Offshore wind resource forecasting method, device, computer device, and storage medium

Abstract

This application relates to the technical field of offshore wind power forecasting and discloses an offshore wind power resource forecasting method, apparatus, computer equipment, and storage medium. The application converts a first offshore meteorological field dataset into a second offshore meteorological field dataset that can identify weather forecast modes, and optionally uses a first wind tower wind speed dataset and a second wind tower wind speed dataset as the basis to build a target offshore wind power resource forecasting model using an advanced distributed gradient boosting library. This not only improves the accuracy of offshore wind power resource forecasting, but also greatly simplifies the data fusion process. The integrated use of multi-source forecast data based on a global combined forecasting system can make up for the deficiencies of different data, fully utilize the advantages of various data, and make the forecast results more reliable.

Description

[Technical Field]

[0001] The present application relates to the technical field of offshore wind forecasting, and in particular to an offshore wind resource forecasting method, apparatus, computer device and storage medium. [Background technology]

[0002] Weather Research and Forecast (WRF) is a numerical model used in meteorological forecasting and research. It is a high-resolution, non-static, non-hierarchical, non-uniform atmospheric numerical model. WRF divides the Earth's atmosphere into horizontal grids and vertical layers, and simulates atmospheric phenomena by solving atmospheric dynamics, thermodynamics, and hydrological recycling equations using discrete equations. This model is widely used in wind resource simulation and wind speed forecasting.

[0003] To improve the accuracy and reliability of the model, a global coupled forecast system is generally adopted to provide accurate and comprehensive meteorological environment driving data for the WRF model. The fusion of multi-source forecast data can make full use of the advantages of each global coupled forecast system, compensate for the shortcomings between different systems, and improve the accuracy, completeness, and reliability of the forecast.

[0004] Although numerous studies have shown that multi-mode forecasts are clearly superior to single-mode forecast systems, several problems still exist in the fusion process. For example, differences exist between the data formats and time resolutions of different forecast systems, necessitating data format conversion and time alignment. At the same time, the physical processes and parameter settings of the forecast models of different systems are different, necessitating constant parameter matching and data calibration. Second, in the case of inconsistencies or discrepancies between the forecast data of different systems, rational data processing and quality control must be carried out, otherwise the forecast results will be adversely affected. Summary of the Invention [Problem to be solved by the invention]

[0005] [Means for solving the problem]

[0006] In view of this, the present application provides an offshore wind resource forecasting method, apparatus, computer device and storage medium to solve the problem of poor forecast results due to the problems existing in the conventional multi-mode forecast fusion process.

[0007] In a first aspect, the present application provides an offshore wind resource forecasting method, the offshore wind resource forecasting method comprising: The method includes the steps of: acquiring a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecast system; using the global coupled forecast system to operate a weather forecast mode and converting the first offshore meteorological field dataset into a second offshore meteorological field dataset based on the weather forecast mode; performing an interpolation process on the second offshore meteorological field dataset to obtain a second wind tower wind speed dataset; establishing a target offshore wind resource forecast model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset using a distributed gradient boosting library; acquiring a third offshore meteorological field dataset and a third wind tower wind speed dataset of a target offshore wind field; and inputting the third offshore meteorological field dataset and the third wind tower wind speed dataset into the target offshore wind resource forecast model to obtain a wind resource forecast result of the target offshore wind field.

[0008] The offshore wind resource forecasting method of the present application converts a first offshore meteorological field dataset into a second offshore meteorological field dataset that can identify weather forecast modes, and optionally uses an advanced distributed gradient boosting library to build a target offshore wind resource forecasting model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset. This not only improves the accuracy of offshore wind resource forecasting, but also greatly simplifies the data fusion process. The comprehensive use of multi-source forecast data based on a global combined forecasting system can make up for the deficiencies of different data, fully utilize the advantages of various data, and make the forecast results more reliable.

[0009] In one alternative embodiment, using the global coupled forecast system to operate a weather forecast mode and converting the first ocean weather field dataset into a second ocean weather field dataset based on the weather forecast mode includes using the global coupled forecast system to operate a weather forecast mode and generate a target simulation domain based on the weather forecast mode, and interpolating the first ocean weather field dataset into the target simulation domain using a preset interpolation method to obtain the second ocean weather field dataset.

[0010] The present application uses the global combined forecasting system to drive the weather forecast mode, converting the first ocean meteorological field dataset into a second ocean meteorological field dataset that can be identified by the weather forecast mode, thereby improving the accuracy of the data; and obtaining the first ocean meteorological field dataset based on the global combined forecasting system can make up for the deficiencies of different data, fully utilizing the advantages of various data, and making the forecast results more reliable.

[0011] In one alternative embodiment, the step of utilizing the global coupled forecast system to drive a weather forecast mode and generating a target simulation domain based on the weather forecast mode includes: The method includes the steps of using the global coupled forecast system to drive a weather forecast mode and generate an initial simulation area, obtaining an area length and an area width of the initial simulation area, dividing the area length into a plurality of longitude segments and dividing the area width into a plurality of latitude segments, determining a grid point longitude and latitude for each grid point based on the plurality of longitude segments and the plurality of latitude segments, wherein the grid points are composed of longitude segments and latitude segments, and generating a target simulation area based on the weather forecast mode based on the initial simulation area and the grid point longitude and latitude for each grid point.

[0012] By determining the grid point latitude and longitude for each grid point, the generated target simulation domain can convert the input data into latitude and longitude data that the weather forecast mode can identify.

[0013] In one alternative embodiment, the step of establishing a target offshore wind resource forecasting model based on the first wind tower wind speed data set and the second wind tower wind speed data set utilizing a distributed gradient boosting library comprises: The method includes processing the first and second wind tower wind speed datasets using a data processing method to obtain first and second target wind tower wind speed datasets; generating a target feature model based on the first and second target wind tower wind speed datasets using a feature engineering process and a cross-validation method; establishing an initial offshore wind resource forecast model based on the second target wind tower wind speed dataset and the target feature model using a distributed gradient boosting library; and adjusting the initial offshore wind resource forecast model using the first wind tower wind speed dataset until the target offshore wind resource forecast model is obtained.

[0014] The present application uses a data processing method to process a first wind tower wind speed dataset and a second wind tower wind speed dataset to achieve the objective of eliminating the random effects of annual wind speed fluctuations, and also uses a distributed gradient boosting library to build a model and uses the second wind tower wind speed dataset to adjust the model and improve the forecast accuracy of the target offshore wind resource forecasting model.

[0015] In one alternative embodiment, the step of establishing an initial offshore wind resource forecast model utilizing a distributed gradient boosting library based on the first target wind tower wind speed dataset and the target feature model comprises: The method includes the steps of obtaining a model feature map of a target feature model, inputting the model feature map into a distributed gradient boosting library to obtain a plurality of mark points corresponding to the model feature map, obtaining a construction tree based on the plurality of mark points, and establishing an initial offshore wind resource forecasting model based on the construction tree.

[0016] This application obtains a construction tree at multiple mark points corresponding to the model feature map, and uses the construction tree to establish an initial offshore wind resource prediction model, thereby improving the prediction accuracy of the model to a certain extent.

[0017] In one alternative embodiment, the method comprises: The method further includes calculating a mean absolute error, a mean square error, and a root mean square error of the target offshore wind resource forecasting model; and evaluating the target offshore wind resource forecasting model based on the mean absolute error, the mean square error, and the root mean square error to obtain an evaluation result.

[0018] The present application utilizes the mean absolute error, mean square error and root mean square error to evaluate the target offshore wind resource forecasting model, and can obtain the forecasting performance of the target offshore wind resource forecasting model, and further provides data support for the forecasting of offshore wind resource.

[0019] In one alternative embodiment, after the step of utilizing the global coupled forecast system to drive a weather forecast mode and generating a target simulation domain based on the weather forecast mode, the method further comprises: The method further includes dividing the target simulation domain into a plurality of sub-domains, and setting the sub-domains as nest domains when the sub-domains perform computing resource configuration.

[0020] The present application allows sub-areas requiring computing resource configuration to be nested, and allows high-resolution grids to be adopted only within the nested areas, thereby balancing the timeliness of computing and the accuracy of weather forecast mode simulations.

[0021] In a second aspect, the present application provides an offshore wind resource prediction apparatus, the offshore wind resource prediction apparatus including: a data acquisition module for acquiring a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecast system; a conversion module for using the global coupled forecast system to operate a weather forecast mode and converting the first offshore meteorological field dataset into a second offshore meteorological field dataset based on the weather forecast mode; a processing module for performing an interpolation process on the second offshore meteorological field dataset to obtain a second wind tower wind speed dataset; an establishment module for establishing a target offshore wind resource prediction model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset using a distributed gradient boosting library; and an acquisition / input module for acquiring a third offshore meteorological field dataset and a third wind tower wind speed dataset of a target wind field, and inputting the second offshore meteorological field dataset and the third wind tower wind speed dataset into the target offshore wind resource prediction model to obtain a wind resource prediction result for the target wind field.

[0022] In a third aspect, the present application provides a computer apparatus comprising a memory and a processor communicatively connected to each other, the memory having computer instructions stored therein, the processor executing the computer instructions to perform the offshore wind resource forecasting method according to the first aspect above or any of its corresponding embodiments.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium having stored thereon computer instructions for causing a computer to execute an offshore wind resource forecasting method according to the first aspect or any of the corresponding embodiments thereof.

[0024] In order to more clearly describe the specific embodiments of the present application or the technical solutions of the prior art, the drawings necessary for describing the specific embodiments or the prior art will be briefly described below. It is obvious that the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without any creative work. [Brief explanation of the drawings]

[0025] [Figure 1] 1 is a flowchart of an offshore wind resource forecasting method according to an embodiment of the present application. [Figure 2] 1 is a flowchart of another offshore wind resource forecasting method according to an embodiment of the present application. [Figure 3] 1 is a flowchart of another offshore wind resource forecasting method according to an embodiment of the present application. [Figure 4] 1 is a flowchart of an offshore wind resource forecasting method using multi-source forecast data according to an embodiment of the present application; [Figure 5] 1 is a structural block diagram of an offshore wind resource prediction device according to an embodiment of the present application. [Figure 6] FIG. 1 is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present application; DETAILED DESCRIPTION OF THE INVENTION

[0026] In order to clarify the objectives, technical solutions and advantages of the embodiments of the present application, the following will clearly and completely describe them with reference to the drawings of the embodiments and the technical solutions of the embodiments of the present application, and it is obvious that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments that a person skilled in the art can obtain based on the embodiments of the present application without any creative work fall within the scope of protection of the present application.

[0027] The offshore wind resource forecasting method according to the embodiment of the present application makes comprehensive use of multi-source forecast data and utilizes a distributed gradient boosting library to establish a target offshore wind resource forecasting model, thereby improving the forecast accuracy of offshore wind resource, compensating for the deficiencies of different data, and fully utilizing the advantages of various data, thereby achieving the effects of making the forecast results more reliable.

[0028] According to an embodiment of the present application, an embodiment of an offshore wind resource forecasting method is provided, and it is noted that the steps illustrated in the flowcharts of the drawings may be performed in a computer system, such as a set of computer-executable instructions, and that although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in an order different from that shown here.

[0029] In this embodiment, an offshore wind resource forecasting method is provided. FIG. 1 is a flowchart of the offshore wind resource forecasting method according to the embodiment of the present application. As shown in FIG. 1, the process includes the following steps: Step S101: Obtain a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecasting system.

[0030] Here, the global coupled forecast system may include the National Centers for Environmental Prediction (NCEP), the European Center for Medium-Range Weather Forecasts (ECMWF), the Japan Meteorological Agency (JMA), and the Meteorological Service of Canada (MSC), among others.

[0031] Optionally, each of the globally coupled forecasting systems has a corresponding data library configured, and meteorological data is stored in the data library.

[0032] The first offshore meteorological field data set may include information such as offshore wind field potential height, horizontal wind speed, temperature, air pressure, humidity, soil water content, and soil temperature.

[0033] The first wind tower wind speed data set is actual measurement data from a wind tower installed at an offshore wind field, and may include measurements corresponding to different heights, wind speeds, wind direction values, ground temperature, humidity, and air pressure.

[0034] Here, the steps for acquiring the first wind tower wind speed data set are as follows: The measurement period is set and marked as T, the actual measurement data within one measurement period T is obtained, the amount of abnormal data in the actual measurement data is obtained and marked as R, and if the amount of abnormal data R exceeds the set period change value, the measurement period T is extended and the actual measurement data within the extended measurement period is continuously monitored in real time until the corresponding first wind speed data set is obtained. In this embodiment, the measurement period is set and whether to extend the measurement period is determined based on the amount of abnormal data, thereby achieving the purpose of eliminating the influence of randomness on annual wind speed fluctuations.

[0035] By selectively combining meteorological data in the data library corresponding to each global coupled forecasting system, multi-source meteorological data, i.e., the first ocean meteorological field data set, can be generated, which can compensate for the shortcomings of different data, fully utilize the advantages of various data, and make the forecast results more reliable.

[0036] Step S102: Utilizing the global combined forecasting system to drive a weather forecast mode, and converting the first ocean meteorological field dataset into a second ocean meteorological field dataset based on the weather forecast mode.

[0037] Here, the weather forecasting mode (WRF mode) is configured with a pre-processing process that converts data generated by the global coupled forecasting system into a data format that the WRF mode can directly recognize and read.

[0038] Specifically, when the WRF mode is driven using a global coupled forecasting system, the pre-processing process converts the first ocean meteorological field dataset into a second ocean meteorological field dataset that can be directly identified and read by the WRF mode.

[0039] In step S103, an interpolation process is performed on the second offshore meteorological field data set to obtain a second wind speed data set from an anemometer tower.

[0040] First, data that does not meet the preset data criteria in the second ocean meteorological field dataset is filtered out, and then the filtered out second ocean meteorological field dataset is interpolated to the corresponding wind tower to generate a refined dataset, i.e., a second wind tower wind speed dataset, including Data-NCEP, Data-ECMWF, Data-JMA, and Data-MSC.

[0041] Step S104: Based on the first wind tower wind speed data set and the second wind tower wind speed data set, a target offshore wind resource forecasting model is established using a distributed gradient boosting library.

[0042] Here, the distributed gradient boosting library (eXtreme Gradient Boosting, XGBoost) represents a machine learning function library that focuses on gradient boosting algorithms and can be used for classification, regression, and sorting.

[0043] In this embodiment, the first wind speed data set from a wind gauge tower is used as the test set, and the second wind speed data set from a wind gauge tower is used as the training set. A regression model, i.e., a target offshore wind resource prediction model, is constructed based on the XGBoost library to realize offshore wind resource prediction, improve the offshore wind resource prediction accuracy, and greatly simplify the data fusion process.

[0044] Step S105: Obtain a third offshore meteorological field dataset and a third wind tower wind speed dataset for the wind field of the forecast target area, input the third offshore meteorological field dataset and the third wind tower wind speed dataset into the target offshore wind resource prediction model, and obtain wind resource prediction results for the wind field of the forecast target area.

[0045] Specifically, the acquired third offshore meteorological field dataset and third wind tower wind speed dataset for the wind field of the predicted target area can be input into the constructed target offshore wind resource prediction model, and the wind resource prediction results for the wind field of the predicted target area can be output.

[0046] The offshore wind resource forecasting method of this embodiment converts a first offshore meteorological field dataset into a second offshore meteorological field dataset that can identify weather forecast modes, and then uses an advanced distributed gradient boosting library to build a target offshore wind resource forecasting model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset, which not only improves the offshore wind resource forecasting accuracy but also greatly simplifies the data fusion process. The comprehensive use of multi-source forecast data based on the global combined forecasting system can make up for the deficiencies of different data, fully utilize the advantages of various data, and make the forecast results more reliable.

[0047] In this embodiment, an offshore wind resource forecasting method is provided. FIG. 2 is a flowchart of the offshore wind resource forecasting method according to the embodiment of the present application. As shown in FIG. 2, the process includes steps S201 to S202: Step S201: obtain a first offshore meteorological field dataset and a first wind speed dataset from a wind gauge tower based on a global combined forecasting system, which can be referred to as step S101 in the embodiment shown in FIG. 1, and the description will not be repeated here. Step S202: Utilizing the global combined forecasting system to drive a weather forecast mode, and converting the first ocean meteorological field dataset into a second ocean meteorological field dataset based on the weather forecast mode.

[0048] Specifically, the above step S202 includes steps S2021 to S2022, Step S2021: Use the global coupled forecasting system to drive the weather forecast mode, and generate a target simulation area based on the weather forecast mode. Here, the target simulation domain converts the input data into a data format that the WRF model can directly recognize and read. Specifically, the corresponding target simulation domain can be obtained according to the pre-processing process of the WRF mode. In step S2022, the first offshore meteorological field data set is interpolated into the target simulation domain using a preset interpolation method to obtain a second offshore meteorological field data set. Here, different interpolation dimensions are set in the target simulation domain, and the first offshore meteorological field dataset has corresponding interpolation feature coefficients. Specifically, after the first ocean meteorological field dataset is input into the target simulation domain, the first ocean meteorological field dataset corresponding to the interpolation feature coefficient can be matched one-to-one with the interpolation dimension in the target simulation domain based on the correspondence relationship in the preset reference interpolation correspondence table. After the matching is successful, the data in the first ocean meteorological field dataset can be converted into the corresponding longitude and latitude, and then converted into a second ocean meteorological field dataset in WRF format, and the WRF mode can directly identify the second ocean meteorological field dataset. Here, the reference interpolation correspondence table records the correspondence between the interpolation feature coefficients and the interpolation dimensions.

[0049] JPEG2025527972000002.jpg132159

[0050] JPEG2025527972000003.jpg35159

[0051] Optionally, get grid point longitude and latitude.

[0052] Specifically, every two adjacent longitude segments and every two adjacent latitude segments form a square grid, and the intersection point formed by the intersection position of the longitude segment and the latitude segment is called a grid point. That is, the initial simulation domain is composed of multiple square grids, and each square grid contains multiple grid points.

[0053] Selectively mark grid points as X, X=<longitude, latitude>, and the grid point longitude and latitude corresponding to each grid point will be X=<w、h> and the value of w ranges from w1 to wN and the value of h ranges from h1 to h M is. Finally, based on the initial simulation area and the grid point longitude and latitude of each grid point, a target simulation area can be determined based on the corresponding weather forecast mode.

[0054] In some alternative embodiments, after step a1, step S2021 further includes steps a6 and a7; Step a6: Divide the target simulation domain into a plurality of sub-domains. Step a7: when the sub-domain performs computing resource configuration, the sub-domain is set as a nest domain.

[0055] Specifically, a plurality of longitude segments and a plurality of latitude segments are grouped together, the longitude segments and the latitude segments intersect to form an equal-longitude / latitude grid, and the simulation domain is covered by the entire equal-longitude / latitude grid.

[0056] Optionally, the equal longitude and latitude grid does not have the capability of local encryption, and although the equal longitude and latitude grid is a high resolution grid, the high resolution grid consumes a large amount of computing resources.

[0057] Therefore, the target simulation domain can be divided into multiple sub-domains, and each sub-domain can configure its computing resources. When a sub-domain configures its computing resources, it obtains the positioning number of the sub-domain and sets the sub-domain with the positioning number as the nest domain.

[0058] In this embodiment, the sub-region requiring computing resource configuration is designated as a nested region, and a high-resolution grid is adopted only within the nested region. The part that belongs to the target simulation region but does not belong to the sub-region where computing resources are configured is divided into a parent region, which provides a coarse resolution and plays a role in balancing the timeliness of computing and the accuracy of the simulation.

[0059] Optionally, the second marine meteorological field dataset obtained by the above step is called downscaled data because of its low spatial resolution.

[0060] Step S203: interpolate the second offshore meteorological field data set to obtain a second wind speed data set from a wind gauge tower. For details, refer to step S103 in the embodiment shown in FIG. 1, and the repeated description will be omitted here.

[0061] Step S204: Establish a target offshore wind resource forecasting model based on the first wind tower wind speed data set and the second wind tower wind speed data set using a distributed gradient boosting library, for details, see step S104 in the embodiment shown in Figure 1, and the repeated description will be omitted here.

[0062] Step S205: obtain a third offshore meteorological field dataset and a third wind speed dataset from a wind gauge tower for the wind field of the target region, input the third offshore meteorological field dataset and the third wind speed dataset from a wind gauge tower into a target offshore wind resource forecasting model to obtain a wind resource forecast result for the wind field of the target region. For details, see step S105 in the embodiment shown in Figure 1, and repeated explanations will be omitted here.

[0063] Step S206: Calculate the mean absolute error, mean square error and root mean square error of the target offshore wind resource forecasting model.

[0064] JPEG2025527972000004.jpg43160

[0065] JPEG2025527972000005.jpg35160

[0066] JPEG2025527972000006.jpg35160

[0067] Step S207: Evaluate the target offshore wind resource forecasting model based on the mean absolute error, the mean square error and the root mean square error to obtain an evaluation result.

[0068] JPEG2025527972000007.jpg25160

[0069] The offshore wind resource forecasting method of this embodiment determines the grid point longitude and latitude of each grid point, and thereby converts the input first offshore meteorological field data set into a second offshore meteorological field data set that can identify the weather forecast mode for the generated target simulation domain, improving data accuracy. Furthermore, the first offshore meteorological field data set is obtained based on a global combined forecasting system, making it possible to compensate for the deficiencies of different data sets, fully utilizing the advantages of various data sets and making the forecast results more reliable. Optionally, the mean absolute error, mean square error, and root mean square error are used to evaluate the model forecasting performance of the target offshore wind resource forecasting model, providing data support for offshore wind resource forecasting.

[0070] In this embodiment, an offshore wind resource forecasting method is provided. FIG. 3 is a flowchart of the offshore wind resource forecasting method according to the embodiment of the present application. As shown in FIG. 3, the process includes steps S301 to S305: Step S301: obtain a first offshore meteorological field dataset and a first wind speed dataset from a wind gauge tower based on a global combined forecasting system, which can be referred to as step S101 in the embodiment shown in FIG. 1, and the detailed description will be omitted here. Step S302: Activate the weather forecast mode using the global combined forecasting system to convert the first ocean meteorological field data set into a second ocean meteorological field data set based on the weather forecast mode. For details, refer to step S202 in the embodiment shown in FIG. 2, and the repeated description will be omitted here. Step S303: interpolate the second offshore meteorological field data set to obtain a second wind speed data set from a wind gauge tower. For details, refer to step S103 in the embodiment shown in Figure 1, and the repeated description will be omitted here. Step S304: Based on the first wind tower wind speed data set and the second wind tower wind speed data set, a target offshore wind resource forecasting model is established using a distributed gradient boosting library.

[0071] Specifically, the above step S304 includes steps S3041 to S3044, In step S3041, the first wind tower wind speed data set and the second wind tower wind speed data set are processed by a data processing method to obtain a first target wind tower wind speed data set and a second target wind tower wind speed data set.

[0072] Here, data processing methods may include data evaluation, missing value treatment, repeated value treatment, outlier treatment and variable transformation. (1) Data evaluation: Data evaluation performs completeness detection on the test set and training set, and screens out missing values, repeated values, and outliers in the test set and training set. (2) Missing Value Handling: Missing value handling identifies the content of missing values ​​and the matrix position where the missing values ​​exist, and performs a filling operation on the missing values. (3) Repeated value processing: Repeated data in the test set (second wind speed data set from wind gauge tower) and the training set (first wind speed data set from wind gauge tower) are detected by repeating value processing, and the data in the test set and the training set are traversed in sequence. The first traversed data is marked as A, and when the data is traversed for the second or subsequent times, it is marked as A'. After the traversal is completed, the number of times the data is acquired is obtained, marked as Y, and Y-1 pieces of the data are removed. (4) Outlier Treatment: Outlier treatment detects abnormal data values ​​in the training set and the test set, and performs retention, deletion, and correction on the abnormal data values. (5) Variable transformation: The modeling format of the training set and the data set is transformed, and the transformed modeling format can be used to build a data model.

[0073] In step S3042, a target feature model is generated based on the first target wind tower wind speed data set and the second target wind tower wind speed data set through feature engineering processing and cross-validation method processing.

[0074] Here, the feature engineering process may include feature extraction, feature transformation and feature construction. Specifically, after obtaining the modeling format, feature extraction, feature conversion and feature construction are performed on the first target wind tower wind speed data set and the second target wind tower wind speed data set corresponding to the modeling format; First, a plurality of valid feature points and invalid noise points are obtained in the first target wind speed data set and the second target wind speed data set by feature extraction, and the invalid noise points are removed. Next, the effective feature points obtained by the feature transformation are transformed into mark points of feature map information, and a model feature map is generated by combining a plurality of mark points. Finally, the model feature maps are modeled into corresponding feature models through feature construction.

[0075] Optionally, in the process of constructing the feature model, cross-validation needs to be performed on a second target wind tower wind speed dataset (training set); First, we use K-set cross-validation to divide the training set into K equal subsets, traverse the K subsets in turn, and use each subset in the traversal as a validation set, while the other K-1 subsets are used as sub-training sets. Next, the feature model is trained K-1 times using the sub-training set, and the performance of the feature model is evaluated using the validation set after each training run. Each subset serves as a validation set once, and training of the feature model on the sub-training set and evaluation of the performance of the feature model on the test set are repeated until optimal model parameters are obtained. Finally, based on the optimal model parameters and the corresponding feature model constructed by feature engineering, a corresponding target feature model can be determined.

[0076] Step S3043: Based on the second target wind tower wind speed dataset and the target feature model, an initial offshore wind resource forecasting model is established using a distributed gradient boosting library. Specifically, the XGBoost library and the second target wind tower wind speed dataset can be used to train the target feature model, and the corresponding initial offshore wind resource forecast model can be obtained.

[0077] In step S3044, the first wind tower wind speed data set is used to adjust the initial offshore wind resource forecast model until a target offshore wind resource forecast model is obtained. Specifically, the first wind tower wind speed data set is used to adjust the model parameters and hyperparameters of the initial offshore wind resource forecasting model, and the corresponding target offshore wind resource forecasting model is output until the model parameters and hyperparameters are optimized.

[0078] Here, model parameters can be selected based on a set of initial parameters, such as the learning rate, the number of trees to build, the maximum depth of the trees to build, the column sampling rate, etc.

[0079] The hyperparameter optimization adjustment can be selectively performed using grid search or random search, and any two different hyperparameters are selected and combined to evaluate the performance of the initial offshore wind resource forecasting model when different hyperparameters are used. The performance of the initial offshore wind resource forecasting model is judged by a performance coefficient, and the hyperparameter combination with the largest performance coefficient is selected as the optimal hyperparameter combination.

[0080] Optionally, without changing the hyperparameters, the model parameters are adjusted to perform optimization tuning to continuously improve the performance of the initial offshore wind resource forecasting model. The optimization tuning of the model parameters may be performed by adjusting the learning rate to different set learning coefficient levels, building the maximum tree depth to different set tree depth levels, and increasing the column sampling rate.

[0081] In some optional embodiments, step S3043 includes steps b1 to b3, Step b1: obtain the model feature map of the target feature model. Step b2, the model feature map is input into the distributed gradient boosting library to obtain multiple mark points corresponding to the model feature map. Step b3, a construction tree is obtained based on the multiple mark points, and an initial offshore wind resource forecasting model is established based on the construction tree.

[0082] Specifically, the model feature map of the obtained target feature model is input into the XGBoost library, and the XGBoost library can obtain multiple mark points contained in the model feature information deconstructed from the model feature map.

[0083] Then, we obtain the construction tree at the mark point.

[0084] Optionally, the construction tree includes multiple tree nodes, where different tree nodes have corresponding split contributions, where the split contributions have corresponding feature importances, where the feature importances have different importance scores, and in this embodiment, the feature subset with the highest value is selected as the final feature subset, and the feature subset is obtained to construct and generate an initial offshore wind resource forecasting model.

[0085] Here, the importance score can be obtained by the feature_import attribute of the XGBoost library.

[0086] Step S305: obtain a third offshore meteorological field dataset and a third wind speed dataset from a wind gauge tower for the wind field of the target region, input the third offshore meteorological field dataset and the third wind speed dataset from a wind gauge tower into a target offshore wind resource forecasting model to obtain a wind resource forecast result for the wind field of the target region. For details, see step S105 in the embodiment shown in Figure 1, and repeated explanations will be omitted here.

[0087] The offshore wind resource forecasting method of this embodiment processes a first wind tower wind speed dataset and a second wind tower wind speed dataset using a data processing method to achieve the objective of eliminating the random influence of annual wind speed fluctuations, and also uses a distributed gradient boosting library to obtain a construction tree at multiple mark points corresponding to the obtained model feature map, and establishes a target offshore wind resource forecasting model using the first wind tower wind speed dataset and the construction tree to improve the forecasting accuracy of the target offshore wind resource forecasting model.

[0088] In one embodiment, an offshore wind resource forecasting method based on multi-source forecast data is provided, which includes steps S1 to S3, as shown in FIG. 4 : Step S1: Access the global coupled forecast system and use the data-driven WRF mode to define the simulation domain and nest domain, calculate the grid point longitude and latitude, obtain the ground data and meteorological field data, and interpolate them within the simulation domain. Step S2: Obtain meteorological field data for the target wind field, interpolate the downscaled data generated in WRF mode to the wind tower to obtain refined wind speed data, and obtain the actual measured data from the wind tower. The refined wind speed data and the actual measured data are used as the training set and test set, respectively, and perform data preprocessing and feature engineering operations. Cross-validation is performed on the training set to select the optimal model parameters. Step S3: Build a regression model using the XGBoost library, train the regression model using the training set, adjust the model parameters and hyperparameters of the regression model according to the test set, calculate the mean absolute error, mean square error and root mean square error of the regression model to evaluate the predictive performance of the regression model, and use the trained regression model to predict the wind speed at the wind field location.

[0089] Optionally, the process of defining the simulation domain and nest domain includes: After the data drives the WRF mode, the ocean measurement area positioned is obtained, the ocean measurement area is marked as a simulation area, the area length and area width of the ocean measurement area are obtained, the area length and area width are respectively divided into multiple longitude segments and latitude segments, the longitude segments and latitude segments intersect to form an equal-longitude and latitude grid, the simulation area is divided into multiple sub-areas, each of which allows computing resource configuration, and when a sub-area performs computing resource configuration, the sub-area is set as a nested area.

[0090] Optionally, the process of obtaining grid point longitude and latitude includes: A plurality of square grids are obtained, each consisting of two adjacent longitude segments and two adjacent latitude segments in the simulation area. The intersections of the longitude and latitude segments constitute grid points, and the grid points have corresponding longitudes and latitudes. The longitude and latitude of each grid point are then used to obtain the grid point longitudes and latitudes for the entire simulation area.

[0091] Optionally, the process of the interpolation operation includes: After obtaining the ground data and meteorological field data, the simulation domain is divided into a ground layer and a meteorological field layer, the ground data and the meteorological field data have corresponding interpolation feature coefficients, and different interpolation dimensions are set for the ground layer and the meteorological field layer. After the ground data and the meteorological field data are input into the simulation domain, the ground data and the meteorological field data related to the interpolation feature coefficients are matched with the interpolation dimensions corresponding to the ground layer and the meteorological field layer.

[0092] Optionally, the process of generating the training and test sets includes: The wind field area of ​​the wind speed to be detected is set as a target area, meteorological field data of the target area is obtained, downscaling data is interpolated to an installed wind tower, the downscaling data is interpolated to the wind tower and then refined data is generated, actual measurement data actually measured by the wind tower is obtained, the obtained refined data is used as a training set, and the obtained actual measurement data is used as a test set.

[0093] Optionally, the process of data pre-processing and feature engineering operations includes: The data preprocessing includes data evaluation, missing value treatment, repeated value treatment, outlier treatment, and variable transformation. Data evaluation is used to screen out missing values, repeated values, and outliers in the test set and the training set. Missing value treatment is used to identify the content of the missing values ​​and the matrix positions where the missing values ​​are located, and fill the missing values. Repeated value treatment is used to detect repeated data in the training set and the test set, and traversal is used to remove the repeated data. Outlier treatment is used to detect abnormal data values ​​in the training set and the test set, and retain, delete, and modify the abnormal data values. Variable transformation is used to convert the modeling format of the training set and the data set. Feature extraction is performed on the training set and the data set corresponding to the modeling format to obtain multiple effective feature points. The effective feature points obtained by feature transformation are converted into mark points of feature map information, and the multiple mark points are combined to generate a model feature map. Feature construction is used to model the model feature map into a corresponding feature model.

[0094] Optionally, the process of cross-validating the training set includes: After obtaining the training set, divide the training set into K equal subsets, traverse the K subsets, and use the subset in each traversal as a validation set in turn, and the other K-1 subsets as sub-training sets. The feature model is trained K-1 times using the sub-training sets, and the performance of the feature model is evaluated using the validation set for each training. The training of the feature model using the sub-training sets and the evaluation of the performance of the feature model using the test set are repeated until each subset has served as a validation set once, and the optimal model parameters are generated after cross-validation is completed.

[0095] Optionally, the process of building a regression model and tuning model parameters and hyperparameters includes: A model feature map is obtained and input into the XGBoost library. The XGBoost library obtains a plurality of mark points contained in the deconstructed feature map information from the model feature map. A construction tree is obtained at the mark points. The construction tree includes a plurality of tree nodes, different tree nodes have corresponding split contributions, the split contributions have different feature importances, and the feature importances have different importance scores. The importance score with the highest value is selected as a feature subset. A regression model is constructed according to the feature subset. The regression model has corresponding model parameters and hyperparameters. The model parameters and hyperparameters are transmitted to the regression model. The regression model is trained using the training set. The model parameters and hyperparameters are optimized and adjusted according to the expression of the test set.

[0096] Optionally, the process of calculating the mean absolute error, mean square error and root mean square error of the regression model refers to the above relations (1) to (3).

[0097] Compared with the prior art, the beneficial effects of the offshore wind resource forecasting method based on multi-source forecast data according to this embodiment are as follows: 1. Based on the results of predicting the target area wind speed in WRF mode and actual measured data, fitting prediction is performed using the advanced XGBoost regression fitting algorithm, which not only improves the forecast accuracy of offshore wind resources, but also greatly simplifies the data fusion process. The comprehensive use of multi-source forecast data can make up for the deficiencies of different data, fully utilize the advantages of various data, and make the forecast results more reliable. 2. The equal-longitude and latitude grid does not have the ability to be locally encrypted. Although the equal-longitude and latitude grid is a high-resolution grid, high-resolution grids consume a large amount of computing resources. Therefore, the sub-areas that require computing resource configuration are nested, and high-resolution grids are only used within the nested areas. The parts that belong to the simulation area but do not belong to the sub-areas where computing resources are configured are divided into parent areas, which provide coarse resolution and play a role in balancing the timeliness of computing and the accuracy of the simulation. By setting the measurement period and determining whether to extend the measurement period based on the amount of abnormal data, the purpose of eliminating the random effects of annual wind speed fluctuations is achieved.

[0098] This embodiment further provides an offshore wind resource forecasting device, which is used to realize the above-described embodiment and optional embodiments, and the parts that have already been described will not be described again. As used below, the term "module" can realize a combination of software and / or hardware for a given function, and the device described in the following embodiment is preferably realized in software, but it is also possible and envisioned to realize it in hardware or a combination of software and hardware.

[0099] This embodiment provides an offshore wind resource prediction device, as shown in FIG. a data acquisition module 501 for acquiring a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecast system; a transformation module 502 for utilizing a global coupled forecast system to drive a weather forecast mode and transforming the first marine weather field dataset into a second marine weather field dataset based on the weather forecast mode; a processing module 503 for performing interpolation processing on the second offshore meteorological field data set to obtain a second wind tower wind speed data set; an establishment module 504 for establishing a target offshore wind resource forecasting model based on the first wind tower wind speed data set and the second wind tower wind speed data set utilizing a distributed gradient boosting library; and an acquisition and input module 505 for acquiring a third offshore meteorological field dataset and a third wind tower wind speed dataset of the target offshore wind field, inputting the third offshore meteorological field dataset and the third wind tower wind speed dataset into a target offshore wind resource prediction model, and obtaining a wind resource prediction result for the target offshore wind field.

[0100] In some alternative embodiments, the conversion module 502: a generating unit for driving a weather forecast mode using a global coupled forecast system and generating a target simulation domain based on the weather forecast mode; and an interpolation unit for interpolating the first offshore meteorological field data set to the target simulation domain using a preset interpolation method to obtain a second offshore meteorological field data set.

[0101] In some alternative embodiments, the generating unit: a first generation subunit for generating an initial simulation domain by utilizing a global coupled forecast system to drive a weather forecast mode; a first acquisition subunit for acquiring an area length and an area width of an initial simulation region; a division subunit for dividing the length of the area into a plurality of longitude segments and dividing the width of the area into a plurality of latitude segments; a determining subunit for determining a grid point longitude and latitude for each grid point based on the plurality of longitude segments and the plurality of latitude segments, such that the grid point is composed of a longitude segment and a latitude segment; and a second generating subunit for generating a target simulation area based on a weather forecast mode, based on the initial simulation area and the grid point longitude and latitude of each grid point.

[0102] In some alternative embodiments, the establishing module 504: a first processing unit for processing the first wind tower wind speed data set and the second wind tower wind speed data set by a data processing method to obtain a first target wind tower wind speed data set and a second target wind tower wind speed data set; a second processing unit for generating a target feature model based on the first target wind tower wind speed data set and the second target wind tower wind speed data set through feature engineering processing and cross-validation method processing; an establishment unit for establishing an initial offshore wind resource forecasting model using a distributed gradient boosting library based on the second target wind tower wind speed dataset and the target feature model; and an adjusting unit for adjusting the initial offshore wind resource forecast model using the first wind tower wind speed data set until a target offshore wind resource forecast model is obtained.

[0103] In some alternative embodiments, the establishing unit: a second acquisition subunit for acquiring a model feature map of the target feature model; an input subunit for inputting the model feature map into a distributed gradient boosting library to obtain a plurality of mark points corresponding to the model feature map; and an establishment subunit for obtaining a construction tree based on the plurality of mark points and establishing an initial offshore wind resource forecasting model based on the construction tree.

[0104] In some alternative embodiments, the offshore wind resource prediction device comprises: a computing module for calculating the mean absolute error, the mean squared error, and the root mean squared error of the target offshore wind resource forecast model; and an evaluation module for evaluating the target offshore wind resource forecasting model based on the mean absolute error, the mean square error, and the root mean square error to obtain an evaluation result.

[0105] In some alternative embodiments, the conversion module 502: a division unit for dividing the target simulation domain into a plurality of sub-domains; The sub-region further includes a setting unit for setting the sub-region to a nested region when the sub-region performs computing resource configuration.

[0106] The further functional description of each of the above modules and units is the same as that of the corresponding embodiment, and the repeated description will be omitted here.

[0107] The offshore wind resource prediction device in this embodiment is presented in the form of a functional unit, where unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory running one or more software or fixed programs, and / or other device capable of providing the above functionality.

[0108] An embodiment of the present application further provides a computer device including the offshore wind resource prediction apparatus shown in FIG. 5 above.

[0109] Referring to FIG. 6, FIG. 6 is a structural diagram of a computer device according to an alternative embodiment of the present application. As shown in FIG. 6, the computer device includes one or more processors 10, memory 20, and interfaces for connecting each component, including high-speed and low-speed interfaces. Each component is communicatively connected to each other via different buses and may be mounted on a common motherboard or in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in memory or memory for displaying GUI graphic information on an external input / output device (e.g., a display device coupled to the interface). In some alternative embodiments, multiple processors and / or multiple buses may be used along with multiple memories as needed. Similarly, multiple computer devices may be connected, each performing a portion of the required operations (e.g., functioning as a server array, a set of blade servers, or a multiprocessor system). FIG. 6 uses one processor 10 as an example.

[0110] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable logic gate array, a generic array logic, or any combination thereof.

[0111] However, the memory 20 stores instructions executable by at least one processor 10, causing the at least one processor 10 to execute and realize the methods shown in the above embodiments.

[0112] Memory 20 may include a program storage area for storing an operating system and application programs necessary for at least one function, and a data storage area for storing data generated in response to use of the computer device. Memory 20 may also include high-speed random access memory and may further include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, memory 20 may optionally include memory located remotely from processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0113] Memory 20 may include volatile memory, such as random access memory, or may include non-volatile memory, such as flash memory, a hard disk, or a solid state drive, or memory 20 may include a combination of the above types of memory.

[0114] The computing device further includes a communications interface 30 that allows the computing device to communicate with other devices or communications networks.

[0115] The present embodiment further provides a computer-readable storage medium, and the methods according to the embodiments of the present application may be implemented in hardware, firmware, recordable on a storage medium, or as computer code downloaded over a network, originally stored on a remote storage medium or a non-transitory machine-readable storage medium, but stored on a local storage medium, whereby the methods described herein may be processed by software stored on a storage medium using a general-purpose computer, a special-purpose processor, or programmable or special-purpose hardware. The storage medium may be a magnetic disk, optical disk, read-only memory, random-access memory, flash memory, hard disk, solid-state drive, etc., and optionally, the storage medium may include a combination of the above types of memory. As will be understood, a computer, processor, microprocessor controller, or programmable hardware may include a storage component capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods described in the above embodiments.

[0116] Although the embodiments of the present application have been described with reference to the drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. 1. An offshore wind resource forecasting method, comprising: obtaining a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecast system; utilizing the global coupled forecast system to operate a weather forecast mode and convert the first marine meteorological field data set into a second marine meteorological field data set based on the weather forecast mode; performing an interpolation process on the second offshore meteorological field data set to obtain a second wind tower wind speed data set; establishing a target offshore wind resource forecasting model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset using a distributed gradient boosting library; acquiring a third offshore meteorological field dataset and a third wind tower wind speed dataset for a target wind field area, inputting the third offshore meteorological field dataset and the third wind tower wind speed dataset into the target offshore wind resource prediction model, and obtaining a wind resource prediction result for the target wind field area.

2. the step of using the global coupled forecast system to operate a weather forecast mode and converting the first marine meteorological field dataset into a second marine meteorological field dataset based on the weather forecast mode, utilizing the global coupled forecast system to drive a weather forecast mode and generate a target simulation domain based on the weather forecast mode; and interpolating the first offshore meteorological field data set to the target simulation domain using a preset interpolation method to obtain the second offshore meteorological field data set.

3. The step of using the global coupled forecast system to drive a weather forecast mode and generating a target simulation domain based on the weather forecast mode includes: using the global coupled forecast system to drive a weather forecast mode and generate an initial simulation domain; obtaining an area length and an area width of the initial simulation domain; Dividing the length of the area into a plurality of longitude segments and the width of the area into a plurality of latitude segments; determining a grid point longitude and latitude for each grid point based on the plurality of longitude segments and the plurality of latitude segments, the grid point being comprised of a longitude segment and a latitude segment; and generating the target simulation area based on the weather forecast mode based on the initial simulation area and the grid point longitude and latitude of each of the grid points.

4. establishing a target offshore wind resource forecasting model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset using a distributed gradient boosting library, processing the first wind tower wind speed data set and the second wind tower wind speed data set using a data processing method to obtain a first target wind tower wind speed data set and a second target wind tower wind speed data set; generating a target feature model based on the first target wind tower wind speed data set and the second target wind tower wind speed data set through feature engineering and cross-validation methods; establishing an initial offshore wind resource forecasting model based on the second target wind tower wind speed dataset and the target feature model using the distributed gradient boosting library; and adjusting the initial offshore wind resource forecast model using the first wind tower wind speed data set until obtaining the target offshore wind resource forecast model.

5. establishing an initial offshore wind resource forecasting model using the distributed gradient boosting library based on the first target wind tower wind speed dataset and the target feature model, obtaining a model feature map of the target feature model; inputting the model feature map into the distributed gradient boosting library to obtain a plurality of mark points corresponding to the model feature map; and obtaining a construction tree based on the plurality of marked points; and establishing the initial offshore wind resource forecast model based on the construction tree.

6. calculating the mean absolute error, mean squared error, and root mean squared error of the target offshore wind resource forecast model; 2. The method of claim 1, further comprising: evaluating the target offshore wind resource forecast model based on the mean absolute error, the mean squared error, and the root mean squared error to obtain an evaluation result.

7. After the step of using the global coupled forecast system to drive a weather forecast mode and generating a target simulation area based on the weather forecast mode, dividing the target simulation domain into a plurality of sub-domains; The method of claim 2 , further comprising: setting a sub-region as a nested region when the sub-region performs computing resource configuration.

8. An offshore wind resource prediction device, a data acquisition module for acquiring a first offshore meteorological field dataset and a first wind tower wind speed dataset based on a global coupled forecast system; a transformation module for utilizing the global coupled forecast system to drive a weather forecast mode and transforming the first marine weather field data set into a second marine weather field data set based on the weather forecast mode; a processing module for performing interpolation processing on the second offshore meteorological field data set to obtain a second wind tower wind speed data set; an establishment module for establishing a target offshore wind resource forecasting model based on the first wind tower wind speed dataset and the second wind tower wind speed dataset using a distributed gradient boosting library; an acquisition and input module for acquiring a third offshore meteorological field dataset and a third wind tower wind speed dataset for a target offshore wind resource wind field, inputting the third offshore meteorological field dataset and the third wind tower wind speed dataset into the target offshore wind resource prediction model, and obtaining a wind resource prediction result for the target offshore wind resource wind field.

9. A computer device comprising: A computer device comprising: a memory and a processor communicatively connected to each other; computer instructions stored in the memory; and the processor executing the computer instructions to perform the offshore wind resource forecasting method according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon computer instructions for causing a computer to execute the offshore wind resource forecasting method according to any one of claims 1 to 7. A computer-readable storage medium.

Images