Real-time hybrid test method and system for ai-augmented floating wind turbine aerodynamic loads

Abstract

The application discloses a kind of based on AI enhancement's floating fan aerodynamic load real-time hybrid test method and system, comprising: based on real-time floating foundation dynamics equation, six-degree-of-freedom motion information of floating foundation is calculated using numerical calculation model;Scale fan model is carried on six-degree-of-freedom motion platform, is placed in wind tunnel atmospheric boundary layer, six-degree-of-freedom motion platform is controlled based on six-degree-of-freedom motion information of floating foundation, drive scale fan model motion, generate aerodynamic load in simulated wind environment, measure original load signal and carry out correction, six-degree-of-freedom platform actual pose is recorded simultaneously;AI correction is carried out to the corrected load signal and six-degree-of-freedom platform actual pose, obtain net aerodynamic load, feedback net aerodynamic load to numerical calculation model, closed loop updates are realized to achieve coupling cycle.The application solves the numerical and physical boundary interaction problem, and can significantly improve the test precision of large floating fan under complex environmental load.

Description

Technical Field

[0001] This invention belongs to the field of floating wind power technology and relates to a real-time hybrid test method and system for aerodynamic loads of floating wind turbines based on AI enhancement. It is mainly applied to the high-precision verification of aerodynamic-servo-hydraulic-mooring fully coupled systems of large floating wind turbines under different operating conditions. Background Technology

[0002] Floating wind turbines are a key technology for the development of offshore wind power in deep water areas, but their complex coupled dynamic characteristics (aerodynamic-servo-hydrodynamic-mooring) pose significant challenges to traditional testing methods. In the field of floating wind turbine design and analysis, existing numerical simulation methods struggle to simultaneously address aerodynamic-hydrodynamic coupled responses and large-scale time-domain response analysis. Numerical simulation alone is insufficient to interpret the complex multiphysics characteristics of offshore wind power across different spatial and temporal scales. With the rapid development of offshore wind power, turbine designs are becoming increasingly larger and more complex. The bending-torsional coupling effect of long, flexible blades in large turbines, the axial expansion and contraction effect under large deformations, blade flutter, and the in-plane negative damping effect of the impeller are becoming more pronounced. These effects lead to structural fatigue and increased difficulty in servo control, necessitating more accurate acquisition of turbine aerodynamic loads and consideration of the turbine's aeroelastic response characteristics. Design methods based on engineering experience and predictive numerical tools require continuous calibration based on experimental data to ensure their reliability and effectiveness. Full-scale experiments are not feasible due to limitations in experimental conditions and costs. Therefore, accurate and reliable wind tunnel experiments on scaled-down models of large wind turbines can not only verify wind turbine parameters but also further improve the reliability of the design.

[0003] Current wind tunnel testing techniques for scaled-down models of large wind turbines still have many shortcomings. While pure physical model tests can reflect the real environment, they suffer from aerodynamic load simulation distortion due to the scaling effect. Most mainstream hybrid testing methods currently employ unidirectional coupling, applying only numerically calculated loads to the physical model for simple numerical verification, without establishing a real-time feedback loop for the physical response, thus preventing dynamic correction of the hydrodynamic model. Furthermore, existing aerodynamic load calculations largely rely on direct measurement using six-component force sensors; however, sensor noise and the response delay of the six-degree-of-freedom platform lead to load decoupling errors, failing to meet high-precision requirements.

[0004] Based on the above considerations, there is an urgent need to develop a bidirectional real-time coupled hybrid test method that dynamically corrects the numerical model using the corrected physical response data, breaks through the bottlenecks of scaling effect and nonlinear load simulation, and provides a highly reliable test verification method for the design of large floating wind turbines. Summary of the Invention

[0005] To address the challenges of real-time correction limitations in unidirectional coupling and aerodynamic load measurement distortion caused by traditional HIL hardware delays in existing technologies, this invention provides an AI-enhanced method and system for real-time hybrid aerodynamic load testing of floating wind turbines. This method offers greater accuracy in multi-physics field verification of large floating wind turbines and more efficient iterative optimization.

[0006] The technical solution adopted in this invention is as follows:

[0007] A real-time hybrid test method for aerodynamic loads on a floating wind turbine based on AI enhancement includes the following steps:

[0008] Based on the real-time floating foundation dynamic equations, a numerical calculation model is used to simulate the six-degree-of-freedom motion of the floating body, wave loads, and mooring system response, and to calculate the six-degree-of-freedom motion information of the floating foundation.

[0009] A scaled-down wind turbine model is mounted on a six-degree-of-freedom motion platform and placed in the atmospheric boundary layer of a wind tunnel. The six-degree-of-freedom motion platform is controlled by acquiring posture commands based on the six-degree-of-freedom motion information of the floating foundation, driving the scaled-down wind turbine model to move. Aerodynamic loads are generated in the simulated wind environment, the original load signals are measured and corrected, and the actual posture of the six-degree-of-freedom platform is recorded simultaneously.

[0010] AI correction is performed on the corrected load signal and the actual pose of the six-degree-of-freedom platform. The net aerodynamic load is obtained based on the correction results. The net aerodynamic load is fed back to the numerical calculation model, and the closed-loop update realizes the coupled cycle.

[0011] Furthermore, the wave load includes hydrodynamic load and hydrostatic restoring force. The hydrodynamic load consists of radiation force, viscous force and diffraction force. The hydrostatic restoring force is solved using a linearized model. The mooring system adopts a lumped mass model and solves for mooring tension through real-time integration.

[0012] Furthermore, the scaled-down wind turbine model tower base is equipped with a six-component force sensor to measure the original load signal, and the nacelle is equipped with an acceleration sensor to measure the hub center height acceleration. The original load signal is corrected based on the hub center height acceleration.

[0013] Furthermore, the AI ​​correction includes noise suppression of the corrected load signal and delay compensation of the actual pose of the six-degree-of-freedom platform.

[0014] Furthermore, the specific steps of noise suppression include: decomposing the corrected load signal using wavelet packet transform, automatically identifying the noise-dominant frequency band and applying adaptive threshold filtering to obtain the noise-reduced load signal.

[0015] Furthermore, the specific steps of the delay compensation include: inputting the actual pose sequence of the six-degree-of-freedom platform, learning the motion features of the six-degree-of-freedom motion platform based on the LSTM network model, and outputting the compensated pose of the six-degree-of-freedom platform.

[0016] Furthermore, obtaining the net aerodynamic load based on the correction result specifically includes: calculating the compensation force based on the compensated pose of the six-degree-of-freedom platform, and calculating the net aerodynamic load based on the load signal after noise reduction of the compensation force.

[0017] A real-time hybrid test system for aerodynamic loads of a floating wind turbine based on AI enhancement, used to implement the above method, includes:

[0018] Numerical Subsystem: Based on the real-time floating foundation dynamic equations, it uses a numerical calculation model to simulate the six-degree-of-freedom motion of the floating body, wave loads, and mooring system response, calculates the six-degree-of-freedom motion information of the floating foundation, and generates pose commands based on the six-degree-of-freedom motion information of the floating foundation.

[0019] The physics subsystem is used to mount the scaled-down wind turbine model on a six-degree-of-freedom motion platform and place it in the atmospheric boundary layer of the wind tunnel. It controls the six-degree-of-freedom motion platform according to the pose command, drives the scaled-down wind turbine model to move, generates aerodynamic loads in the simulated wind environment, measures and corrects the original load signals, and records the actual pose of the six-degree-of-freedom platform simultaneously.

[0020] AI correction module: used to perform AI correction on the corrected load signal and the actual pose of the six-degree-of-freedom platform, obtain the net aerodynamic load based on the correction result, feed the net aerodynamic load back to the numerical calculation model, and realize the coupled loop through closed-loop update.

[0021] A computer device, the computer device comprising:

[0022] One or more processors;

[0023] Memory, used to store one or more programs;

[0024] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-described AI-enhanced real-time hybrid test method for aerodynamic loads on floating wind turbines.

[0025] A computer-readable storage medium storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform the steps in the method described above.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0027] The AI-enhanced real-time hybrid test method and system for aerodynamic loads of floating wind turbines provided by this invention can be used for high-precision verification of aerodynamic-servo-hydraulic-mooring fully coupled floating wind turbines under different operating conditions. It constructs a fully closed-loop bidirectional coupling system, that is, establishes a bidirectional real-time feedback channel of "numerical command → physical execution" and "physical response → numerical model correction", dynamically iterates until convergence, breaks through the limitations of traditional unidirectional coupling, and significantly improves the test fidelity. AI correction solves the problems of sensor noise and hardware delay, continuously reduces the RMSE of aerodynamic load decoupling, and significantly improves data reliability. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the method flow in an embodiment of the present invention.

[0029] Figure 2 This is a schematic diagram of the physical subsystem principle in an embodiment of the present invention.

[0030] Figure 3 This is a schematic diagram of the principle of the numerical subsystem in an embodiment of the present invention.

[0031] Figure 4 This is a schematic diagram of the AI ​​correction module in an embodiment of the present invention.

[0032] Figure 5 This is a schematic diagram of the noise suppression processing unit in an embodiment of the present invention.

[0033] Figure 6 This is a schematic diagram of the delay compensation unit in an embodiment of the present invention.

[0034] Figure 7 This is a comparison chart showing the difference in dynamic response of a floating wind turbine platform under surge conditions between the AI-enhanced real-time hybrid test method for aerodynamic loads of a floating wind turbine (HIL-AI) and the FAST numerical model in this embodiment of the invention. Detailed Implementation

[0035] The technical solution of the present invention will be further described clearly and in detail below with reference to the accompanying drawings and specific embodiments.

[0036] A real-time hybrid test method for aerodynamic loads on a floating wind turbine based on AI enhancement includes the following steps:

[0037] Based on the real-time floating foundation dynamic equations, a numerical calculation model is used to simulate the six-degree-of-freedom motion of the floating body, wave loads, and the mooring system response, calculating the six-degree-of-freedom motion information of the floating foundation. The wave loads include hydrodynamic loads and hydrostatic restoring forces. The hydrodynamic loads consist of radiation forces, viscous forces, and diffraction forces, while the hydrostatic restoring forces are solved using a linearized model. The mooring system employs a lumped mass model, and the mooring tensions are solved through real-time integration.

[0038] A scaled-down wind turbine model was mounted on a six-degree-of-freedom (DOF) motion platform and placed in the atmospheric boundary layer of a wind tunnel. Position and attitude commands were acquired based on the six-DOF motion information from the floating foundation to control the six-DOF motion platform, driving the scaled-down wind turbine model's motion and generating aerodynamic loads in a simulated wind environment. Six force sensors were installed at the base of the scaled-down wind turbine model to measure the original load signal, and an acceleration sensor was installed in the nacelle to measure the hub center height acceleration. The original load signal was corrected based on the hub center height acceleration, and the actual position and attitude of the six-DOF platform were recorded simultaneously.

[0039] AI correction is performed on the corrected load signal and the actual pose of the six-degree-of-freedom platform. The net aerodynamic load is obtained based on the correction results. The net aerodynamic load is fed back to the numerical calculation model, and the closed-loop update realizes the coupled cycle.

[0040] The AI ​​correction includes noise suppression of the corrected load signal and delay compensation of the actual pose of the six-degree-of-freedom platform. The noise suppression steps specifically include: decomposing the corrected load signal using wavelet packet transform, automatically identifying the noise-dominant frequency band, and applying adaptive threshold filtering to obtain the denoised load signal. The delay compensation steps specifically include: inputting the actual pose sequence of the six-degree-of-freedom platform, learning the motion characteristics of the six-degree-of-freedom motion platform based on an LSTM network model, and outputting the compensated pose of the six-degree-of-freedom platform to compensate for the response hysteresis of the six-degree-of-freedom platform. Then, the compensation force is calculated based on the compensated pose of the six-degree-of-freedom platform, and the net aerodynamic load is calculated based on the denoised load signal after compensation force reduction.

[0041] In another embodiment of the present invention, an AI-enhanced real-time hybrid test system for aerodynamic loads of a floating wind turbine is provided to implement the above method, the flowchart of which is shown below. Figure 1 As shown. The system includes a numerical subsystem, a physical subsystem, and an AI correction module. During initialization, the numerical subsystem uses a numerical calculation model to simulate the six-degree-of-freedom motion of the floating body, wave loads, and mooring system response. It solves the six-degree-of-freedom dynamic equations of the floating foundation in real time, outputs the six-degree-of-freedom motion information of the floating foundation, and generates the six-degree-of-freedom platform pose command q. s .

[0042] The six-degree-of-freedom dynamic equations of the floating foundation are:

[0043]

[0044] Among them, F hst For the restoring force of still water, F moor For mooring tension, F hydro For hydrodynamic loads, F aero This is the net aerodynamic load.

[0045] like Figure 3 As shown, the external forces acting on the six-degree-of-freedom platform consist of three parts: still water restoring force, mooring force, and hydrodynamic loads. The still water restoring force is generated by the balance of buoyancy and gravity, and its solution uses a linearized model. The mooring force uses a lumped mass model to discretize the mooring cables, and is solved by coupling it with the platform dynamic model to calculate the anchor chain dynamic response in real time and obtain the mooring tension. The hydrodynamic loads consist of radiation force, viscous force, and diffraction force, and their calculation formulas are as follows:

[0046] F hydro =F rad +F visc +F diff

[0047] Among them, F rad The radiation force is obtained by solving for the platform velocity and delay matrix; F visc For viscous forces, the Morrison equation and other semi-empirical formulas are used for calculation; F diff To determine the diffraction force, the frequency domain transfer function is obtained through pre-simulation using the three-dimensional surface element method, and then solved in real time by combining this with the wave power spectral density. For example... Figure 2 As shown, the physical subsystem uses a scaled-down wind turbine model placed in the atmospheric boundary layer of a wind tunnel, and transmits the pose command q through a six-degree-of-freedom platform. s The spatial motion of the scaled-down wind turbine model is converted into aerodynamic loads in a wind tunnel. Six force sensors are placed at the base of the scaled-down wind turbine model to measure the original load signal F. bal An acceleration sensor was installed in the nacelle of a scaled-down wind turbine model to measure the acceleration 'a' at the center height of the hub. hub Due to weight limitations of the nacelle's mechanical structure, the mass of the blade-hub-nacelle in the scaled-down model will often exceed the theoretical value calculated for scaled-down. This introduces additional inertial forces into the structural load, requiring inertial force correction. The corrected tower base six-component load signal The calculation method is as follows:

[0048]

[0049] in, The raw load signal measured by the six-component force sensor of the tower base. This is the corrected six-component force load signal for the tower base. M is the acceleration vector at the center height of the wheel hub. model M0 and M1 are the actual inertia tensor of the scaled-down wind turbine model and the theoretical inertia tensor calculated based on the scaled-down prototype, respectively.

[0050] The actual pose q of the six-DOF platform is recorded synchronously, and the read data is transmitted to the AI ​​correction module.

[0051] like Figure 4The diagram shown illustrates the principle of the AI ​​correction module. The AI ​​correction module employs multi-source data fusion to synchronously receive the corrected load signal from the physical subsystem. And the actual pose q of the six-degree-of-freedom platform, and then the corrected load signal is processed by a noise suppression unit. Wavelet packet decomposition is performed, and the noise-dominant frequency band is automatically identified and adaptive threshold filtering is applied to obtain the denoised load signal. The displacement at future time Δt is predicted using an LSTM network via a delay compensation unit, generating a compensation pose command to compensate for the response hysteresis of the six-DOF platform, and thus obtaining the compensated pose of the six-DOF platform.

[0052] Figure 5 The diagram shown is a schematic of the noise suppression processing unit in this embodiment. The noise suppression processing uses wavelet packet transform to modify the corrected load signal. Decomposed into 32 sub-bands, the coefficients C of each band are obtained. i (t)(i=1, 2, …, 32); Automatic identification of noise-dominant frequency bands is used to calculate the energy proportion of each frequency band:

[0053]

[0054] Among them, C i (t) is the wavelet packet coefficient sequence of the i-th sub-band, C k (t) is the coefficient sequence of the k-th subband after wavelet packet decomposition.

[0055] The frequency bands are sorted from high to low, and energy accumulation is performed. Where m is the cumulative cutoff frequency band number (accumulating to the m-th frequency band from front to back), j is the frequency band traversal index (the index after sorting by frequency from high to low), and E j This represents the energy percentage of the j-th frequency band after sorting.

[0056] When E cum When (m) > η, the first m high-frequency bands are determined to be noise-dominant bands (η is the energy accumulation threshold, which can be adaptively adjusted according to signal characteristics); adaptive threshold filtering is performed on the noise-dominant bands (threshold). σ is the noise standard deviation, N is the signal length; the output is the noise-reduced load signal F. bal_clean .

[0057] Figure 6 This is a schematic diagram of the delay compensation unit in this embodiment, illustrating the working mechanism of an LSTM-based delay compensation unit for response lag compensation of a six-degree-of-freedom platform. The LSTM network model receives a historical pose sequence [q] containing three time points over the past 2Δt time intervals. t-2Δt ,qt-Δt ,q t As input, the platform's motion features are learned through an LSTM network model to predict its pose at the future time t+Δt. Δt is dynamically determined by the platform's response delay characteristics.

[0058] The core processing of the LSTM network model includes three key steps:

[0059] First, the gating mechanism is calculated, including the forgetting gate f. t =σ(W f ·[h t-1 ,q t ]+b f The percentage of historical information retained is determined (0 = complete forgetting, 1 = complete retention), where σ is the Sigmoid activation function and W... f Here is the weight matrix of the forget gate (f is a special parameter indicating that this parameter belongs to the forget gate), h t-1 The hidden state from the previous moment, q t Let b be the actual pose input vector of the six-DOF platform at the current moment. f Here is the bias vector for the forget gate; the input vector is gate i. t =σ(W i ·[h t-1 ,q t ]+b i ) controls the degree of updating of new information, where W i Let be the weight matrix of the vector input gate (i is a special parameter indicating that this parameter belongs to the vector input gate), b i The bias vector of the input gate; candidate state Where tanh is the hyperbolic tangent activation function, W c Let c be the weight matrix of the candidate state (c is a special parameter indicating that the parameter belongs to the candidate state), and b be the weight matrix of the candidate state. c This is the bias vector for the candidate state. Next is the state update, the cell state. Where f t For the forget gate at the current time step, C t-1 i represents the cell state at the previous time step. t This is the input step for the current time step. Given the candidate cell state at the current time step; integrate historical and current information, output gate o t =σ(W o ·[h t-1 ,q t ]+b o Adjust the output ratio, W o Here, is the weight matrix of the output gate, where 'o' represents a parameter specific to the output gate, and 'b' is the weight matrix. oThe bias vector of the output gate; hidden state Generate the current output (including compressed memory information), where o t C is the output gate for the current time step. t The cell state at the current time step is defined; three time steps are processed cyclically (t-2Δt→t-Δt→t), where Δt is the time step size, and the final output h is calculated. t C t Save this as a new initial state for prediction calculations in subsequent time steps, ultimately yielding the predicted pose. The calculation formula is as follows:

[0060]

[0061] Where g is a nonlinear activation function, W out h is the weight matrix of the output layer (usually a two-dimensional matrix). t To represent the hidden state at the current time step, b out This is the bias vector (one-dimensional vector). Finally, the compensated pose is obtained. It is then passed to the fusion processor.

[0062] The load is decoupled and calculated, and the pose is compensated using a six-degree-of-freedom platform obtained from the delay compensation unit. Calculate the compensating force Among them, [M] t Let be the inertia tensor matrix of the scaled-down wind turbine model. Let [K] be the acceleration vector for compensating the pose of a six-DOF platform. t This is the gravity stiffness matrix of the scaled-down wind turbine model. The denoised load signal F is obtained by incorporating a noise suppression unit. bal_clean The net aerodynamic load F was extracted. aero =F bal_clean -F corr ; the net aerodynamic load F aero Feedback is sent to the numerical subsystem to resolve the platform dynamics equations, completing the real-time coupled loop. The six-DOF platform pose command q... s The calculation formula is as follows:

[0063]

[0064] Where, q s =[x,y,z,ρ,θ,σ] T (x represents sway, y represents yaw, z represents heave, ρ represents roll, θ represents sway, θ represents pitch), F hst For the restoring force of still water, F moor For mooring tension, F hydro For hydrodynamic loads, F aero This is the net aerodynamic load.

[0065] Figure 7 To verify the consistency between the method of the present invention (HIL-AI) and FAST simulation in the dynamic response of surge. By comparing the power spectral density (PSD) of the platform motion under the Operational sea condition (Hs = 7.10 m, Tp = 12.10 s), it is shown that this method can reproduce the wave-dominated dynamic characteristics with high precision. It can be seen from the image that in the low-frequency band (f < 0.01 Hz), the PSD curves of the coupled calculation method (red line) and the numerical simulation (blue line) coincide highly near the natural frequency of the platform surge. The peak position and amplitude error are less than 5%, and this coupled calculation method can accurately reproduce the low-frequency coupling dynamics; in the middle-frequency band (0.1 Hz < f < 0.3 Hz), the trends of the two curves are similar, and the difference in PSD amplitude is less than 10%. Especially near the main wave frequency, the energy matches well, which proves that this coupled calculation method can accurately simulate the linear response under wave excitation and verifies the rationality of HIL-AI.

[0066] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0067] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / blocks Figure 1 one block or multiple blocks.

[0068] These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including instruction means, and the instruction means implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] The above description is merely a preferred embodiment of the present invention. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the technical solutions of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall still fall within the protection scope of the technical solutions of the present invention.

Claims

1. A real-time hybrid test method for aerodynamic loads on a floating wind turbine based on AI enhancement, characterized in that, Includes the following steps: Based on the real-time floating foundation dynamic equations, a numerical calculation model is used to simulate the six-degree-of-freedom motion of the floating body, wave loads, and mooring system response, and to calculate the six-degree-of-freedom motion information of the floating foundation. A scaled-down wind turbine model is mounted on a six-degree-of-freedom motion platform and placed in the atmospheric boundary layer of a wind tunnel. The six-degree-of-freedom motion platform is controlled by acquiring posture commands based on the six-degree-of-freedom motion information of the floating foundation, driving the scaled-down wind turbine model to move. Aerodynamic loads are generated in the simulated wind environment, the original load signals are measured and corrected, and the actual posture of the six-degree-of-freedom platform is recorded simultaneously. AI correction is performed on the corrected load signal and the actual pose of the six-degree-of-freedom platform. The net aerodynamic load is obtained based on the correction results. The net aerodynamic load is fed back to the numerical calculation model, and the closed-loop update realizes the coupled cycle. The AI ​​correction includes noise suppression of the corrected load signal and delay compensation of the actual pose of the six-degree-of-freedom platform. The noise suppression steps include: decomposing the corrected load signal using wavelet packet transform; automatically identifying the noise-dominant frequency band and applying adaptive threshold filtering to obtain the noise-reduced load signal; The specific steps of the delay compensation include: inputting the actual pose sequence of the six-degree-of-freedom platform, learning the motion features of the six-degree-of-freedom motion platform based on the LSTM network model, and outputting the compensated pose of the six-degree-of-freedom platform; wherein, the LSTM network model receives data containing the past two... Historical pose sequence at 3 moments within a time period As input, the platform's motion features are learned through an LSTM network model to predict the future. position t+Δt ,in The response latency is dynamically determined by the platform's response latency characteristics; Net aerodynamic load ,in This is the noise-reduced load signal. For compensating force.

2. The real-time mixed test method for aerodynamic loads of a floating wind turbine based on AI enhancement according to claim 1, characterized in that, The wave loads include hydrodynamic loads and hydrostatic restoring forces. The hydrodynamic loads consist of radiation forces, viscous forces, and diffraction forces. The hydrostatic restoring forces are solved using a linearized model. The mooring system uses a lumped mass model, and the mooring tensions are solved through real-time integration.

3. The real-time mixed test method for aerodynamic loads of a floating wind turbine based on AI enhancement according to claim 1, characterized in that, The scaled-down wind turbine model tower base is equipped with a six-component force sensor to measure the original load signal, and the nacelle is equipped with an acceleration sensor to measure the hub center height acceleration. The original load signal is corrected based on the hub center height acceleration.

4. A real-time hybrid test system for aerodynamic loads of a floating wind turbine based on AI enhancement, characterized in that, For implementing the method as described in any one of claims 1 to 3, comprising: Numerical Subsystem: Based on the real-time floating foundation dynamic equations, it uses a numerical calculation model to simulate the six-degree-of-freedom motion of the floating body, wave loads, and mooring system response, calculates the six-degree-of-freedom motion information of the floating foundation, and generates pose commands based on the six-degree-of-freedom motion information of the floating foundation. The physics subsystem is used to mount the scaled-down wind turbine model on a six-degree-of-freedom motion platform and place it in the atmospheric boundary layer of the wind tunnel. It controls the six-degree-of-freedom motion platform according to the pose command, drives the scaled-down wind turbine model to move, generates aerodynamic loads in the simulated wind environment, measures and corrects the original load signals, and records the actual pose of the six-degree-of-freedom platform simultaneously. AI correction module: used to perform AI correction on the corrected load signal and the actual pose of the six-degree-of-freedom platform, obtain the net aerodynamic load based on the correction result, feed the net aerodynamic load back to the numerical calculation model, and realize the coupled loop through closed-loop update.

5. A computer device, characterized in that, The computer device includes: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the AI-enhanced real-time hybrid test method for aerodynamic loads of floating wind turbines as described in any one of claims 1 to 3.

6. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed by one or more processors, the one or more processors are caused to perform the steps of the method according to any one of claims 1 to 3.

Images