Mountain wind field prediction method and device based on CFD and Kriging model, terminal and medium

Abstract

The invention discloses a CFD and Kriging model-based mountain wind field prediction method and device, a terminal and a medium. The method comprises the following steps of establishing a three-dimensional model of a target mountain terrain; establishing a computational domain of the three-dimensional model and setting boundary conditions of the computational domain; carrying out gridding processing on the computational domain, and obtaining a certain amount of sample data by adopting a computational fluid dynamics numerical simulation method; establishing a Kriging prediction model based on the obtained sample data; the Kriging prediction model is trained and tested; and adopting the trained Kriging prediction model to carry out mountain wind field prediction at other positions to obtain a mountain wind field distribution prediction result. According to the invention, the existing wind field data can be accurately fitted, and the wind field of the unknown point in the parameter range can be accurately predicted; mountain wind fields of other terrains can be researched, and a basis is provided for structural wind resistance design in the mountain terrains.

Description

Technical Field

[0001] The present invention relates to the technical fields of wind-resistant design of mountainous structures and wind energy site selection, and specifically designs a mountainous wind field prediction method, device, terminal and medium based on CFD and Kriging models. Background Art

[0002] Ultra-high voltage transmission lines are often laid out along mountainous terrain. Topographical features such as hills and valleys can significantly alter the wind speed distribution and turbulence characteristics within the atmospheric boundary layer near the ground, and can also easily lead to extreme weather conditions such as sudden high winds. Mountain wind farms have a significant impact on the design of ultra-high voltage transmission lines in mountainous areas and the safe and stable operation of power grid systems.

[0003] The main research methods for mountain wind fields include analytical methods, numerical simulations, and wind tunnel tests. Early research was mainly based on theoretical analysis and wind tunnel experiments. In terms of theoretical analysis, Jackson and Hunt first proposed a theoretical model for the turbulent boundary layer on two-dimensional shallow hill terrain in 1975. This model linearized the motion equations of the background logarithmic wind profile. Subsequently, Mason and Sykes expanded the applicability of the Jackson and Hunt model to three-dimensional terrain in 1979. In 1988, Hunt et al. further refined the inner and outer layers of the atmospheric boundary layer in the Jackson and Hunt model and divided them into two layers, proposing a more complex model.

[0004] In terms of wind tunnel experiments, Bowen and Lindley conducted wind tunnel experiments on forward steep slopes of different slopes in 1977. Cao and Tamura's research in 2006-2007 showed that the roughness of the hill surface significantly affects the velocity ratio and turbulence intensity. Lubitz and White proved in 2007 that the incident wind direction has a significant effect on the acceleration factor above the hill. Based on particle image velocimetry (PIV) technology, Kamada et al. explored the flow field characteristics around a two-dimensional single-hill model through wind tunnel experiments in 2019, while Li et al. conducted flow experiments on a two-dimensional multi-hill model in 2017, revealing the large clockwise circulation formed between the hills. These experiments not only provide valuable measured data, but also provide a basis for the verification and development of numerical models. These works have laid the foundation for understanding the basic principles of flow over hills and valleys.

[0005] With the improvement of computer capabilities, numerical simulation has become a more powerful tool that can capture more complex flow phenomena. Liu et al. (2020) used large eddy simulation to find that the wind direction behind the hill is not parallel to the incident wind direction, resulting in the formation of organized flow structures and changes in the turbulent characteristics of the flow field. In 2022, Zhou et al. used the DDES method to study the flow characteristics of the turbulent boundary layer on steep mountain terrain. In 2023, Deng et al. used the LES method to simulate the impact of various saddle-shaped micro-topography on the wind field of transmission lines. The RANS model is widely used in wind field research due to its simplicity and computational efficiency, but the LES and DDES models have become increasingly popular in recent years because they can analyze transient and unsteady flows (Yang et al., 2021).

[0006] Furthermore, field experiments enable observations under natural conditions, providing direct evidence of atmospheric boundary layer flows in real environments. These experimental data can be used to validate theoretical predictions and numerical simulations. However, field experiments can only be conducted at specific locations and are affected by factors such as topography and weather variations, making data collection difficult and requiring significant resource investment (Finnigan et al., 2020).

[0007] Considering the limitations of theoretical and experimental methods in capturing the complexity of real-world terrain, high-precision numerical simulations require extremely high computing resources, while field experiments are constrained by multiple factors. Therefore, a mountain wind field prediction method is needed that can avoid complex experimental steps and huge time costs, providing a reference for the wind-resistant design of transmission structures in mountainous areas and the selection of wind energy sites. Summary of the Invention

[0008] The present invention provides a mountain wind field prediction method, device, terminal and medium based on CFD and Kriging models. First, a certain amount of sample data is obtained by using a computational fluid dynamics (CFD) numerical simulation method. Then, a Kriging model is established based on the obtained sample data and trained and tested. In this way, a prediction model for mountain wind fields is established, which can overcome existing technical defects.

[0009] In a first aspect, the present invention proposes a mountain wind field prediction method based on CFD and Kriging model, comprising the following steps:

[0010] Establish a three-dimensional model of the target mountainous terrain;

[0011] Establishing a computational domain of the three-dimensional model and setting its boundary conditions;

[0012] The computational domain is gridded and a certain amount of sample data is obtained using computational fluid dynamics numerical simulation methods;

[0013] Establish a Kriging prediction model based on the obtained sample data;

[0014] Train and test the Kriging prediction model;

[0015] The trained Kriging prediction model is used to predict mountain wind fields at other locations to obtain mountain wind field distribution prediction results.

[0016] Furthermore, the size of the calculation domain is 20L×20D×6H, where L is the length of the target mountain terrain three-dimensional model, D is the width of the target mountain terrain three-dimensional model, and H is the height of the target mountain terrain three-dimensional model.

[0017] Furthermore, the boundary conditions of the computational domain include: blockage rate, inlet wind speed, outlet pressure, top boundary and two side boundaries, wherein the expression of the inlet wind speed is:

[0018]

[0019] Where, u is the inlet wind speed; u ref is the reference height z ref The wind speed value at the location; z is the actual height; z ref is the reference height; α is the surface roughness category parameter.

[0020] Furthermore, the control equation for obtaining a certain amount of sample data using the computational fluid dynamics numerical simulation method is:

[0021]

[0022] Among them, x i (i=1, 2 and 3) represent the three coordinate directions of x, y and z respectively; u i Corresponding to the speed in the three coordinate directions; p is pressure; ρ is fluid density; μ is dynamic viscosity; S ij is the average strain rate tensor; is the Reynolds stress tensor.

[0023] Furthermore, the mathematical expression of the Kriging prediction model is:

[0024]

[0025] Among them, w is the weight parameter; n is the number of sample data; y (i) is the response value of the i-th sample data; For the prediction model.

[0026] Furthermore, the mathematical expression of the basis function of the Kriging prediction model is:

[0027]

[0028] Among them, θ j is an undetermined parameter, which allows the bandwidth of the basis function to change with the variable; The i-th sample data when x is the j-th basis function; j is the predicted position when it is the jth basis function; p j is the basis function index.

[0029] Furthermore, the specific steps of training and testing the Kriging prediction model are as follows:

[0030] Randomly dividing the sample data into a training set and a test set;

[0031] Normalize the range of the variable and set the upper and lower limits of the search;

[0032] Inputting the training set into the Kriging prediction model for training;

[0033] The test set is input into the Kriging prediction model, and the prediction accuracy of the model is evaluated using the maximum likelihood function. If the prediction accuracy meets the standard, the required Kriging prediction model is obtained, otherwise the training is repeated.

[0034] In a second aspect, the present invention proposes a mountain wind field prediction device based on CFD and Kriging model, comprising:

[0035] The first model building module is used to build a three-dimensional model of the target mountain terrain;

[0036] A calculation domain establishment module, used to establish the calculation domain of the three-dimensional model and set its boundary conditions;

[0037] The sample data acquisition module is used to perform grid processing on the computational domain and obtain a certain amount of sample data using computational fluid dynamics numerical simulation methods;

[0038] The second model building module is used to build a Kriging model prediction model based on the obtained sample data;

[0039] Model training module, used to train and test the Kriging prediction model;

[0040] The prediction module is used to use the trained Kriging prediction model to predict mountain wind fields in other locations and obtain prediction results.

[0041] In a third aspect, the present invention proposes a computer terminal, comprising a processor, a memory and a communication interface; the memory and the communication interface are coupled to the processor, and the memory is used to store computer program instructions; wherein, when the processor executes the computer program instructions, the computer terminal implements the steps of the method described in the first aspect.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0043] Beneficial effects:

[0044] 1. The present invention proposes a method for predicting wind fields in mountainous terrain based on computational fluid dynamics and the Kriging proxy model. First, CFD numerical simulation is used to provide a certain amount of wind field sample data. Then, a Kriging prediction model is established based on the wind field sample data and trained and tested to obtain a prediction model for predicting wind field data with other unknown parameters, thereby quickly and efficiently providing data support for structural wind resistance design and wind energy site selection in mountainous terrain.

[0045] 2. The present invention predicts mountain wind fields based on CFD methods and Kriging proxy models, avoiding complex experimental steps and huge time costs, effectively overcoming the limitations of existing technologies in capturing the complexity of real-world terrain, and eliminating the need for extremely high computing resources and field experiments constrained by multiple factors in high-precision numerical simulations. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the method described in Example 1 of the present invention;

[0047] Figure 2 is a schematic diagram of the watershed model in Example 1 of the present invention;

[0048] Figure 3 is a schematic diagram of the calculation domain described in Example 1 of the present invention;

[0049] Figure 4 This is a schematic diagram of the arrangement of measuring points in Example 1 of the present invention;

[0050] Figure 5 is a principle block diagram of the device described in Example 2 of the present invention;

[0051] Figure 6 This is a principle block diagram of embodiment 3 of the present invention;

[0052] Figure 7It is a comparison chart of the wind field prediction value and the simulation value of the present invention;

[0053] Figure 8 It is a wind field prediction structure diagram of the present invention at full attack angle. DETAILED DESCRIPTION

[0054] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0055] Example 1:

[0056] like Figure 1 As shown, this embodiment provides a mountain wind field prediction method based on CFD and Kriging model, and the specific steps are as follows:

[0057] Step 1: Establish a three-dimensional model of the target mountain terrain;

[0058] This embodiment uses the watershed terrain as an example to illustrate that the watershed terrain will increase the wind speed at the top of the mountain and may also cause an increase in ice on the transmission lines in the mountainous area in winter. Figure 2 As shown, the cross-sectional profile is expressed using the cosine function:

[0059]

[0060] If the height of the selected watershed mountain is 100m and the diameter of the mountain bottom is 300m, then

[0061] Step 2: Establishing the computational domain of the three-dimensional model and setting its boundary conditions;

[0062] To eliminate the influence of the computational domain boundary on the flow field near the mountain, the computational domain needs to be large enough. According to the requirements of the computational wind engineering, the blockage rate needs to be less than 3%. Therefore, a computational domain with a size of 40D×20D×6H (length, width, height) is established, that is, the length is twice the width. Figure 3 As shown in the figure, the computational domain inlet uses a velocity inlet, the outlet uses a pressure outlet, and the boundary parameters are set the same as for the inlet. Symmetric boundaries are used at the top and sides of the computational domain. No-slip walls are used on the mountain surface and the ground.

[0063] Among them, the change of the average inlet wind speed u along the height z is expressed by the exponential function

[0064]

[0065] Among them, z ref =10 is the reference height, uref is the wind speed value at the reference height; α=0.15 is the surface roughness category parameter.

[0066] Step 3: Grid the computational domain. In this example, the entire computational domain uses a hexahedral structured grid. To improve computational accuracy, the grid around the mountain is encrypted. A certain amount of sample data is obtained using computational fluid dynamics numerical simulation methods.

[0067] In this example, the numerical simulation is based on the general commercial software ANSYS FLUENT, and the governing equations are the three-dimensional incompressible unsteady Reynolds-averaged Navier-Stokes (URANS) equations, which are as follows:

[0068]

[0069] Among them, x i (i=1, 2 and 3) represent the three coordinate directions of x, y and z respectively; u i Corresponding to the speed in the three coordinate directions; p is pressure; ρ is fluid density; μ is dynamic viscosity; S ij is the average strain rate tensor; is the Reynolds stress tensor.

[0070] In addition, the turbulence model adopted is SST k-ω; the SIMPLE scheme is used for the pressure-velocity coupling equation during the solution process, and the second-order implicit format is used for the transient equation; the second-order upwind format is used for the turbulent kinetic energy and specific dissipation rate.

[0071] In this embodiment, the influence of position (x, y and z coordinates) and wind direction angle (α) on the wind field is taken into account during the numerical simulation. For this purpose, the full factorial experimental design method is used to obtain sample data. The data measurement points in the numerical simulation are arranged as follows: Figure 4 As shown, -150<<x<<150m, Δx=150m, -150<<z<<150m, Δz=75m, the height range of the measuring points is 0<<yt<<300m, Δy=20m, and no measuring points are arranged inside the mountain. The range of wind direction angle is 0<<α<<90°, Δα=15°, Figure 4 The wind direction shown is 0°, and the wind direction rotates clockwise.

[0072] Step 4: Establish a Kriging prediction model based on the obtained sample data;

[0073] Generally, A radial basis function approximation of is:

[0074]

[0075] Among them, c (i) Indicates n c The i-th of the basis functions, ψ is n including the value of the basis function ψ itselfc dimensional vector, between the predicted position x and the basis function center c (i) The Euclidean distance between them is calculated.

[0076] Common basis functions include linear basis functions, cubic basis functions, and Gauss basis functions. This embodiment uses the following basis functions:

[0077]

[0078] Among them, θ j is an undetermined parameter, which allows the bandwidth of the basis function to change with the variable; The i-th sample data when x is the j-th basis function; j is the predicted position when it is the jth basis function; p j is the basis function index.

[0079] The basis function of the model brought by Kriging has a vector θ={θ1,θ2,…θ k} T , so that the bandwidth of the basis function can change with the variable. Kriging allows the exponent (p i ={p1, p2, ..., p k} T ) varies with each dimension of x (usually p j ∈[1,2]).

[0080] Common parameter estimation evaluation criteria include maximum likelihood estimation and cross validation. This embodiment adopts maximum likelihood estimation. Given a set of weight parameters w model Assuming that the random distribution of error ∈ satisfies the normal distribution with standard deviation σ, then The generated dataset {(x (1) ,y (1) ±∈),(x (2) ,y (2) ±∈),…,(x (n) ,y (n) ±∈)} is:

[0081]

[0082] Negate the natural logarithm of equation (7) and minimize it:

[0083]

[0084] Assuming a constant σ and ∈, equation (8) can be simplified to the least squares criterion, that is, the mathematical expression of the Kriging prediction model is:

[0085]

[0086] Among them, w is the weight parameter; n is the number of sample data; y (i) is the response value of the i-th sample data; For the prediction model.

[0087] Step 5: Train and test the Kriging prediction model;

[0088] In this example, the specific steps of training and testing the Kriging prediction model are as follows:

[0089] Step 5.1, randomly dividing the sample data into a training set and a test set;

[0090] Step 5.2: Normalize the range of the variables to [0, 1]. This simplifies mathematical operations and avoids the problem of inconsistent ranges of multiple variables. Set the search limits for θ to 2 and -3 (fix p = 2), which correspond to the upper and lower bounds of 102 and 10-3, respectively.

[0091] Step 5.3: input the training set into the Kriging prediction model for training, and use a genetic algorithm to search for likelihood values during training;

[0092] Step 5.4: Input the test set into the Kriging prediction model and use the maximum likelihood function to evaluate the prediction accuracy of the model. If the prediction accuracy meets the standard, the required Kriging prediction model and its optimal model parameter θ are obtained; otherwise, repeat the training.

[0093] Step 6: Based on the range of parameter θ, the trained Kriging prediction model is used to predict mountain wind fields at other locations to obtain mountain wind field distribution prediction results.

[0094] Example 2:

[0095] like Figure 5 As shown, this embodiment provides a mountain wind field prediction device based on CFD and Kriging model, the device comprising:

[0096] A first model building module is used to build a three-dimensional model of the target mountain terrain. The process of building the three-dimensional model is described in step 1 of Example 1.

[0097] A computational domain establishment module, used to establish the computational domain of the three-dimensional model and set its boundary conditions. For the process of establishing the computational domain and determining its boundary conditions, please refer to the content described in step 2 of embodiment 1;

[0098] A sample data acquisition module is used to perform grid processing on the computational domain and obtain a certain amount of sample data using a computational fluid dynamics numerical simulation method. The sample data acquisition process is described in step 3 of Example 1.

[0099] The second model building module is used to build a Kriging prediction model based on the obtained sample data. For the process of building the Kriging prediction model, please refer to the content described in step 4 of Example 1;

[0100] The model training module is used to train and test the Kriging prediction model. For the specific process, please refer to the content described in step 5 of Example 1;

[0101] The prediction module is used to use the trained Kriging prediction model to predict mountain wind fields at other locations to obtain prediction results. For the mountain wind field prediction process, please refer to the content described in step 6 of Example 1.

[0102] Example 3:

[0103] like Figure 6 As shown, this embodiment provides a computer terminal based on the above-mentioned mountain wind field prediction method, and the computer terminal includes a processor, a memory and a communication interface; the memory and the communication interface are coupled to the processor, and the memory is used to store computer program instructions; wherein, when the processor executes the computer program instructions, the computer terminal implements the steps of the method described in the above-mentioned embodiment 1.

[0104] Example 4:

[0105] This embodiment provides a computer-readable storage medium, which stores computer program instructions. When the computer program instructions are executed by a processor, the steps of the method described in the above embodiment 1 are implemented.

[0106] Next, we take the watershed with a mountain height of 100m and a mountain bottom diameter of 300m as an example to predict the wind field. The comparison between the wind field prediction value and the simulation value is as follows: Figure 7 As shown, Figure 7 The scatter plot in the figure shows the wind field distribution when different mountain locations are at different wind direction angles. Figure 7 At point 1 shown in (a), the wind speed decreases as the wind direction increases from 0° to 90°. This is because point 1 is located at the foot of a mountain. As the wind direction increases, point 1 is on the windward side, and the mountain has a blocking effect on the wind speed. Figure 7 Point 7 (b) is located at the top of a mountain. Wind speed increases as it approaches the mountain, rather than slowing down due to friction with the ground. The change in wind speed with wind direction generally follows the rule that the greater the wind direction, the greater the wind speed. Figure 7Although the wind speed variation at point 13 shown in (c) is similar to that at point 1, it is located on the leeward side, so the wind field variation becomes smaller when the wind direction angle is larger. Figure 7 The curve in FIG is the wind field prediction value obtained by training the Kriging model, and the model parameters are θ = {0.7450, 1.9390, 0.7500, 1.2187}.

[0107] It can be seen from this that the prediction model obtained by the present invention can accurately fit the wind field data of each monitoring point and provide data of non-monitoring points.

[0108] Figure 8 Further given Figure 7 The wind field prediction values of the three measuring points at full angle of attack are consistent with the wind field laws analyzed in the previous article, indicating that the established wind field prediction model is accurate and reliable.

[0109] In summary, the present invention combines CFD and Kriging proxy models to study the wind field characteristics of watershed terrain. First, CFD numerical simulation is used to obtain wind field data at different spatial positions of the mountain under wind direction angles of 0-90°. Then, a Kriging prediction model is established and trained based on the obtained wind field data, and the accuracy of the model is evaluated using the maximum likelihood function. The wind field prediction model finally obtained can accurately fit the existing wind field data and accurately predict the wind field of unknown points within the parameter range. It can also study mountain wind fields of other terrains, providing a basis for the wind-resistant design of structures in mountainous terrain.

[0110] Finally, it should be noted that the above description is only a preferred embodiment of the present invention. Under the guidance of the present invention, ordinary technicians in this field can make various similar expressions without violating the purpose and claims of the present invention. Such changes fall within the scope of protection of the present invention.

Claims

1. A mountain wind field prediction method based on CFD and Kriging model, characterized by: The steps include: Establish a three-dimensional model of the target mountainous terrain; Establishing a computational domain of the three-dimensional model and setting its boundary conditions; The computational domain is gridded and a certain amount of sample data is obtained using computational fluid dynamics numerical simulation methods; Establish a Kriging prediction model based on the obtained sample data; Train and test the Kriging prediction model; The trained Kriging prediction model is used to predict mountain wind fields at other locations to obtain mountain wind field distribution prediction results.

2. The mountain wind field prediction method based on CFD and Kriging model according to claim 1 is characterized in that: The size of the calculation domain is 20L×20D×6H, where L is the length of the target mountain terrain 3D model, D is the width of the target mountain terrain 3D model, and H is the height of the target mountain terrain 3D model.

3. The mountain wind field prediction method based on CFD and Kriging model according to claim 1 is characterized in that: The boundary conditions of the computational domain include: blockage rate, inlet wind speed, outlet pressure, top boundary and two side boundaries, wherein the expression of the inlet wind speed is: Where, u is the inlet wind speed; u ref is the reference height z ref The wind speed value at the location; z is the actual height; z ref is the reference height; α is the surface roughness category parameter.

4. The mountain wind field prediction method based on CFD and Kriging model according to claim 1 is characterized in that: The control equation for obtaining a certain amount of sample data using the computational fluid dynamics numerical simulation method is: Among them, x i (i=1, 2 and 3) represent the three coordinate directions of x, y and z respectively; u i Corresponding to the speed in the three coordinate directions; p is pressure; ρ is fluid density; μ is dynamic viscosity; S ij is the average strain rate tensor; is the Reynolds stress tensor.

5. The mountain wind field prediction method based on CFD and Kriging model according to claim 1 is characterized in that: The mathematical expression of the Kriging prediction model is: Among them, w is the weight parameter; n is the number of sample data; y (i) is the response value of the i-th sample data; For the prediction model.

6. The mountain wind field prediction method based on CFD and Kriging model according to claim 5 is characterized in that: The mathematical expression of the basis function of the Kriging prediction model is: Among them, θ j is an undetermined parameter, which allows the bandwidth of the basis function to change with the variable; The i-th sample data when x is the j-th basis function; j is the predicted position when it is the jth basis function; p j is the basis function index.

7. The mountain wind field prediction method based on CFD and Kriging model according to claim 1 is characterized in that: The specific steps of training and testing the Kriging prediction model are as follows: Randomly dividing the sample data into a training set and a test set; Normalize the range of the variable and set the upper and lower limits of the search; Inputting the training set into the Kriging prediction model for training; The test set is input into the Kriging prediction model, and the prediction accuracy of the model is evaluated using the maximum likelihood function. If the prediction accuracy meets the standard, the required Kriging prediction model is obtained, otherwise the training is repeated.

8. A mountain wind field prediction device based on CFD and Kriging model, characterized in that: include: The first model building module is used to build a three-dimensional model of the target mountain terrain; A calculation domain establishment module, used to establish the calculation domain of the three-dimensional model and set its boundary conditions; The sample data acquisition module is used to perform grid processing on the computational domain and obtain a certain amount of sample data using computational fluid dynamics numerical simulation methods; The second model building module is used to build a Kriging prediction model based on the obtained sample data; Model training module, used to train and test the Kriging prediction model; The prediction module is used to use the trained Kriging prediction model to predict mountain wind fields in other locations and obtain prediction results.

9. A computer terminal, characterized in that: The computer terminal includes a processor, a memory, and a communication interface; the memory and the communication interface are coupled to the processor, and the memory is used to store computer program instructions; wherein, when the processor executes the computer program instructions, the computer terminal implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program instructions, which implement the steps of the method according to any one of claims 1 to 7 when the computer program instructions are executed by a processor.

Images