Real-time analysis field construction method for electric power meteorological large model reasoning and related device

Abstract

The invention belongs to the technical field of electric power weather forecast, and particularly relates to a real-time analysis field construction method for electric power weather large model reasoning and a related device. The method comprises the following steps: acquiring multi-source electric power meteorological real-time observation data; constructing a real-time analysis model of electric power meteorological large model reasoning; and inputting the multi-source electric power meteorology real-time observation data into a constructed real-time analysis model for electric power meteorology big model reasoning to obtain real-time analysis field data for electric power meteorology big model reasoning. According to the invention, the contradiction of training-reasoning data consistency in the prior art is solved; the method is completely aligned with ERA5 reanalysis data in the aspects of generation logic, temporal-spatial resolution and statistical characteristics, and error offset of model input data is eliminated. The real-time bottleneck is broken through, a continuous analysis framework based on rapid updating circulation and station approximation is designed, a real-time analysis field is generated through model continuous analysis and high-frequency circulation rolling, hourly updating is achieved, and the requirement of a large meteorological model for a high-timeliness initial condition is met.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power meteorological forecasting, and in particular relates to a real-time analysis field construction method and related devices for electric power meteorological large model reasoning. Background Art

[0002] With the in-depth application of artificial intelligence (AI) in weather forecasting, large-scale deep learning-based meteorological models (such as Pangu-Weather and GraphCast) have become a vital addition to global numerical weather forecasting. These models rely on high-quality historical reanalysis data (such as ERA5 (Global Forecast System)) released by the European Centre for Medium-Range Weather Forecasts (ECMWF) during training. However, in actual forecasting operations, they require real-time observations or rapid analysis data (such as the Global Forecast System (GFS) and High Resolution Forecast (HRES)) as inference input. However, due to significant differences in the generation logic, spatiotemporal resolution, and statistical properties of training data and real-time inference data, these models face core challenges in practical applications, including loss of accuracy and limited generalization capabilities. Therefore, a data consistency assurance technology that can bridge both training and inference is urgently needed.

[0003] Existing large-scale meteorological models generally employ a "data substitution" strategy for inference data, directly using real-time analysis fields from global numerical forecast systems (such as the ECMWF High Resolution Area (HRES) and the NCEP (National Centers for Environmental Prediction) GFS) as model input. For example, Pangu-Weather uses GFS 0.25° real-time analysis data for operational inference, and GraphCast relies on ECMWF HRES real-time analysis fields to generate initial conditions. While these real-time analysis data can meet the timeliness requirements of large-scale model inference, their generation relies on complex assimilation algorithms (such as 4D-Var (Four Dimensional Variational Data Assimilation)). These data differ from ERA5 reanalysis data in terms of data source (observation type, spatial coverage), temporal resolution (hourly for ERA5 and 6-hourly for GFS), and variable definition (such as vertical layer distribution), adversely affecting the accuracy and reliability of large-scale model inference.

[0004] Specifically, current large-scale meteorological model inference data suffers from the following issues: First, there are inconsistencies between training and inference data. Differences in the generation processes of ERA5 reanalysis data and real-time analysis fields (GFS / HRES) lead to systematic deviations in the distribution characteristics of temperature, humidity, and other factors. This statistical bias in the input data during inference amplifies forecast errors, a problem known as statistical bias accumulation. Furthermore, while the model learns the characteristics of ERA5 data during training, it must adapt to the different data characteristics of real-time analysis fields during inference. This reduces its ability to capture extreme weather events (such as typhoons and severe convection), creating a bottleneck in generalization. Second, real-time analysis data is not timely enough. ERA5 reanalysis data undergoes a complex quality control and post-processing process, resulting in a delay of approximately one week, making it unsuitable for direct use in real-time forecasts. While real-time analysis fields like GFS / HRES are updated more frequently, their spatial resolution (typically ≥0.25°) does not match that of ERA5 (0.1°), limiting the model's performance in refined forecasting scenarios. Summary of the Invention

[0005] The purpose of the present invention is to provide a real-time analysis field construction method and related devices for power meteorological large-scale model reasoning, so as to solve at least one of the problems existing in the prior art: the inconsistency between training and reasoning data and the lack of timeliness of real-time analysis data.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for constructing a real-time analysis field for power meteorological large-scale model reasoning, comprising: Obtain real-time observation data of power meteorology from multiple sources; Build a real-time analysis model for power and meteorological large-scale model reasoning; Inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning; Among them, the time and space resolution of the multi-source power meteorological real-time observation data are consistent with the real-time analysis model.

[0007] A further improvement of the present invention is that the step of obtaining multi-source electric power meteorological real-time observation data specifically includes: Acquire observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA, and pre-process them to obtain multi-source real-time power meteorological observation data.

[0008] A further improvement of the present invention is that the step of performing pretreatment specifically includes: Use statistical methods to screen out abnormal values, missing data and erroneous data; Convert observation data into a format supported by real-time analysis models; The observation data are spatially interpolated to make the temporal and spatial resolution of the observation data consistent with the real-time analysis model.

[0009] A further improvement of the present invention is that the step of constructing a real-time analysis model for the power meteorological large model inference specifically includes: Obtain the division of rolling time windows, rolling updates and feedback mechanisms, continuous analysis strategies, analysis parameters and weight settings, multi-factor collaborative analysis and iterative correction, and build a real-time analysis model.

[0010] The present invention is further improved in that: in the steps of obtaining the division of the rolling time window, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and constructing and generating a real-time analysis model: The rolling time window is divided into 1-hour intervals. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection observation, and when the deviation is found to be greater than the set threshold, feedback is provided to adjust the update frequency. The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

[0011] A further improvement of the present invention is that the step of inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning specifically includes: By adopting the rolling time window division, rolling update and feedback mechanism, combined with the continuous analysis strategy, analysis parameters and weight setting, real-time analysis based on the station observation data approximating the model state is carried out, and the analysis field is continuously numerically integrated to generate the initial field of the next time window; The wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

[0012] In a second aspect, the present invention provides a real-time analysis field construction device for power meteorological large-scale model reasoning, comprising: Acquisition module, used to obtain multi-source power meteorological real-time observation data; Building modules for constructing real-time analysis models for power and meteorological large-scale model reasoning; A generation module, configured to input the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning, and obtain real-time analysis field data for electric power meteorological large model reasoning; Among them, the time and space resolution of the multi-source power meteorological real-time observation data are consistent with the real-time analysis model.

[0013] A further improvement of the present invention is that the step of obtaining multi-source electric power meteorological real-time observation data specifically includes: Acquire observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA, and pre-process them to obtain multi-source real-time power meteorological observation data.

[0014] A further improvement of the present invention is that the step of performing pretreatment specifically includes: Use statistical methods to screen out abnormal values, missing data and erroneous data; Convert observation data into a format supported by real-time analysis models; The observation data are spatially interpolated to make the temporal and spatial resolution of the observation data consistent with the real-time analysis model.

[0015] A further improvement of the present invention is that the step of constructing a real-time analysis model for the power meteorological large model inference specifically includes: Obtain the division of rolling time windows, rolling updates and feedback mechanisms, continuous analysis strategies, analysis parameters and weight settings, multi-factor collaborative analysis and iterative correction, and build a real-time analysis model.

[0016] The present invention is further improved in that: in the steps of obtaining the division of the rolling time window, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and constructing and generating a real-time analysis model: The rolling time window is divided into 1-hour intervals. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection observation, and when the deviation is found to be greater than the set threshold, feedback is provided to adjust the update frequency. The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

[0017] A further improvement of the present invention is that the step of inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning specifically includes: By adopting the rolling time window division, rolling update and feedback mechanism, combined with the continuous analysis strategy, analysis parameters and weight setting, real-time analysis based on the station observation data approximating the model state is carried out, and the analysis field is continuously numerically integrated to generate the initial field of the next time window; The wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

[0018] In a third aspect, the present invention provides an electronic device comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the method for constructing a real-time analysis field for large-scale power meteorological model reasoning.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium storing at least one instruction, which, when executed by a processor, implements the real-time analysis field construction method for power meteorological large model reasoning.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for constructing a real-time analysis field for electric power meteorological large model reasoning, comprising: obtaining multi-source electric power meteorological real-time observation data; constructing a real-time analysis model for electric power meteorological large model reasoning; inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning, and obtaining real-time analysis field data for electric power meteorological large model reasoning. The present invention first solves the contradiction between training and inference data consistency. The present invention is fully aligned with ERA5 reanalysis data in terms of generation logic, spatiotemporal resolution and statistical characteristics, eliminating the error offset of model input data. Second, it breaks through the real-time bottleneck and designs a continuous analysis framework based on fast update cycles and site approximation. The real-time analysis field is generated through continuous model analysis and high-frequency cyclic rolling, achieving hourly updates and meeting the requirements of meteorological large models for high-time initial conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: Figure 1 A flowchart of a method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to an embodiment of the present invention; Figure 2 A detailed flow chart of a method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to an embodiment of the present invention; Figure 3 This is a flow chart of another method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to an embodiment of the present invention; Figure 4 This is a structural diagram of a real-time analysis field construction device for power meteorological large-scale model reasoning according to an embodiment of the present invention; Figure 5 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other.

[0023] The following detailed description is an exemplary description and is intended to provide further detailed description of the present invention. Unless otherwise indicated, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0024] This paper proposes a real-time analysis field construction method for power meteorological large-scale model reasoning, which can be divided into three major steps: First, real-time input data is collected and pre-processed for the generation of real-time analysis fields. The output will serve as the input for the real-time analysis model constructed in the second step. Secondly, carry out customized configuration and optimization from different angles to build a real-time analysis model for power and meteorological large-scale model reasoning; Finally, taking the data from the first link as input, the real-time analysis model constructed in the second link adopts the configuration of continuous analysis and high-frequency cyclic rolling, and finally constructs the real-time analysis field data for the inference of the large-scale power meteorological model.

[0025] The present invention provides a real-time analysis field construction method for power meteorological large model reasoning, and the specific steps are as follows: step S1 is the first link, completing the collection and preprocessing of input data; step S2 and step S3 belong to the second link, respectively constructing an analysis model for generating a real-time analysis field from different levels of rolling time window division, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction; step S4 is the third link, taking the data of step S1 as input, adopting the real-time analysis model constructed in steps S2 and S3, and finally generating real-time analysis field data that can be used for power meteorological large model reasoning through specific real-time continuous analysis configuration.

[0026] See also Figure 1 As shown, an embodiment of the present invention provides a real-time analysis field construction method for power meteorological large model reasoning, comprising the following steps: S1: Collection and preprocessing of multi-source power meteorological real-time observation data. Collect multi-source power meteorological real-time observation data and perform preprocessing for quality control, formatting, and time-space alignment to prepare for the continuous absorption of observations by the numerical model to generate real-time analysis fields.

[0027] S2: Rapid Update Cycle and Continuous Analysis Framework Construction. Rapidly update the numerical model on an hourly or shorter cycle. Based on this, a continuous analysis framework is constructed, including rolling time window division, rolling update strategy, and feedback mechanism for continuous analysis. This allows the numerical model to continuously analyze real-time power and meteorological observation data.

[0028] S3: Continuous analysis strategy and weight setting and iterative correction. Through continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis, and iterative correction, the accuracy of the analysis after the power meteorological observation data is absorbed into the numerical model is guaranteed, ensuring that the real-time analysis field data is closer to actual observations.

[0029] S4: Real-time continuous analysis and high-frequency rolling inference data generation. Through real-time continuous analysis configuration of numerical models and high-frequency rolling numerical integration, a real-time analysis field is generated for inference of large power meteorological models. This includes wind, temperature, and radiation field data, meeting the needs of power meteorological forecasting and early warning.

[0030] In one embodiment, see Figure 2 As shown, step S1 specifically includes: S11: Collection of multi-source real-time power meteorological observation data. This involves real-time collection of observation data from multiple sources, including ground-based meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at renewable energy sites, and power system SCADA. Observational elements include wind speed, wind direction, irradiance, temperature, humidity, and precipitation. The real-time requirement for data collection is a delay of no more than one hour, meaning that the collection of this multi-source real-time power meteorological observation data must be completed before the next hour.

[0031] In a specific embodiment, the multi-source electric power meteorological real-time observation data adopts the ERA5 data of the European Center.

[0032] S12: Real-time data quality control, data format conversion, and temporal and spatial alignment. Statistical methods (such as climate mean and standard deviation) are used to filter out anomalous values, missing data, and erroneous data to ensure data consistency. Multi-source power and meteorological real-time observation data is converted into a format supported by model assimilation (such as the Little_R format supported by the real-time analysis model). If the temporal and spatial resolutions of multi-source power and meteorological real-time observation data are inconsistent, spatial interpolation is performed to align the observation data with the model grid and vertical levels. The "model grid and vertical levels" referred to here refer to the real-time analysis model constructed in the second step of this invention, which is essentially a customized numerical weather forecast model.

[0033] In one embodiment, see Figure 2 As shown, step S2 specifically includes: S21: Rolling Time Window Division for Continuous Analysis. Based on the temporal resolution of real-time multi-source observation data, the real-time analysis process is divided into multiple rolling time windows (e.g., one window every hour or 30 minutes). Each window includes three steps: observation injection, model integration, and short-term prediction. The multi-source observation data in this application has a temporal resolution of 5 minutes, 10 minutes, 15 minutes, and 1 hour. Considering these factors, the time window is divided into 1 hour windows.

[0034] S22: Construction of a rolling update strategy and feedback mechanism. At the end of each time window simulation, the model state is saved and used as the initial condition for the next window. Multiple iterative updates are proposed, with 1–2 refinements made within the window to further reduce model simulation bias. A real-time feedback mechanism is also established to evaluate the simulation results after each observation injection. If abnormally large deviations are detected, feedback is provided to adjust the update frequency.

[0035] In one embodiment, see Figure 2 As shown, step S3 specifically includes: S31: Determine the continuous analysis strategy. To continuously introduce observational data during the model integration process to effectively correct the model state and thereby improve the accuracy of real-time analysis field data, a reasonable continuous analysis strategy is required to ensure optimal analysis results. First, the correction increment of the introduced observations is calculated within each integration time step. Second, a larger correction weight is applied in the lower boundary layer of the model (0-2 km), while the effect of observation calibration in the upper layer is weakened to ensure that the analysis process conforms to the characteristics of the vertical structure of the atmosphere. This is achieved by determining whether to analyze observational data within the boundary layer and the corresponding vertical weight coefficient.

[0036] S32: Customized analysis parameters and weights. In the model configuration, the site observation approximation method is enabled to calibrate the model field variables, and independent correction coefficients are set for variables such as wind speed, temperature, and humidity. This application, through sensitivity testing, sets the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as parameters such as the horizontal influence radius, vertical influence range, and influence time window of the observation data.

[0037] S33: Multi-factor collaborative analysis and iterative correction. Cross-validation methods are used to timely adjust the weights of different observation data (such as ground station and radar data) to improve overall system adaptability. One to two iterative corrections are allowed within each time window. When the initial observation injection increment is large, rapid feedback is used to correct the background field, ensuring that the updated output is closer to the actual observation.

[0038] In one embodiment, see Figure 2 As shown, step S4 specifically includes: S41: Configuration of real-time continuous analysis model. In order to balance the real-time nature of model analysis, the continuity of integration and the computational cost, and to ensure the stability and accuracy of continuous analysis results, the prediction time after each time window analysis is set in the model configuration file, usually 1-6 hours. At the same time, it is necessary to ensure that the time resolution of the model output meets the actual business needs, that is, the output result is used as the initial condition for the next analysis window, usually every 10 minutes or every 1 hour. This application determines the prediction time to be 1 hour, and the output time resolution is also 1 hour.

[0039] S42: High-frequency cyclic rolling numerical integration. Using the continuous analysis time window divided in step S2 and the constructed rolling update strategy, combined with the continuous analysis strategy and weight setting determined in the previous steps, a real-time analysis based on the site observation data approximating the model state is achieved, and then the analysis field is used to continue numerical integration to generate the initial field for the next window. The time window for this application is 1 hour, that is, the model needs to analyze the site observation data every hour, and then generate the initial field for the next hour, and so on.

[0040] S43: Generate real-time analysis field data for large-scale model inference. The wind, temperature, and radiation field data for each analysis time window in the above process are saved as a NetCDF file. As can be seen from the previous steps, the analysis time is hourly, and through a loop, continuous analysis data is generated. This data is accumulated in chronological order to form the real-time analysis field data for the large-scale power meteorology model inference.

[0041] See also Figure 3 As shown, an embodiment of the present invention provides a real-time analysis field construction method for power meteorological large model reasoning, including: S100, obtaining multi-source power meteorological real-time observation data; S200, building a real-time analysis model for power meteorological large-scale model reasoning; S300, inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning; Among them, the time and space resolution of the multi-source power meteorological real-time observation data are consistent with the real-time analysis model.

[0042] In a specific embodiment, the step of obtaining multi-source electric power meteorological real-time observation data specifically includes: Acquire observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA, and pre-process them to obtain multi-source real-time power meteorological observation data.

[0043] In a specific embodiment, the multi-source electric power meteorological real-time observation data adopts the ERA5 data of the European Center.

[0044] In a specific embodiment, the step of performing pretreatment specifically includes: Use statistical methods to screen out abnormal values, missing data and erroneous data; Convert observation data into a format supported by real-time analysis models; The observation data are spatially interpolated to make the temporal and spatial resolution of the observation data consistent with the real-time analysis model.

[0045] In a specific embodiment, the step of constructing a real-time analysis model for power meteorological large model reasoning specifically includes: Obtain the division of rolling time windows, rolling updates and feedback mechanisms, continuous analysis strategies, analysis parameters and weight settings, multi-factor collaborative analysis and iterative correction, and build a real-time analysis model.

[0046] In a specific embodiment, in the steps of obtaining the division of the rolling time window, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and constructing and generating a real-time analysis model: The rolling time window is divided into 1-hour intervals. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection observation, and when the deviation is found to be greater than the set threshold, feedback is provided to adjust the update frequency. The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

[0047] In a specific embodiment, the step of inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning specifically includes: By adopting the rolling time window division, rolling update and feedback mechanism, combined with the continuous analysis strategy, analysis parameters and weight setting, real-time analysis based on the station observation data approximating the model state is carried out, and the analysis field is continuously numerically integrated to generate the initial field of the next time window; The wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

[0048] This invention provides a method for constructing a real-time analysis field for inference of a large-scale power and meteorological model. First, it ensures consistency between the training and inference data of the large-scale power and meteorological model. The real-time analysis field generated by this invention has a correlation coefficient of 0.98 or greater with ERA5 reanalysis data in terms of variable definition, vertical layer structure, and statistical error distribution. This ensures a high degree of consistency between the training dataset and inference data of the large-scale model in terms of generation mode, spatiotemporal resolution, element definition, and deviation characteristics. Second, it addresses the bottleneck issue of real-time inference data. Based on this real-time analysis field construction technology, the initial analysis field data required for large-scale model inference can be rapidly generated, meeting real-time requirements.

[0049] See also Figure 4As shown, an embodiment of the present invention provides a real-time analysis field construction device for power meteorological large model reasoning, comprising: Acquisition module, used to obtain multi-source power meteorological real-time observation data; In a specific embodiment, observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA are obtained and preprocessed to obtain multi-source power meteorological real-time observation data.

[0050] In a specific embodiment, preprocessing specifically includes: using statistical methods to screen out abnormal values, missing and erroneous data; converting the observation data into a format supported by the real-time analysis model; and performing spatial interpolation on the observation data to make the time and spatial resolution of the observation data consistent with the real-time analysis model.

[0051] Building modules for constructing real-time analysis models for power and meteorological large-scale model reasoning; In a specific implementation, it specifically includes: obtaining the division of rolling time windows, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and building a real-time analysis model. Specifically, for example: The rolling time window is divided into 1-hour intervals. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection observation, and when the deviation is found to be greater than the set threshold, feedback is provided to adjust the update frequency. The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

[0052] A generation module, configured to input the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning, and obtain real-time analysis field data for electric power meteorological large model reasoning; Among them, the time and space resolution of the multi-source power meteorological real-time observation data are consistent with the real-time analysis model.

[0053] In a specific implementation, a rolling time window division, rolling update and feedback mechanism are adopted, combined with a continuous analysis strategy, analysis parameters and weight setting, to conduct real-time analysis based on site observation data to approximate the model state, and continue numerical integration of the analysis field to generate the initial field of the next time window; the wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

[0054] This paper proposes a real-time analysis field construction method for power meteorological large-scale model inference: (1) The real-time analysis field construction method dynamically integrates ERA5 reanalysis data with real-time observations to generate a high-precision real-time analysis field with temporal and spatial resolution and variable definitions that are completely consistent with ERA5. (2) A continuous analysis and hierarchical weight setting scheme designs differentiated site observation approximation time windows and weight coefficients based on the vertical structure characteristics of the atmosphere (such as the boundary layer and free troposphere) to optimize analysis efficiency and accuracy. (3) A high-frequency rolling generation of inference datasets under a fast update loop framework.

[0055] The present invention proposes a method for constructing a real-time analysis field for inference of a large power meteorological model: (1) a fast update cycle and continuous analysis framework is constructed, and continuous analysis of real-time power meteorological observation data is achieved through rolling time window division; (2) a continuous analysis strategy and weight setting are determined, and dynamic correction of layered approximation of temperature, humidity, and wind field is achieved; (3) real-time continuous analysis is achieved through high-frequency cycle rolling, and a real-time analysis field for inference of a large power meteorological model is accumulated.

[0056] See also Figure 5 As shown, an embodiment of the present invention provides an electronic device 100 for implementing a real-time analysis field construction method for power meteorological large model reasoning; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0057] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for constructing a real-time analysis field for large-scale power meteorological model inference described in the embodiment by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area. The program storage area can store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data (such as audio data) created based on the use of the electronic device 100. In addition, the memory 101 can include non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device.

[0058] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100 and connects various parts of the entire electronic device 100 using various interfaces and lines.

[0059] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a real-time analysis field construction method for power meteorological large model reasoning, and the processor 102 can execute the plurality of instructions to implement: Obtain real-time observation data of power meteorology from multiple sources; Build a real-time analysis model for power and meteorological large-scale model reasoning; The multi-source electric power meteorological real-time observation data is input into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning.

[0060] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0061] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0062] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0063] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A real-time analysis field construction method for large-scale power meteorological model reasoning, characterized in that: include: Obtain real-time observation data of power meteorology from multiple sources; Build a real-time analysis model for power and meteorological large-scale model reasoning; Inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning; The time and space resolutions of the multi-source electric power meteorological real-time observation data are consistent with the real-time analysis model.

2. The method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to claim 1 is characterized in that: The step of obtaining multi-source electric power meteorological real-time observation data specifically includes: Acquire observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA, and pre-process them to obtain multi-source real-time power meteorological observation data.

3. The method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to claim 2 is characterized in that: The steps of performing pre-processing specifically include: Use statistical methods to screen out abnormal values, missing data and erroneous data; Convert observation data into a format supported by real-time analysis models; The observation data are spatially interpolated to make the temporal and spatial resolution of the observation data consistent with the real-time analysis model.

4. The method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to claim 1, characterized in that: The steps of constructing a real-time analysis model for the power meteorological large model reasoning specifically include: Obtain the division of rolling time windows, rolling updates and feedback mechanisms, continuous analysis strategies, analysis parameters and weight settings, multi-factor collaborative analysis and iterative correction, and build a real-time analysis model.

5. The method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to claim 4 is characterized in that: In the steps of obtaining the division of rolling time windows, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and constructing and generating a real-time analysis model: The rolling time window is divided into 1 hour periods. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection of observations, and when the deviation is found to be greater than the set threshold, feedback is given to adjust the update frequency; The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

6. The method for constructing a real-time analysis field for power meteorological large-scale model reasoning according to claim 4, characterized in that: The step of inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning specifically includes: By adopting the rolling time window division, rolling update and feedback mechanism, combined with the continuous analysis strategy, analysis parameters and weight setting, real-time analysis based on the station observation data approximating the model state is carried out, and the analysis field is continuously numerically integrated to generate the initial field of the next time window; The wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

7. A real-time analysis field construction device for power meteorological large-scale model reasoning, characterized in that: include: Acquisition module, used to obtain multi-source power meteorological real-time observation data; Building modules for constructing real-time analysis models for power and meteorological large-scale model reasoning; A generation module, configured to input the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning, and obtain real-time analysis field data for electric power meteorological large model reasoning; The time and space resolutions of the multi-source electric power meteorological real-time observation data are consistent with the real-time analysis model.

8. The real-time analysis field construction device for electric power meteorological large-scale model reasoning according to claim 7 is characterized in that: The step of obtaining multi-source electric power meteorological real-time observation data specifically includes: Acquire observation data from multiple sources including ground meteorological stations, sounding meteorological stations, weather radars, meteorological satellites, wind and light measurement stations at new energy sites, and power system SCADA, and pre-process them to obtain multi-source real-time power meteorological observation data.

9. The real-time analysis field construction device for electric power meteorological large model reasoning according to claim 8 is characterized in that: The steps of performing pre-processing specifically include: Use statistical methods to screen out abnormal values, missing data and erroneous data; Convert observation data into a format supported by real-time analysis models; The observation data are spatially interpolated to make the temporal and spatial resolution of the observation data consistent with the real-time analysis model.

10. The real-time analysis field construction device for electric power meteorological large model reasoning according to claim 7, characterized in that: The steps of constructing a real-time analysis model for the power meteorological large model reasoning specifically include: Obtain the division of rolling time windows, rolling updates and feedback mechanisms, continuous analysis strategies, analysis parameters and weight settings, multi-factor collaborative analysis and iterative correction, and build a real-time analysis model.

11. The real-time analysis field construction device for electric power meteorological large model reasoning according to claim 10, characterized in that: In the steps of obtaining the division of rolling time windows, rolling update and feedback mechanism, continuous analysis strategy, analysis parameter and weight setting, multi-factor collaborative analysis and iterative correction, and constructing and generating a real-time analysis model: The rolling time window is divided into 1 hour periods. The rolling update strategy in the rolling update and feedback mechanism is to save the model state at the end of each time window simulation and use it as the initial condition for the next window. The feedback mechanism strategy in the rolling update and feedback mechanism is to evaluate the simulation results after each injection of observations, and when the deviation is found to be greater than the set threshold, feedback is given to adjust the update frequency; The continuous analysis strategy includes whether to analyze the observation data within the boundary layer and the vertical weight coefficient corresponding to the boundary layer; The analysis parameters and weight settings include the wind speed observation approximation weight system, temperature observation approximation weight coefficient, humidity observation approximation weight coefficient, as well as the horizontal influence radius, vertical influence range and influence time window of the observation data; Multi-factor collaborative analysis and iterative correction include adjusting the weights of different observation data through cross-validation methods, and performing 1 to 2 iterative corrections in each time window to make the updated output closer to the actual observation.

12. The real-time analysis field construction device for electric power meteorological large-scale model reasoning according to claim 10, characterized in that: The step of inputting the multi-source electric power meteorological real-time observation data into the constructed real-time analysis model for electric power meteorological large model reasoning to obtain real-time analysis field data for electric power meteorological large model reasoning specifically includes: By adopting the rolling time window division, rolling update and feedback mechanism, combined with the continuous analysis strategy, analysis parameters and weight setting, real-time analysis based on the station observation data approximating the model state is carried out, and the analysis field is continuously numerically integrated to generate the initial field of the next time window; The wind field, temperature field and radiation field data at the analysis moment corresponding to each time window are saved as a preset format file to obtain real-time analysis field data for the power meteorological large model inference.

13. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the real-time analysis field construction method for power meteorological large model reasoning as claimed in any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the real-time analysis field construction method for power meteorological large model reasoning according to any one of claims 1 to 6 is implemented.

Images