CFD and AI coupling general fan blade optimization method and system

Abstract

The invention discloses a CFD and AI coupled universal fan blade optimization method and system, and the method comprises the steps: selecting a blade shape, determining a design space, generating alternative fan blade design schemes in the design space through employing a Latin hypercube sampling method, building a CFD-PINN model, simulating the wind capturing performances of different alternative design schemes, and carrying out the optimization of the fan blade. Generating a plurality of wind catching performance indexes of the fan; the fan blade shape design schemes and the multi-dimensional fan wind capture performance indexes are stored in a cloud platform CFD simulation scheme library, an adaptive multi-core Gaussian process regression AMGPR model is established based on a sample database, an improved wave search algorithm IWSA is utilized to optimize hyper-parameters of the AMGPR model, and the wind capture performance indexes of different fan blade shape design schemes are predicted; calculating a performance index corresponding to the optimized design scheme by using a CFD-PINN model, and comparing a deviation between an AMGPR model prediction index and a CFD-PINN model simulation index; according to the method, intelligent and efficient optimization design of the fan blades is effectively achieved, and the wind catching performance of the fan is greatly improved.

Description

Technical Field

[0001] The present invention relates to a method and system for optimizing fan blades, and particularly to a method and system for optimizing general fan blades by coupling CFD and AI. Background Art

[0002] With the adjustment of the global energy structure, renewable energy has attracted much attention due to its advantages of being clean and environmentally friendly. Among them, wind energy, as a clean energy with great potential, has been highly regarded by various countries. Fans are important devices for efficient conversion of wind energy, and blades, as the key components for capturing wind energy by fans, optimizing the blade shape is of great significance for improving the efficiency of fans.

[0003] Traditional fan blade design mainly relies on experience accumulation and the trial-and-error method, with low optimization efficiency. It usually depends on trial calculations and iterations, involving a large amount of calculation and long time consumption, and it is difficult to achieve personalized optimization for complex actual working conditions. At the same time, the traditional scheme lacks effective consideration of the randomness and variability of the wind field, resulting in poor optimization effects and difficulty in meeting the actual application requirements. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide a method for optimizing general fan blades by coupling CFD and AI to improve the wind capture performance and design efficiency of fans, thereby reducing costs and optimizing the design scheme. On the other hand, a system for optimizing general fan blades by coupling CFD and AI is provided.

[0005] Technical Solution: A method for optimizing general fan blades by coupling CFD and AI according to the present invention includes the following steps:

[0006] (1) Select the leaf shape to be optimized, and determine the design space and the shape of the fan blade backbone line;

[0007] (2) Use the Latin hypercube sampling method to generate several alternative fan blade design schemes within the design space;

[0008] (3) Construct a CFD-PINN model to simulate the wind capture performance of fans with different alternative design schemes, and generate fan wind capture performance indicators. The CFD-PINN model is based on CFD of PINN coupled with SST k-w;

[0009] (4) Save the fan blade design scheme and the fan wind capture performance indicators into the CFD simulation scheme library;

[0010] (5) Based on the alternative fan blade design schemes and fan wind capture performance indicators in the CFD simulation scheme library, establish an adaptive multi-core Gaussian process regression AMGPR model, and use the improved wave search algorithm IWSA to optimize the hyperparameters of the AMGPR model;

[0011] (6) Calculate the wind capture performance index corresponding to the optimized blade design scheme using the CFD-PINN model;

[0012] (7) Compare the deviation between the wind capture performance index predicted by the AMGPR model and the wind capture performance index simulated by the CFD-PINN model. If the end condition is met, save the hyperparameters of the AMGPR model. If not, continue to optimize until the required conditions are reached.

[0013] Preferably, the blade shape in step 1 includes an S-shaped fan blade, an airfoil blade, and a turbine blade. The S-shaped fan blade determines the shape of the blade backbone line using a Bezier curve. The airfoil blade determines the shape of the blade backbone line using a CST function. The turbine blade determines the shape of the blade backbone line using B-Spline characterization.

[0014] Preferably, the wind capture performance index of the fan in step 3 includes the average moment coefficient, pressure distribution, and velocity distribution.

[0015] Preferably, the specific steps of step 3 are as follows:

[0016] (31) Determine the geometric parameters to be optimized for the fan blade, perform geometric construction and adaptive mesh division;

[0017] (32) Set the physical properties, initial conditions, and boundary conditions, and perform CFD simulation based on PINN coupled with SST k-w;

[0018] (33) Output the simulation results corresponding to specific geometric parameters and calculate the wind capture performance index.

[0019] Preferably, the specific calculation formula of step 32 is as follows:

[0020]

[0021] minξ = ψ 1 ξ data +ψ 2 ξ ph +ψ 3 ξ d ;

[0022] para = {σ k1 ,σ ω1 ,β 1 ,σ k2 ,σ ω2 ,β 2 ,β * ,κ,α 1};

[0023]

[0024] ξ ph =ξPDE +ξ IC +ξ BC ;

[0025] ξ d = 0.5KL(Dis CFD ||Dis exp ));

[0026] Among them, ρ, u, k, ω, and P represent the pressure, velocity, turbulent kinetic energy, unit turbulent kinetic energy dissipation, and pressure corresponding to each grid of the model at different times, ψ 1 , ψ 2 and ψ 3 represent the importance parameters for measuring three different loss functions, ξ data represents the eddy viscosity μ t for calculating the data loss function, ξ ph represents the loss function for simplifying the physical mechanism, ξ PDE represents the partial differential equation loss, ξ IC represents the initial condition loss, ξ BC represents the boundary condition loss, ξ d represents the CFD simulation eddy viscosity μ t distribution Dis CFD and the experimental eddy viscosity μ t distribution Dis exp deviation.

[0027] Preferably, the calculation formula of the AMGPR model described in step 5 is as follows:

[0028]

[0029] Among them, X and X' represent the fan design schemes, and λ x represent the variance and length factor of the squared exponential kernel function, represents the hyperparameters of the AMGPR model obtained by optimizing with IWSA, represents the variance of the squared exponential kernel function, λ x represents the length factor of the squared exponential kernel function, and c represent the parameters of the linear kernel function, represents the parameter of the stochastic kernel function, reflecting the noise level, δ ij represents the Kronecker delta function, δ represents the adjustment factor for the wind capture performance deviation, λ 1 , λ 2 , λ 3 represent the importance parameters for the wind capture performance deviation.

[0030] Preferably, the formula for optimizing the hyperparameters of the AMGPR model using the wave search algorithm IWSA described in step 5 is as follows:

[0031]

[0032] W ij = [θ 1j , θ 2j , …, θ pj T ;

[0033]

[0034] W i = W i - αg i ;

[0035] g i = (f(W +εi ) - f(W -εi )) / 2ε;

[0036] Among them, W ij represents the position of the electromagnetic wave particle encoded by the hyperparameters of the AMGPR model, f(W i ) represents the wind capture performance deviation corresponding to the i-th group of hyperparameters, up j and low j represent the upper and lower limits of the search space of the j-th dimensional hyperparameters, f max represents the wind capture performance deviation corresponding to the optimal hyperparameters, f mean represents the mean value of the wind capture performance deviation, represents the updated hyperparameters, W up represents the vector composed of the maximum values in each dimension of the alternative hyperparameter group, W low represents the vector composed of the minimum values in each dimension of the alternative hyperparameter group, σ represents the electromagnetic wave waveform control parameter, m i represents a random number obeying the Gaussian distribution, a column vector arranged in ascending order, W best represents the current optimal hyperparameters, W re represents the goodness and badness matrix, β represents the reflection intensity coefficient, n w2 represents the number of particles simulating the reflected electromagnetic waves, W fit the hyperparameter matrix rearranged in ascending order of fitness value, r represents a random number, ε = 10 -6 , g represents the hyperparameter optimization search direction, and α represents the step size coefficient.

[0037] A CFD and AI coupled general fan blade optimization system according to the present invention includes:

[0038] ​The fan blade airfoil decision-making module is used to select the airfoil to be optimized and the geometric dimension range, determine the design space, and determine the shape of the fan blade backbone curve according to the representation of the S-shaped fan blade using Bezier curves, the airfoil blade using CST functions, and the turbine blade using B-Spline;

[0039] The lightweight CFD-PINN module is used to simulate the wind-catching performance of the fan for different alternative design schemes, generate multi-dimensional fan wind-catching performance indicators, and save the design schemes and the corresponding multi-dimensional fan wind-catching performance indicators in the CFD simulation scheme library;

[0040] The design scheme and performance index library module specifically includes a CFD simulation scheme library, an AI intelligent prediction library, a turbulence model hyperparameter library, and an AI model hyperparameter library. The CFD simulation scheme library is used to store the airfoil design schemes and the corresponding wind-catching performance indicators of the CFD-PINN simulation. The AI intelligent prediction library is used to store the airfoil design schemes and the corresponding wind-catching performance indicators predicted by the AMGPR model. The turbulence model hyperparameter library is used to store the optimal turbulence model hyperparameters identified under different working conditions of the CFD-PINN;

[0041] The AI intelligent rapid design module is used to construct an AMGPR model, use the data in the CFD simulation scheme library as the modeling input, optimize the AMGPR model hyperparameters using IWSA, save the optimal AMGPR model hyperparameters in the AI model hyperparameter library, and use them to predict the wind-catching performance indicators corresponding to the new design scheme of the fan blade, store them in the AI intelligent prediction library, and store the optimal AMGPR model hyperparameters.

[0042] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the CFD and AI coupled general fan blade optimization method according to any one of claims 1 to 7.

[0043] An electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, it implements the CFD and AI coupled general fan blade optimization method according to any one of claims 1 to 7.

[0044] Advantages: Compared with the prior art, the present invention has the following remarkable advantages: 1. Using the CFD-PINN model to simulate the wind-catching performance of the fan for different design schemes, improving the accuracy of turbulence simulation; 2. Constructing the IWSA-AMGPR model, using the improved wave search algorithm to optimize the hyperparameters of the Gaussian process regression model, and quickly predicting the wind-catching performance of the fan blade; 3. Establishing a cloud platform CFD simulation scheme library to store different design schemes and the corresponding simulation data and performance indicators, facilitating users to query and compare. Brief Description of the Drawings

[0045] Figure 1 Schematic diagram of the optimization method for the general fan blade of the present invention;

[0046] Figure 2 Schematic diagram of the polygon mesh divided for the airfoil fan blade of the present invention. Detailed implementation manners

[0047] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0048] (1) Select the airfoil to be optimized, determine the design space and the shape of the fan blade backbone line, where the S-shaped fan blade is characterized by a Bezier curve, the airfoil blade is characterized by a CST function, and the turbine blade is characterized by a B-Spline;

[0049] (2) Use the Latin hypercube sampling method to generate several alternative fan airfoil design schemes within the design space;

[0050] (3) Construct a CFD-PINN model to simulate the wind capture performance of the fan for different alternative design schemes, and generate the wind capture performance index of the fan. The CFD-PINN model is based on CFD of PINN coupled with SST k-w;

[0051] 1. Determine the geometric parameters to be optimized for the fan blade, including the fan rotor diameter, the fan blade diameter, the moving domain diameter, the overlap ratio, the static domain length, the static domain width, and the number of blades;

[0052] 2. Geometric construction and adaptive mesh division;

[0053] 3. Set physical properties, initial conditions, and boundary conditions;

[0054] 4. CFD simulation based on PINN coupled with SST k-w. The calculation formula of the SST k-w turbulence model is as follows:

[0055]

[0056] μ eff,k = μ l + σ k μ t ;

[0057] μ eff,ω = μ l + σ ω μ t ;

[0058]

[0059] φ = F 1 φ 1 +(1 - F 1 )φ 2 ;

[0060] F 1 = tanh(arg 1 4 );

[0061]

[0062] F 2 = tanh(arg 2 2 );

[0063]

[0064] Among them, ρ represents density, u represents velocity, k represents turbulent kinetic energy, ω represents unit turbulent kinetic energy dissipation, Ω represents vorticity function, P represents pressure, S k represents the source term of the k equation, S ω represents the source term of the ω equation, F 1 and F 2 represent the blending function, which is related to the distance from the grid point to the wall;

[0065] γ 1 represents the first set of parameters of the SST k-ω turbulence model, γ 2 represents the second set of parameters of the SST k-ω turbulence model, β 1 , β 2 , σ ω1 , σ ω2 , k 2 , α 1 , β * , σ k1 and σ k2 represent the nine empirical parameters of the SST-ω turbulence model. The nine empirical parameters are dynamically identified using the PINN model, and then μ t is substituted into the NS equation to calculate the moment coefficient, pressure distribution, and velocity distribution of the entire research area;

[0066] The calculation formula for dynamically identifying the nine empirical parameters of the SST k-ω turbulence model using the PINN model is:

[0067]

[0068] minξ = ψ 1 ξ data + ψ 2 ξ ph + ψ 3 ξ d ;

[0069] para = {σ k1 , σ ω1 , β 1, σ k2 , σ ω2 , β 2 , β * , κ, α 1};

[0070]

[0071] ξ ph = ξ PDE + ξ IC + ξ BC ;

[0072] ξ d = 0.5KL(Dis CFD ||Dis exp );

[0073] Among them, ρ, u, k, ω, P represent the pressure, velocity, turbulent kinetic energy, unit turbulent kinetic energy dissipation, and pressure corresponding to each grid of the model at different times, ψ 1 , ψ 2 and ψ 3 represent the importance parameters for measuring three different loss functions, ξ data represents the eddy viscosity μ t for calculating the data loss function, ξ ph represents the loss function for simplifying the physical mechanism, ξ PDE represents the partial differential equation loss, ξ IC represents the initial condition loss, ξ BC represents the boundary condition loss, ξ d represents the CFD simulation eddy viscosity μ t distribution Dis CFD and the experimental eddy viscosity μ t distribution Dis exp deviation;

[0074] 5. Output the simulation results corresponding to specific geometric parameters and statistically calculate the wind capture performance indicators;

[0075] (4) Save the fan blade shape design scheme and the fan wind capture performance indicators into the CFD simulation scheme library;

[0076] Based on the alternative fan blade shape design schemes and fan wind capture performance indicators in the CFD simulation scheme library, establish an Adaptive Multi-kernel Gaussian Process Regression (AMGPR) model, and use the Improved Wave Search Algorithm (IWSA) to optimize the hyperparameters of the AMGPR model. The calculation formula of the AMGPR model is as follows:

[0077]

[0078] Among them, X and X' represent the fan design schemes, and λ xrepresent the variance and length scale of the squared exponential kernel function represent the hyperparameters of the AMGPR model optimized by IWSA represent the variance of the squared exponential kernel function, λ x represent the length scale of the squared exponential kernel function and c represent the parameters of the linear kernel function represent the parameter of the stochastic kernel function, reflecting the noise level, δ ij represent the Kronecker delta function, δ represents the adjustment factor for the wind capture performance deviation, λ 1 、λ 2 、λ 3 represent the importance parameter of the wind capture performance deviation

[0079] (5) Calculate the wind capture performance index corresponding to the optimized blade shape design scheme using the CFD-PINN model, and the hyperparameters of the AMGPR model The optimization process simulates the behavior of a radar by transmitting radio waves, receiving the reflected echoes, and processing and analyzing the echoes to simulate the hyperparameter optimization process. The specific steps are as follows:

[0080] 1. Encode the hyperparameters of the AMGPR model as the position of the electromagnetic wave particle W ij , and use f(W i ) to represent the wind capture performance deviation corresponding to the i-th group of hyperparameters

[0081]

[0082] W ij =[θ 1j ,θ 2j ,…,θ pj T ;

[0083] where up j and low j represent the upper and lower limits of the search space for the j-th dimensional hyperparameter

[0084] 2. Optimize the hyperparameters over the entire search space:

[0085]

[0086] where f mean represents the mean wind capture performance deviation represents the updated hyperparameters, W up represents the vector composed of the maximum values in each dimension of the alternative hyperparameter group, W low represents the vector composed of the minimum values in each dimension of the alternative hyperparameter group, r 1 represents a random number

[0087] ​3. Electromagnetic wave transmission process:

[0088] Emission process: The hyperparameters are diffused outward in the form of electromagnetic waves, and it is necessary to ensure that the deviation of the wind-catching performance of the new hyperparameters during diffusion shall not be greater than the wind-catching performance deviation f corresponding to the current optimal hyperparameters max ;

[0089]

[0090] Among them, σ represents the electromagnetic wave waveform control parameter, m i represents a random number obeying the Gaussian distribution, a column vector arranged in ascending order, W best represents the current optimal hyperparameters, W re represents the goodness matrix, and the hyperparameter matrix after arranging W in ascending order of the degree of closeness to W best ;

[0091] Reflection process: The hyperparameters with lower wind-catching performance deviation are reflected towards the optimal hyperparameters W best ; The hyperparameters with larger wind-catching performance deviation move away from the optimal hyperparameters W best and expand outward. The calculation formula is:

[0092]

[0093] Among them, β represents the reflection intensity coefficient, r 2 represents a random value from 0 to 1, n w2 represents the number of particles simulating the reflected electromagnetic waves, W fit is the hyperparameter matrix rearranged in ascending order of fitness value;

[0094] Receiving process: The hyperparameters search in the optimal direction and are interfered with with a certain probability. The calculation formula:

[0095]

[0096] Parameter update and search direction:

[0097] W i = W i - αg i ;

[0098] g i = (f(W +εi ) - f(W -εi )) / 2ε;

[0099] Among them, ε = 10 -6 , g represents the hyperparameter optimization search direction, and α represents the step size coefficient;

[0100] The calculation formula for the wind-catching performance deviation is:

[0101] min Error = λ 1 Hb Cm +λ 2 Hb press +λ 3 Hb speed ;

[0102]

[0103] Among them, Error represents the total wind capture performance deviation, Hb Cm 、Hb press 、Hb speed represent the blade moment coefficient, pressure distribution error and velocity distribution error respectively, δ represents the wind capture performance deviation adjustment factor, λ 1 、λ 2 、λ 3 are the importance parameters of the wind capture performance deviation.

[0104] (6) Compare the deviation between the wind capture performance index predicted by the AMGPR model and the wind capture performance index simulated by the CFD - PINN model. If the end condition is met, save the hyperparameters of the AMGPR model. If not, continue to optimize until the required conditions are reached.

Claims

1. A CFD and AI coupled general wind turbine blade optimization method, characterized in that: The following steps are involved: (1) Select the blade shape to be optimized, determine the design space and the blade bone line shape; (2) Generate several alternative fan blade design schemes in the design space using the Latin hypercube sampling method; (3) constructing a CFD-PINN model to simulate the wind capture performance of the fan under different alternative design schemes and generate wind capture performance indicators of the fan, wherein the CFD-PINN model is based on the CFD of PINN coupled SSTk-w; (4) Saving the fan blade design scheme and the fan wind capture performance index into the CFD simulation scheme library; (5) Based on the alternative fan blade design schemes and fan wind capture performance indicators in the CFD simulation solution library, an adaptive multi-core Gaussian process regression AMGPR model is established, and the wave search algorithm IWSA is used to optimize the AMGPR model hyperparameters; (6) using the CFD-PINN model to calculate the wind-catching performance index corresponding to the optimized blade design; (7) Compare the deviations between the wind capture performance indicators predicted by the AMGPR model and those simulated by the CFD-PINN model. If the end conditions are met, save the AMGPR model hyperparameters. If not, continue to optimize until the required conditions are met.

2. The fan blade optimization method according to claim 1, characterized in that: The blade shape in step 1 includes an S-shaped fan blade, an airfoil blade and a turbine blade. The S-shaped fan blade uses a Bezier curve to determine the fan blade bone line shape, the airfoil blade uses a CST function to determine the fan blade bone line shape, and the turbine blade uses a B-Spline characterization to determine the fan blade bone line shape.

3. The fan blade optimization method according to claim 1, characterized in that: The wind-catching performance indicators of the fan described in step 3 include average torque coefficient, pressure distribution, and speed distribution.

4. The fan blade optimization method according to claim 1, characterized in that: Step 3 The specific steps are as follows: (31) Determine the geometric parameters of the fan blade to be optimized, geometry construction and adaptive meshing; (32) Set physical properties, initial conditions, boundary conditions, and CFD simulation based on PINN coupled SSTk-w; (33) Output the simulation results corresponding to specific geometric parameters and calculate the wind capture performance indicators.

5. The fan blade optimization method according to claim 4, characterized in that: The specific calculation formula of step 32 is as follows: minξ=ψ1ξ data +ψ2ξ ph +ψ3ξ d ; para={σ k1 ,s ω1 ,β1,σ k2 ,s ω2 ,β2,β * ,k,a1}; x ph =ξ PDE +ξ IC +ξ BC ; ξ d =0.5KL(Dis CFD ||Dis exp ); Among them, ρ, u, k, ω, and P represent the pressure, velocity, turbulent kinetic energy, unit turbulent kinetic energy dissipation, and pressure of each grid at different times in the model. ψ1, ψ2, and ψ3 represent the importance parameters of three different loss functions. ξ data represents the eddy viscosity μ t Calculate the data loss function, ξ ph represents the physical mechanism simplified loss function, ξ PDE represents the partial differential equation loss, ξ IC represents the initial condition loss, ξ BC represents the boundary condition loss, ξ d Indicates the CFD simulation eddy viscosity μ t Distribution CFD and experimental eddy viscosity μ t Distribution exp deviation.

6. The fan blade optimization method according to claim 1, characterized in that: The calculation formula of the AMGPR model in step 5 is as follows: Among them, X and X' represent the fan design scheme, and λ x represents the variance and length factor of the squared exponential kernel function, represents the AMGPR model hyperparameters obtained by optimizing using IWSA. represents the variance of the squared exponential kernel function, λ x represents the length factor of the squared exponential kernel function, and c represent the parameters of the linear kernel function, Represents the parameters of the random kernel function, reflecting the noise level, δ ij represents the Kronecker delta function, δ represents the wind-catching performance deviation adjustment factor, and λ1, λ2, and λ3 represent the wind-catching performance deviation importance parameters.

7. The fan blade optimization method according to claim 1, characterized in that: The formula for optimizing the hyperparameters of the AMGPR model using the wave search algorithm IWSA described in step 5 is as follows: W ij =[θ 1j ,i 2j ,…,θ pj ] T ; IN i =In i -αg i ; g i =f(W +εi )-f(W -εi )) / 2ε; Among them, W ij represents the electromagnetic wave particle position encoded by the AMGPR model hyperparameter, f(W i ) represents the wind-catching performance deviation corresponding to the i-th group of hyperparameters, up j and low j represents the upper and lower limits of the j-th dimension hyperparameter search space, f max represents the wind-catching performance deviation corresponding to the optimal hyperparameters, f mean represents the mean deviation of wind-catching performance, represents the updated hyperparameters, W up Represents the vector composed of the maximum values ​​in each dimension of the candidate hyperparameter group, W low represents the vector composed of the minimum values ​​in each dimension of the candidate hyperparameter group, σ represents the electromagnetic wave waveform control parameter, m i Represents random numbers that follow a Gaussian distribution, a column vector arranged in ascending order, W best represents the current optimal hyperparameter, W re represents the quality matrix, β represents the reflection intensity coefficient, n w2 Represents the number of particles simulating reflected electromagnetic waves, W fit The hyperparameter matrix is ​​rearranged in order of fitness value from small to large, r represents a random number, ε = 10 -6 , g represents the search direction of hyperparameter optimization, and α represents the step size coefficient.

8. A CFD and AI coupled universal fan blade optimization system, characterized in that: include: The fan blade shape decision module is used to select the blade shape to be optimized, the geometric size range, determine the design space, and determine the fan blade bone line shape based on the Bezier curve for S-type fan blades, the CST function for airfoil blades, and the B-Spline for turbine blades; Lightweight CFD-PINN module, used to simulate the wind-capturing performance of different alternative design schemes, generate multi-dimensional wind-capturing performance indicators of the fan, and the design schemes and corresponding multi-dimensional wind-capturing performance indicators of the fan are saved in the CFD simulation scheme library; The design scheme and performance index library module specifically includes a CFD simulation scheme library, an AI intelligent prediction library, a turbulence model hyperparameter library and an AI model hyperparameter library. The CFD simulation scheme library is used to store the blade shape design schemes and corresponding wind-catching performance indicators simulated by CFD-PINN. The AI ​​intelligent prediction library is used to store the blade shape design schemes and corresponding wind-catching performance indicators predicted by the AMGPR model. The turbulence model hyperparameter library is used to store the optimal turbulence model hyperparameters identified by CFD-PINN under different working conditions; The AI ​​intelligent rapid design module is used to build the AMGPR model, use the data in the CFD simulation solution library as the modeling input, and use IWSA to optimize the AMGPR model hyperparameters. The optimal AMGPR model hyperparameters are saved in the AI ​​model hyperparameter library and used to predict the wind capture performance indicators corresponding to the new design scheme of the fan blade shape, which are stored in the AI ​​intelligent prediction library to store the optimal AMGPR model hyperparameters.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the CFD and AI coupled universal wind turbine blade optimization method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the CFD and AI coupled universal wind turbine blade optimization method according to any one of claims 1 to 7 is implemented.

Images