forecast data.

[0065] The calculation process of inputting the timestamp, temperature value, station power and ultra-short-term meteorological forecast data into the pre-established time series forecast model and outputting the ultra-short-term power forecast data is as follows:

[0066] The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction.

[0067] Loss f3=mse(p_f,p_c)

[0068] Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the model loss function, and mse represents the mean square error.

[0069] Embodiment 2: In the second aspect, as Figure 4 As shown, the present invention discloses a photovoltaic station ultra-short-term power prediction system based on a multi-modal large model, comprising:

[0070] A primary prediction module is used to receive short-term meteorological data and satellite cloud images, input the short-term meteorological data and satellite cloud images into a pre-established first multi-modal large model, and output corrected ultra-short-term meteorological forecast data;

[0071] The secondary prediction module is used to receive the base cloud map and the detection information of the meteorological detection station, input the base cloud map, the satellite cloud map, the detection information of the meteorological detection station and the revised ultra-short-term meteorological forecast data into the pre-established second multimodal large model, and output the ultra-short-term meteorological forecast data;

[0072] The power prediction module is used to obtain timestamps, temperature values ​​and site power, input the timestamps, temperature values, site power and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and output ultra-short-term power forecast data.

[0073] Alternatively, in a single forecast module, short-term meteorological data and satellite cloud images are input into a pre-established first multimodal large model, and the calculation process of outputting the corrected ultra-short-term meteorological forecast data is as follows:

[0074] The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows;

[0075] Loss f 1=mse(x_f1,x_c1)

[0076] Where x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error;

[0077] Alternatively, in the secondary prediction module, the base cloud map, satellite cloud map, meteorological detection station detection information and the revised ultra-short-term meteorological forecast data are input into the pre-established second multimodal large model, and the calculation process of outputting the ultra-short-term meteorological forecast data is as follows:

[0078] The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows:

[0079] Loss f2=mse(x_f2,x_c2)

[0080] Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error;

[0081] Or in the power prediction module, the timestamp, temperature value, station power and ultra-short-term meteorological forecast data are input into the pre-established time series prediction large model, and the calculation process of outputting the ultra-short-term power prediction data is obtained:

[0082] The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction.

[0083] Loss f3=mse(p_f,p_c)

[0084] Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the model loss function, and mse represents the mean square error.

[0085] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0086] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0087] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0088] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. The ultra-short-term power prediction method of photovoltaic stations based on multi-modal large model is characterized by: The method comprises the following steps: Receiving short-term meteorological data and satellite cloud images, inputting the short-term meteorological data and satellite cloud images into a pre-established first multimodal large model, and outputting corrected ultra-short-term meteorological forecast data; The calculation process of inputting the short-term meteorological data and the satellite cloud image into the pre-established first multi-modal large model and outputting the corrected ultra-short-term meteorological forecast data is as follows: The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows; Loss(f 1) = mse(x_f1,x_c1) Where x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error; Receive base cloud images and meteorological detection station detection information, input the base cloud images, satellite cloud images, meteorological detection station detection information and revised ultra-short-term meteorological forecast data into a pre-established second multimodal large model, and output the ultra-short-term meteorological forecast data; The calculation process of inputting the base cloud map, satellite cloud map, meteorological detection station detection information and modified ultra-short-term meteorological forecast data into the pre-established second multimodal large model and outputting the ultra-short-term meteorological forecast data is as follows: The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows: Loss(f2)=mse(x_f2,x_c2) Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error; The timestamp, temperature value and station power are obtained, and the timestamp, temperature value and station power as well as the ultra-short-term meteorological forecast data are input into a pre-established time series forecasting large model, and the ultra-short-term power forecast data is outputted.

2. The method for ultra-short-term power prediction of photovoltaic stations based on a multi-modal large model according to claim 1 is characterized in that: The calculation process of inputting the timestamp, temperature value, station power and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model and outputting the ultra-short-term power forecast data: The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction; Loss(f3)=mse(p_f,p_c) Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the loss function of the third model, and mse represents the mean square error.

3. The ultra-short-term power prediction system for photovoltaic stations based on multi-modal large models is characterized by: include: A primary prediction module is used to receive short-term meteorological data and satellite cloud images, input the short-term meteorological data and satellite cloud images into a pre-established first multimodal large model, and output corrected ultra-short-term meteorological forecast data; The calculation process of inputting short-term meteorological data and satellite cloud images into the pre-established first multi-modal large model in the primary prediction module and outputting the corrected ultra-short-term meteorological forecast data is as follows: The target loss function is set as the cloud cover prediction accuracy for the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output as the revised ultra-short-term meteorological forecast data. The formula is as follows; Loss f 1=mse(x_f1,x_c1) Where x_f1 represents the first predicted cloud cover, x_c1 represents the first actual cloud cover, Loss(f1) represents the first model loss function, and mse represents the mean square error; The secondary prediction module is used to receive the base cloud map and the detection information of the meteorological detection station, input the base cloud map, satellite cloud map, the detection information of the meteorological detection station and the revised ultra-short-term meteorological forecast data into the pre-established second multimodal large model, and output the ultra-short-term meteorological forecast data; The secondary prediction module inputs the base cloud map, satellite cloud map, meteorological detection station detection information and the revised ultra-short-term meteorological forecast data into the pre-established second multimodal large model, and the calculation process of outputting the ultra-short-term meteorological forecast data is as follows: The target loss function is set as the prediction accuracy of solar radiation intensity in the next four hours, and the ultra-short-term meteorological forecast data for the next four hours is output. The formula is as follows: Loss f2=mse(x_f2,x_c2) Where x_f2 represents the second predicted cloud cover, x_c2 represents the second actual cloud cover, Loss(f2) represents the second model loss function, and mse represents the mean square error; The power prediction module is used to obtain the timestamp, temperature value and site power, input the timestamp, temperature value and site power as well as the ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and output the ultra-short-term power forecast data.

4. The ultra-short-term power prediction system for photovoltaic stations based on a multi-modal large model according to claim 3 is characterized in that: The power prediction module inputs the timestamp, temperature value, station power and ultra-short-term meteorological forecast data into a pre-established time series forecasting large model, and outputs the calculation process of obtaining ultra-short-term power forecast data: The target loss function is set to the power prediction accuracy for the next four hours to achieve ultra-short-term power prediction; Loss f3=mse(p_f,p_c) Where p_f represents the predicted power value, p_c represents the actual power value, Loss(f3) represents the model loss function, and mse represents the mean square error.

5. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on a processor. When the processor loads and executes the computer program, the ultra-short-term power prediction method for a photovoltaic station based on a multi-modal large model described in any one of claims 1 to 2 is adopted.

Images