Seamless fine forecasting method, device and equipment for marine meteorology of offshore wind plant and medium

Abstract

The invention discloses a seamless fine forecasting method and device for marine meteorology of an offshore wind plant, equipment and a medium, and the method comprises the steps: fusing multi-source data, preprocessing the data, constructing a Transform model, training and optimizing the Transform model, deploying the trained Transform model to carry out time sequence prediction, and carrying out the seamless fine forecasting of the marine meteorology of the offshore wind plant based on the time sequence prediction and spatial interpolation. And seamless refined meteorological ocean forecasting is realized. According to the invention, through an advanced artificial intelligence network algorithm, in combination with multi-source data fusion and a meteorological ocean numerical mode module, a meteorological ocean forecasting system with high timeliness, high resolution and high accuracy is constructed, seamless fusion with user side data is realized, and reliable support is provided for meteorological ocean APP, data API service and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of marine meteorological forecasting for offshore wind farms, and in particular to a method, device, equipment and medium for seamless and precise forecasting of marine meteorological conditions for offshore wind farms. Background Art

[0002] Offshore wind resources are abundant, and power companies are gradually shifting their focus from onshore wind power to offshore wind power. Eastern coastal regions, in particular, are endowed with offshore wind energy resources, have a strong industrial base, and strong power consumption capacity, but face a shortage of conventional energy sources, resulting in a strong demand for wind energy development.

[0003] However, offshore wind power projects are typically weather-dependent. Severe weather conditions such as severe convection, cold front winds, sea fog, and typhoons significantly impact project construction, production, and operations. This is especially true given the increasing frequency, intensity, and concurrence of extreme weather events, which pose significant challenges to many aspects of offshore wind power projects. Consequently, offshore wind power projects are experiencing a surge in demand for meteorological services, placing extremely high demands on seamless and precise forecasting capabilities for marine weather at offshore wind farms.

[0004] Currently, marine meteorological forecasts for offshore wind farms primarily rely on traditional numerical forecasting methods: A variety of observational data (including radar and satellite data) is integrated into gridded meteorological variables such as temperature, pressure, humidity, and wind speed. Partial differential equations describing the interactions between these elements are then developed based on atmospheric dynamics. Finally, their evolution is simulated numerically. This approach has the following problems:

[0005] (1) Poor prediction accuracy. According to ECMWF data, the 3- to 7-day forecast error for several major meteorological elements decreased by less than 5% from 2012 to 2022. This is because traditional numerical meteorological forecasting algorithms rely heavily on approximations and parameterization, which can introduce cumulative errors. (2) Low resolution and insufficient accuracy. Most existing meteorological and oceanographic forecasting technologies rely on numerical weather prediction (NWP) models, which typically have low spatial and temporal resolution and cannot provide sufficiently detailed meteorological and oceanographic data. This results in the inability of wind farms to obtain high-resolution data that is sensitive to small meteorological changes, thus affecting the energy forecasting and equipment maintenance decisions of wind farms; (3) Lack of seamless forecasting capabilities. Traditional forecasting systems usually have a "time gap" problem, that is, the forecast results cannot cover every time point, resulting in mutations between different time periods; (4) Ignoring nonlinear relationships and complex meteorological and ocean phenomena. Traditional meteorological forecasting models are mostly based on physical equations or linear assumptions. Although they can provide a relatively wide range of meteorological data forecasts, it is difficult to accurately capture the complex nonlinear relationships in the meteorological and oceanic environment; (5) Lack of multi-source data fusion. Existing technologies often rely on only a single data source (such as satellite remote sensing data, weather station data or buoy data). This limitation makes the data coverage insufficient.

[0006] Invention application number 202210146830.X discloses a method for short-term wind speed forecasting for offshore wind farms. This method obtains a short-term wind speed forecast for the offshore wind farm by inputting multiple meteorological factors affecting wind speed into a short-term wind speed forecast model based on a BP neural network. This method focuses solely on meteorological factors affecting wind speed and does not incorporate multiple factors such as satellite remote sensing data, buoy observation data, radar data, and numerical weather forecast data.

[0007] The rapid growth of the offshore wind power industry has led to complex and demanding meteorological and oceanographic environments that pose serious risks to wind turbine equipment, operators, and vessels. This has also increased the demand for meteorological and oceanographic forecasting and warning services for safety-critical conditions such as high waves, storm surges, severe convection, and sea fog. Therefore, a new, seamless forecasting approach is needed to address the limited availability of specialized meteorological and oceanographic services for offshore wind farms and the limited sophistication of high-impact weather monitoring and warning. Summary of the Invention

[0008] In response to the above-mentioned problems, the purpose of the present invention is to provide a method, device, equipment and medium for seamless and fine forecasting of marine meteorology for offshore wind farms, which integrates multi-source data to achieve seamless and fine meteorological and ocean forecasting of marine meteorology for offshore wind farms.

[0009] The embodiments of the present invention provide a method, device, equipment and medium for seamless and precise forecasting of marine meteorology in offshore wind farms.

[0010] A first aspect: A method for seamless and precise forecasting of marine meteorology for offshore wind farms, comprising:

[0011] S1, integration of multi-source data;

[0012] S2. Preprocess the data;

[0013] S3. Build the Transformer model and perform training and optimization.

[0014] S4. Deploy the trained Transformer model for time series forecasting.

[0015] S5. Based on time series prediction and spatial interpolation, seamless and refined meteorological and ocean forecasts are achieved.

[0016] Optionally, the multi-source data includes:

[0017] Satellite remote sensing data, used to provide large-scale meteorological and oceanographic information;

[0018] Buoy observation data, used to provide real-time ocean environment data;

[0019] Radar data, which provides high-resolution precipitation and wind speed information;

[0020] Numerical weather prediction data is used to provide global or regional weather forecast data.

[0021] Optionally, the data is preprocessed, including normalizing the multi-source data, and then performing Gaussian filtering to remove noise and bilinear interpolation to fill missing data, wherein:

[0022] Gaussian filtering denoising, the formula is expressed as:

[0023]

[0024] in, is the filtered data, is the Gaussian kernel function, is the original data;

[0025] Bilinear interpolation fills missing data, and the formula is expressed as:

[0026]

[0027] in, is the interpolated data, and is the known data around the missing point;

[0028] After data preprocessing, the training set, validation set, and test set were constructed in a ratio of 7:2:1 for subsequent model training and evaluation.

[0029] Optionally, the Transformer model includes a multi-head attention mechanism module and a position encoding module, wherein:

[0030] The multi-head attention mechanism module adopts the multi-head attention mechanism, which is expressed as follows:

[0031]

[0032] in, represent the query matrix, key matrix and value matrix respectively, is the dimension of the key vector; the head attention mechanism captures the complex nonlinear relationship between meteorological elements at different time scales through multiple attention heads.

[0033] Position encoding module, the formula is expressed as:

[0034]

[0035]

[0036] in, is the position index, is the dimension index, The position encoding module introduces explicit time order information for time series data, ensuring that the model can capture time dependencies.

[0037] Optionally, the Transformer model is trained and optimized using a mean square error loss function and an Adam optimizer, where:

[0038] The mean square error loss function is expressed as:

[0039]

[0040] in, is the actual observed value, is the model prediction value, is the sample size;

[0041] Adam optimizer, the formula is expressed as:

[0042]

[0043]

[0044]

[0045] in, and are the first and second moment estimates of the gradient, is the learning rate, is the smoothing term.

[0046] Optionally, the Transformer model performs time series prediction, and the formula is expressed as:

[0047]

[0048] in, is the meteorological and oceanographic data at the current moment, is the input data at the current moment.

[0049] Optionally, the spatial interpolation formula is expressed as:

[0050]

[0051] in, is the interpolated time series, and are the known time series points around the missing point;

[0052] The results of spatial interpolation are combined with the time series prediction results to generate spatiotemporally continuous refined forecast data.

[0053] The second aspect: A seamless and precise forecasting device for marine meteorology at an offshore wind farm, comprising:

[0054] Acquisition module, collects multi-source data and performs pre-processing;

[0055] The prediction module deploys a trained Transformer model to obtain time series predictions based on multi-source data;

[0056] The interpolation module realizes seamless and refined meteorological and ocean forecasts based on time series prediction and spatial interpolation technology.

[0057] A third aspect: An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method provided in the first aspect are implemented.

[0058] A fourth aspect: A non-transitory computer-readable storage medium having a computer program stored thereon, which implements the steps of the method provided in the first aspect when executed by a processor.

[0059] Beneficial effects of the present invention:

[0060] 1. The self-attention mechanism of the Transformer large model can capture the complex nonlinear interactions between multiple factors such as wind speed, waves and temperature, thereby improving the prediction accuracy. Combining bilinear interpolation with multi-source data fusion, the spatial resolution is improved to meet the needs of refined forecasting at the wind turbine point level, and the forecast accuracy is significantly improved.

[0061] 2. This invention adopts a parallelized Transformer architecture and Adam optimizer to shorten regional forecasting time, meet minute-level response requirements for scenarios such as sudden changes in typhoon paths, and improve real-time forecasting capabilities. This solution reduces energy consumption costs and resource consumption for computing at the same scale.

[0062] 3. The present invention achieves continuous forecasting by integrating time series prediction with spatial interpolation technology, eliminating the "blind spots" of traditional model update intervals and enhancing temporal continuity. At the same time, it integrates multi-source data such as satellites, buoys, and radars to build a three-dimensional seamless monitoring network for land, sea, and air, with high recognition accuracy.

[0063] 4. The present invention adopts multi-parameter joint forecasting, supports synchronous prediction of multiple parameters such as wind speed, wave height, visibility, etc., meets the complex decision-making needs of scenarios such as wind turbine start and stop, can automatically adapt the forecast model for different wind farm locations, and shortens the user-customized response time.

[0064] 5. This technical solution, through the deep integration of AI large-scale models and meteorological and oceanographic disciplines, addresses the industry pain points of "inaccurate measurements, slow calculations, and incomplete information" in meteorological services for the offshore wind power sector, providing reliable technical support for the full life cycle management of offshore wind power. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 Schematic diagram of the process of the seamless fine forecasting method of the present invention;

[0066] Figure 2 This is a principle flow chart of the seamless fine forecasting method of the present invention;

[0067] Figure 3 This is a schematic structural diagram of the seamless fine forecasting device of the present invention;

[0068] Figure 4 Schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0069] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0070] Traditional meteorological and ocean forecasting technologies have problems such as low resolution, poor real-time performance, and insufficient accuracy, making it difficult to meet the offshore wind farms' requirements for high-precision and high-timeliness meteorological and ocean data.

[0071] In order to solve the above problems, the present invention provides a method for seamless and precise forecasting of marine meteorology for offshore wind farms. Figure 1 A flow chart of a method for seamless and precise forecasting of marine meteorology for offshore wind farms provided by an embodiment of the present invention, the method comprising:

[0072] S1. Fusion of multi-source data.

[0073] Traditional meteorological and oceanographic forecasting systems rely primarily on single data sources (such as satellite remote sensing, weather stations, or buoy data). The limitations of these data sources prevent traditional models from fully reflecting the complex meteorological and oceanographic environments of offshore wind farms. Existing technologies often overlook the complementary nature of different data sources, severely impacting forecast accuracy when data is missing or inaccurate.

[0074] Multi-source data collection is the basis of the present invention. The forecast accuracy is improved by fusing the collected multi-source data. Specifically, the sources of multi-source data include:

[0075] Satellite remote sensing data (provides a wide range of meteorological and oceanographic information, such as sea surface temperature, wind speed, cloud distribution, etc.); buoy observation data (provides real-time ocean environment data, such as wave height, current speed, etc.); radar data (provides high-resolution precipitation, wind speed, etc. information); numerical weather forecast data (provides global or regional weather forecast data).

[0076] This invention overcomes the limitations of a single data source by fusing data from multiple sources, including satellite remote sensing, buoy observations, radar data, and numerical weather forecasts. Multi-source data fusion not only increases data coverage but also integrates data from different sources into a single system, thereby improving the reliability and accuracy of forecast results. This approach enables the system to obtain more information in complex environments, resulting in more comprehensive and accurate forecasts of meteorological and oceanographic conditions.

[0077] Multi-source data fusion enhances the model's ability to perceive diverse meteorological and ocean phenomena, especially in special environments such as offshore wind farms. It can better respond to changes in complex data such as sea surface, wind speed, and waves, and provide more accurate predictions for real-time decision-making.

[0078] S2. Preprocess the data.

[0079] Data preprocessing is to unify the format of multi-source data, remove noise, fill missing values, and provide high-quality data for model training. Data preprocessing includes standardizing the multi-source data using a normalization formula, and then performing Gaussian filtering to remove noise and bilinear interpolation to fill missing data.

[0080] Gaussian filtering denoising, the formula is expressed as:

[0081]

[0082] in, is the filtered data, is the Gaussian kernel function, is the original data;

[0083] Bilinear interpolation fills missing data, and the formula is expressed as:

[0084]

[0085] in, is the interpolated data, and is the known data around the missing point.

[0086] After preprocessing, construct the data set and divide the data set into

[0087] S3. Build the Transformer model and perform training and optimization.

[0088] Traditional meteorological and oceanographic forecasting models are typically based on physical equations. While these models have a long history of application, they often struggle to capture nonlinear relationships due to the complexities of meteorological and oceanographic conditions. While common machine learning models (such as support vector machines and random forests) can capture certain nonlinear characteristics, they suffer from low efficiency and accuracy when processing time series and large-scale data.

[0089] like Figure 2 As shown in Figure 1, the present invention uses the Transformer model to model meteorological and oceanographic data. Through its self-attention and multi-head attention mechanisms, the Transformer can flexibly capture complex nonlinear relationships in time series data. It is particularly powerful when processing multidimensional data (such as wind speed, temperature, and waves). Furthermore, the Transformer's parallel computing capabilities significantly improve training and inference efficiency, overcoming the computational bottlenecks of traditional models.

[0090] The Transformer model includes a multi-head attention mechanism module and a position encoding module, where:

[0091] The multi-head attention mechanism module adopts the multi-head attention mechanism, which is expressed as follows:

[0092]

[0093] in, represent the query matrix, key matrix and value matrix respectively, is the dimension of the key vector;

[0094] Position encoding module, the formula is expressed as:

[0095]

[0096]

[0097] in, is the position index, is the dimension index, is the model dimension.

[0098] The position encoding module introduces explicit temporal order information for time series data, ensuring that the model can capture temporal dependencies.

[0099] The Transformer model accurately captures complex spatiotemporal dependencies, playing a crucial role in the refined forecasting of meteorological and oceanographic data. Its efficient computing power and nonlinear modeling capabilities make this method superior to traditional technologies in terms of both accuracy and real-time performance in meteorological and oceanographic forecasting.

[0100] By collecting historical data from multiple sources, we construct training sets, test sets, and validation sets to train and optimize the Transformer model. The mean square error loss function and Adam optimizer are used for Transformer model training and optimization.

[0101] The mean square error loss function is expressed as:

[0102]

[0103] in, is the actual observed value, is the model prediction value, is the sample size;

[0104] Adam optimizer, the formula is expressed as:

[0105]

[0106]

[0107]

[0108] in, and are the first and second moment estimates of the gradient, is the learning rate, is the smoothing term.

[0109] Traditional meteorological and ocean forecasting systems require powerful computing resources and long computing times, making it difficult to meet the needs of real-time forecasting, especially in application scenarios with high real-time requirements such as offshore wind farms.

[0110] This paper improves the computational efficiency of the Transformer model by using a high-performance computing platform and the Adam optimizer optimization algorithm, and significantly shortens the model training and inference time through parallelization technology; even in the case of massive data, the model can complete predictions in a relatively short time and provide real-time meteorological and ocean forecast results.

[0111] Through efficient calculations, the present invention can meet the needs of real-time meteorological and ocean forecasts, help offshore wind farms respond quickly in complex and changing marine environments, and optimize the scheduling, maintenance and safety warnings of wind farms.

[0112] S4. Deploy the trained Transformer model for time series forecasting.

[0113] The Transformer model performs time series prediction, and the formula is expressed as:

[0114]

[0115] in, is the meteorological and oceanographic data at the current moment, . is the input data at the current moment.

[0116] S5. Based on time series prediction and spatial interpolation, seamless and refined meteorological and ocean forecasts are achieved.

[0117] Traditional forecasting methods often face problems such as low temporal and spatial resolution and large time gaps. Especially when forecasting in highly dynamic environments such as offshore wind farms, they often result in forecast gaps or insufficient spatial accuracy, affecting the operation and decision-making of wind farms.

[0118] This paper combines Transformer model time series prediction with spatial interpolation technology to achieve seamless and refined meteorological and oceanographic forecasts. The Transformer model's time series modeling capabilities enable this paper to accurately predict future meteorological and oceanographic conditions, and spatial interpolation technology improves spatial resolution, ensuring the continuity and high resolution of forecast results.

[0119] Since the time series prediction results may be spatially discontinuous, it is necessary to combine the interpolated results with the time series prediction results to generate spatiotemporally continuous refined forecast data.

[0120] Spatial interpolation, the formula is expressed as:

[0121]

[0122] in, is the interpolated time series, and are the known time series points around the missing point.

[0123] Seamless and refined forecasts can provide wind farms with more accurate real-time meteorological data, eliminating the "time gaps" in traditional forecasts, while improving spatial resolution, making forecast results more in line with the needs of wind farms, thereby improving the production efficiency and safety of wind farms.

[0124] The present invention also provides a device for seamless and precise forecasting of marine meteorology at offshore wind farms, such as Figure 3 As shown, the device includes:

[0125] Acquisition module, collects multi-source data and performs pre-processing;

[0126] The prediction module deploys a trained Transformer model to obtain time series predictions based on multi-source data;

[0127] The interpolation module realizes seamless and refined meteorological and ocean forecasts based on time series prediction and spatial interpolation technology.

[0128] The present invention utilizes an acquisition module to collect offshore wind power meteorological service data sets, including satellite remote sensing data, buoy observation data, radar data, and numerical weather forecast data. A prediction module, incorporating a trained Transformer model, then performs time series forecasting based on the acquired multi-source data. The interpolation module utilizes time series prediction and spatial interpolation techniques to achieve seamless and refined meteorological and oceanographic forecasts.

[0129] Compared with the resource-intensive characteristics of traditional numerical forecasts, the seamless and fine forecasting device for marine meteorology in offshore wind farms of the present invention adopts a large AI model to perform seamless and fine forecasting of meteorology and oceanography in offshore wind farms, which significantly reduces the consumption of computing resources and improves the timeliness of business applications. It can provide customization and construction services of forecast service systems tailored to the personalized needs of different users, provide high-precision, high-timeliness and adaptable meteorological and ocean forecasts, and provide users with accurate and reliable meteorological and ocean information services.

[0130] The present invention also provides an electronic device, Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 4 As shown, the electronic device may include: a processor, a communications interface, a memory, and a communication bus, wherein the processor, the communications interface, and the memory communicate with each other via the communication bus. The processor may call logic instructions in the memory, for example, to execute the following method:

[0131] S1, integration of multi-source data;

[0132] S2. Preprocess the data;

[0133] S3. Build the Transformer model and perform training and optimization.

[0134] S4. Deploy the trained Transformer model for time series forecasting.

[0135] S5. Based on time series prediction and spatial interpolation, seamless and refined meteorological and ocean forecasts are achieved.

[0136] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0137] An embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method provided in each of the above embodiments is implemented, for example, including:

[0138] S1, integration of multi-source data;

[0139] S2. Preprocess the data;

[0140] S3. Build the Transformer model and perform training and optimization.

[0141] S4. Deploy the trained Transformer model for time series forecasting.

[0142] S5. Based on time series prediction and spatial interpolation, seamless and refined meteorological and ocean forecasts are achieved.

[0143] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0144] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A seamless and precise forecasting method for marine meteorology at offshore wind farms, characterized in that: include: S1, integration of multi-source data; S2. Preprocess the data; S3. Build the Transformer model and perform training and optimization. S4. Deploy the trained Transformer model for time series forecasting. S5. Based on time series prediction and spatial interpolation, seamless and refined meteorological and ocean forecasts are achieved.

2. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The multi-source data includes: Satellite remote sensing data, used to provide large-scale meteorological and oceanographic information; Buoy observation data, used to provide real-time ocean environment data; Radar data, which provides high-resolution precipitation and wind speed information; Numerical weather prediction data is used to provide global or regional weather forecast data.

3. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The data preprocessing includes normalizing the multi-source data, then performing Gaussian filtering to remove noise and bilinear interpolation to fill in missing data, wherein: Gaussian filtering denoising, the formula is expressed as: ； in, is the filtered data, is the Gaussian kernel function, is the original data; Bilinear interpolation fills missing data, and the formula is expressed as: ； in, is the interpolated data, and is the known data around the missing point; After data preprocessing, the training set, validation set, and test set were constructed in a ratio of 7:2:1 for subsequent model training and evaluation.

4. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The Transformer model includes a multi-head attention mechanism module and a position encoding module, where: The multi-head attention mechanism module adopts the multi-head attention mechanism, which is expressed as follows: ； in, represent the query matrix, key matrix and value matrix respectively, is the dimension of the key vector; the head attention mechanism captures the complex nonlinear relationship between meteorological elements at different time scales through multiple attention heads; Position encoding module, the formula is expressed as: ； ； in, is the position index, is the dimension index, The position encoding module introduces explicit time order information for time series data, ensuring that the model can capture time dependencies.

5. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The Transformer model is trained and optimized using the mean square error loss function and the Adam optimizer, where: The mean square error loss function is expressed as: ； in, is the actual observed value, is the model prediction value, is the sample size; Adam optimizer, the formula is expressed as: ； ； ； in, and are the first and second moment estimates of the gradient, is the learning rate, is the smoothing term.

6. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The Transformer model performs time series prediction, and the formula is expressed as: ； in, is the meteorological and oceanographic data at the current moment, is the input data at the current moment.

7. The method for seamless and precise forecasting of marine meteorology for offshore wind farms according to claim 1, characterized in that: The spatial interpolation formula is expressed as: ； in, is the interpolated time series, and are the known time series points around the missing point; The results of spatial interpolation are combined with the time series prediction results to generate spatiotemporally continuous refined forecast data.

8. A seamless and precise forecasting device for marine meteorology in an offshore wind farm, applied to the method according to any one of claims 1 to 7, characterized in that: The device comprises: Acquisition module, collects multi-source data and performs pre-processing; The prediction module deploys a trained Transformer model to obtain time series predictions based on multi-source data; The interpolation module realizes seamless and refined meteorological and ocean forecasts based on time series prediction and spatial interpolation technology.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method for seamless and precise forecasting of marine meteorology for offshore wind farms according to any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for seamless and precise forecasting of marine meteorology for an offshore wind farm as claimed in any one of claims 1 to 7 are implemented.

Images