Cross-scale wind power plant wind resource evaluation method based on numerical mode

Abstract

The invention relates to the technical field of wind power generation, and discloses a cross-scale wind power plant wind resource evaluation method based on a numerical mode, comprising the following steps: step 1, acquiring multi-source observation data of a target area; step 2, screening a WRF optimal configuration scheme adaptive to the regional terrain; step 3, through a WRF power downscaling method, generating annual scale wind resource data as initial wind resource data; 4, performing partition correction on the initial wind resource data to obtain wind measurement data; step 5, constructing a WRF-WFP model coupling the WRF mode and the wind power plant parameterized model; step 6, optimizing WRF lower boundary conditions; 7, simulating the atmospheric flow of the target area, and quantifying the wake flow evolution characteristics of the wind power plant; 8, constructing a wake flow analysis model; and step 9, formulating a collaborative productivity plan of the wind power plant group. The power generation capacity of the wind power plant to be developed can be accurately predicted, and a reliable basis is provided for scientific decision-making of resource distribution development and planning of the million-kilowatt wind power plant.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power generation technology, and particularly relates to a cross-scale wind farm wind resource assessment method based on a numerical model. BACKGROUND

[0002] Wind resource assessment is one of the most important reference bases in the macro and micro site selection stages of a wind farm. Only by accurately assessing the wind resource can the power generation potential of the target region for building a wind farm be determined, the investment risk of building a wind farm be quantified, and the economic benefits of the wind farm be ensured. With the rapid development of the wind power industry, the development of conventional types of wind resource-rich areas tends to be saturated, which promotes the development of the wind power industry in regions such as sand deserts or the sea, which are wide and difficult to develop, and have low economic development utilization rates.

[0003] The characteristics of the wind are time-varying (daily variation, seasonal variation, intergenerational variation), and to assess the wind resource conditions of a candidate wind farm, long-term wind speed historical observation data generally need to be collected, so as to reduce the uncertainty caused by wind speed variation. In actual engineering, the general practice is to use the measure-correlate-predict (MCP) algorithm to perform wind resource assessment when there is a lack of long-term observation data of the wind farm. The basic idea of the MCP algorithm is to first set up a wind measurement tower at the candidate wind farm site to measure the wind speed, requiring at least one year of observation data, and the position of the wind measurement tower must be representative, which can represent the wind resource conditions of the entire wind farm. Since the one-year observation data of the wind measurement tower can only represent the short-term wind resource conditions of the wind farm, in order to obtain the long-term wind resource conditions of the wind farm, long-term observation data is obtained from long-term observation stations such as reference meteorological stations near the wind farm, and these data are used to correct the short-term wind measurement data of the wind measurement tower to a set of wind speed data that can reflect the long-term characteristics of the wind farm. However, in the sand desert region of Alashan League, due to the rapid development of wind power construction, and the large area and complex terrain of the sand desert region, it is often difficult to obtain representative observation data. In the traditional MCP method, the observation data of the first step cannot be obtained, and the process of wind resource assessment cannot be effectively performed. Therefore, a numerical simulation method is needed to obtain wind resource simulation data sets instead of observation data.

[0004] However, the application of current numerical simulation methods faces multiple technical challenges. At the physical mechanism level, first, real onshore wind farms are mostly located in complex mountainous or hilly terrains and show large-scale and diversified layout. Although some research has proposed wind farm parameterization models, their universality under different scales, atmospheric environment and geographical conditions still needs to be systematically verified. Second, existing research mainly focuses on the wake effect of a single wind turbine or an independent wind farm, and the mechanism of mutual interference between wind farm groups is seriously insufficient, which is a key factor for large-scale wind farm planning. With the continuous large-scale of wind turbine single capacity, wind wheel diameter and installed capacity, accurate and efficient evaluation of large-scale wind farm wake characteristics and power output has become a core challenge for industrial development, and is also a basic prerequisite for exploring the atmospheric feedback effect of wind farms.

[0005] On the other hand, the feedback effect of wind farm operation on the atmospheric environment is still in its infancy. Existing research mainly focuses on climate effect analysis in Europe and the United States, and there is a severe lack of research on wind farms in China. Key scientific issues include: (1) The influence of wind farms on meteorological elements (such as turbulent flux) has significant spatial heterogeneity and uncertainty, and the environmental effect strength and range of wind farms of different scales need to be quantified; (2) The potential correlation between wind farms and atmospheric pollutants (such as haze) has not been clearly defined, and typical cases such as the controversy over the causes of haze in the Beijing-Tianjin-Hebei region and wind farm construction need to be scientifically responded; (3) The mechanism of the indirect mesoscale climate effect caused by wind farms by changing the surface-atmosphere turbulent flux has not been clarified, and its indirect impact may be much greater than the direct local effect.

[0006] In summary, it is urgent to develop a cross-scale numerical evaluation system that integrates high-resolution terrain analysis, wind turbine group wake coupling and atmospheric environment feedback, to solve the three industrial needs of evaluation feasibility in data-free areas, scientificity of large-scale base planning and controllability of environmental effects. SUMMARY

[0007] The present application aims to provide a cross-scale wind resource evaluation method based on numerical model, which can accurately predict the power generation capacity of the wind farm to be developed, effectively reduce the risk of wind power development, and provide a reliable basis for scientific decision-making of large-scale wind farm resource planning.

[0008] The basic scheme provided by the present application is: a cross-scale wind resource evaluation method based on numerical model, comprising the following steps: Step 1, obtaining multi-source observation data of the target area; Step 2, based on the meteorological conditions and underlying surface characteristics of the target area, screening the optimal configuration scheme of WRF suitable for the regional terrain; Step 3, for the target area, combining digital elevation data and surface roughness data, generating time and space resolution adjustable annual wind resource data as initial wind resource data through WRF dynamic downscaling method; Step 4, partition correction is carried out on the initial wind resource data, and wind measurement data is obtained; Step 5, based on the wind measurement data, according to the wind farm planning information, combined with the regional terrain characteristics and regional distribution characteristics, the WRF-WFP model coupled with the WRF model and the wind farm parameterization model is constructed; Step 6, based on the terrain database of the target area, the lower boundary condition of WRF is optimized; Step 7, using the WRF optimal configuration scheme in step 2, cooperating with the WRF-WFP model constructed in step 5, and the WRF lower boundary condition in step 7, the atmospheric flow of the target area is simulated, and the wind farm wake evolution characteristics are quantified; Step 8, based on the wake evolution characteristics, the wind turbine level and wind farm level wake analysis model is constructed; Step 9, using the machine learning automatic optimization method, combining the wake analysis model to optimize the wind farm machine arrangement scheme, and analyzing the wake interference effect of the adjacent wind farm, the collaborative capacity planning of the wind farm group is made.

[0009] The working principle and advantages of the application are: The cross-scale wind resource assessment method of the wind farm based on the numerical model can accurately predict the power generation capacity of the to-be-developed wind farm, effectively reduce the risk of wind power development, and provide reliable basis for scientific decision-making of large-scale wind farm resource planning. The key points are: Firstly, steps 1-3 use multi-source observation data (satellite, reanalysis data, etc.) to drive WRF dynamic downscaling, directly solve the problem of traditional MCP failure caused by lack of observation data in Shagehuang area, generate time and space adjustable annual scale wind field data, and avoid the engineering bottleneck of long-term wind tower deployment. This process completely avoids the dependence on field wind tower, replaces physical observation with numerical simulation, and can provide basic input meeting the engineering precision requirements for data-free areas. Compared with the traditional MCP method, this scheme can not only cover the evaluation blind area of Shagehuang area, but also can significantly reduce the wind speed time-varying uncertainty caused by insufficient representativeness of short-term observation, and provide a universal solution path for scenarios such as offshore wind power which also face observation difficulties.

[0010] Secondly, to address the issues of insufficient accuracy in complex terrain and lack of wind farm parameterization verification, steps 5-7 construct a WRF-WFP coupled model. First, this approach incorporates wind farm planning information into the parameterized model, using this information (machine site coordinates, wind turbine parameters) to calculate the source term contribution of wind turbine thrust to the atmospheric momentum equation in real time, thereby specifically quantifying wind turbine-atmosphere interactions. This tight coupling approach overcomes the limitations of existing research that focuses solely on the wake of independent wind farms, enabling simulation of the dynamic interactions of wind farms of varying sizes (from single farms to wind farm clusters) during the evolution of the atmospheric boundary layer. Second, by optimizing the lower boundary conditions of the terrain database in step 6, this approach improves the accuracy of flow simulations in complex terrain and overcomes the local flow field distortion caused by terrain smoothing in traditional models. This helps to improve the simulation accuracy and realism of complex terrain.

[0011] Finally, steps 8-9 establish a multi-level wake model (single unit → wind farm → wind farm cluster) and introduce machine learning to optimize the layout. This can systematically quantify the wake interference effect between wind farm clusters under a unified framework, filling the key gap in large-scale base planning in background technology and providing decision-making recommendations for the large-scale development of the Shagohuang area. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a first schematic diagram of a method flow in Embodiment 1 of a cross-scale wind farm wind resource assessment method based on a numerical model according to the present invention; Figure 2 This is a second schematic diagram of a method flow in Embodiment 1 of a cross-scale wind farm wind resource assessment method based on a numerical model according to the present invention; Figure 3 This is a schematic diagram of the height layer setting of Example 1 of a cross-scale wind farm wind resource assessment method based on a numerical model of the present invention. DETAILED DESCRIPTION

[0013] The following is a further detailed description through specific implementation methods: Example 1 The embodiment is basically as shown in the attached Figure 1 and Figure 2 A wind resource assessment method for wind farms across scales based on numerical models is shown, comprising the following steps: Step 1: Obtain multi-source observation data of the target area.

[0014] The multi-source observation data includes data from wind towers, weather stations, lidar, and wind power towers. This data has high wind speed accuracy but limited spatial coverage. After acquiring the multi-source observation data, it is also sorted and cleaned according to geographic location, different altitudes, and data availability time periods. In the collation and cleaning process, outliers are removed, time alignment and data interpolation are performed, and missing data in the Shaguo barren area are reconstructed. The data interpolation includes interpolation based on wind shear curve for short-time missing values and interpolation based on Markov method for long-time missing values.

[0015] Optionally, a variety of reanalysis data (such as ERA5, ERA-Interim, CMFD, MERRA-2, JAR-55 and GLDAS) and LiGA database can also be synchronously collected to further improve the data perfection. This part of data has the characteristics of wide spatial coverage but relatively low precision.

[0016] In this embodiment, the selected target area is the Shaguo barren area (such as the Shaguo barren area of the Ordos large base), and FNL (i.e., global analysis data published by the U.S. National Environmental Forecasting Center) and ERA5 (i.e., the fifth generation global atmospheric reanalysis data published by the European Medium-term Weather Forecasting Center) are selected as driving data to drive the WRF model to simulate the Shaguo barren area. According to the comparison of the simulation results and the nearby meteorological observation station, the reanalysis data with higher precision is selected as the final driving data.

[0017] Step 2: Based on the meteorological conditions and the characteristics of the underlying surface of the target area, the WRF optimal configuration scheme suitable for the terrain of the region is screened.

[0018] The WRF model is composed of a radiation transmission scheme (long-wave radiation scheme and short-wave radiation scheme), cumulus parameterization scheme, microphysical model, land surface model, atmospheric boundary layer parameterization scheme and surface parameterization scheme. Different parameterization schemes have obvious differences in simulation effect for different regions and different weather processes.

[0019] In the screening of the WRF optimal configuration scheme suitable for the terrain of the region, a plurality of physical parameterization schemes are set, and then based on the reanalysis data, sensitivity analysis is performed on different physical parameterization schemes, and the physical parameterization scheme whose analysis index meets the preset standard is selected as the WRF optimal configuration scheme.

[0020] The plurality of physical parameterization schemes include four groups of physical parameterization schemes in combination, which are composed of PBL parameterization schemes MYJ, MYNN2.5, QNSE and YSU, and corresponding SL schemes ETA, MYNN, QNSE and MM5. As shown in Table 1.

[0021] The other physical parameterization settings of the four sets of physical parameterization schemes are: long-wave radiation selects RRTM scheme, short-wave radiation selects Dudhia scheme, microphysical process selects WSM6 scheme, land surface mode selects Noah scheme, and cumulus scheme selects Kain-Fritsch scheme, and the cumulus parameterization scheme is not used for the area with a resolution below 3km (because the vertical flux caused by the updraft and downdraft and the compensatory motion outside the cloud can be explicitly resolved in about 5-10km grid size).

[0022] Table 1 Parameterization scheme sensitivity test table

[0023] The analysis indicators include mean bias error MBE, root mean square error RMSE, relative root mean square error RRMSE and consistency IA; and the preset standard includes: , .

[0024] In this embodiment, sensitivity analysis of horizontal grid and vertical grid is also carried out synchronously: A nested domain is set for the target area - in this embodiment, three nested domains are set, with horizontal resolutions of 9km, 3km and 1km respectively, and the corresponding horizontal grid layer numbers are 180*180, 180*180 and 180*180 respectively, wherein the innermost 1km domain (dx03) covers the area centered on the simulated wind farm. The vertical resolution is defined as 10m (dz10) within the range of 300m near the ground, and then stretched vertically. Then the horizontal and vertical grid sensitivity tests are carried out by changing the horizontal resolution (dx) and the vertical resolution (dz) respectively. For example, the innermost high-resolution nested grid is set to 1km, 3km and 9km (using 3 times nested) respectively to test the sensitivity of the mode to the horizontal grid resolution; further, the vertical grid resolution near the ground is increased from 10m to 20m and 30m respectively to test the sensitivity of the mode to the vertical grid resolution. Further, according to the sensitivity analysis results, the most suitable grid resolution is selected; for example, by comparing the errors of the simulated key meteorological elements (such as wind speed, turbulent kinetic energy and wind direction frequency distribution) under different resolutions (9km / 3km / 1km), when the resolution is improved to 1km, the error improvement rate is <3%, and the next level resolution (3km) is selected as the optimal solution of the horizontal grid in terms of performance and cost.

[0025] By comparing the simulation accuracy of the boundary layer height and the wind speed vertical shear under the resolutions of 10m / 20m / 30m, if the R² (i.e. the fitting degree of the wind speed profile) decreases by >5% when the resolution is reduced to a certain level (such as 20m), and the R² of the next level (10m) is ≥0.95, then the next level (10m) is determined as the optimal solution of the vertical grid.

[0026] Step 3, for the target area, combined with digital elevation data and surface roughness data, the WRF dynamic downscaling method is used to generate wind resource data with adjustable spatial and temporal resolution at the annual scale as the initial wind resource data.

[0027] Specifically, in this embodiment, the WRF dynamic downscaling method used includes an ndown one-way nested downscaling method and a weather forecast model spatial downscaling method based on a terrain classification super-resolution model; the mesoscale wind resource data is processed by the above method to obtain high-precision simulated wind resource data with different spatial and temporal resolutions and different height layers.

[0028] The ndown one-way nested downscaling method has a running process of first running simulation of a coarse resolution grid, then inputting the simulation results as initial field and boundary conditions into a fine resolution grid, and then running simulation of the fine grid. This method can be applied to multi-nested grid downscaling, can meet the requirements of the finest downscaling in this embodiment, and has shorter running time and higher calculation efficiency. At the same time, in order to ensure the accuracy of the simulation data, the WRF-LES data of a part of the representative time period is used to verify the results of the ndown downscaling.

[0029] The weather forecast model spatial downscaling method based on the terrain classification super-resolution model includes the following sub-steps: S3.1, under m different terrains, n (n>1) layers of nested dynamic downscaling calculation are performed by a weather forecast model to obtain meteorological element samples under m terrains, wherein the regions represented by the 1st, 2nd, …, n layers of data gradually decrease, and the resolution gradually increases; S3.2, different meteorological element data samples of different resolutions in the n-layer region under the same terrain in S3.1 are extracted, two groups of data with different resolutions are selected, and they are divided into a training set, a validation set and a test set according to the time dimension; S3.3, the training set and the validation set in S3.2 are used to train the super-resolution model respectively, wherein the high-resolution data sample is used as the label of the super-resolution model, the low-resolution data is interpolated into data with the same resolution as the label as the input of the super-resolution model, and the test set is used for verification, and finally the super-resolution model under m different terrains is obtained; S3.4, terrain data under m terrains is obtained, a terrain classification model is obtained through m classification models, when the terrain classification model is used, the similarity probability between the predicted terrain output through the softmax layer and the m terrains is selected as the weight of the fusion multi-terrain prediction result; S3.5, using the super-resolution model obtained in S3.3 to respectively predict the meteorological data to be predicted under m different terrains; using the terrain classification model obtained in S3.4 to output the similarity probability between the terrain data of the region to be predicted and the m different terrains; and using the similarity probability between the m terrains to weight and sum the m prediction results of the super-resolution model to obtain the fusion prediction result of the m super-resolution models.

[0030] The fusion prediction result is the final prediction value of the high-resolution meteorological element (such as wind speed, temperature) of the target region.

[0031] Through the above method, the intelligence and high-efficiency replacement and enhancement of dynamic downscaling can be realized. First, the super-resolution model itself is trained using high-resolution results generated by nested dynamic downscaling (such as WRF multi-layer calculation) as "labels", so it learns and internalizes the physical laws (such as terrain-atmosphere interaction) captured by the downscaling process; second, in the application stage, this fusion model can quickly "super-resolution" the lower resolution meteorological input data (such as global model output) to the required high resolution (such as 1km) without the need to re-run the time-consuming complete nested dynamic downscaling calculation, thereby greatly improving the efficiency, while maintaining the physical reasonableness and regional adaptability of the downscaling through terrain adaptive fusion; significantly improving the resolution and accuracy of weather prediction in complex terrain areas (through terrain-specific models), greatly reducing the calculation cost and time (replacing repeated downscaling), and achieving the smoothness of terrain transition zone prediction (through probability fusion weights).

[0032] In particular, the super-resolution model not only learns resolution enhancement in the training process, but also implicitly captures and quantifies the physical influence of specific terrain (such as sand dunes, valleys) on local meteorological elements (such as turbulence, wind shear), which can even be inverted and reused to form a valuable "terrain-climate response" knowledge base, going beyond the scope of image super-resolution and providing a new approach to understanding regional climate. At the same time, the probability weight output by the terrain classification model can also serve as an effective indicator of the complexity and mixing degree of regional terrain.

[0033] Step 4: Subdivision correction of initial wind resource data to obtain wind measurement data.

[0034] The subdivision correction includes a first-level correction based on machine learning, including the following sub-steps: S4.1, determining the height layer wind speed that needs to be corrected according to the numerical simulation and statistical downscaling verification results; in this embodiment, the 10m wind speed and the 30m wind speed are selected; S4.2, obtaining the wind measurement data and simulation data corresponding to the height layer to be corrected, and extracting a 24h length time series from the original data with a step of 1h to generate a data set; S4.3, using random forest simulation training machine learning model, and applied to the site to be corrected, check the effect of correction.

[0035] Secondary correction based on vertical extrapolation of wind speed, including the following sub-steps: S4.4, taking the high-level wind speed as the reference point for vertical extrapolation, fitting to obtain the wind profile; In this embodiment, 100m height wind speed is selected; If the 100m height data is missing, the nearest wind data from 100m height can be selected for extrapolation; S4.5, taking the wind profile as the reference for correction, if the relative error of the simulation result is greater than 3%, set the correction coefficient as the wind profile value / simulation value; If the relative error of the simulation result is less than 3%, set the correction coefficient as 1; S4.6, multiply the correction coefficient by the time series data of the corresponding height of the corresponding station to realize the correction of the initial wind resource data.

[0036] Step 5, based on the wind measurement data, according to the wind farm planning information (such as the layout of wind turbines in the wind farm to be developed, the planned installed capacity, and the power curve and thrust curve of the wind turbine, etc.), combined with regional terrain characteristics and regional distribution characteristics, a WRF-WFP model coupling WRF model and wind farm parameterization model (WFD) is constructed.

[0037] The coupling WRF model and wind farm parameterization model include: for different wind turbine layouts and installed capacities, the wind turbine fitting power curve and fitting thrust curve are parameterized into the WRF model in the form of momentum sink.

[0038] The coupling WRF model and wind farm parameterization model further include: coupling the aerodynamic characteristics of the super flexible blade wind wheel.

[0039] Further, in this embodiment, the coupling of the WRF model and the wind farm parameterization model is realized based on Fortran code; In the coupling process, the dragforce subroutine of the Fortran code is called to realize the calculation of the interaction between the wind turbine and the atmosphere, including the following steps: Design a vertical resolution scheme for the wind turbine action layer: adjust the vertical coordinate system of the model iteratively to ensure that at least 25 vertical layers cover the wind turbine wind wheel sweeping height range to capture the blade-atmosphere momentum exchange process. Specifically, when setting the vertical layers, first set a group of near-surface layer encryption height layers, and then run in the WRF model, then export the height from the ground of different vertical layers (here, the altitude minus the terrain height is needed), compare the height from the ground with the height swept by the wind wheel, and adjust the eta_levels. Then, repeat the above process until there are enough vertical layers within the height swept by the wind wheel, such asFigure 3 eta_levels is a parameter used to define the vertical staggering of the grid; the WRF model defines the number of vertical grid levels by setting the eta_levels parameter (e.g., 35 levels, 50 levels, etc.), each level corresponding to a specific value of eta; eta is a dimensionless vertical coordinate, which usually ranges from 0 to 1.

[0040] The more the vertical layers through which the wind turbine impeller passes, the more accurate the analysis of the interaction between the wind turbine and the atmosphere, and the more accurate the calculated wind power and wake. In the embodiment, 65 height layers are provided, and 30 height layers below 300 m are provided to capture the corresponding wind parameter changes. 25 vertical layers are selected within the swept range of the wind wheel diameter, which can balance the accuracy and computational efficiency.

[0041] The wind speed at the hub height of the wind turbine is calculated; in the specific calculation, firstly, the upper and lower two height layers closest to the hub height of the wind turbine are found, and the wind speed is calculated respectively, and then the wind speed at the hub height is obtained by linear interpolation.

[0042] The dragcof subroutine of the Fortran code is called to calculate the turbulent kinetic energy coefficient, power coefficient and thrust coefficient of the wind turbine. Firstly, the program calculates the area swept by the turbine blades through the calculation of area. Then, according to the segmentation of the cut-in wind speed and the cut-out wind speed, the thrust coefficient is calculated respectively. When the wind speed is less than the cut-in wind speed or greater than the cut-out wind speed, the static thrust coefficient value is assigned respectively; when in the working wind speed section, the value of the thrust coefficient is calculated by interpolation. Then, for the power coefficient, the same segmentation method is adopted, and 0 is assigned to the non-working wind speed section, and interpolation calculation is performed in the working wind speed section combined with the input data. Finally, the turbulent kinetic energy coefficient is calculated by the difference between the thrust coefficient and the power coefficient.

[0043] The power generated by the wind turbine is calculated: The power generated by the wind turbine in a grid (i, j) is: ; wherein, is the standard air density ; is the power coefficient; is the hub height wind speed; is the wind wheel rotation area; is the number of wind turbines in a grid.

[0044] The influence of the wind turbine on the turbulent kinetic energy and horizontal momentum is calculated—specifically, based on the calculated turbulent kinetic energy coefficient, power coefficient and thrust coefficient of the wind turbine, the momentum and turbulent action term of the wind turbine in the WRF grid unit (i.e., the influence of the wind turbine on the turbulent kinetic energy and horizontal momentum) is calculated according to the following formula: The momentum balance equation is: ; If the momentum change in the vertical direction is ignored, then: ; where u, v are the horizontal velocity wind speed.

[0045] Similarly, the turbulent kinetic energy balance equation is: ; ; where, is the TKE coefficient, TKE is the turbulent kinetic energy, , is the power coefficient, is the thrust coefficient; that is, the part not converted into electrical energy is converted into turbulent kinetic energy.

[0046] (i, j) is the grid coordinate to which the wind turbine belongs; is the average wind speed of the grid (i, j); k is the number of vertical layers; z is the height of the vertical layer; is the area of the wind turbine intercepted by the kth and k+1th vertical layers.

[0047] At this point, the calculation of the wind turbine parameters is completed, and is coupled into the WRF model.

[0048] where, when calculating the influence of the wind turbine on the turbulent kinetic energy, the turbulent kinetic energy horizontal transmission mechanism is activated in the planetary boundary layer scheme, so that the wind farm wake turbulence can spread horizontally between grid cells, to configure the turbulent kinetic energy horizontal advection function.

[0049] The turbulent kinetic energy horizontal advection function is realized by the following means: Select the MYNN2 boundary layer scheme containing the TKE prognostic variable; Enable the TKE scalarization processing mechanism to convert the turbulent kinetic energy field into a scalar field that can be subjected to horizontal convection and diffusion; Correlate the surface parameterization scheme and the boundary layer scheme to ensure the conservation of turbulent kinetic energy.

[0050] Through the above settings, the turbulent kinetic energy (TKE) horizontal advection function can be activated in the program calculation, thereby significantly improving the simulation accuracy of the wind farm wake turbulence. Compared with the traditional MYNN2 boundary layer scheme, the traditional scheme predicts the evolution of TKE in the vertical direction of each node by a one-dimensional TKE equation, which only relates to the vertical coordinate. That is, there is no further design of the TKE transport scheme (i.e., horizontal convection, vertical convection, and diffusion) in the traditional MYNN2 scheme, so there is no 1TKE horizontal advection from one horizontal node to another, resulting in an underestimated turbulent intensity in the wake area and not conforming to the real wind farm wake characteristics.

[0051] In this scheme, the horizontal transmission mechanism (corresponding to the TKE scalarization processing mechanism) is introduced to make the turbulent energy diffuse horizontally between grids, accurately reproduce the wake sector expansion effect, and improve the simulation accuracy and reality. Secondly, the MYNN2 boundary layer scheme (containing TKE prognostic variable) is adopted and associated with the surface parameterization scheme, which can ensure the strict conservation of turbulent energy in the horizontal and vertical directions, avoid numerical dissipation, and reduce the prediction error of turbulent sudden change (such as the enhancement of turbulent flow on the leeward slope of sand dunes) under complex terrain.

[0052] Step 6, optimizing the lower boundary condition of WRF based on the terrain database of the target area.

[0053] Specifically, in this embodiment, when optimizing the lower boundary condition of WRF, a public high-precision satellite terrain database is used, such as STER, MODIS, SRTM, etc. Through the WRF preprocessing tool WPS (WRF Preprocessing System), the geogrid program of WPS filters out the noise of satellite data and retains the true terrain fluctuations. Then, the filtered terrain field is coupled with the reanalysis data (such as ERA5) to generate a thermodynamic-dynamic coordinated lower boundary, thereby constructing the fine lower boundary terrain condition of WRF.

[0054] Preferably, the WRF-WFP model in step 5 is used to draw a high-precision wind farm site wind resource map considering the wake characteristics under different atmospheric stability and wind speed and direction by considering the changes in atmospheric flow caused by wind turbine operation, for intuitive analysis of wind farm evolution.

[0055] The wake characteristics include turbulent intensity, wake recovery coefficient, wake influence range, etc. In this embodiment, when drawing the wind resource map, existing drawing software can be called to convert physical quantities into a map.

[0056] Further, when optimizing the lower boundary condition of WRF, the subgrid terrain drag force parameterization scheme is integrated synchronously, including: The terrain-induced turbulent deformation drag force is parameterized as an explicit terrain stress term, and the turbulent deformation drag force is coupled to the atmospheric motion equation as a stress profile to correct the momentum loss caused by terrain flow; Specifically, based on the terrain gradient tensor Calculate the turbulent deformation drag force: ; Where, is the turbulent deformation drag force; is the terrain drag coefficient, which is classified and calibrated according to the surface roughness; is the air density; is the horizontal wind speed; The turbulent deformation drag force term is directly embedded in the momentum equation source term of WRF, explicitly representing the terrain resistance to airflow.

[0057] Through the terrain stress-wind speed feedback mechanism, the systematic deviation of the near-surface wind resource simulation is reduced.

[0058] Specifically, the stress transfer function is constructed in the MYNN2 boundary layer scheme: ; wherein, is the terrain stress turbulent diffusion coefficient, which is vertically diffused to the full layer of the boundary layer (up to 2 km) through nonlinear iteration, correcting the momentum loss caused by terrain flow.

[0059] Further, the terrain stress-wind speed feedback mechanism is introduced, including: Introducing a dynamic terrain drag coefficient : .

[0060] wherein, is the feedback gain factor; and are the observed wind speed and the simulated wind speed, respectively. In this embodiment, the measured data is assimilated every 6 hours, thereby reducing the systematic deviation of the near-surface wind resource simulation in real time.

[0061] Through the above settings, the terrain stress term can be explicitly quantified and vertically coupled to the atmospheric motion equation, directly correcting the momentum loss caused by terrain flow, and breaking through the limitations of the traditional WRF model in which the terrain drag force is not parameterized. In particular, the scheme establishes a dynamic feedback mechanism of terrain stress-wind speed, dynamically calibrates the drag coefficient using real-time observation data, which can systematically reduce the simulation deviation of near-surface wind speed in complex terrain areas (such as steep slopes and valleys), significantly improve the reliability of desert / mountain wind farm power generation prediction, and provide more reliable wind shear input for super-soft blade anti-turbulence design.

[0062] Secondly, the vertical diffusion of the terrain stress term can stimulate the breaking effect of gravity waves, significantly enhancing the turbulent mixing intensity at the top of the boundary layer (300-500m height) at night, which is completely missing in the traditional parameterization scheme; and the setting of the stress-wind speed feedback mechanism helps to capture the periodic oscillation characteristics of the terrain wake vortex, which can provide a new frequency domain analysis dimension for the turbulent fatigue load evaluation of wind farms.

[0063] Step 7, using the WRF optimal configuration scheme in step 2, combining the WRF-WFP model built in step 5, and the WRF lower boundary condition in step 7, simulating the atmospheric flow in the target area under different wind turbine layouts and different wind turbine specifications, and then obtaining the wind flow characteristics of different sectors of the wind turbine, such as wind speed attenuation distance and attenuation size, and quantifying the wind farm wake evolution characteristics.

[0064] The wind farm wake evolution characteristics include wake velocity loss rate, turbulence enhancement index, wake deflection angle, wake recovery distance, etc.

[0065] Step 8, based on the wake evolution characteristics, build a wind turbine level and wind farm level wake analysis model.

[0066] The wake analysis model includes the wind turbine Jensen wake model and the Gaussian wake model.

[0067] Specifically, in this embodiment, based on the atmospheric flow field simulated in step 7 and the wind farm wake evolution characteristics extracted therefrom, the downstream main wind direction profile of the target wind turbine is extracted, and the wind speed evolution law and the relationship between wind speed and distance are obtained, and then the wind turbine Jensen wake model and the Gaussian wake model for the target area are fitted.

[0068] Step 9, using machine learning automatic optimization method, combining the wake analysis model to optimize the wind farm layout scheme, and analyzing the wake interference effect of the adjacent wind farm, and making the coordinated capacity planning of the wind farm group.

[0069] Specifically, in this embodiment, through the wake analysis model, the best wind turbine layout scheme is automatically optimized by using the machine learning method based on genetic algorithm. The wake influence area of the wind farm is determined, the influence of the wake influence area on the power output of the adjacent wind farm is analyzed, and an engineering scheme is given on how to further plan the wind energy output of the wind farm to make the overall wind power system have higher power generation efficiency.

[0070] The cross-scale wind farm wind resource assessment method based on numerical mode provided in this embodiment can accurately predict the power generation capacity of the wind farm to be developed, effectively reduce the risk of wind power development, and provide a reliable basis for scientific decision-making of large-scale wind farm resource planning.

[0071] Embodiment Two A cross-scale wind farm wind resource assessment method based on numerical mode, based on embodiment one, makes the following adjustments.

[0072] In step 3, after generating the initial wind resource, the following steps are also performed: Step 3-a, collect the measured wind speed data (wind tower / SCADA data) of the target area in the historical period (e.g., the last 10 years), and calculate the system deviation (e.g., wind speed distribution, turbulence intensity) between the simulation value and the measured value.

[0073] In this embodiment, the measured wind speed data specifically includes: wind tower data (e.g., wind speed, wind direction, turbulence intensity at different height layers); wind turbine SCADA operation data (e.g., hub height wind speed, power curve, etc.).

[0074] Step 3-b, based on the deviation distribution, inverse the sensitivity parameters (e.g., mixing length in the boundary layer scheme, surface roughness scaling factor) of the WRF physical parameterization scheme, and update the parameter set using the ensemble Kalman filter method.

[0075] In step 7, an iterative verification step is also included, including: Use the updated WRF-WFP model to re-simulate the historical wind field, and compare the new simulation value with the measured value again. If the error is higher than the threshold, return to step 3-b to adjust the parameters again until convergence.

[0076] Specifically, for an existing wind farm: call the historical database of the wind farm (in this embodiment, the actual operation data of the last 3 years is selected), and extract the following key parameters as verification benchmarks: average wind speed (based on wind tower retrieval), wind speed profile, turbulence intensity, power output (actual power generation curve of the wind turbine), etc.

[0077] Take the measured value of the key parameters as the benchmark, calculate the deviation between the measured value and the new simulation value, and if the deviation is higher than the threshold, return to step 3-b to adjust the parameters again until convergence.

[0078] In this embodiment, the deviation higher than the threshold includes: wind speed RMSE>1.5m / s, power MAE>6%.

[0079] The convergence condition of the convergence is: wind speed RMSE≤0.5m / s and power MAE≤3%.

[0080] For a wind farm to be planned: call the historical wind measurement database of the target area, and extract the following key parameters as verification benchmarks: average wind speed, wind speed profile, turbulence intensity, etc.

[0081] Take the key parameters as the measured value, calculate the deviation between the measured value and the simulation value, and if the deviation is higher than the threshold, return to step 3-b to adjust the parameters again until convergence. The deviation higher than the threshold includes: wind speed RMSE>1.5m / s, and the convergence condition of the convergence is: wind speed RMSE≤0.5m / s.

[0082] The cross-scale wind farm wind resource assessment method based on the numerical mode provided in the embodiment can form a closed loop of "initial simulation → measured comparison → parameter correction → secondary simulation", and through the simulation correction of the closed loop iteration, the accuracy of the resource assessment can be effectively improved.

[0083] Embodiment three The cross-scale wind farm wind resource assessment method based on the numerical mode makes the following adjustments on the basis of embodiment one.

[0084] In step 1, the geographic information and multi-source observation data around the target area are synchronously collected. In the embodiment, the information within 100 km around the target area is collected.

[0085] The geographic information around the target area is traversed, and if there is a built wind farm therein, the basic information (including the coordinate position of the wind farm, the fan model, the hub height, etc.), the historical wind measurement database and the historical operation database of the built wind farm are further collected; if there is no built wind farm therein, no processing is performed.

[0086] In order to reduce the calculation amount, the data within a specified period can be collected for subsequent verification analysis.

[0087] In step 7, the edge verification step is further included, which includes: (a) Extracting simulation data: from the WRF-WFP simulation field output in step 7, extracting the time series wind speed at the coordinate of the wind measurement tower in the target area; extracting the hub height simulation wind speed at the GPS coordinate of the surrounding wind farm, and when there is no built wind farm around, a plurality of surrounding coordinate points are randomly extracted, and the simulation wind speed at different heights is extracted.

[0088] (b) Calculate the fitting degree: For the target area, the correlation coefficient R² of the simulation value and the measured value, and the absolute error |Δk| of the Weibull shape parameter k are calculated.

[0089] For the surrounding wind farm, the relative error RE of the simulation value and the measured value is calculated and is counted according to the wind direction sector.

[0090] For the surrounding area where there is no built wind farm, the correlation coefficient R² of the simulation value and the measured value is calculated and is counted according to the coordinate point.

[0091] (c) Threshold judgment: If the R² of the target area is less than 0.99 or the |Δk| is greater than 0.1, the terrain parameter correction is triggered.

[0092] If the RE of any sector of the surrounding wind farm is greater than 5%, the wake parameter correction is triggered.

[0093] If R2 of the surrounding area is less than 0.95, the wind field initialization correction is triggered.

[0094] Specifically, in the terrain parameter correction, step 6 is returned, and the terrain parameters of the target area are corrected. In this embodiment, the corrected terrain parameter is the terrain dynamic roughness length (z0). The larger the z0 value, the rougher the ground surface (such as mountains, forests, and urban building groups), the stronger the hindering effect of the wind, and the more obvious the near-surface wind speed decay; on the contrary, the smaller the z0 value (such as flat grassland and sea surface), the smaller the influence of the ground surface on the wind, and the more uniform the wind speed profile.

[0095] In the correction, the target area is divided into 1km×1km grids, the measured-simulated wind speed deviation of each grid is calculated, and a regression model of the deviation and the terrain parameters (slope and curvature) is established; the z0 is corrected in reverse based on the regression coefficient, and the lower boundary condition of the terrain database is updated.

[0096] In the wake parameter correction, the wake attenuation coefficient (corresponding to the Jensen model) and the turbulence intensity correction factor (corresponding to the Fitch source term) in the WRF-WFP model are corrected.

[0097] For the wind direction sector that exceeds the RE limit, the measured wake loss value at the dominant wind direction downstream of the sector is extracted, and the relative error between the simulated wake loss value and the simulated wake loss value is calculated, and the error proportion is adjusted. That is: ; , is the turbulence intensity deviation value.

[0098] Further, the parameters of the WRF-WFP model are updated.

[0099] In the wind field initialization, the correction object is the WRF initial boundary field (such as ERA5 reanalysis data). Specifically, the deviation value of the simulation value and the measured value in the wind direction frequency distribution is calculated, and a multi-dimensional wind direction frequency difference vector is formed, which is then converted into a tendency term in the Nudging algorithm to relax the adjustment of the initial wind field (such as converting into the adjustment requirement of the wind field component; for example, if the measured frequency of a wind direction sector is higher than the simulation value, the wind field strength in that direction needs to be enhanced).

[0100] The embodiment provides a cross-scale wind farm wind resource evaluation method based on a numerical mode, compared with the first embodiment, the surrounding area of the target area is additionally introduced for cooperative simulation analysis, so that the simulation data obtained by the model of the scheme not only reaches more than 99% fitting with the wind field operation data of the target area, but also reaches more than 95% fitting with the data of the surrounding wind field, has extremely high simulation precision, and can meet the engineering landing demand.

[0101] Embodiment four A cross-scale wind farm wind resource evaluation method based on a numerical mode, on the basis of the first embodiment, the following adjustments are made.

[0102] If the target area contains a built wind farm, the following adjustment steps are introduced.

[0103] The following step is inserted before step 5: constructing a "no wind farm" underlying surface scene, running a control test based on the same WRF configuration to generate background wind resource data not affected by wind power development.

[0104] In step 7, simulation under two kinds of conditions is performed, including: Scenario A: Load the actual wind farm layout (including built / new sites) and run the WRF-WFP model.

[0105] Load wind farm parameters, including: wind turbine geometric parameters - collect detailed parameters of built and new wind turbines, including hub height, blade radius, tower height and diameter, etc., to ensure accurate parameters. Wind turbine aerodynamic parameters - obtain thrust coefficient curve, power curve and other data of each wind turbine, which can be obtained through technical manual or measured data provided by wind turbine manufacturers. For new wind turbines, if there is a lack of measured data, the typical parameters of the same type of wind turbine can be referred to. Layout parameters - determine the accurate coordinates (longitude and latitude or projection coordinates) of built and new wind turbines, as well as the arrangement angle. And run the WRF-WFP model.

[0106] Scenario B: Load only the built wind farm layout and run the WRF-WFP model.

[0107] Select the relevant data of the built wind farm from the wind farm parameters of scenario A, including wind turbine parameters and layout information, and remove the new wind farm part. Use the same WRF-WFP model configuration and running settings as scenario A to ensure that the simulation conditions of the two scenarios are consistent, and only the wind farm layout is different.

[0108] Calculate the wind speed difference between scenario A and the control test (i.e. background wind resource data) to quantify the superimposed effect of the new wind farm on the target area. This difference can quantify the additional wind speed attenuation caused by the new wind farm based on the built farm, helping to evaluate the incremental influence of the new planning project on the regional wind resource: .

[0109] Calculate the difference between Scenario B and the control test (i.e. the background wind resource data) to quantify the existing impact of the built wind farm: .

[0110] And the cross-influence between the two: . The cross-influence can be used to capture the synergistic effect of the wind farm cluster. When , it indicates that the wake of the built and new wind farms superimpose on each other, resulting in more severe wind speed decay; when , there may be partial offsetting effect of the wake interference.

[0111] is the wind speed containing both the built and new wind farm scenarios (Scenario A). is the wind speed containing only the built wind farm scenario (Scenario B). is the wind speed of the background wind resource data (i.e. the wind speed without wind farms).

[0112] Optionally, based on and , a distribution map can be drawn, with icons added at the location of the new wind farm site, and key areas (such as grid points with more than 5% decay) marked; and an contour map can be drawn, with annotations added in areas with dense contours (indicating intense cross-influence), which helps to analyze the causes of synergistic effect (such as wake superposition direction, terrain acceleration / deceleration effect) in combination with wind farm layout.

[0113] Through the above analysis, the incremental impact of the new wind farm can be confirmed, and the complex impact of the wind farm cluster can be accurately quantified, which helps to optimize the layout scheme of the new wind farm.

[0114] In step 9, in the machine learning automatic optimization, a wind resource change threshold constraint is also set, such as wind speed decay rate ≤ 5%, to ensure that the new planned wind farm cluster does not significantly worsen the operating environment of the built wind farm.

[0115] In this embodiment, a machine learning method based on genetic algorithm is used for automatic optimization, and the total power generation of the wind farm cluster is taken as the optimization objective, and a wind speed decay rate penalty term is set.

[0116] The cross-scale wind resource assessment method for wind farms based on numerical mode provided in this embodiment is suitable for the evaluation scenario of wind farm clusters. This scheme particularly decouples the influence of wind farms into independent influence of built farms, superimposed influence of new farms, and cross-influence, which can accurately quantify each part of the effect and make up for the defects of the traditional evaluation method in considering the synergistic effect of wind farm clusters.

[0117] Embodiment Five A cross-scale wind farm wind resource assessment method based on a numerical model, on the basis of embodiment one, the following adjustments are made.

[0118] The coupling of the WRF model and the wind farm parameterization model further comprises: coupling the aerodynamic characteristics of the super-flexible blade wind wheel.

[0119] In this embodiment, a super-flexible blade solver is constructed.

[0120] Specifically, the super-flexible blade solver is provided with an aerodynamic-structural coupling equation: ; Wherein, EI is the bending stiffness, y is the blade transverse displacement, is the relative wind speed, is the flexibility index (blade tip displacement / impeller diameter), is the real-time aerodynamic twist angle; is the dynamic lift coefficient; is the aerodynamic force.

[0121] x refers to the coordinate along the blade span (the length direction from the blade root to the blade tip, similar to the one-dimensional coordinate "from the handle to the tip"), which is used to describe the mechanical response of different span positions of the blade (such as the blade root x=0, the blade tip x=R, R is the half length of the blade). m is the mass per unit span length of the blade.

[0122] is the initial angle of attack of the blade; is the effective angle of attack of the blade; is the air density; and are the first-order time derivative of the blade transverse displacement (i.e. transverse vibration velocity) and the second-order time derivative of the blade transverse displacement (i.e. transverse acceleration), respectively.

[0123] In specific applications, the grid point wind speed is transmitted to the super-flexible blade solver. Based on the grid point wind speed, the aerodynamic-structural coupling equation is solved to obtain the aerodynamic force . represents the wind speed value at the (x, y) horizontal position, z height layer, and t time.

[0124] is weighted and averaged according to the impeller swept area , and is injected into the WRF momentum equation to update the atmospheric flow field.

[0125] And dynamic roughness correction is carried out to calculate the equivalent roughness to update the WRF underlying surface. Further, the aerodynamic characteristics of the super-flexible blade wind wheel are coupled.

[0126] Wherein, ; Z0=Z0+Z1 Z1=Z1+Z2 Z2=Z2+Z3

[0127] The method provided by the embodiment one is a cross-scale wind resource assessment method based on a numerical model. Compared with the embodiment one, the method can more truly reflect the flow field in the wind farm and achieve more accurate wind resource assessment.

[0128] Specifically, the super-flexible blade will produce a large elastic deformation under the action of airflow, and its actual aerodynamic shape will change constantly. Coupling this feature into the model can more accurately simulate the airflow flow in the wind farm. For example, the traditional rigid blade model assumes that the blade shape is fixed, while the super-flexible blade may be significantly bent under strong wind, causing the direction and speed distribution of the incoming flow to change. After considering this feature, the model can more accurately predict the wind speed, wind direction distribution in the wind farm, including the range and intensity of the wake effect, so that the simulation results are closer to the actual operation of the wind farm.

[0129] The above is only an embodiment of the present application, and the common knowledge of specific structures and characteristics in the scheme is not described in detail. The person skilled in the art knows all the common technical knowledge in the field of the present application before the filing date or the priority date, can know all the prior art in this field, and has the ability to apply conventional experimental means before that date. The person skilled in the art can improve and implement the present scheme based on the disclosure given in this application, and some typical known structures or known methods should not be an obstacle for the person skilled in the art to implement the present application. It should be noted that for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should be considered as the protection scope of the present application. These will not affect the effect and practicality of the present application.

Claims

1. A cross-scale wind farm wind resource assessment method based on numerical model, characterized by: The following steps are involved: Step 1: Obtain multi-source observation data of the target area; Step 2: Based on the meteorological conditions and underlying surface characteristics of the target area, select the optimal WRF configuration scheme that is suitable for the regional terrain; Step 3: For the target area, digital elevation data and surface roughness data are combined to generate annual wind resource data with adjustable temporal and spatial resolution using the WRF dynamic downscaling method as the initial wind resource data. Step 4: Perform partition correction on the initial wind resource data to obtain wind measurement data; Step 5: Based on the wind measurement data, according to the wind farm planning information, and combined with the regional terrain characteristics and regional distribution characteristics, a WRF-WFP model is constructed that couples the WRF model with the wind farm parameterization model; Step 6: Optimize the WRF lower boundary condition based on the terrain database of the target area; In step 7, the WRF optimal configuration scheme in step 2 is used in conjunction with the WRF-WFP model constructed in step 5 and the WRF lower boundary condition in step 7 to simulate the atmospheric flow in the target area and quantify the evolution characteristics of the wind farm wake. Step 8: Constructing wind turbine-level and wind farm-level wake analysis models based on the wake evolution characteristics; Step 9: Use machine learning automatic optimization methods combined with wake analysis models to optimize the wind farm layout plan, analyze the wake interference effect of adjacent wind farms, and formulate a coordinated production capacity plan for the wind farm group.

2. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 1, the multi-source observation data includes wind tower data, weather station data, lidar data, and wind power tower data; after obtaining the multi-source observation data, the multi-source observation data is also sorted and cleaned according to geographical location, different altitude layers, and data available time period; The sorting and cleaning process includes outlier removal, time alignment and data interpolation, and reconstruction of missing data in the Shagohuang area; the data interpolation includes interpolation based on wind shear curves for short-term missing values ​​and interpolation based on Markov methods for long-term missing values.

3. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 2, when selecting the optimal WRF configuration scheme that is suitable for the regional terrain, multiple physical parameterization schemes are first set, and then sensitivity analysis is performed on different physical parameterization schemes. The physical parameterization scheme whose analysis indicators meet the preset standards is selected as the optimal WRF model configuration scheme; The multiple physical parameterization schemes include: four groups of combined physical parameterization schemes consisting of PBL parameterization schemes - MYJ, MYNN2.5, QNSE, YSU, and their corresponding SL schemes - ETA, MYNN, QNSE, MM5; The analysis indicators include mean deviation MBE, root mean square error RMSE, relative root mean square error RRMSE and consistency IA; the preset standards include: 、 .

4. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 4, the partition correction includes: a first-level correction based on machine learning, including the following sub-steps: S4.

1. Determine the wind speed at the altitude that needs to be corrected based on the results of numerical simulation and statistical downscaling verification; S4.2, obtain the wind measurement data and simulation data corresponding to the altitude layer to be corrected, and extract a 24-hour time series from the original data with a step length of 1 hour to generate a data set; S4.3, use random forest simulation to train a machine learning model, apply it to the site to be corrected, and check the correction effect; The second-level correction based on vertical extrapolation of wind speed includes the following sub-steps: S4.4, using the upper-level wind speed as the reference point for vertical extrapolation, the wind profile is obtained by fitting; S4.5, using this wind profile as the basis for correction, if the relative error of the simulation result is greater than 3%, set the correction factor to be the wind profile value divided by the simulation value; if the relative error of the simulation result is less than 3%, set the correction factor to 1; S4.6, multiply the correction coefficient by the time series data of the corresponding height of the corresponding station to correct the initial wind resource data.

5. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 4, characterized in that: The WRF dynamic downscaling method includes: The spatial downscaling method of weather forecast model based on terrain classification super-resolution model includes the following sub-steps: S3.1, perform n-layer nested dynamic downscaling calculations using a weather forecast model under m different terrains to obtain meteorological element samples under m terrains, where the areas represented by the 1st, 2nd, …, nth layers of data gradually decrease while the resolution gradually increases; S3.2, extract data samples of different meteorological elements at different resolutions for the nth layer of the same terrain in S3.1, select two sets of data at different resolutions, and divide them into a training set, a validation set, and a test set according to the time dimension; S3.3: Use the different meteorological element data in the training and validation sets in S3.2 to train the super-resolution model. The high-resolution data samples are used as labels for the super-resolution model. The low-resolution data is interpolated to have the same resolution as the labels as the input of the super-resolution model. The model is then tested on the test set. Finally, super-resolution models for m different terrains are obtained. S3.4, obtaining terrain data under m types of terrain, passing it through m classification models to obtain a terrain classification model. When using this terrain classification model, the similarity probability between the terrain to be predicted and the m types of terrain is selected as the weight for fusing the multi-terrain prediction results. S3.5, using the super-resolution models obtained in S3.3 to predict the meteorological data under m different terrains separately; using the terrain classification model obtained in S3.4 to output the similarity probability between the area to be predicted and the m different terrains; using the similarity probabilities between the m terrains to perform weighted summation on the m prediction results of the super-resolution model, to obtain the fusion prediction result of the m super-resolution models.

6. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 5, coupling the WRF model with the wind farm parameterized model includes: parameterizing the wind turbines into the WRF model in the form of momentum sinks using wind turbine fitting power curves and fitting thrust curves for different wind turbine layouts and installed capacities.

7. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 6, characterized in that: In step 5, coupling the WRF model with the wind farm parameterized model further includes: coupling the aerodynamic characteristics of the ultra-flexible blade wind wheel.

8. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 5, the coupling of the WRF model and the wind farm parameterized model is implemented based on Fortran code; During the coupling process, the dragforce subroutine of the Fortran code is called to calculate the interaction between the wind turbine and the atmosphere, which includes the following steps: Design a vertical resolution scheme for the wind turbine action layer: by iteratively adjusting the model's vertical coordinate system, ensure that at least 25 vertical layers cover the wind turbine rotor sweep height range to capture the rotor-atmosphere momentum exchange process; Calculate wind speed at wind turbine hub height; Call the dragcof subroutine to calculate the turbulent kinetic energy coefficient, power coefficient and thrust coefficient of the wind turbine; Calculate the power generated by the wind turbine; then calculate the effect of the wind turbine on the turbulent kinetic energy and horizontal momentum. This completes the calculation of the wind turbine parameters and couples them to the WRF model. Among them, when calculating the impact of wind turbines on turbulent kinetic energy, the horizontal transmission mechanism of turbulent kinetic energy is activated in the planetary boundary layer scheme, so that the wind farm wake turbulence can diffuse laterally between grid cells to configure the horizontal advection function of turbulent kinetic energy.

9. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 8, characterized in that: The turbulent kinetic energy horizontal advection function is achieved by: The MYNN2 boundary layer scheme including the TKE prognostic variable was selected; Enable the TKE scalar processing mechanism to convert the turbulent kinetic energy field into a scalar field that can undergo horizontal convection and diffusion; The surface parameterization scheme is linked to the boundary layer scheme to ensure conservation of turbulent energy.

10. The method for wind resource assessment of a cross-scale wind farm based on a numerical model according to claim 1, characterized in that: In step 6, when optimizing the WRF lower boundary conditions, the subgrid terrain drag parameterization scheme is integrated simultaneously, including: The turbulent deformation drag force caused by terrain is transformed into an explicit terrain stress term; the turbulent deformation drag force is coupled to the atmospheric motion equations as a stress profile to correct the momentum loss caused by the flow around the terrain; The systematic deviation of near-surface wind resource simulation is reduced through the terrain stress-wind speed feedback mechanism.

Images