A wind power generation amount prediction method, device, equipment and storage medium

Abstract

The application discloses a wind power generation amount prediction method, device and equipment and a storage medium, and is applied to the field of wind power generation evaluation. The slope and curvature of a basic grid are determined, and a target area is divided into a flat area, a transition area and a mutation area based on the slope and curvature. The basic grid in the mutation area is subjected to encryption processing based on a preset grid size, and the basic grid in the transition area is subjected to gradual encryption processing. An optimal performance turbulence model of each area is set, a simulation is performed based on the optimal performance turbulence model, and a wind resource evaluation result of the target area is obtained. A power curve of a wind turbine is determined, and a wind power generation amount evaluation result of the target area is determined based on the power curve and the wind resource evaluation result. The application realizes more accurate wind resource evaluation by recognizing and encrypting the grid of the topographic mutation area and setting the optimal performance turbulence model for each type of area, and thus more accurate wind power generation amount prediction results are obtained.

Description

Technical Field

[0001] This invention relates to the field of wind power generation assessment, and in particular to a wind power generation prediction method, a wind power generation prediction device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] Wind farm development is highly dependent on accurate power generation assessment. Assessment errors directly lead to problems such as deviations in project investment returns, failure of wind turbine selection and layout schemes, and inaccurate assessment of financing and insurance risks. As wind power development expands into complex terrain areas (mountains, coastal cliffs, plateau valleys), areas with abrupt terrain changes account for more than 60% of newly planned wind farms worldwide. Traditional wind power generation assessment models suffer from wind shear simulation distortion in areas with abrupt terrain changes and are not suitable for power generation assessment in such areas. The wind power generation assessment results in areas with abrupt terrain changes deviate significantly from the actual results, resulting in huge economic losses and wasted human and material resources. Summary of the Invention

[0003] The purpose of this invention is to provide a wind power generation prediction method, a wind power generation prediction device, an electronic device, and a computer-readable storage medium, which are applied in the field of wind power generation assessment. This method identifies and densifies the grid in areas of abrupt terrain changes and sets up turbulence models with optimal performance for each type of area to achieve more accurate wind resource assessment, thereby obtaining more accurate wind power generation prediction results.

[0004] To solve the above-mentioned technical problems, the present invention provides a method for predicting wind power generation, comprising:

[0005] Generate a base grid for the target region, determine the slope and curvature of each base grid, and divide the target region into flat regions, transition regions, and abrupt change regions based on the slope and curvature;

[0006] The basic grid in the abrupt change region is encrypted based on a preset grid size, and the basic grid in the transition region is gradually encrypted.

[0007] After the grid is refined, an optimal performance turbulence model is set for each region, and the wind resource assessment results for the target region are obtained by simulation based on the optimal performance turbulence model.

[0008] The power curve of the wind turbine is determined, and the wind power generation assessment result of the target area is determined based on the power curve and the wind resource assessment result.

[0009] Optionally, the method further includes:

[0010] A dataset is constructed based on historical wind speed prediction data, historical wind direction prediction data, historical terrain parameters, and historical stability. The difference between the historical wind speed prediction data and the historical wind speed measured data is used as the label of the dataset.

[0011] The LSTM model is trained based on the dataset, and the model parameters are updated based on the loss value of the LSTM model until the trained LSTM model is obtained.

[0012] Based on the wind resource assessment results, real-time wind speed prediction data and real-time wind direction prediction data are determined. The real-time wind speed prediction data, the real-time wind direction prediction data, terrain parameters and stability are input into the trained LSTM model to obtain the wind speed correction amount output by the model.

[0013] The wind resource assessment result is corrected based on the wind speed correction amount to obtain the corrected wind resource assessment result.

[0014] Accordingly, the wind power generation assessment result for the target area is determined based on the power curve and the wind resource assessment result, including:

[0015] The wind power generation assessment result for the target area is determined based on the power curve and the wind resource assessment correction result.

[0016] Optionally, the loss function of the LSTM model includes a momentum conservation loss term, the expression of which is:

[0017] ;

[0018] In the formula, L physics The momentum conservation loss term is ΔU, and the wind speed correction amount output by the model is ΔU. u Let ΔU be the component of the wind speed correction in the u direction. v Let be the component of the wind speed correction in the v direction, t be the time step, x be the horizontal spacing between adjacent grids, y be the vertical spacing between adjacent grids, ρ be the air density, and λ be a constant term.

[0019] Optionally, the basic grid within the mutation region is densified based on a preset grid size, including:

[0020] The grid size threshold of the abrupt change region is determined based on the windward curvature.

[0021] The minimum grid size among the grid size threshold and the preset grid size is determined as the encrypted grid size of the mutation region;

[0022] The base grid within the mutation region is encrypted based on the encrypted grid size.

[0023] Optionally, the basic mesh within the transition region undergoes a gradient densification process, including:

[0024] Determine the shortest distance from the target mesh within the transition region to the boundary of the abrupt change region;

[0025] The shortest distance is input into the exponential decay grid size control function to obtain the transitional refinement grid size of the target grid;

[0026] Based on the transition encryption grid size, the target grid in the transition area is encrypted, and the base grid in the transition area is determined as the target grid in turn, until all the base grids in the transition area are encrypted.

[0027] The expression for the exponentially decaying grid size control function is as follows:

[0028] ;

[0029] In the formula, Δx b Δx is the transition refinement mesh size of the target mesh within the transition region. a Let Δx be the base grid size within the flat region. c d is the size of the encrypted grid within the mutation region, d is the shortest distance, and l is the length of the transition feature.

[0030] Optionally, set the optimal performance turbulence model for each region, including:

[0031] The optimal performance turbulence model for the flat region is set to the standard k-ε model;

[0032] The optimal performance turbulence model for the transition region is set as the RNG k-ε model;

[0033] The optimal performance turbulence model for the abrupt change region is set as the SST k-ω model.

[0034] Optionally, the wind resource assessment results for the target area are obtained by simulation based on the optimal performance turbulence model, including:

[0035] The wind resource assessment results for the target region are obtained by simulation based on the optimal performance turbulence model; during the turbulence model simulation, the turbulence viscosity coefficient in the optimal performance turbulence model for the abrupt change region is corrected based on the radius of curvature.

[0036] To solve the above-mentioned technical problems, the present invention provides a wind power generation prediction device, comprising:

[0037] The first module is used to generate the basic grid of the target area, determine the slope and curvature of each basic grid, and divide the target area into flat areas, transition areas and abrupt change areas based on the slope and curvature.

[0038] The second module is used to perform encryption processing on the basic grid in the abrupt change region based on a preset grid size, and to perform gradual encryption processing on the basic grid in the transition region.

[0039] The third module is used to set the optimal performance turbulence model for each region after grid refinement, and to perform simulation based on the optimal performance turbulence model to obtain the wind resource assessment results for the target region.

[0040] The fourth module is used to determine the power curve of the wind turbine, and based on the power curve and the wind resource assessment results, to determine the wind power generation assessment results for the target area.

[0041] To solve the above-mentioned technical problems, the present invention provides an electronic device, comprising:

[0042] Memory, used to store computer programs;

[0043] A processor is used to implement the wind power generation prediction method described above when executing the computer program.

[0044] To address the aforementioned technical problems, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aforementioned wind power generation prediction method.

[0045] As can be seen, this invention generates a basic grid for the target area, determines the slope and curvature of each basic grid, and divides the target area into flat areas, transition areas, and abrupt change areas based on the slope and curvature; it refines the basic grid in the abrupt change areas based on a preset grid size, and performs gradual refinement on the basic grid in the transition areas; after grid refinement, it sets the optimal performance turbulence model for each area, and performs simulation based on the optimal performance turbulence model to obtain the wind resource assessment results for the target area; it determines the power curve of the wind turbine, and determines the wind power generation assessment results for the target area based on the power curve and the wind resource assessment results.

[0046] This invention achieves more accurate wind resource assessment by identifying and refining the grid in areas of abrupt terrain changes and setting optimal turbulence models for each type of area, thereby obtaining more accurate wind power generation prediction results. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0048] Figure 1 A flowchart of a wind power generation prediction method provided in an embodiment of the present invention;

[0049] Figure 2 This is a structural block diagram of a wind power generation prediction device provided in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Wind farm development heavily relies on accurate power generation assessment, and errors in this assessment directly lead to: deviations in project investment returns (an error of >5% can cause economic losses in the tens of millions); failure of wind turbine selection and layout schemes; and inaccurate assessment of financing and insurance risks. As wind power development expands into complex terrain areas (mountains, coastal cliffs, plateau valleys), areas with abrupt terrain changes now account for over 60% of newly planned wind farms globally. The limitations of traditional assessment models in such areas are becoming increasingly apparent, making the accuracy of model calculations ever more crucial.

[0052] Current mainstream wind power assessment models, such as WAsP (Wind Atlas Analysis and Application Program), OpenFOAM (Open Source Field Operation and Manipulation), and standard CFD (Computational Fluid Dynamics) solutions, have three core problems:

[0053] Topographic simplification leads to distortion of wind shear physics: In order to control computational costs, a fixed coarse grid is often used for the entire domain. The grid resolution is insufficient and cannot resolve the micro-scale airflow separation of abrupt terrain such as steep slopes, ridges, and valleys. As a result, the vertical extrapolation error of the wind speed profile in abrupt terrain reaches 15-25%.

[0054] Limitations of turbulence models: Traditional k-ε / k-ω models are based on the assumption of isotropic turbulence, but abrupt topographic changes can trigger strong shear layer separation, unsteady eddies (such as leeward backflow), and lateral transport of secondary flows, resulting in turbulent kinetic energy prediction errors >40% and distortion of wind speed temporal fluctuations. The k-ε model is a turbulence model based on turbulent kinetic energy (k) and turbulent dissipation rate (ε), while the k-ω model is a turbulence model based on turbulent kinetic energy (k) and turbulent frequency (ω).

[0055] Computational accuracy and efficiency are mutually exclusive: if a high-resolution mesh is used across the entire area, the wind field simulation time increases, which cannot meet the requirements of engineering research and development cycles. Furthermore, local manual densification relies on human experience, and numerical oscillations are prone to occur in mesh transition zones. The wind speed simulation error on steep slopes is more than three times that of flat areas, and existing models lack specific terrain adaptability.

[0056] This invention proposes a multi-parameter fusion-based terrain abrupt change criterion, simultaneously capturing slope abrupt changes and curvature features to identify terrain abrupt change regions. The grid within these regions is then refined (mesh size reduced), and an optimal turbulence model is applied to each terrain region, resulting in more accurate wind resource assessments and consequently, more precise wind power generation assessments. Furthermore, this invention can incorporate an LSTM (Long Short-Term Memory) model to refine the wind resource assessment results, further enhancing the accuracy of wind power generation assessments.

[0057] The following combination Figure 1 , Figure 1 A flowchart of a wind power generation prediction method provided in an embodiment of the present invention, the method may include:

[0058] S101: Generate the base grid of the target area, determine the slope and curvature of each base grid, and divide the target area into flat areas, transition areas and abrupt change areas based on the slope and curvature.

[0059] This embodiment can perform wind resource assessment and wind power generation assessment within a target area. First, a basic grid for the target area can be constructed. This embodiment does not limit the tools used to generate the basic grid or its size; these can be set based on the actual application. For example, this embodiment can use Delaunay triangulation to construct an unstructured basic grid, with a grid size of 50m × 50m.

[0060] Traditional CFD tools have two major problems in mesh generation: fixed resolution across the entire domain, using the same mesh density for flat and abrupt change regions, resulting in wasted computing resources or insufficient accuracy in steep areas; and strong reliance on manual intervention, requiring engineers to manually annotate the densified areas, which is inefficient and prone to subjective errors.

[0061] This embodiment proposes a multi-parameter fusion criterion for terrain abrupt change, determining the slope and curvature of each basic grid, and dividing the target area into flat regions, transition regions, and abrupt change regions based on the slope and curvature. Flat regions are terrainally flat areas, abrupt change regions are terrainally abrupt change regions, and transition regions are transitional areas between flat and abrupt change regions, where the terrain changes from steep to gentle.

[0062] This embodiment does not limit the calculation method of the slope and curvature of the basic grid. Generally, the slope value of each grid point can be calculated based on the Digital Elevation Model (DEM). The calculation formula is as follows:

[0063] ;

[0064] In the formula, S is the slope in degrees, z is the elevation, x is the horizontal spacing between adjacent grids, and y is the vertical spacing between adjacent grids.

[0065] Elevation represents the altitude of a point on the Earth's surface, measured in meters, and is directly derived from the input digital elevation model data. z / x is the horizontal (east-west) gradient, which represents the rate of change of elevation along the east-west direction. It is a dimensionless quantity and is calculated by the central difference method. Specifically, it is the average of the ratio of the elevation difference between adjacent grid points to the horizontal spacing. z / y is the vertical (north-south) gradient, which represents the rate of change of elevation along the north-south direction. It is also a dimensionless quantity, and its calculation method can be the same as that of the horizontal gradient.

[0066] In this embodiment, the second derivative of the elevation can generally be calculated using the Laplacian operator to characterize the curvature, as shown in the following formula:

[0067] ;

[0068] In the formula, C represents curvature, with units of meters (m). -1 .

[0069] After calculating the slope and curvature, decision rules can be set to divide the target area into flat regions, transition regions, and abrupt change regions based on the slope and curvature of the base grid. This embodiment does not limit the specific setting method of the decision rules. Generally, slope threshold ranges and curvature threshold ranges can be set for flat regions, transition regions, and abrupt change regions, and the division of flat regions, transition regions, and abrupt change regions can be based on the slope threshold ranges and curvature threshold ranges.

[0070] This embodiment does not limit the setting method of the slope threshold range and curvature threshold range, and can be set according to the actual application. In one embodiment, a slope > 30° is generally considered a potential abrupt change zone, and a curvature > 0.1m is considered a potential abrupt change zone. -1 This area can be marked as a region sensitive to airflow distortion; therefore, in this embodiment, the base grid can be used when the slope is >30° and the curvature is >0.1m. -1 The region is defined as a mutation region; the base grid can be defined with a slope of 30° ≥ 10° and a height of 0.1m. -1 Curvature ≥ 0.05m -1 The area is defined as a transition zone; the base grid can be set at a slope greater than 10° and 0.05m. -1 Regions with curvature greater than or equal to the curvature are defined as flat regions.

[0071] Furthermore, this embodiment can generate flat region masks, abrupt region masks, and transition region masks by determining the threshold range of each grid, and then mark each region using the masks.

[0072] Compared to existing technologies, this embodiment exhibits superior performance in terms of terrain representation capabilities, airflow distortion correlation, accuracy, and engineering applicability by simultaneously capturing slope and curvature features. It can accurately locate airflow distortion zones such as ridgelines (accelerating flow), leeward slopes (separated vortices), and narrow valleys (channel effects), with a recognition coverage rate exceeding 95%. Optimized computational resource allocation, with mesh densification only applied to real abrupt change areas and avoiding over-densification in flat regions, improves CFD computation efficiency by 40%. Furthermore, the mask supporting subsequent physical correction output directly drives turbulence model switching and eddy viscosity correction, forming a closed-loop technology.

[0073] S102: Based on the preset mesh size, the basic mesh in the mutation region is densified, and the basic mesh in the transition region is gradually densified, that is, the closer to the mutation region, the smaller the mesh size.

[0074] In existing technologies, the same mesh density is used for flat and abrupt change regions, leading to wasted computational resources or insufficient accuracy in steep slope areas. This embodiment refines the base mesh in abrupt change and transition regions to improve simulation efficiency while maintaining accuracy.

[0075] This embodiment does not limit the specific method of mesh encryption. Generally, the basic mesh in the abrupt change area can be encrypted based on the preset mesh size, and the basic mesh in the transition area can be gradually encrypted.

[0076] Specifically, in this embodiment, the grid size threshold of the abrupt change region can be determined based on the curvature along the wind direction. The minimum grid size among the grid size threshold and preset grid sizes is determined as the encrypted grid size of the abrupt change region. The basic grid within the abrupt change region is then encrypted based on the encrypted grid size. The formula for determining the encrypted grid size of the abrupt change region can be shown below:

[0077] ;

[0078] In the formula, Δx c Δx is the size of the encrypted grid within the mutation region, f is the wind direction vector, and s is the curvature tensor. 2 f / s 2 Windward curvature, specifically referring to the topographic curvature along the prevailing wind direction, is measured in meters (m). -1 The physical correlation of airflow acceleration / deceleration intensity is obtained through a quadratic projection operation of the unit wind direction vector and the curvature tensor, with min being the minimum value function.

[0079] This embodiment can determine a target mesh from the base mesh of the transition region, and determine the shortest distance from the target mesh in the transition region to the boundary of the abrupt change region; input the shortest distance into the exponential decay mesh size control function to obtain the transition refinement mesh size of the target mesh; based on the transition refinement mesh size, refine the target mesh in the transition region, and sequentially determine the base mesh in the transition region as the target mesh until all the base meshes in the transition region are refined, with only one target mesh existing in the transition region in each refinement process; wherein, the expression of the exponential decay mesh size control function is:

[0080] ;

[0081] In the formula, Δx b Δx represents the transition mesh size of the target mesh within the transition region. a Δx represents the base grid size within the flat region. c Let d be the size of the encrypted mesh within the mutation region, d be the shortest distance, and l be the transition feature length, which can generally be taken as Δx. a 1 to 2 times the size to ensure a smooth transition.

[0082] In this embodiment, the basic grid size in flat areas is generally 50m×50m, and the preset grid size is generally 5m×5m. Through densification, the total number of grids is reduced to 32% of the 5m grids in the entire area. The grid resolution in key areas of airflow distortion such as ridgelines and canyon bottoms reaches 5m×5m, which can resolve vortex structures with a scale of ≥10m, ensuring simulation accuracy.

[0083] S103: After mesh refinement, set the optimal performance turbulence model for each region, and perform simulation based on the optimal performance turbulence model to obtain the wind resource assessment results for the target region.

[0084] Currently, there are limitations in the selection of turbulence models in commercial CFD software: a single global model cannot simultaneously balance efficiency in flat areas and accuracy in abrupt changes; the terrain-turbulence coupling is ignored, and the standard model does not consider the influence of curvature on eddy viscosity.

[0085] In this embodiment, after mesh refinement, the optimal performance turbulence model for each region can be set. Based on the optimal performance turbulence model, simulation is performed to obtain the wind resource assessment results of the target region. Specifically, the optimal performance turbulence model for flat regions can be set as the standard k-ε model; the optimal performance turbulence model for transition regions can be set as the RNG (Renormalization Group) k-ε model; and the optimal performance turbulence model for abrupt change regions can be set as the SST (Shear Stress Transport) k-ω model.

[0086] In this embodiment, efficiency is prioritized in flat regions, and the k-ε model satisfies the isotropic flow requirements with a complexity of 2 equations, avoiding overcomputation; in transition regions, a balance between accuracy and efficiency is achieved, and the RNG k-ε model improves physical realism through mathematical renormalization, with a manageable increase in cost; in abrupt change regions, accuracy is prioritized, and the SST k-ω model solves the problem of simulating separated flow with a hybrid framework and curvature-endogenous coupling.

[0087] Furthermore, in this embodiment, the wind resource assessment results for the target area can be obtained through simulation based on the optimal performance turbulence model. During the turbulence model simulation, the turbulent viscosity coefficient in the optimal performance turbulence model for abrupt changes can be corrected based on the radius of curvature. This embodiment does not limit the specific correction method; it can be as shown in the following formula:

[0088] ;

[0089] In the formula, μ t,new μ is the corrected turbulent viscosity coefficient. t The original turbulent viscosity coefficient is α, which is the sensitivity coefficient calibrated by wind tunnel testing. 2 z) represents the curvature sign, which can be 1 at ridges and -1 at valleys, used to distinguish between accelerating and decelerating flow. R c The radius of curvature is used to quantify the degree of terrain curvature.

[0090] S104: Determine the power curve of the wind turbine, and based on the power curve and wind resource assessment results, determine the wind power generation assessment results for the target area.

[0091] After obtaining the wind resource assessment results for the region, the power curve of the wind turbine can be determined, and the wind power generation assessment results for the target region can be determined based on the power curve and the wind resource assessment results.

[0092] Furthermore, this embodiment can train an LSMT model using historical wind field data and then refine the wind resource assessment results using an LSTM model to obtain more accurate wind resource assessment results. The LSTM model in this embodiment can be replaced with other types of neural network models, and no restrictions are placed here.

[0093] This embodiment does not limit the training method of the LSTM model. Generally, a dataset can be constructed based on historical wind speed prediction data, historical wind direction prediction data, historical terrain parameters, and historical stability. The difference between historical wind speed prediction data and historical measured wind speed data can be used as the label of the dataset. The LSTM model is trained based on the dataset, and the model parameters are updated based on the loss value of the LSTM model until the trained LSTM model is obtained.

[0094] In this embodiment, during the model training process, the input dimension of the LSTM model can be 5-dimensional time series data, such as historical wind speed prediction data, historical wind direction prediction data, historical terrain parameters (including historical slope and historical curvature) and historical stability (Richardson number Ri).

[0095] This embodiment can perform feature enhancement processing on the input data, such as constructing a slope-curvature cross term: S×C to quantify the intensity of terrain abrupt changes; it can also construct a stability set-wind direction coupling term: Ri×cos(θ-θ0), where θ is the wind direction and θ0 is the ridge orientation.

[0096] This embodiment can physically constrain the LSTM unit. The momentum conservation constraint can be expressed as follows. This embodiment can force the momentum conservation to be satisfied by embedding a loss function:

[0097] ;

[0098] In this embodiment, the loss function of the model can include a momentum conservation loss term, the expression of which is:

[0099] ;

[0100] In the formula, L physics The momentum conservation loss term is represented by ΔU, which is the wind speed correction term in the model output. u Let ΔU be the component of the wind speed correction in the u direction. vdenoted as the component of the wind speed correction in the v direction, t as the time step, x as the horizontal spacing between adjacent grids, y as the vertical spacing between adjacent grids, ρ as the air density, and λ as a constant term, which can be taken as 0.1.

[0101] In this embodiment, the total model loss function Loss can contain two terms: the first term is used to minimize the correction error, and the second term is used to enforce momentum conservation, as shown in the following equation:

[0102] .

[0103] After completing the LTSM model training, this embodiment can determine the real-time wind speed prediction data U based on the wind resource assessment results. CFD Real-time wind direction prediction data, topographic parameters (slope and curvature), and stability (Ri) are input into the trained LSTM model to obtain the wind speed correction ΔU output by the model. The wind resource assessment results are then corrected based on this correction to obtain the corrected wind resource assessment result (U). CFD +ΔU); Based on the power curve and the wind resource assessment correction results, the wind power generation assessment results for the target area are determined.

[0104] In this embodiment, to avoid the problem of model overfitting caused by the lack of engineering site data, the LTSM model can be pre-trained using LES (Large Eddy Simulation) data, and the pre-trained LSTM model can be transferred to the wind speed correction task using engineering site data.

[0105] Based on the above embodiments, the present invention identifies and densifies the grid in areas of abrupt terrain change, and sets up turbulence models with optimal performance for each type of area, thereby achieving more accurate wind resource assessment and obtaining more accurate wind power generation prediction results.

[0106] The following combination Figure 2 , Figure 2 This is a structural block diagram of a wind power generation prediction device provided in an embodiment of the present invention. The device may include:

[0107] The first module 100 is used to generate the basic grid of the target area, determine the slope and curvature of each basic grid, and divide the target area into flat areas, transition areas and abrupt change areas based on the slope and curvature.

[0108] The second module 200 is used to refine the basic grid in the abrupt change region based on the preset grid size, and to perform gradual refinement of the basic grid in the transition region.

[0109] The third module 300 is used to set the optimal performance turbulence model for each region after grid refinement, and to obtain the wind resource assessment results of the target region by simulation based on the optimal performance turbulence model;

[0110] The fourth module 400 is used to determine the power curve of the wind turbine and, based on the power curve and the wind resource assessment results, to determine the wind power generation assessment results for the target area.

[0111] Based on the above embodiments, the present invention identifies and densifies the grid in areas of abrupt terrain change, and sets up turbulence models with optimal performance for each type of area, thereby achieving more accurate wind resource assessment and obtaining more accurate wind power generation prediction results.

[0112] Based on the above embodiments, the device may further include:

[0113] The fifth module is used to construct a dataset based on historical wind speed prediction data, historical wind direction prediction data, historical terrain parameters, and historical stability. The difference between historical wind speed prediction data and historical measured wind speed data is used as the label of the dataset.

[0114] The sixth module is used to train an LSTM model based on the dataset, update the model parameters based on the loss value of the LSTM model, until the trained LSTM model is obtained;

[0115] The seventh module is used to determine real-time wind speed prediction data and real-time wind direction prediction data based on wind resource assessment results. It inputs the real-time wind speed prediction data, real-time wind direction prediction data, terrain parameters and stability into the trained LSTM model to obtain the wind speed correction amount output by the model.

[0116] The eighth module is used to correct the wind resource assessment results based on the wind speed correction amount, and obtain the corrected wind resource assessment results.

[0117] Correspondingly, the fourth module 400 may include:

[0118] The first unit is used to determine the wind power generation assessment results for the target area based on the power curve and the wind resource assessment correction results.

[0119] Based on the above embodiments, the loss function of the LSTM model includes a momentum conservation loss term, the expression of which is:

[0120] ;

[0121] In the formula, L physics The momentum conservation loss term is represented by ΔU, which is the wind speed correction term in the model output. u Let ΔU be the component of the wind speed correction in the u direction. vdenoted as the component of the wind speed correction in the v direction, t as the time step, x as the horizontal spacing between adjacent grids, y as the vertical spacing between adjacent grids, ρ as the air density, and λ as a constant term.

[0122] Based on the above embodiments, the second module 200 may include:

[0123] The second unit is used to determine the grid size threshold for abrupt regions based on the curvature along the wind direction.

[0124] The third unit is used to determine the minimum grid size among the grid size threshold and preset grid sizes as the encrypted grid size for the mutation region;

[0125] The fourth unit is used to encrypt the base grid within the mutation region based on the encrypted grid size.

[0126] Based on the above embodiments, the second module 200 may include:

[0127] The fifth unit is used to determine the shortest distance from the target mesh within the transition region to the boundary of the abrupt change region;

[0128] The sixth unit is used to input the shortest distance into the exponential decay grid size control function to obtain the transitional refinement grid size of the target grid;

[0129] The seventh unit is used to densify the target mesh in the transition area based on the transition densification mesh size, and to determine the base mesh in the transition area as the target mesh in turn, until all the base mesh in the transition area is densified.

[0130] The expression for the exponentially decaying grid size control function is as follows:

[0131] ;

[0132] In the formula, Δx b Δx represents the transition mesh size of the target mesh within the transition region. a Δx represents the base grid size within the flat region. c d is the size of the encrypted grid within the mutation region, d is the shortest distance, and l is the length of the transition feature.

[0133] Based on the above embodiments, the third module 300 may include:

[0134] Unit 8 is used to set the optimal performance turbulence model for flat regions as the standard k-ε model;

[0135] Unit 9 is used to set the optimal performance turbulence model for the transition region as the RNG k-ε model;

[0136] Unit 10 is used to set the optimal performance turbulence model for the abrupt change region as the SST k-ω model.

[0137] Based on the above embodiments, the third module 300 may include:

[0138] Unit 11 is used to obtain wind resource assessment results for the target area based on the simulation of the optimal performance turbulence model; during the turbulence model simulation, the turbulence viscosity coefficient in the optimal performance turbulence model of the abrupt change region is corrected based on the radius of curvature.

[0139] Based on the above embodiments, the present invention also provides an electronic device, which may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the steps provided in the above embodiments. Of course, the device may also include various necessary network interfaces, a power supply, and other components.

[0140] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by an execution terminal or processor, can implement the method provided in the embodiments of the present invention; the storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0141] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for predicting wind power generation, characterized in that, include: Generate a base grid for the target region, determine the slope and curvature of each base grid, and divide the target region into flat regions, transition regions, and abrupt change regions based on the slope and curvature; The basic grid in the abrupt change region is encrypted based on a preset grid size, and the basic grid in the transition region is gradually encrypted. After the grid is refined, an optimal performance turbulence model is set for each region, and the wind resource assessment results for the target region are obtained by simulation based on the optimal performance turbulence model. Determine the power curve of the wind turbine, and based on the power curve and the wind resource assessment results, determine the wind power generation assessment results for the target area; The basic mesh within the transition region undergoes gradient densification processing, including: Determine the shortest distance from the target mesh within the transition region to the boundary of the abrupt change region; The shortest distance is input into the exponential decay grid size control function to obtain the transitional refinement grid size of the target grid; Based on the transition encryption grid size, the target grid in the transition area is encrypted, and the base grid in the transition area is determined as the target grid in turn, until all the base grids in the transition area are encrypted. The expression for the exponentially decaying grid size control function is as follows: ； In the formula, Δx b Δx is the transition refinement mesh size of the target mesh within the transition region. a Let Δx be the base grid size within the flat region. c d is the size of the encrypted grid within the mutation region, d is the shortest distance, and l is the length of the transition feature.

2. The wind power generation prediction method according to claim 1, characterized in that, Also includes: A dataset is constructed based on historical wind speed prediction data, historical wind direction prediction data, historical terrain parameters, and historical stability. The difference between the historical wind speed prediction data and the historical wind speed measured data is used as the label of the dataset. The LSTM model is trained based on the dataset, and the model parameters are updated based on the loss value of the LSTM model until the trained LSTM model is obtained. Based on the wind resource assessment results, real-time wind speed prediction data and real-time wind direction prediction data are determined. The real-time wind speed prediction data, the real-time wind direction prediction data, terrain parameters and stability are input into the trained LSTM model to obtain the wind speed correction amount output by the model. The wind resource assessment result is corrected based on the wind speed correction amount to obtain the corrected wind resource assessment result. Accordingly, the wind power generation assessment result for the target area is determined based on the power curve and the wind resource assessment result, including: The wind power generation assessment result for the target area is determined based on the power curve and the wind resource assessment correction result.

3. The wind power generation prediction method according to claim 2, characterized in that, The loss function of the LSTM model includes a momentum conservation loss term, the expression of which is: ； In the formula, L physics The momentum conservation loss term is ΔU, and the wind speed correction amount output by the model is ΔU. u Let ΔU be the component of the wind speed correction in the u direction. v Let be the component of the wind speed correction in the v direction, t be the time step, x be the horizontal spacing between adjacent grids, y be the vertical spacing between adjacent grids, ρ be the air density, and λ be a constant term.

4. The wind power generation prediction method according to claim 1, characterized in that, The basic grid within the mutation region is densified based on a preset grid size, including: The grid size threshold of the abrupt change region is determined based on the windward curvature. The minimum grid size among the grid size threshold and the preset grid size is determined as the encrypted grid size of the mutation region; The base grid within the mutation region is encrypted based on the encrypted grid size.

5. The wind power generation prediction method according to claim 1, characterized in that, Set the optimal performance turbulence model for each region, including: The optimal performance turbulence model for the flat region is set to the standard k-ε model; The optimal performance turbulence model for the transition region is set as the RNG k-ε model; The optimal performance turbulence model for the abrupt change region is set as the SST k-ω model.

6. The wind power generation prediction method according to claim 1, characterized in that, The wind resource assessment results for the target area are obtained by simulation based on the optimal performance turbulence model, including: The wind resource assessment results for the target region are obtained by simulation based on the optimal performance turbulence model; during the turbulence model simulation, the turbulence viscosity coefficient in the optimal performance turbulence model for the abrupt change region is corrected based on the radius of curvature.

7. A wind power generation prediction device, characterized in that, include: The first module is used to generate the basic grid of the target area, determine the slope and curvature of each basic grid, and divide the target area into flat areas, transition areas and abrupt change areas based on the slope and curvature. The second module is used to perform encryption processing on the basic grid in the abrupt change region based on a preset grid size, and to perform gradual encryption processing on the basic grid in the transition region. The third module is used to set the optimal performance turbulence model for each region after grid refinement, and to perform simulation based on the optimal performance turbulence model to obtain the wind resource assessment results for the target region. The fourth module is used to determine the power curve of the wind turbine, and based on the power curve and the wind resource assessment results, to determine the wind power generation assessment results for the target area. The basic mesh within the transition region undergoes gradient densification processing, including: Determine the shortest distance from the target mesh within the transition region to the boundary of the abrupt change region; The shortest distance is input into the exponential decay grid size control function to obtain the transitional refinement grid size of the target grid; Based on the transition encryption grid size, the target grid in the transition area is encrypted, and the base grid in the transition area is determined as the target grid in turn, until all the base grids in the transition area are encrypted. The expression for the exponentially decaying grid size control function is as follows: ； In the formula, Δx b Δx is the transition refinement mesh size of the target mesh within the transition region. a Let Δx be the base grid size within the flat region. c d is the size of the encrypted grid within the mutation region, d is the shortest distance, and l is the length of the transition feature.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the wind power generation prediction method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the wind power generation prediction method as described in any one of claims 1 to 6.

Images