Cold start wind power prediction method based on multi-source weather fusion and representation learning

Abstract

The present disclosure provides a cold start wind power prediction method based on multi-source meteorological fusion and feature learning. In some embodiments of the present disclosure, multi-source meteorological prediction data of a new wind farm station is obtained; the multi-source meteorological prediction data is input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; the fused meteorological features are input into a pre-trained power prediction module to obtain a wind power prediction result of the new wind farm station; wherein, according to sample multi-source meteorological prediction data and sample historical power and meteorological data of a general wind farm station, an initial multi-source meteorological feature fusion module and an initial power prediction module are jointly trained; the weight of the trained multi-source meteorological feature fusion module is migrated to the new wind farm station, and through the training of the general wind farm station and the migration mechanism of the new wind farm station, the new wind farm station can achieve high-precision prediction without a large amount of data, thereby improving the precision of wind power prediction.

Description

Technical Field

[0001] This disclosure relates to the field of new energy power generation technology, and in particular to a cold start wind power prediction method based on multi-source meteorological fusion and representation learning. Background Technology

[0002] As the global energy transition accelerates, wind power, as a core component of clean and renewable energy, continues to see growth in installed capacity and grid connection. Short-term wind power forecasting is a key technology for ensuring the safe and stable operation of the power system and improving wind power absorption efficiency; its forecasting accuracy directly depends on the precise perception and modeling of meteorological elements.

[0003] Numerical Weather Prediction (NWP) is the core meteorological data source for wind power prediction. Traditional NWP models, such as the IFS of the European Centre for Medium-Range Weather Forecasts (ECMWF), the GFS of the National Oceanic and Atmospheric Administration (NOAA) in the United States, and the ICON of the German Meteorological Office (DWD), achieve forecasts by solving atmospheric dynamics and thermodynamic equations. They have the advantages of strong physical consistency and high interpretability, but they have limitations in computational efficiency and localized fine-grained forecasting. In recent years, AI-driven large-scale meteorological models such as FengWu, Pangu-Weather, and GraphCast have emerged. They learn atmospheric motion patterns from massive amounts of historical meteorological data using data-driven deep learning methods, showing significant advantages in forecast accuracy and computational efficiency. However, different models still have problems such as systematic bias and the accumulation of localized scale errors.

[0004] Currently, the accuracy of wind power prediction is relatively low. Summary of the Invention

[0005] This disclosure provides a cold-start wind power prediction method based on multi-source meteorological fusion and representation learning, which at least solves the problem of low accuracy in existing wind power prediction methods.

[0006] The technical solution disclosed herein is as follows: This disclosure provides a cold-start wind power prediction method based on multi-source meteorological fusion and representation learning, including: Acquire multi-source meteorological forecast data for new wind farms; The multi-source meteorological forecast data is input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features. The fused meteorological features are input into the pre-trained power prediction module to obtain the wind power prediction results of the new wind farm. Specifically, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are then transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0007] Optionally, the method further includes: Collect sample multi-source meteorological forecast data and sample historical power and meteorological data of the ordinary wind farms; Based on the sample multi-source meteorological forecast data and the sample historical power and meteorological data, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, thus obtaining the trained multi-source meteorological feature fusion module and the trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are transferred to the new wind farm; and Based on the multi-source meteorological forecast training data and historical power and meteorological training data of the new wind farm, the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0008] Optionally, the collection of sample multi-source meteorological forecast data and sample historical power and meteorological data of the ordinary wind farm includes: Collect initial multi-source weather forecast data and initial historical power and meteorological data; Bilinear interpolation is used to map the gridded initial multi-source weather forecast data and initial historical power and meteorological data to ordinary wind farm sites. Spline interpolation is used to unify data sources with different time resolutions to the same sampling frequency, resulting in data-aligned sample multi-source weather forecast data and sample historical power and meteorological data.

[0009] Optionally, the method further includes: During the training of the completed power prediction module, the parameters of the forward layer of the power prediction module remain unchanged, while the weights and biases of the output layer are updated.

[0010] Optionally, the step of inputting the multi-source weather forecast data into a pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features includes: Use the multi-source meteorological forecast feature sequence corresponding to the multi-source meteorological forecast data as the key vector and value vector, and use the historical power and meteorological feature sequence as the query vector. By generating features using a cross-attention mechanism, fused meteorological features are obtained.

[0011] Optionally, the power prediction module includes a forward layer and an output layer; the step of inputting the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction result of the new wind farm includes: The fused meteorological features are input into the forward layer to obtain a high-dimensional representation; The high-dimensional representation is input into the output layer to obtain the wind power prediction result of the new wind farm.

[0012] Optionally, the power prediction module is a multilayer sensor.

[0013] This disclosure also provides a wind power prediction device, including: The acquisition module is used to acquire multi-source meteorological forecast data for new wind farms; The fusion module is used to input the multi-source weather forecast data into the pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features; The prediction module is used to input the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction results of the new wind farm. Specifically, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are then transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0014] This disclosure also provides an electronic device, including: processor; Memory used to store processor-executable instructions; The processor is configured to execute instructions to implement the steps in the above method.

[0015] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0016] This disclosure also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method described above.

[0017] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: In some embodiments of this disclosure, multi-source meteorological forecast data of new wind farms are acquired; the multi-source meteorological forecast data are input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; the fused meteorological features are input into a pre-trained power prediction module to obtain the wind power prediction result of the new wind farm; wherein, based on sample multi-source meteorological forecast data and sample historical power and meteorological data of ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module; the weights of the trained multi-source meteorological feature fusion module are transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module. Through the training of ordinary wind farms and the transfer mechanism of new wind farms, the new wind farm can achieve high-precision prediction without a large amount of data, thus improving the accuracy of wind power prediction.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0020] Figure 1 A flowchart illustrating a cold-start wind power prediction method based on multi-source meteorological fusion and representation learning, provided as an exemplary embodiment of this disclosure; Figure 2 A schematic diagram of the structure of a multi-source meteorological feature fusion module provided as an exemplary embodiment of this disclosure; Figure 3 A schematic diagram of the structure of a wind power prediction device provided for an exemplary embodiment of this disclosure; Figure 4 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of this disclosure. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0022] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0023] It should be noted that the user information involved in this disclosure includes, but is not limited to, user device information and user personal information; the collection, storage, use, processing, transmission, provision and disclosure of user information in this disclosure all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0024] The above-mentioned defects are particularly prominent in the cold start scenario of new wind farms in the early stage of operation: new stations usually only have a very small amount of real-time meteorological data and scattered historical power data, lacking the time series data of the correlation between meteorological elements and power that meet the requirements of traditional modeling. As a result, wind power prediction models that rely on a large amount of historical data for training cannot be directly and effectively modeled, and the prediction accuracy drops significantly.

[0025] To address the aforementioned technical issues, in some embodiments of this disclosure, multi-source meteorological forecast data for new wind farms are acquired; the multi-source meteorological forecast data is input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; the fused meteorological features are input into a pre-trained power prediction module to obtain the wind power prediction result for the new wind farm; wherein, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module; the weights of the trained multi-source meteorological feature fusion module are transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module. Through the training mechanism of ordinary wind farms and the transfer mechanism of new wind farms, high-precision prediction can be achieved for new wind farms without a large amount of data, thus improving the accuracy of wind power prediction.

[0026] The technical solutions provided by the embodiments of this disclosure are described in detail below with reference to the accompanying drawings.

[0027] Figure 1 This is a flowchart illustrating a cold-start wind power prediction method based on multi-source meteorological fusion and representation learning, provided as an exemplary embodiment of this disclosure. Figure 1 As shown, the method includes: S101: Acquire multi-source meteorological forecast data for new wind farms; S102: Input multi-source weather forecast data into the pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features; S103: Input the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction results of the new wind farm. Specifically, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are then transferred to new wind farms, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0028] In this embodiment, the execution subject of the above method is a terminal device or a server.

[0029] The terminal device includes, but is not limited to, mobile stations (MS), mobile terminals, mobile phones, handsets, and portable equipment. This terminal device can communicate with one or more core networks via a radio access network (RAN). For example, the terminal device can be a mobile phone (or "cellular" phone), a computer with wireless communication capabilities, a computer with wireless transceiver capabilities, a virtual reality (VR) terminal device, an AR terminal device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical care, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. The operating systems installed on the terminal device include, but are not limited to, iOS, Android, Windows, Linux, and Mac OS. In different networks, terminals may be called by different names, such as: user equipment, mobile station, user unit, station, cellular phone, personal digital assistant, wireless modem, wireless communication device, handheld device, laptop, cordless phone, wireless local loop station, television, etc. For ease of description, this embodiment will simply refer to it as terminal device.

[0030] In this embodiment, the implementation form of the server is not limited. For example, the server can be a conventional server, a cloud server, a cloud host, a virtual center, or other server devices. The server mainly consists of a processor, hard disk, memory, system bus, and other common computer architecture types.

[0031] In this embodiment, multi-source meteorological forecast data of the new wind farm is acquired; the multi-source meteorological forecast data is input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; the fused meteorological features are input into a pre-trained power prediction module to obtain the wind power prediction result of the new wind farm; wherein, based on the sample multi-source meteorological forecast data and sample historical power and meteorological data of ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module; the weights of the trained multi-source meteorological feature fusion module are transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module. Through the training of ordinary wind farms and the transfer mechanism of new wind farms, the new wind farm can achieve high-precision prediction without a large amount of data, thus improving the accuracy of wind power prediction.

[0032] This disclosure constructs an end-to-end transfer learning framework for multi-source meteorological fusion, meteorological characterization extraction, and power mapping learning. The framework is divided into two stages: at stations with sufficient data, the multi-source meteorological fusion module and the MLP mapping module are jointly trained to learn generalized fusion rules and power mapping relationships; at new stations, the pre-trained fusion module weights are directly reused, and only the MLP is lightly fine-tuned to quickly adapt to the characteristics of the station.

[0033] This embodiment of the disclosure takes multi-source weather forecast data as input, fuses multi-source meteorological features to generate a unified representation, then learns the power mapping through an MLP (Multi-Level Processing) to finally output the predicted power. The weights of the trained multi-source meteorological fusion module are directly transferred and frozen, with only the terminal MLP being fine-tuned to adapt to the characteristics of the new site, enabling rapid deployment in cold-start scenarios. The two links achieve technology reuse through weight transfer, allowing the pre-trained fusion capabilities to take effect directly at the new site, significantly reducing the dependence on historical data of the new site.

[0034] In some embodiments of this disclosure, before using the pre-trained multi-source meteorological feature fusion module and the pre-trained power prediction module, it is necessary to first train the pre-trained multi-source meteorological feature fusion module and the pre-trained power prediction module. One possible approach is to collect sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms; based on the sample multi-source meteorological forecast data and sample historical power and meteorological data, jointly train the initial multi-source meteorological feature fusion module and the initial power prediction module to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, thereby obtaining the trained multi-source meteorological feature fusion module and the trained power prediction module; transfer the weights of the trained multi-source meteorological feature fusion module to new wind farms; and train the trained power prediction module based on the multi-source meteorological forecast training data and historical power and meteorological training data from the new wind farms to obtain the pre-trained multi-source meteorological feature fusion module and the pre-trained power prediction module. In the pre-training phase for ordinary wind farms, an initial multi-source meteorological feature fusion module and an initial power prediction module are jointly trained. The former learns the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, while the latter learns the mapping relationship between meteorological characteristics and wind power based on the fused meteorological characteristics, while implicitly capturing the wind turbine characteristics and terrain influences of older wind farms. When entering a cold start scenario for a new wind farm, the weights of the pre-trained multi-source meteorological feature fusion module are directly transferred and frozen, and the top layer of the trained power prediction module is lightly fine-tuned. Using a small amount of real-time data from the new wind farm, it can quickly adapt to the characteristics of the new wind farm, avoiding the dependence of traditional models on a large amount of historical data.

[0035] In the above embodiments, sample multi-source meteorological forecast data and sample historical power and meteorological data of ordinary wind farms are collected. One possible approach is to collect initial multi-source meteorological forecast data and initial historical power and meteorological data; use bilinear interpolation to map the gridded initial multi-source meteorological forecast data and initial historical power and meteorological data to the ordinary wind farm site locations, and use spline interpolation to unify data sources with different time resolutions to the same sampling frequency, resulting in data-aligned sample multi-source meteorological forecast data and sample historical power and meteorological data. Specifically, firstly, multi-source meteorological forecast data is collected, including multiple numerical meteorological forecasts and AI meteorological forecasts, while simultaneously acquiring historical power data from wind farms with sufficient historical data, including key meteorological elements such as wind speed (m), wind direction, 2m temperature, and sea level pressure, as well as wind turbine output records for the corresponding time periods. To ensure data consistency, the collected heterogeneous data undergoes spatiotemporal alignment processing: bilinear interpolation is used to map the gridded meteorological forecast data to specific wind farm site locations, and spline interpolation is used to unify data sources with different time resolutions to the same sampling frequency, ultimately constructing a complete training dataset.

[0036] It should be noted that the power prediction module is a multilayer perceptron (MLP).

[0037] In the above embodiments, the pre-training of the ordinary battlefield includes cross-attention fusion module training and MLP joint training. Cross-attention fusion module training uses multi-source meteorological forecast feature sequences and historical power and meteorological sequences from stations with sufficient historical data as input. It trains the cross-attention fusion module to dynamically model the spatiotemporal correlation between multi-source meteorological and historical power data, generating unified fused meteorological features. This makes the fusion result closer to the actual meteorological state of the stations, fully leveraging the complementary advantages of multi-source data. MLP joint training inputs the fused meteorological features output from the cross-attention fusion module into the MLP, using historical power data from stations with sufficient historical data as labels. It trains the MLP's meteorological representation extraction and power mapping capabilities. During training, it simultaneously learns the relevant characteristics of stations with sufficient historical data, ensuring the model has generalizable power prediction capabilities. After training, the weight parameters of both the cross-attention fusion module and the MLP are saved to prepare for subsequent migration to new stations.

[0038] In the above embodiments, the cold start fine-tuning of the new power station includes two stages: weight transfer and freezing, and MLP lightweight fine-tuning. Weight transfer and freezing involves transferring the weights of the trained multi-source meteorological feature fusion module to the new power station deployment environment and freezing all parameters of the module to ensure its generalizable multi-source meteorological fusion capability can be directly reused without retraining. MLP lightweight fine-tuning uses a small amount of power and meteorological data from the new power station to fine-tune the parameters of the MLP module's output layer, quickly adapting it to the personalized characteristics of the new power station.

[0039] In some embodiments of this disclosure, during the training of the trained power prediction module, the parameters of the feedforward layer of the power prediction module remain unchanged, while the weights and biases of the output layer are updated. During fine-tuning, the parameters of the feedforward layer of the MLP remain unchanged, and only the weights of the output layer are updated. With bias This enables the model to converge quickly in cold start scenarios.

[0040] In some embodiments of this disclosure, multi-source weather forecast data is input into a pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features. One possible approach is to use the multi-source weather forecast feature sequence corresponding to the multi-source weather forecast data as the key vector and value vector, and the historical power and weather feature sequence as the query vector; the features are generated according to a cross-attention mechanism to obtain the fused weather features.

[0041] Figure 2 This is a schematic diagram of the structure of a multi-source meteorological feature fusion module provided as an exemplary embodiment of this disclosure. Figure 2As shown, meteorological data from points around the station are first extracted from multiple numerical weather forecasts and AI weather forecasts. Feature sequences are then constructed by combining the station's historical power with the meteorological data of the same period. Among them, the multi-source weather forecast feature sequences serve as the key vector (K) and value vector (V) of the cross-attention module, while the historical measured power and meteorological feature sequences serve as the query vector (Q). The spatiotemporal relationship between the two is modeled through the cross-attention mechanism, and finally a unified fused meteorological feature containing complementary information from multiple meteorological sources and historical power correlation features is generated.

[0042] The formula for calculating cross attention is: ; in Using the dimension of the key vector, this mechanism can dynamically capture the nonlinear dependence between meteorological elements and power, such as automatically strengthening the weight of high-resolution meteorological sources in complex weather scenarios to improve the accuracy of fused features.

[0043] In some embodiments of this disclosure, the power prediction module includes a forward layer and an output layer; fused meteorological features are input into the pre-trained power prediction module to obtain the wind power prediction result of the new wind farm. One possible approach is to input the fused meteorological features into the forward layer to obtain a high-dimensional representation; and input the high-dimensional representation into the output layer to obtain the wind power prediction result of the new wind farm.

[0044] The fused meteorological features are input into the forward layer of the MLP, and a high-dimensional representation is generated through multi-layer nonlinear transformation. The calculation of the l-th layer is as follows: , in, For the first The hidden layer output, where GELU is the activation function. , This represents the weights and biases of this layer.

[0045] The MLP output layer maps the high-dimensional representation to the predicted power, as shown in the formula: ; in, To predict power, This is the output of the last layer of the MLP. Output layer weights, For bias.

[0046] In some embodiments of this disclosure, fused meteorological features are input into a pre-trained power prediction module to obtain wind power prediction results for new wind farms. Specifically, high-dimensional representations are extracted from the fused meteorological features, and their nonlinear mapping to wind power is learned, while implicitly capturing farm characteristics. For example, short-term wind power predictions for the new farm over the next 72 hours are output, and the prediction accuracy is verified using indicators such as root mean square error (RMSE) and mean absolute error (MAE) by comparing with measured power data. The prediction results can be directly integrated into the power grid dispatch system, providing data support for grid connection scheduling and power consumption of the new farm.

[0047] Figure 3 This is a schematic diagram of the structure of a wind power prediction device 30 provided for an exemplary embodiment of this disclosure. Figure 3 As shown, the wind power prediction device 30 includes: an acquisition module 31, a fusion module 32, and a prediction module 33.

[0048] Among them, the acquisition module 31 is used to acquire multi-source meteorological forecast data of the new wind farm station; The fusion module 32 is used to input multi-source meteorological forecast data into the pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; Prediction module 33 is used to input the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction results of the new wind farm. Specifically, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are then transferred to new wind farms, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0049] Optionally, the prediction module 33 can also be used for: Collect sample multi-source meteorological forecast data and sample historical power and meteorological data of ordinary wind farms; Based on the sample multi-source meteorological forecast data and the sample historical power and meteorological data, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, thus obtaining the trained multi-source meteorological feature fusion module and the trained power prediction module. The weights of the trained multi-source meteorological feature fusion module will be transferred to the new wind farm; and Based on the multi-source meteorological forecast training data and historical power and meteorological training data of the new wind farm, the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

[0050] Optionally, when collecting sample multi-source meteorological forecast data and sample historical power and meteorological data for ordinary wind farms, the prediction module 33 is used for: Collect initial multi-source weather forecast data and initial historical power and meteorological data; Bilinear interpolation is used to map the gridded initial multi-source weather forecast data and initial historical power and meteorological data to ordinary wind farm sites. Spline interpolation is used to unify data sources with different time resolutions to the same sampling frequency, resulting in data-aligned sample multi-source weather forecast data and sample historical power and meteorological data.

[0051] Optionally, the prediction module 33 can also be used for: During the training of the power prediction module, the parameters of the forward layer of the power prediction module remain unchanged, while the weights and biases of the output layer are updated.

[0052] Optionally, when the fusion module 32 inputs multi-source weather forecast data into the pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features, it is used for: Use the multi-source meteorological forecast feature sequence corresponding to the multi-source meteorological forecast data as the key vector and value vector, and use the historical power and meteorological feature sequence as the query vector; By generating features using a cross-attention mechanism, fused meteorological features are obtained.

[0053] Optionally, the power prediction module includes a forward layer and an output layer; the prediction module 33, when inputting the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction result for the new wind farm, is used for: By inputting fused meteorological features into the forward layer, a high-dimensional representation is obtained; By inputting the high-dimensional representation into the output layer, the wind power prediction results of the new wind farm are obtained.

[0054] Optionally, the power prediction module is a multilayer sensor.

[0055] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0056] Figure 4 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of the present disclosure. For example... Figure 4As shown, the electronic device includes a memory 41 and a processor 42. Additionally, the electronic device also includes a power supply component 43 and a communication component 44.

[0057] Memory 41 is used to store computer programs and can be configured to store various other data to support operation on the electronic device. Examples of this data include instructions for any application or method used to operate on the electronic device.

[0058] The memory 41 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0059] Communication component 44 is used for data transmission with other devices.

[0060] The processor 42 can execute computer instructions stored in the memory 41 to: acquire multi-source meteorological forecast data of the new wind farm; input the multi-source meteorological forecast data into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features; input the fused meteorological features into a pre-trained power prediction module to obtain the wind power prediction result of the new wind farm; wherein, based on the sample multi-source meteorological forecast data and sample historical power and meteorological data of ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, so as to obtain the trained multi-source meteorological feature fusion module and the trained power prediction module; transfer the weights of the trained multi-source meteorological feature fusion module to the new wind farm, and train the trained power prediction module to obtain the pre-trained multi-source meteorological feature fusion module and the pre-trained power prediction module.

[0061] Accordingly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program. When the computer-readable storage medium stores a computer program, and the computer program is executed by one or more processors, it causes one or more processors to perform... Figure 1 Each step in the method embodiment.

[0062] Accordingly, embodiments of this disclosure also provide a computer program product, which includes a computer program / instructions that are executed by a processor. Figure 1 Each step in the method embodiment.

[0063] The above Figure 4The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component also includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0064] The above Figure 4 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0065] The aforementioned electronic devices also include a display screen and audio components.

[0066] The display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions, but also the duration and pressure associated with the touch or swipe operation.

[0067] An audio component may be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals may be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0068] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0070] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0071] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0072] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0073] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0074] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0075] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0076] The above are merely specific embodiments of this disclosure, enabling those skilled in the art to understand or implement this disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A cold-start wind power prediction method based on multi-source meteorological fusion and representation learning, characterized in that, include: Acquire multi-source meteorological forecast data for new wind farms; The multi-source meteorological forecast data is input into a pre-trained multi-source meteorological feature fusion module for feature fusion to obtain fused meteorological features. The fused meteorological features are input into the pre-trained power prediction module to obtain the wind power prediction results of the new wind farm. Specifically, based on sample multi-source meteorological forecast data and sample historical power and meteorological data from ordinary wind farms, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, resulting in a trained multi-source meteorological feature fusion module and a trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are then transferred to the new wind farm, and the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

2. The method according to claim 1, characterized in that, The method further includes: Collect sample multi-source meteorological forecast data and sample historical power and meteorological data of the ordinary wind farms; Based on the sample multi-source meteorological forecast data and the sample historical power and meteorological data, the initial multi-source meteorological feature fusion module and the initial power prediction module are jointly trained to learn the fusion rules of multi-source meteorological forecast data and the mapping relationship between meteorological characteristics and wind power, thus obtaining the trained multi-source meteorological feature fusion module and the trained power prediction module. The weights of the trained multi-source meteorological feature fusion module are transferred to the new wind farm; and Based on the multi-source meteorological forecast training data and historical power and meteorological training data of the new wind farm, the trained power prediction module is trained to obtain a pre-trained multi-source meteorological feature fusion module and a pre-trained power prediction module.

3. The method according to claim 2, characterized in that, The collection of sample multi-source meteorological forecast data and sample historical power and meteorological data for the ordinary wind farms includes: Collect initial multi-source weather forecast data and initial historical power and meteorological data; Bilinear interpolation is used to map the gridded initial multi-source weather forecast data and initial historical power and meteorological data to ordinary wind farm sites. Spline interpolation is used to unify data sources with different time resolutions to the same sampling frequency, resulting in data-aligned sample multi-source weather forecast data and sample historical power and meteorological data.

4. The method according to claim 2, characterized in that, The method further includes: During the training of the completed power prediction module, the parameters of the forward layer of the power prediction module remain unchanged, while the weights and biases of the output layer are updated.

5. The method according to claim 1, characterized in that, The step of inputting the multi-source weather forecast data into a pre-trained multi-source weather feature fusion module for feature fusion to obtain fused weather features includes: Use the multi-source meteorological forecast feature sequence corresponding to the multi-source meteorological forecast data as the key vector and value vector, and use the historical power and meteorological feature sequence as the query vector. By generating features using a cross-attention mechanism, fused meteorological features are obtained.

6. The method according to claim 1, characterized in that, The power prediction module includes a forward layer and an output layer; the step of inputting the fused meteorological features into the pre-trained power prediction module to obtain the wind power prediction result of the new wind farm includes: The fused meteorological features are input into the forward layer to obtain a high-dimensional representation; The high-dimensional representation is input into the output layer to obtain the wind power prediction result of the new wind farm.

7. The method according to claim 1, characterized in that, The power prediction module is a multilayer sensor.

Images