A floating wind turbine semi-physical model delay compensation closed-loop control method and system

Abstract

The application discloses a kind of floating wind turbine semi-physical model delay compensation closed-loop control method and system, belong to offshore floating wind power semi-physical simulation test technical field;System includes physical submodule, numerical submodule and control submodule;The present application is aimed at the problems of load loading lag, timing mismatch, low simulation accuracy caused by actuator physical delay, through timing prediction model, platform motion response is predicted in advance, target aerodynamic load is calculated in advance to offset hardware delay, high-precision tracking and closed-loop correction of load are realized through composite controller, while continuously optimizing the precision of prediction model;The present application not only avoids the inherent similarity criterion conflict problem of floating wind turbine pool test from the root, but also can improve the real-time performance of floating wind turbine semi-physical simulation and the fidelity of coupled dynamics simulation, and can provide high reliability technical support for floating wind turbine pool model test.

Description

Technical Field

[0001] This invention belongs to the field of semi-physical simulation test technology for offshore floating wind power, and particularly relates to a delayed compensation closed-loop control method and system for a semi-physical model of a floating wind turbine. Background Technology

[0002] Deep-sea floating wind power has become the core direction for the large-scale development of renewable energy. Wave pool testing is a key means to explore the dynamic characteristics of the coupling of wind turbine aerodynamics, platform hydrodynamics, and mooring system.

[0003] Traditional all-physical pool tests have inherent, irreconcilable flaws. Hydrodynamic simulation of floating platforms must adhere to the Froude similarity criterion, while aerodynamic load simulation of wind turbines must follow the Reynolds similarity criterion. These two criteria have completely contradictory requirements regarding scale and test parameters, which cannot be simultaneously satisfied in a single test. This directly leads to distortions in aerodynamic load simulation, large deviations in platform motion response, and the inability to accurately reproduce multi-field coupling characteristics, making it difficult for test data to guide engineering practice. Semi-physical real-time hybrid model tests offer a new approach to solving this problem. By coupling the physical simulation of floating platform motion, numerical calculation of wind turbine aerodynamic loads, and real-time load application by actuators, the conflict of similarity criteria is theoretically avoided. However, this technology is constrained by inherent hardware latency. The cumulative delays in data acquisition and transmission, command processing, and actuator response easily lead to problems such as load loading lag, timing mismatch, and low simulation fidelity, directly reducing the effectiveness of the test.

[0004] However, existing semi-physical real-time hybrid model systems lack standardized design schemes, and most existing technical solutions have certain limitations: existing semi-physical thrust actuators are mostly based on open-loop control, whose control accuracy relies excessively on prior calibration, resulting in weak anti-interference capabilities; closed-loop control schemes mostly use traditional PID single control, which has poor load tracking adaptability and insufficient accuracy; the thrust reproduction of thrust actuators suffers from unavoidable hardware delays, and existing control schemes often employ passive control based on real-time data without an advance prediction mechanism, failing to fundamentally offset hardware delays and making it difficult to meet the requirements of high real-time performance and high fidelity semi-physical simulation. Existing technologies complete experiments by building a multi-fan drive system, modifying the OpenFAST numerical model, and acquiring platform motion in real time while loading aerodynamic loads. However, this technology uses an open-loop control method and does not set up a real-time closed-loop feedback loop for fan thrust or motor speed, making it difficult to adapt to dynamic marine environments, easily causing output deviations from expectations, and reducing experimental reliability.

[0005] Existing technology employs seven sets of variable-speed fans as pneumatic load actuators; it establishes a synchronous acquisition system, combining a visual camera and a six-dimensional force sensor to acquire platform motion and actual load data; and it uses closed-loop control to adjust the motor speed to achieve target load tracking. However, this method uses traditional PID single control, lacks adaptive parameter adjustment capability, has insufficient load tracking accuracy, and does not consider the inherent physical delays of the actuators and data transmission. It relies solely on real-time acquired data for passive feedback, which easily leads to load loading lag and timing mismatch problems.

[0006] Therefore, there is an urgent need for a delay-compensated closed-loop control technology for semi-physical real-time hybrid models to overcome the aforementioned core bottlenecks and improve the reliability of experimental data and its engineering application value. Summary of the Invention

[0007] The purpose of this invention is to provide a semi-physical model delay compensation closed-loop control method and system for floating wind turbines, which solves the technical defects in the prior art, such as load lag, timing mismatch, and insufficient control accuracy caused by the physical delay of the actuator.

[0008] To achieve the above objectives, this invention provides a semi-physical model delay compensation closed-loop control method for floating wind turbines, comprising the following steps:

[0009] S1. Real-time acquisition of motion time-series data generated by the floating platform based on wave excitation;

[0010] S2. Based on historical data and real-time collected platform motion time series data, a time series prediction model is constructed to predict the motion response of the floating platform in advance.

[0011] S3. Based on the advanced prediction results of S2, perform aerodynamic load calculations for the full-size wind turbine in advance, and perform scale conversion from the full-size load to the model scale to obtain the target aerodynamic load at the model scale.

[0012] S4. Input the target aerodynamic load obtained in S3 into the composite controller to drive the actuator of the semi-physical fan to output the corresponding load; collect the actual output load of the actuator in real time and feed it back to the composite controller for high-precision load tracking and closed-loop correction.

[0013] S5. Based on the platform motion time series data collected in real time in S1, the training dataset of the time series prediction model is updated synchronously to continuously optimize the prediction accuracy of the time series prediction model.

[0014] S6. After loading, the floating platform generates a new round of motion response, repeating S1 to S5, and performing a real-time closed-loop cycle of the floating wind turbine semi-physical simulation test.

[0015] Preferably, the time series prediction model in S2 adopts an LSTM (Long Short-Term Memory) network model or a Transformer time series prediction model; the model inputs a preset length of historical platform six-degree-of-freedom motion time series data and a real-time acquired motion time series data sequence. The expression is as follows:

[0016] ;

[0017] In the formula, The length of the sliding window; For a moment Observational data;

[0018] The model outputs a sequence of predicted six-DOF motion responses for a floating platform at one or more future time steps. The expression is as follows:

[0019] ;

[0020] In the formula, This is a predicted value; To predict the step size.

[0021] Preferably, in S3, the dynamic link library (DLL) interface of the FAST wind turbine aerodynamic simulation software is called through the Python SDK to input the predicted platform motion response into the FAST software for advance numerical calculation of the full-scale wind turbine aerodynamic load.

[0022] Preferably, S3 is based on the Froude similarity criterion. A scaled-down conversion from full-scale load to model scale is performed to obtain the target aerodynamic load directly used for control. F target (t), the expression is as follows:

[0023] ;

[0024] In the formula, The characteristic velocity of the prototype; The feature length of the prototype; The characteristic velocity of the model; The feature length of the model; It is the acceleration due to gravity; For model-scale target loads; This is a full-size prototype aerodynamic load; It is a geometric scaling ratio;

[0025] Target aerodynamic load for The specific time-varying form in the time dimension.

[0026] Preferably, the composite controller in S4 adopts a fuzzy PID and feedforward composite controller, including a feedforward control channel and a fuzzy PID closed-loop control channel, and the overall control output expression is as follows:

[0027] ;

[0028] In the formula, Forward control variable, where For feedforward gain; For fuzzy PID control;

[0029] Generate feedforward control quantity based on target aerodynamic load. The target load is tracked rapidly through the feedforward control channel;

[0030] Calculate the tracking error between the target aerodynamic load and the actual output load. The expression is as follows:

[0031] ;

[0032] In the formula, This refers to the actual aerodynamic load;

[0033] By using a fuzzy PID closed-loop control channel, the PID control parameters are adaptively adjusted to achieve dynamic correction of the load tracking error, as shown in the following expression:

[0034] ;

[0035] In the formula, This is the proportionality coefficient; The integral coefficient; is the differential coefficient.

[0036] Preferably, in S5, a sliding window mechanism is used to update the training dataset of the time series prediction model online, remove historical data with a time span exceeding a preset length, and simultaneously supplement the platform's six-degree-of-freedom motion time series data collected in real time, so as to continuously iterate and optimize the time series prediction model.

[0037] Preferably, a semi-physical model delay compensation closed-loop control system for a floating wind turbine, applying the above-mentioned semi-physical model delay compensation closed-loop control method for a floating wind turbine, includes: a physical submodule, a numerical submodule, and a control submodule;

[0038] The physics submodule is used to collect the motion timing data generated by the wave-excited floating platform in real time, as well as the actual output load of the semi-physical wind turbine actuator, and at the same time execute the aerodynamic load loading action to drive the floating platform to generate a new round of motion response.

[0039] The control submodule is used to receive real-time platform motion and load data from the physics submodule and interact with the numerical submodule to complete the delay compensation and high-precision closed-loop control of the semi-physical real-time hybrid model of the floating wind turbine.

[0040] The numerical submodule is used to receive the predicted platform motion response in advance, complete the aerodynamic load numerical calculation of the full-size wind turbine in advance, and complete the scale conversion of the full-size load to the model scale, output the target aerodynamic load at the model scale, and offset the inherent physical delay of the actuator.

[0041] The system completes the real-time closed-loop simulation of the floating wind turbine through data interaction between the physical submodule, control submodule, numerical submodule, composite control unit, and physical submodule.

[0042] Preferably, the physical submodule includes a floating platform, a mooring system, data acquisition equipment, and a semi-physical wind turbine actuator;

[0043] The floating platform and mooring system are deployed in a wave pool to generate a coupled dynamic motion response of the floating wind turbine under wave excitation;

[0044] The data acquisition equipment includes an optical six-degree-of-freedom motion acquisition device and a six-dimensional force sensor. The optical six-degree-of-freedom motion acquisition device is used to acquire the six-degree-of-freedom motion time-series data of the floating platform in real time, and the six-dimensional force sensor is used to acquire the actual aerodynamic load output by the semi-physical fan actuator in real time.

[0045] The semi-physical fan actuator is used to output the corresponding aerodynamic load according to the control command of the composite control unit.

[0046] Preferably, the control submodule includes a time-series prediction unit, a composite control unit, and an online model optimization unit;

[0047] The timing prediction unit is used to receive historical and real-time platform motion timing data collected by the physical submodule and predict the floating platform motion response in advance at the next moment.

[0048] The composite control unit is used to receive the target aerodynamic load, generate control commands to drive the actuator of the semi-physical fan to output the corresponding load, receive the actual output load collected by the physical submodule, and complete the closed-loop correction of load tracking.

[0049] The online model optimization unit receives real-time platform motion time-series data collected by the physics submodule, synchronously updates the training dataset of the time-series prediction unit, and continuously optimizes the prediction accuracy of the time-series prediction model.

[0050] Preferably, the semi-physical fan actuator includes three identical radial support rods evenly arranged at 120° circumference and six sets of identical drive actuators; among the six sets of identical drive actuators, D1, D2 and D3 are arranged in the YZ plane of the support rod and perpendicular to the plane, with an arrangement radius of L1, and D4, D5 and D6 are arranged perpendicularly on the support rod in a clockwise direction, with an arrangement radius of L2, and output upward, diagonally upward and diagonally downward thrust respectively.

[0051] Therefore, the beneficial effects of the above-mentioned semi-physical model delay compensation closed-loop control method and system for floating wind turbines are as follows:

[0052] By using a time-series prediction model to anticipate platform motion and preload target loads, hardware delays are offset, thus resolving load lag and timing mismatch issues. Fuzzy PID + feedforward composite control is employed to improve load tracking accuracy, and a sliding window mechanism is used to achieve online optimization of the prediction model, ensuring long-term experimental stability. A real-time closed-loop control system is constructed to significantly improve the real-time performance of semi-physical simulation and the fidelity of coupled dynamics simulation, providing highly reliable technical support for floating wind turbine pool model tests.

[0053] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall process of a semi-physical model delay compensation closed-loop control method for floating wind turbines according to the present invention;

[0055] Figure 2 This is a schematic diagram of the time series prediction model structure and data input / output of the present invention;

[0056] Figure 3 A schematic diagram of the fuzzy PID + feedforward composite controller structure of the invention;

[0057] Figure 4 This is a front view of the semi-physical fan actuator of the present invention;

[0058] Figure 5 This is a side view of the semi-physical fan actuator of the present invention;

[0059] Figure 6 This is an overall layout diagram of the water tank test of the present invention;

[0060] Reference numerals: 1. Semi-physical wind turbine actuator; 2. Six-dimensional force sensor; 3. Tower; 4. Floating platform; 5. Mooring system; 6. Six-degree-of-freedom motion acquisition instrument; 7. Wave pool; 101. Cabin; 102. Carbon fiber support rod; 103. Propeller mount; 104. Propeller; 105. Brushless motor. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0062] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0063] The following is combined Figures 1-6 The embodiments of the present invention will be described in detail below.

[0064] Example 1

[0065] A semi-physical model delay compensation closed-loop control method for floating wind turbines is proposed. A full-scale numerical model of the wind turbine is built in FAST software. The model reads the six-degree-of-freedom motion of the platform, calculates the six-degree-of-freedom load of the wind turbine, and encapsulates it into a dynamic link library (DLL) interface, which can be called externally through Python SDK.

[0066] A semi-physical fan actuator 1 containing 6 sets of drive actuators was constructed, and a six-dimensional force sensor 2 was installed. The specific layout of the actuators was determined based on the requirements of multi-degree-of-freedom load simulation. The drive actuators were calibrated, and the correspondence between the PWM control signal and the output thrust of the actuator was established.

[0067] Physical delay calibration is performed on the semi-physical wind turbine actuator 1. By issuing step control commands and synchronously collecting load response timing data, the inherent delay covering the entire process of command transmission, motor response, thrust establishment and load data feedback is calculated. After taking the average value through multiple parallel tests, the calibration delay duration is pre-stored in the control module for matching the subsequent timing prediction step size.

[0068] Based on Freud's similarity The criteria determine the scale ratio of the water tank test model, fabricate the floating platform 4 and mooring system 5, complete the sample layout, leveling and debugging in the test water tank, and integrate and assemble the semi-physical fan actuator 1, motion acquisition equipment and load acquisition equipment to form a complete water tank test physical system.

[0069] Based on the motion time series dataset generated by the multi-condition simulation platform, an LSTM (Long Short-Term Memory) or Transformer time series prediction model is constructed and pre-trained and its accuracy is verified. The trained model weights are then loaded into the control submodule to complete the pre-training and initialization of the time series prediction model.

[0070] Formal scale-down model test of the water tank was carried out. Through the optical six-degree-of-freedom motion acquisition device and the six-dimensional force sensor 2, the six-degree-of-freedom motion data of the wave-excited floating platform 4 and the actual output load data of the actuator were collected simultaneously. After filtering and noise reduction, standardized time series data were formed.

[0071] Using historical and real-time platform motion data as input, a sliding window input sequence is constructed: The pre-trained temporal model is used to predict the platform's motion response in advance, and the predicted sequence is output: The predicted step size K is matched with the physical delay of the actuator.

[0072] The predicted platform motion data is input into the FAST numerical model to complete the full-scale wind turbine aerodynamic load calculation, and then the load is converted to the model scale according to the Froude similarity criterion. The target aerodynamic load that can be directly used for control is obtained. .

[0073] A fuzzy PID + feedforward composite control strategy is used to generate the control signal, and the total control output is: The feedforward control variable enables rapid load tracking. The fuzzy PID control quantity is based on the load tracking error. Complete adaptive load correction: In the formula This is the feedforward control variable. Here, e(t) represents the feedforward gain, and e(t) represents the load tracking error. and The load is distributed as real-time target aerodynamic load and actual aerodynamic load. , , PID parameters are adaptively adjusted online for fuzzy inference.

[0074] A fixed-length sliding window mechanism is adopted to update the training dataset of the time series prediction model in real time, remove redundant historical data that has exceeded the expiration period, supplement with new motion data collected in the current period, and optimize the model parameters through incremental fine-tuning, so as to maintain the prediction accuracy of the model continuously without interfering with the real-time performance of the experiment.

[0075] The entire process of data acquisition, prediction, calculation, loading, and optimization is completed with a fixed control cycle. After the actuator applies the load, the platform generates a new round of motion response and triggers another data acquisition, forming a continuously iterative real-time closed loop to ensure the temporal synchronization and dynamic fidelity of the semi-physical simulation.

[0076] A semi-physical model delay-compensated closed-loop control system for floating wind turbines includes a physical submodule, a control submodule, and a numerical submodule.

[0077] The physical submodule includes a floating platform 4, a mooring system 5, data acquisition equipment, and a semi-physical wind turbine actuator 1, all of which are installed in the wave pool 7.

[0078] The floating platform 4 and mooring system 5 are fabricated and deployed according to the Froude similarity criterion; the data acquisition unit includes an optical six-degree-of-freedom motion acquisition device and a six-dimensional force sensor 2. The former acquires the six-degree-of-freedom motion timing data of the platform under wave excitation in real time through an optical high-precision camera deployed on the shore, while the latter is installed between the floating wind turbine model tower 3 and the upper semi-physical wind turbine actuator 1 to synchronously acquire the actual output load data of the actuator. The two types of acquisition units maintain timing synchronization to ensure that there is no timing deviation in data transmission and feedback; the semi-physical wind turbine actuator 1, as the core execution component, is equipped with multiple sets of drive execution elements and receives control commands to complete the target load output.

[0079] The physical fan actuator includes three radial support rods evenly arranged at 120° circumference and six sets of independent drive actuators. The entire assembly is divided into two groups: the first group, D1, D2, and D3, is located at the outer end of the support rods in the XZ plane, with a radius of L1, and outputs a backward sway thrust in forward rotation; the second group, D4, D5, and D6, is located near the center end of the support rods in the YZ plane, with a radius of L2, where D4 outputs a vertically upward thrust in forward rotation, and D5 and D6 output obliquely upward and obliquely downward thrusts, respectively.

[0080] The loads include sway force Fx, vertical force Fz, roll moment Mx, and pitch moment My.

[0081] The independent drive actuators include a brushless motor 105 electronic speed controller and a variable speed fan. The brushless motor 105 electronic speed controller consists of six groups, each containing three three-phase AC power lines (U, V, W) connected to the motor, and two signal lines (SGN, GND). The three-phase AC power lines U, V, and W are connected to the A, B, and C phase lines from the A, B, and C coils of the brushless motor 105. Forward rotation uses a UA, VB, WC reversing connection, while reverse rotation uses a UB, VA, WC reversing connection, thus achieving the rotation of the brushless motor 105 and generating the required thrust in each direction. The six SGN lines are connected to the PWM pin of the controller; the six GND lines are connected in parallel and simultaneously connected to the GND pin of the controller. The variable speed fan consists of an external rotor brushless motor 105, a propeller 104, and a propeller mount 103. The propeller 104 is connected to the motor in a forward rotation and to the propeller 104 in a reverse rotation. This avoids the propeller 104 from disengaging due to the high speed of the motor. The propeller 104 is driven to generate varying thrust or lift by the change in the speed of the brushless motor 105.

[0082] The control submodule has a built-in timing prediction unit, a composite control unit, and an online model optimization unit, and has three core functions: prediction, control, and optimization, enabling full-process delay compensation and closed-loop correction.

[0083] The timing prediction unit is equipped with a pre-trained LSTM or Transformer timing prediction model. Based on the collected historical and real-time platform motion data, it predicts the subsequent motion response of the platform in advance and matches the predicted step size with the inherent delay of the actuator.

[0084] The composite control unit adopts a fuzzy PID + feedforward composite control architecture, which takes into account both rapid load tracking and dynamic error correction, thereby improving loading accuracy.

[0085] Furthermore, the online model optimization unit updates the prediction model dataset in real time through a sliding window mechanism, enabling incremental fine-tuning of the model and ensuring the stability of long-term experimental predictions. W t for t The sliding window data set at any given moment, which is the latest piece of historical data.

[0086] ;

[0087] The numerical submodule is built on FAST software and encapsulated as a callable numerical computing unit. It achieves low-latency data interaction with the control submodule through a high-speed communication interface. It can receive platform motion prediction data transmitted by the control submodule, quickly solve the full-size wind turbine aerodynamic load, and then complete the scale transformation according to the Froude similarity criterion, and send the model-scale target load back to the control submodule.

[0088] Example 2

[0089] like Figure 1 As shown, the semi-physical real-time hybrid model delay-compensated closed-loop control system for floating wind turbines of this invention includes a physical submodule, a control submodule, and a numerical submodule. Its real-time closed-loop control flow is as follows:

[0090] S1. The floating platform in the physical submodule generates a six-degree-of-freedom motion response under wave excitation, and the platform's motion timing data is collected in real time by an optical six-degree-of-freedom motion acquisition device.

[0091] S2. To compensate for the inherent physical delay of hardware actuators such as motors, the control submodule uses LSTM or Transformer timing prediction models to predict the platform motion response at the next moment based on historical and real-time collected platform motion timing data.

[0092] S3. The predicted platform motion data is input into the numerical submodule by calling the FAST software dynamic link library (DLL) interface through the Python SDK. The numerical calculation of the full-size wind turbine aerodynamic load is completed in advance, and the full-size load is scaled down to the model scale based on the Froude similarity criterion to obtain the target aerodynamic load at the model scale.

[0093] S4. Input the target aerodynamic load at the model scale to the fuzzy PID + feedforward composite controller to drive the semi-physical fan actuator (brushless motor and propeller) to output the corresponding load, thereby offsetting the hardware execution delay and improving the real-time performance of the semi-physical simulation.

[0094] S5. The actual output load of the actuator is collected in real time by a six-dimensional force sensor and fed back to the composite controller to complete the closed-loop correction of the load tracking error.

[0095] S6. After loading, the floating platform generates a new round of motion response, which is transmitted back to the control submodule to form a real-time closed loop of "acquisition-prediction-calculation-loading-feedback". On the other hand, it is synchronously updated to the training dataset of the time series prediction model to realize online iterative optimization of the model and continuously improve the prediction accuracy.

[0096] The structure of the time series prediction model and the data input / output process in the control submodule are as follows: Figure 2 As shown.

[0097] First, the data preprocessing module performs sliding window sampling and standardization / normalization on the collected platform motion data to generate a fixed-length sequence. This sequence is then input to the input layer, characterized by the platform's six degrees of freedom (6DoF) motion, in the form of [Time Step: tN,…, t-2, t-1, t]. The sliding window length N can be adjusted according to the experimental conditions. The core network module provides two optional architectures: LSTM and Transformer, used to capture the long-term and short-term dependencies or global temporal correlations of the motion sequence, respectively. After feature extraction, the output layer predicts the platform's six degrees of freedom motion coordinates (Surge, Sway, Heave, Roll, Pitch, Yaw) in advance for the next K time steps. The prediction step size K can be adaptively adjusted according to the actuator delay characteristics. Finally, the prediction results are input to the downstream module to pre-calculate the aerodynamic loads using FAST software, thereby offsetting the inherent physical delays of the actuators and data transmission, and ensuring the temporal matching and real-time performance of the semi-physical simulation.

[0098] The composite controller structure and control logic in the control submodule are as follows: Figure 3 As shown.

[0099] The controller employs a dual-channel composite architecture of feedforward control and fuzzy PID closed-loop control: the target aerodynamic thrust signal is divided into two paths. One path is directly input to the feedforward controller, generating feedforward control parameters and outputting a feedforward control quantity to achieve rapid response and tracking of the target load. The other path is subtracted from the filtered actual thrust signal to obtain the tracking error, which is then input to the fuzzy PID controller. The fuzzy inference mechanism adaptively adjusts the PID control parameters and outputs a closed-loop correction control quantity. The feedforward control quantity and the fuzzy PID closed-loop control quantity are superimposed and input to the actuator to drive the actuator to output the corresponding thrust. The actual output thrust is acquired by the sensor, fed back to the input, and filtered to form a closed-loop correction loop.

[0100] like Figure 4 and Figure 5 As shown, the semi-physical wind turbine actuator 1 includes a nacelle 101, a carbon fiber support rod 102, a propeller base 103, a propeller 104, and a brushless motor 105, and the whole is arranged in a modular manner.

[0101] The semi-physical wind turbine actuator 1 is fixed to the top of the floating platform 4 via the nacelle 101. The carbon fiber support rod 102 and the propeller mount 103 provide rigid support for the actuator. When the motor rotates forward, the propeller 104 is connected to the propeller mount with a positive thread. When the motor rotates in reverse, the propeller 104 is connected to the propeller mount with a negative thread. By matching the thread direction with the motor rotation direction, the propeller 104 is prevented from coming loose when the motor rotates at high speed. The controller adjusts the speed of the brushless motor 105 through the PWM signal, which drives the propeller 104 to rotate, thereby generating thrust with adjustable amplitude.

[0102] The semi-physical fan actuator 1 includes three carbon fiber support rods 102 evenly arranged at 120° circumferential angles, and six sets of independent drive actuators, which are arranged in two groups:

[0103] The first set of actuators, D1, D2, and D3, are located at the outer end of the carbon fiber support rod 102 with a radius of L1. They adopt a forward rotation drive mode and output a backward swaying thrust.

[0104] The second set of actuators, D4, D5, and D6, are located near the center of the carbon fiber support rod 102, with a radius of L2. D4, rotating clockwise, outputs a vertically upward lift force; D5, rotating counter-clockwise, and D6, rotating clockwise, output obliquely upward and obliquely downward thrust forces, respectively. The formulas used to calculate the loads in the Fx, Fz, Mx, and My directions are as follows:

[0105] ;

[0106] Figure 6 This is a schematic diagram of the overall layout for the real-time hybrid model water tank test of this invention. It mainly includes: a semi-physical wind turbine actuator 1, a six-dimensional force sensor 2, a tower 3, a floating platform 4, a mooring system 5, a six-degree-of-freedom motion acquisition instrument 6, and a wave pool 7.

[0107] The floating platform 4 is installed in the wave pool 7 and is anchored to the bottom of the pool by the mooring system 5 to simulate the floating foundation constraints in a real marine environment.

[0108] The tower 3 is vertically installed on the top of the floating platform 4. A semi-physical fan actuator 1 is installed on the top of the tower to simulate the aerodynamic load of the fan and load it onto the floating platform 4.

[0109] The six-dimensional force sensor 2 is located between the tower 3 and the semi-physical fan actuator 1 to collect the actual six-degree-of-freedom load output by the actuator in real time, providing feedback data for closed-loop control;

[0110] Six-degree-of-freedom motion acquisition instrument 6 is deployed on both sides of wave pool 7. It captures the six-degree-of-freedom motion response of floating platform 4 under the combined action of wave excitation and aerodynamic load in real time through optical measurement, providing raw motion data for time series prediction model and numerical submodule.

[0111] Wave pool 7 generates target wave conditions through a wave-generating system, providing a controllable marine environmental stimulus for the experiment.

[0112] The components work together to complete the semi-physical real-time hybrid model test of the floating wind turbine, realizing the full-process simulation of wave excitation, platform motion, aerodynamic load loading and closed-loop control.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A delayed-compensation closed-loop control method for a semi-physical model of a floating wind turbine, characterized in that, Includes the following steps: S1. Real-time acquisition of motion time-series data generated by the floating platform based on wave excitation; S2. Based on historical data and real-time collected platform motion time series data, a time series prediction model is constructed to predict the motion response of the floating platform in advance. The time series prediction model employs either an LSTM (Long Short-Term Memory) network model or a Transformer time series prediction model. The model inputs are historical six-DOF motion time series data of a pre-defined platform and real-time acquired motion time series data sequences. The expression is as follows: ； In the formula, The length of the sliding window; For a moment Observational data; The model outputs a sequence of predicted six-DOF motion responses for a floating platform at one or more future time steps. The expression is as follows: ； In the formula, This is a predicted value; To predict the step size; S3. Based on the advanced prediction results of S2, perform aerodynamic load calculations for the full-size wind turbine in advance, and perform scale conversion from the full-size load to the model scale to obtain the target aerodynamic load at the model scale. By calling the dynamic link library (DLL) interface of the FAST wind turbine aerodynamic simulation software through the Python SDK, the predicted platform motion response is input into the FAST software to perform advance numerical calculation of the full-scale wind turbine aerodynamic load. Based on Froude's similarity criterion A scaled-down conversion from full-scale load to model scale is performed to obtain the target aerodynamic load directly used for control. F target (t), the expression is as follows: ； In the formula, The characteristic velocity of the prototype; The feature length of the prototype; The characteristic velocity of the model; The feature length of the model; It is the acceleration due to gravity; For model-scale target loads; This is a full-size prototype aerodynamic load; It is a geometric scaling ratio; Target aerodynamic load for The specific time-varying form in the time dimension; S4. Input the target aerodynamic load obtained in S3 into the composite controller, and drive the actuator of the semi-physical fan to output the corresponding load. The actual output load of the actuator is collected in real time and fed back to the composite controller for high-precision load tracking and closed-loop correction. S5. Based on the platform motion time series data collected in real time in S1, the training dataset of the time series prediction model is updated synchronously to continuously optimize the prediction accuracy of the time series prediction model. S6. After loading, the floating platform generates a new round of motion response, repeating S1 to S5, and performing a real-time closed-loop cycle of the floating wind turbine semi-physical simulation test.

2. The method for delayed compensation closed-loop control of a floating wind turbine using a semi-physical model according to claim 1, characterized in that: The composite controller in S4 uses a combination of fuzzy PID and feedforward controllers, including a feedforward control channel and a fuzzy PID closed-loop control channel. The overall control output expression is as follows: ； In the formula, Forward control variable, where For feedforward gain; For fuzzy PID control; Generate feedforward control quantity based on target aerodynamic load. The target load is tracked rapidly through the feedforward control channel; Calculate the tracking error between the target aerodynamic load and the actual output load. The expression is as follows: ； In the formula, This refers to the actual aerodynamic load; By using a fuzzy PID closed-loop control channel, the PID control parameters are adaptively adjusted to achieve dynamic correction of the load tracking error, as shown in the following expression: ； In the formula, This is the proportionality coefficient; The integral coefficient; is the differential coefficient.

3. The method for delayed compensation closed-loop control of a floating wind turbine using a semi-physical model according to claim 2, characterized in that: S5 employs a sliding window mechanism to update the training dataset of the time series prediction model online, removing historical data with a time span exceeding a preset length and simultaneously supplementing it with real-time collected platform six-degree-of-freedom motion time series data to continuously iterate and optimize the time series prediction model.

4. A floating wind turbine semi-physical model delay compensation closed-loop control system, employing the floating wind turbine semi-physical model delay compensation closed-loop control method as described in any one of claims 1-3, characterized in that, include: Physics submodule, numerical submodule, and control submodule; The physics submodule is used to collect the motion timing data generated by the wave-excited floating platform in real time, as well as the actual output load of the semi-physical wind turbine actuator, and at the same time execute the aerodynamic load loading action to drive the floating platform to generate a new round of motion response. The control submodule is used to receive real-time platform motion and load data from the physics submodule and interact with the numerical submodule to complete the delay compensation and high-precision closed-loop control of the semi-physical real-time hybrid model of the floating wind turbine. The numerical submodule is used to receive the predicted platform motion response in advance, complete the aerodynamic load numerical calculation of the full-size wind turbine in advance, and complete the scale conversion of the full-size load to the model scale, output the target aerodynamic load at the model scale, and offset the inherent physical delay of the actuator. The system completes the real-time closed-loop simulation of the floating wind turbine through data interaction between the physical submodule, control submodule, numerical submodule, composite control unit, and physical submodule.

5. The floating wind turbine semi-physical model delay compensation closed-loop control system according to claim 4, characterized in that: The physical submodule includes the floating platform, mooring system, data acquisition equipment, and semi-physical wind turbine actuators; The floating platform and mooring system are deployed in a wave pool to generate a coupled dynamic motion response of the floating wind turbine under wave excitation; The data acquisition equipment includes an optical six-degree-of-freedom motion acquisition device and a six-dimensional force sensor. The optical six-degree-of-freedom motion acquisition device is used to acquire the six-degree-of-freedom motion time-series data of the floating platform in real time, and the six-dimensional force sensor is used to acquire the actual aerodynamic load output by the semi-physical fan actuator in real time. The semi-physical fan actuator is used to output the corresponding aerodynamic load according to the control command of the composite control unit.

6. The floating wind turbine semi-physical model delay compensation closed-loop control system according to claim 5, characterized in that: The control submodule includes a timing prediction unit, a composite control unit, and an online model optimization unit; The timing prediction unit is used to receive historical and real-time platform motion timing data collected by the physical submodule and predict the floating platform motion response in advance at the next moment. The composite control unit is used to receive the target aerodynamic load, generate control commands to drive the actuator of the semi-physical fan to output the corresponding load, receive the actual output load collected by the physical submodule, and complete the closed-loop correction of load tracking. The online model optimization unit receives real-time platform motion time-series data collected by the physics submodule, synchronously updates the training dataset of the time-series prediction unit, and continuously optimizes the prediction accuracy of the time-series prediction model.

7. The floating wind turbine semi-physical model delay compensation closed-loop control system according to claim 6, characterized in that: The semi-physical fan actuator includes three identical radial support rods evenly arranged at 120° circumference and six identical drive actuators. Among the six identical drive actuators, D1, D2, and D3 are arranged in the YZ plane of the support rod and perpendicular to the plane, with a radius of L1. D4, D5, and D6 are arranged perpendicularly to the support rod in a clockwise direction, with a radius of L2, and output upward, diagonally upward, and diagonally downward thrust, respectively.

Images