Wind power prediction method and device based on meteorological large model, server and medium

Abstract

The invention provides a wind power prediction method and device based on a meteorological large model, a server and a medium. The method comprises the following steps: constructing a multi-scale prediction model; acquiring historical actual power of the wind power station, and acquiring multi-scale historical meteorological elements through the large meteorological model; training the multi-scale prediction model, and performing hyper-parameter optimization on the multi-scale prediction model through an improved arithmetic optimization algorithm; and obtaining multi-scale instant meteorological elements through the meteorological large model, inputting the multi-scale instant meteorological elements into the multi-scale prediction model after hyper-parameter optimization, and generating a wind power prediction result based on the meteorological large model. According to the wind power prediction method and device based on the meteorological large model, the server and the medium, the limitation of a traditional wind power prediction method can be overcome, and the accuracy and the stability of a wind power prediction result are improved.

Description

Technical Field

[0001] The invention belongs to the field of wind power prediction, and in particular relates to a wind power prediction method, device, server and medium based on a large meteorological model. Background Art

[0002] In the field of renewable energy, wind power generation has received widespread attention due to its environmental friendliness and sustainability. However, the instability and intermittency of wind power generation pose challenges to grid dispatching. Therefore, accurate prediction of wind power is of great significance for improving wind power grid connection efficiency and reducing operating costs.

[0003] Traditional wind power forecasting methods mostly rely on physical models, statistical models or traditional machine learning models. The physical model is based on meteorological principles, taking into account the impact of various meteorological factors such as wind speed, wind direction, temperature, and air pressure on wind power, and can reflect the physical characteristics of wind farms. However, the physical model requires a large amount of real-time meteorological data as input, the model calculation is complex, and it is sensitive to initial conditions. It is easy to ignore the spatial impact of wind farm layout, and it cannot quickly respond to instantaneous changes in wind speed when making short-term and ultra-short-term forecasts. Statistical models (such as time series analysis (ARIMA), regression analysis, etc.) can predict future trends by analyzing historical wind power data. They are easy to implement and have high computational efficiency. However, statistical models mainly rely on historical data, cannot fully consider sudden changes or seasonal changes in weather conditions, and are not effective in predicting extreme weather events. Although traditional machine learning models can learn complex data relationships and have good prediction effects on nonlinear and time series data, they usually require manual selection and construction of features. This process is time-consuming and relies on the knowledge of domain experts. In addition, when dealing with high-dimensional features, the performance of traditional machine learning models will drop significantly, resulting in problems such as long training time, overfitting or underfitting, which affects their performance in practical applications.

[0004] Therefore, how to overcome the limitations of traditional wind power prediction methods and improve the accuracy and stability of wind power prediction results has become an urgent problem that needs to be solved in this field. Summary of the invention

[0005] In view of this, the present invention aims to propose a wind power forecasting method, device, server and medium based on a large meteorological model, so as to overcome the limitations of traditional wind power forecasting methods and improve the accuracy and stability of wind power forecasting results.

[0006] To achieve the above object, the technical solution created by the present invention is implemented as follows:

[0007] In a first aspect, an embodiment of the present invention provides a wind power prediction method based on a large meteorological model, comprising:

[0008] Constructing a multi-scale prediction model, the multi-scale prediction model includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer, the AE feature acquisition layer is used to mine the correlation between input data and generate feature data, the LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series, and the attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, thereby improving the training speed and generalization of the multi-scale prediction model;

[0009] Obtain the historical actual power generated by wind power stations and obtain multi-scale historical meteorological elements through large meteorological models;

[0010] The multi-scale prediction model is trained based on the historical actual power generation and multi-scale historical meteorological elements, the hyperparameters of the multi-scale prediction model are determined, and the hyperparameters of the multi-scale prediction model are optimized through an improved arithmetic optimization algorithm;

[0011] The multi-scale real-time meteorological elements are obtained through the large meteorological model, and the multi-scale real-time meteorological elements are input into the multi-scale prediction model after hyperparameter optimization to generate wind power prediction results based on the large meteorological model.

[0012] Furthermore, the multi-scale prediction model is trained according to the historical actual power and multi-scale historical meteorological elements, including:

[0013] The multi-scale prediction model is trained using multi-scale historical meteorological elements as input data and historical actual power generation as training targets.

[0014] Furthermore, the training of the multi-scale prediction model according to the historical actual power generation and the multi-scale historical meteorological elements also includes:

[0015] Initialize the parameters of the encoder and decoder in the AE feature acquisition layer; perform dimensionality reduction processing on the input data through the encoder of the AE feature acquisition layer to generate reduced-dimensional data, and reconstruct the reduced-dimensional data through the decoder of the AE feature acquisition layer to generate reconstructed data; calculate the reconstruction error between the reconstructed data and the reduced-dimensional data, and update the parameters of the encoder and decoder through the gradient descent algorithm to minimize the reconstruction error; repeat the iteration until convergence.

[0016] Furthermore, the improved arithmetic optimization algorithm includes an AMOA acceleration function for determining the optimization strategy, and the mathematical expression of the AMOA acceleration function is:

[0017] AMOA(t)=Min+(Max-Min)*(t / T) 1 / t

[0018] Where: t represents the current number of iterations, T represents the maximum number of iterations, Min represents the minimum value of the acceleration function, and Max represents the maximum value of the acceleration function;

[0019] When the value of AMOA is less than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts a global exploration optimization strategy; when the value of AMOA is greater than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts a local development optimization strategy.

[0020] Furthermore, before training the multi-scale prediction model according to the historical actual power generation and the multi-scale historical meteorological elements, the wind power prediction method based on the large meteorological model includes:

[0021] Data cleaning, data standardization and data normalization are carried out on historical actual power generation and multi-scale historical meteorological elements.

[0022] Furthermore, before inputting the multi-scale instantaneous meteorological elements into the multi-scale prediction model after the hyperparameter optimization, the wind power prediction method based on the large meteorological model includes:

[0023] Data cleaning, data standardization and data normalization are performed on multi-scale real-time meteorological elements.

[0024] Furthermore, after the hyperparameter optimization of the multi-scale prediction model is performed by the improved arithmetic optimization algorithm, the wind power prediction method based on the large meteorological model further includes:

[0025] The optimized multi-scale prediction model is cross-validated to improve the generalization ability and prediction accuracy of the optimized multi-scale prediction model on unseen data.

[0026] In a second aspect, an embodiment of the present invention further provides a wind power prediction device based on a large meteorological model, comprising:

[0027] A model building module is used to build a multi-scale prediction model, which includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer. The AE feature acquisition layer is used to mine the correlation between input data and generate feature data. The LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series. The attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, thereby improving the training speed and generalization of the multi-scale prediction model.

[0028] The acquisition module is used to obtain the historical actual power generated by the wind power station and obtain multi-scale historical meteorological elements through the large meteorological model;

[0029] A training optimization module is used to train the multi-scale prediction model based on the historical actual power generation and multi-scale historical meteorological elements, determine the hyperparameters of the multi-scale prediction model, and optimize the hyperparameters of the multi-scale prediction model through an improved arithmetic optimization algorithm;

[0030] The prediction module is used to obtain multi-scale real-time meteorological elements through the large meteorological model, input the multi-scale real-time meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate wind power prediction results based on the large meteorological model.

[0031] In a third aspect, an embodiment of the present invention further provides a server, including:

[0032] one or more processors;

[0033] A storage device for storing one or more programs;

[0034] When the one or more programs are executed by the one or more processors, the one or more processors implement the wind power prediction method based on the large meteorological model as provided in the above embodiment.

[0035] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute a wind power prediction method based on a large meteorological model as provided in the above embodiment.

[0036] Compared with the prior art, the wind power prediction method, device, server and medium based on the meteorological big model created by the present invention have the following advantages:

[0037] The invention creates a wind power prediction method, device, server and medium based on a large meteorological model, which can construct a multi-scale prediction model including an AE feature acquisition layer, an LSTM layer and an attention mechanism layer, train the multi-scale prediction model according to the historical actual power and multi-scale historical meteorological elements, and optimize the hyperparameters of the multi-scale prediction model through an improved arithmetic optimization algorithm, and can obtain multi-scale instant meteorological elements through the large meteorological model, input the multi-scale instant meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate a wind power prediction result based on the large meteorological model. Compared with the prior art, the invention can obtain meteorological elements through a large meteorological model, improve the diversity and stability of meteorological elements, and can effectively mine potential high-level effective features between input data through the cooperation of the AE feature acquisition layer, the LSTM layer and the attention mechanism layer, capture the long-term dependency of feature data in the time series and improve the training speed and generalization of the model, so as to overcome the limitations of traditional wind power prediction methods and improve the accuracy and stability of wind power prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings constituting part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation on the present invention. In the drawings:

[0039] Figure 1 A flow chart of the wind power prediction method based on a large meteorological model as described in Example 1 of the present invention;

[0040] Figure 2 A schematic diagram of the structure of a wind power prediction device based on a large meteorological model according to the second embodiment of the present invention;

[0041] Figure 3 This is a structural diagram of the server described in Example 3 of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0043] Embodiment 1

[0044] Figure 1 1 is a flow chart of a wind power prediction method based on a large meteorological model provided in the first embodiment of the present invention. As shown in the figure, the photovoltaic short-term power prediction method specifically includes the following steps:

[0045] Step 110, construct a multi-scale prediction model, the multi-scale prediction model includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer, the AE feature acquisition layer is used to mine the correlation between input data and generate feature data, the LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series, and the attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, so as to improve the training speed and generalization of the multi-scale prediction model.

[0046] In order to realize the prediction of wind power generation, this embodiment will construct a multi-scale prediction model including an AE feature acquisition layer, an LSTM layer and an attention mechanism layer. The AE feature acquisition layer is used to mine the correlation between various meteorological elements and generate feature data. The LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series. The attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, thereby improving the training speed and generalization of the multi-scale prediction model.

[0047] It should be noted that the AE feature acquisition layer is composed of an encoder and a decoder. Compared with the traditional convolutional neural network (CNN), the AE feature acquisition layer is better able to process multimodal time series data. It is an unsupervised learning method that does not require a large amount of labeled data for training, and can learn the intrinsic structure and feature representation of the data. The encoder and decoder of the AE feature acquisition layer are both feedforward neural networks. When working, the encoder is responsible for mapping the input data to a low-dimensional vector, and the decoder is responsible for reconstructing the encoded low-dimensional vector back to the original data space, thereby capturing the important features of the data, removing a large amount of noise in the data, and digging deeper into the correlation between various meteorological elements.

[0048] The LSTM layer is a time-recurrent neural network. Although the traditional LSTM model can handle the long-distance dependency problem in the sequence data, it may encounter the problem of gradient disappearance or gradient explosion when processing very long sequences. In this embodiment, the attention mechanism layer and the LSTM layer will cooperate to enable the multi-scale prediction model to pay attention to more distant historical information when processing long sequences, so as to better capture long-distance dependencies. The attention mechanism layer can assign a weight to each element in the sequence, so that the multi-scale prediction model can clearly point out which parts contribute more to the final output. At the same time, the attention mechanism layer can also enable the multi-scale prediction model to process different parts of the sequence in parallel during training, which can significantly improve the training speed. In addition, the existence of the attention mechanism layer can also enable the multi-scale prediction model to focus on the key parts of the sequence, so that the multi-scale prediction model has better generalization when facing new data. When working, the attention mechanism layer can obtain the attention score by calculating the similarity between the Query and the Key, and apply the softmax function to the calculated attention score to convert it into a probability distribution. The probability distribution obtained is then multiplied by the value matrix to obtain a weighted and represented, so that the multi-scale prediction model can automatically learn which time points are more important to the prediction results.

[0049] Step 120: Obtain the historical actual power generated by the wind power station, and obtain multi-scale historical meteorological elements through a large meteorological model.

[0050] In order to facilitate the training of the multi-scale prediction model so that the multi-scale prediction model can predict wind power generation according to meteorological elements, this embodiment will obtain the historical actual power generation of the wind power station and the multi-scale historical meteorological elements as training samples of the multi-scale prediction model. In the prior art, the meteorological elements used for wind power generation prediction are generally obtained through meteorological forecasts. However, since the diversity and stability of meteorological elements will directly affect the prediction results of wind power generation, this embodiment will obtain multi-scale historical meteorological elements through a large meteorological model, thereby improving the diversity and stability of meteorological elements.

[0051] It should be noted that the meteorological big model described in this embodiment is a high-precision AI prediction model for predicting meteorological elements in the prior art, such as the Pangu meteorological big model, the Fengwu meteorological big model, the Fuxi meteorological big model, the SwinRDM meteorological big model, the Aurora meteorological big model, the NeuralGCM meteorological big model, and the GraphCast meteorological big model, etc. The meteorological big model can accurately and stably predict meteorological elements on various time and space scales, and can output various types of meteorological elements according to actual needs, thereby meeting the use requirements in this embodiment.

[0052] Step 130: train the multi-scale prediction model according to the historical actual power generation and the multi-scale historical meteorological elements, determine the hyperparameters of the multi-scale prediction model, and optimize the hyperparameters of the multi-scale prediction model by using an improved arithmetic optimization algorithm.

[0053] When training the multi-scale prediction model, the multi-scale historical meteorological elements can be used as input data, and the historical actual power generation can be used as the training target to train the multi-scale prediction model.

[0054] Specifically, when training the multi-scale prediction model, the training process of the AE feature acquisition layer can be carried out according to the following steps:

[0055] Initialize the parameters of the encoder and decoder in the AE feature acquisition layer; perform dimensionality reduction processing on the input data through the encoder of the AE feature acquisition layer to generate reduced-dimensional data, and reconstruct the reduced-dimensional data through the decoder of the AE feature acquisition layer to generate reconstructed data; calculate the reconstruction error between the reconstructed data and the reduced-dimensional data, and update the parameters of the encoder and decoder through the gradient descent algorithm to minimize the reconstruction error; repeat the iteration until convergence.

[0056] In addition, when training the LSTM layer and the attention mechanism layer, the input data (i.e., the data set consisting of multi-scale historical meteorological elements and historical actual power) should be divided to obtain the training set, validation set, and test set. The training set is used to train the model, and during the training process, a suitable loss function (such as the MSE function) and optimizer (such as the Adam optimizer) should be selected to update the model parameters. The validation set is used to fine-tune the model's hyperparameters and to verify the current model's generalization ability during iterative model training to determine whether to stop training.

[0057] After completing the training of the multi-scale prediction model, in order to further improve the prediction accuracy of the multi-scale prediction model, this embodiment will also perform hyperparameter optimization on the multi-scale prediction model through an improved arithmetic optimization algorithm.

[0058] Arithmetic Optimization Algorithm (AOA) is a heuristic optimization algorithm that explores the solution space by simulating basic arithmetic operations (addition, subtraction, multiplication and division) and gradually approaches the optimal solution. When executing the arithmetic optimization algorithm, the population should be initialized first to generate the initial solution set, then the fitness value of each solution is calculated, and the arithmetic operation is selected according to a certain probability. Next, the solution is updated according to the selected arithmetic operation. If the current solution is better than the global optimal solution, the global optimal solution is updated. If the current solution is not the global optimal solution, the above steps are repeated until the preset maximum number of iterations is reached to complete the optimization process.

[0059] However, the acceleration function (MOA) used in the traditional arithmetic optimization algorithm shows a linear increasing trend, so its convergence speed and accuracy are limited. Therefore, in this embodiment, the acceleration function of the traditional arithmetic optimization algorithm is improved to form an improved arithmetic optimization algorithm, and the multi-scale prediction model is optimized for hyperparameters by the improved arithmetic optimization algorithm, thereby improving the convergence speed and accuracy of hyperparameter optimization.

[0060] The improved arithmetic optimization algorithm includes an AMOA acceleration function for determining the optimization strategy, and the mathematical expression of the AMOA acceleration function is:

[0061] AMOA(t)=Min+(Max-Min)*(t / T) 1 / t

[0062] Where: t represents the current number of iterations, T represents the maximum number of iterations, Min represents the minimum value of the acceleration function, and Max represents the maximum value of the acceleration function.

[0063] When the value of AMOA is less than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts a global exploration optimization strategy; when the value of AMOA is greater than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts a local development optimization strategy.

[0064] Similar to the traditional arithmetic optimization algorithm, when the improved arithmetic optimization algorithm adopts the global exploration optimization strategy, it will realize global search through multiplication and division operations, improve the dispersion of solutions, enhance the global optimization ability of the algorithm, avoid falling into local optimal ability, and realize global exploration optimization. When the improved arithmetic optimization algorithm adopts the local development optimization strategy, it will realize local development through addition and subtraction operations, thereby reducing the dispersion of solutions, which is conducive to the full development of the population in the local range and enhancing the local optimization ability of the algorithm.

[0065] As an optional implementation of this embodiment, in order to improve the training effect of the multi-scale prediction model, before training the multi-scale prediction model based on the historical actual power and the multi-scale historical meteorological elements, the historical actual power and the multi-scale historical meteorological elements should be subjected to data cleaning, data standardization and data normalization, so as to avoid improper impact on the training effect of the multi-scale prediction model.

[0066] In addition, in order to further improve the generalization ability of the multi-scale prediction model on unseen data and ensure the prediction accuracy, after completing the hyperparameter optimization of the multi-scale prediction model, this embodiment can also cross-validate the optimized multi-scale prediction model through a test set.

[0067] Step 140: Acquire multi-scale real-time meteorological elements through the large meteorological model, input the multi-scale real-time meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate wind power prediction results based on the large meteorological model.

[0068] After completing the training and optimization of the multi-scale prediction model, the multi-scale prediction model will be applied to the actual wind power prediction work. At this time, the multi-scale real-time meteorological elements can be obtained through the large meteorological model, and the multi-scale real-time meteorological elements can be input into the multi-scale prediction model after hyperparameter optimization, so as to stably generate high-precision wind power prediction results based on the large meteorological model.

[0069] Accordingly, in order to further improve the prediction accuracy of the multi-scale prediction model, before inputting the multi-scale real-time meteorological elements into the multi-scale prediction model after hyperparameter optimization, the present embodiment may also perform data cleaning, data standardization and data normalization on the multi-scale real-time meteorological elements to avoid the prediction result accuracy of the multi-scale prediction model being improperly affected by the accuracy of the input data.

[0070] It should be noted that the multi-scale historical meteorological elements and multi-scale real-time meteorological elements in this embodiment refer to the meteorological elements at multiple time scales. Generally, the time scales of meteorological elements can be divided into four types: long-term, medium-term, short-term and ultra-short-term. Meteorological elements at different time scales can be used to predict wind power generation at the corresponding time scale.

[0071] This embodiment can construct a multi-scale prediction model including an AE feature acquisition layer, an LSTM layer, and an attention mechanism layer, train the multi-scale prediction model according to historical actual power and multi-scale historical meteorological elements, and can also optimize the hyperparameters of the multi-scale prediction model through an improved arithmetic optimization algorithm, and can obtain multi-scale instant meteorological elements through a large meteorological model, input the multi-scale instant meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate a wind power prediction result based on the large meteorological model. In this way, the combination of the large meteorological model and the wind power prediction model is realized, the diversity and stability of meteorological elements are improved, and the potential high-level effective features between meteorological elements can be effectively mined through the cooperation of the AE feature acquisition layer, the LSTM layer, and the attention mechanism layer, the long-term dependency of feature data in the time series is captured, and the training speed and generalization of the model are improved, thereby overcoming the limitations of traditional wind power prediction methods and improving the accuracy and stability of wind power prediction results.

[0072] Embodiment 2

[0073] Figure 2 Schematic diagram of the structure of a wind power prediction device based on a large meteorological model provided in the second embodiment of the present invention. Figure 2 As shown, the wind power prediction device based on the large meteorological model includes:

[0074] The model construction module 210 is used to construct a multi-scale prediction model, which includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer. The AE feature acquisition layer is used to mine the correlation between input data and generate feature data. The LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series. The attention mechanism layer is used to make the multi-scale prediction model focus on the key part in the time series, thereby improving the training speed and generalization of the multi-scale prediction model.

[0075] An acquisition module 220 is used to acquire the historical actual power generated by the wind power station and acquire multi-scale historical meteorological elements through a large meteorological model;

[0076] The training optimization module 230 is used to train the multi-scale prediction model according to the historical actual power generation and the multi-scale historical meteorological elements, determine the hyperparameters of the multi-scale prediction model, and optimize the hyperparameters of the multi-scale prediction model by using an improved arithmetic optimization algorithm;

[0077] The prediction module 240 is used to obtain multi-scale real-time meteorological elements through the large meteorological model, input the multi-scale real-time meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate wind power prediction results based on the large meteorological model.

[0078] The wind power prediction device based on the meteorological big model provided in this embodiment can construct a multi-scale prediction model including an AE feature acquisition layer, an LSTM layer and an attention mechanism layer through a model construction module, and can train and optimize the multi-scale prediction model according to historical actual power and multi-scale historical meteorological elements through the cooperation of the acquisition module and the training optimization module, and can also input multi-scale instant meteorological elements into the multi-scale prediction model after hyperparameter optimization through the prediction module to generate wind power prediction results based on the meteorological big model. Compared with the prior art, it can realize the combination of the meteorological big model and the wind power prediction model, improve the diversity and stability of meteorological elements, overcome the limitations of traditional wind power prediction, and improve the accuracy and stability of wind power prediction results.

[0079] On the basis of the above-mentioned embodiment, the wind power prediction device based on the large meteorological model further includes:

[0080] The first preprocessing module is used to perform data cleaning, data standardization and data normalization on the historical actual power and multi-scale historical meteorological elements;

[0081] The second preprocessing module is used to perform data cleaning, data standardization and data normalization on multi-scale real-time meteorological elements;

[0082] The post-processing module is used to cross-validate the optimized multi-scale prediction model to improve the generalization ability and prediction accuracy of the optimized multi-scale prediction model on unseen data.

[0083] The wind power prediction device based on the meteorological big model provided in the embodiment of the present invention can execute the wind power prediction method based on the meteorological big model provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0084] Embodiment 3

[0085] Figure 3 A schematic diagram of the structure of a server provided in Embodiment 3 of the present invention. Figure 3 A block diagram of an exemplary server 12 suitable for use in implementing embodiments of the present invention is shown. Figure 3The server 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0086] like Figure 3 As shown, the server 12 is in the form of a general purpose computing device. The components of the server 12 may include, but are not limited to: one or more processors or processing units 16, server memory 28, and a bus 18 connecting different server components (including server memory 28 and processing unit 16).

[0087] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0088] The server 12 typically includes a variety of computer server readable media. These media can be any available media that can be accessed by the server 12, including volatile and non-volatile media, removable and non-removable media.

[0089] Server memory 28 may include computer server readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Server 12 may further include other removable / non-removable, volatile / non-volatile computer server storage media. By way of example only, storage server 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 3 not shown, usually called a "hard drive"). Although Figure 3 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The server memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0090] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the server memory 28, such program modules 42 including, but not limited to, operating a server, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0091] The server 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), may communicate with one or more devices that enable a user to interact with the server 12, and / or may communicate with any device that enables the server 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the server 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the server 12 via a bus 18. It should be understood that, although not shown in the figure, the server 12 may be assembled to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID servers, tape drives, and data backup storage servers, etc.

[0092] The processing unit 16 executes various functional applications and data processing by running the programs stored in the server memory 28, such as implementing the wind power prediction method based on the large meteorological model provided in the embodiment of the present invention.

[0093] Embodiment 4

[0094] Embodiment 4 of the present invention further provides a storage medium comprising computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, are used to execute any of the wind power prediction methods based on a large meteorological model as provided in the above embodiments.

[0095] The computer storage medium of the embodiment of the present invention 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, electromagnetic, 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 this document, 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.

[0096] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0097] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0098] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or device. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0099] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A wind power forecasting method based on a large meteorological model, characterized in that include: Constructing a multi-scale prediction model, the multi-scale prediction model includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer, the AE feature acquisition layer is used to mine the correlation between input data and generate feature data, the LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series, and the attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, thereby improving the training speed and generalization of the multi-scale prediction model; Obtain the historical actual power generated by wind power stations and obtain multi-scale historical meteorological elements through large meteorological models; The multi-scale prediction model is trained based on the historical actual power generation and multi-scale historical meteorological elements, the hyperparameters of the multi-scale prediction model are determined, and the hyperparameters of the multi-scale prediction model are optimized through an improved arithmetic optimization algorithm; The multi-scale real-time meteorological elements are obtained through the large meteorological model, and the multi-scale real-time meteorological elements are input into the multi-scale prediction model after hyperparameter optimization to generate wind power prediction results based on the large meteorological model.

2. The wind power prediction method based on the meteorological large model according to claim 1 is characterized in that: The multi-scale prediction model is trained according to the historical actual power and multi-scale historical meteorological elements, including: The multi-scale prediction model is trained using multi-scale historical meteorological elements as input data and historical actual power generation as training targets.

3. The wind power prediction method based on the meteorological large model according to claim 1 is characterized in that: The multi-scale prediction model is trained based on the historical actual power and multi-scale historical meteorological elements, including: Initialize the parameters of the encoder and decoder in the AE feature acquisition layer; perform dimensionality reduction processing on the input data through the encoder of the AE feature acquisition layer to generate reduced-dimensional data, and reconstruct the reduced-dimensional data through the decoder of the AE feature acquisition layer to generate reconstructed data; calculate the reconstruction error between the reconstructed data and the reduced-dimensional data, and update the parameters of the encoder and decoder through the gradient descent algorithm to minimize the reconstruction error; repeat the iteration until convergence.

4. The wind power prediction method based on the meteorological large model according to claim 1 is characterized in that: The improved arithmetic optimization algorithm includes an AMOA acceleration function for determining an optimization strategy, and the mathematical expression of the AMOA acceleration function is: AMOA(t)=Min+(Max-Min)*(t / T) 1 / t Where: t represents the current number of iterations, T represents the maximum number of iterations, Min represents the minimum value of the acceleration function, and Max represents the maximum value of the acceleration function; When the value of AMOA is less than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts the global exploration optimization strategy; When the value of AMOA is greater than the random number r1 between 0 and 1, the improved arithmetic optimization algorithm adopts the optimization strategy of local development.

5. The wind power prediction method based on the meteorological large model according to claim 1 is characterized in that: Before training the multi-scale prediction model according to the historical actual power generation and the multi-scale historical meteorological elements, the wind power prediction method based on the large meteorological model includes: Data cleaning, data standardization and data normalization are carried out on historical actual power generation and multi-scale historical meteorological elements.

6. The wind power prediction method based on a large meteorological model according to claim 1 is characterized in that: Before inputting the multi-scale instantaneous meteorological elements into the multi-scale prediction model after the hyperparameter optimization, the wind power prediction method based on the meteorological large model includes: Data cleaning, data standardization and data normalization are performed on multi-scale real-time meteorological elements.

7. The wind power prediction method based on a large meteorological model according to claim 1 is characterized in that: After the hyperparameter optimization of the multi-scale prediction model is performed by the improved arithmetic optimization algorithm, the wind power prediction method based on the large meteorological model further includes: The optimized multi-scale prediction model is cross-validated to improve the generalization ability and prediction accuracy of the optimized multi-scale prediction model on unseen data.

8. A wind power prediction device based on a large meteorological model, characterized in that: include: A model building module is used to build a multi-scale prediction model, which includes: an AE feature acquisition layer, an LSTM layer and an attention mechanism layer. The AE feature acquisition layer is used to mine the correlation between input data and generate feature data. The LSTM layer is used to receive the feature data output by the AE feature acquisition layer and capture the long-term dependency of the feature data in the time series. The attention mechanism layer is used to make the multi-scale prediction model focus on the key parts in the time series, thereby improving the training speed and generalization of the multi-scale prediction model. The acquisition module is used to obtain the historical actual power generated by the wind power station and obtain multi-scale historical meteorological elements through the large meteorological model; A training optimization module is used to train the multi-scale prediction model based on the historical actual power generation and multi-scale historical meteorological elements, determine the hyperparameters of the multi-scale prediction model, and optimize the hyperparameters of the multi-scale prediction model through an improved arithmetic optimization algorithm; The prediction module is used to obtain multi-scale real-time meteorological elements through the large meteorological model, input the multi-scale real-time meteorological elements into the multi-scale prediction model after hyperparameter optimization, and generate wind power prediction results based on the large meteorological model.

9. A server, characterized in that: The server comprises: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the wind power prediction method based on the large meteorological model as described in any one of claims 1-7.

10. A storage medium comprising computer executable instructions, wherein the computer executable instructions are used to execute the wind power prediction method based on a large meteorological model as claimed in any one of claims 1 to 7 when executed by a computer processor.

Images