Energy feedback based dynamic loading test system and method for gearbox

Abstract

The application discloses a gear box dynamic loading test system and method based on energy feedback, which is used for dynamic simulation loading test of a measured gear box under multiple working conditions. The system comprises a main driving module, a composite loading module, an intelligent energy management unit, a multi-dimensional sensor array and a central cooperative controller. The composite loading module comprises a multi-degree-of-freedom loading mechanism and a load motor, which can respectively exert non-rotating force on the gear box body and rotating force on the output shaft, thereby realizing multi-degree-of-freedom composite loading. The intelligent energy management unit realizes efficient bidirectional energy flow between the system internal direct current and the power grid alternating current through a super capacitor module and a four-quadrant PWM converter, thereby effectively suppressing power fluctuation and improving energy feedback efficiency. The central cooperative controller cooperatively controls the main driving module, the composite loading module and the energy management unit according to real-time monitoring data of the multi-dimensional sensor array, thereby significantly improving loading accuracy.

Description

Technical Field

[0001] This invention belongs to the field of gearbox test loading technology, and more specifically, relates to a gearbox dynamic loading test system and method based on energy feedback. Background Technology

[0002] As the core transmission component of wind turbine generators, the reliability testing of wind turbine gearboxes is crucial. Currently, the mainstream testing equipment mainly includes two types: mechanically enclosed and electrically enclosed. Mechanically enclosed test benches achieve power cycling through mechanical structures, but they have inherent drawbacks such as poor loading flexibility and difficulty in simulating complex dynamic loads. Electrically enclosed test benches typically adopt a "motor-gearbox-generator" paired structure and use frequency converters to adjust the load. Some systems achieve power feedback to the grid through uncontrolled rectification or unidirectional PWM rectification. Although this improves load adjustability to some extent, it still suffers from limited multi-degree-of-freedom loading capacity and slow dynamic response.

[0003] In terms of energy utilization, existing closed-loop power test benches generally adopt a unidirectional rectification feedback scheme. The energy feedback path is singular and irreversible, making it difficult to smooth out power fluctuations caused by frequent changes in operating conditions. This easily leads to power surges on the grid side, resulting in low energy recovery efficiency. Furthermore, to meet peak power demands, grid-side converters must be configured according to maximum power capacity, resulting in large equipment capacity and high costs. Energy requires multiple AC-DC-AC conversions, leading to numerous conversion stages and significant losses. The overall energy efficiency of the measured system under dynamic operating conditions is only 65%-75%, indicating significant energy waste. Regarding operating condition simulation performance, existing multi-degree-of-freedom loading devices are mostly based on series mechanical structures, exhibiting significant kinematic and dynamic coupling. Interference between degrees of freedom is severe (>10%), far exceeding the ≤5% coupling interference requirement in actual wind turbine gearbox operation, resulting in insufficient loading accuracy. Meanwhile, due to the inherent high inertia and low stiffness of the series mechanism and the response lag of traditional hydraulic / electric cylinders, its dynamic response time is usually greater than 100ms, which cannot meet the rapid response requirement of ≤50ms under extreme wind conditions, making it difficult to accurately reproduce transient impact loads such as gusts and shearing. In addition, existing systems mostly adopt a distributed control architecture, with each subsystem operating independently and lacking unified coordinated scheduling. The classic PID control strategy is difficult to handle multivariable, strongly coupled nonlinear control problems, affecting the accuracy and reliability of the test process.

[0004] Therefore, those skilled in the art urgently need to invent a test system and method that can solve the problems of low energy feedback efficiency, insufficient accuracy of multi-degree-of-freedom dynamic loading, and lack of coordinated technology in the decentralized system control of existing wind turbine gearbox test benches. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention, by introducing an intelligent energy management unit and integrating a supercapacitor module and a four-quadrant PWM converter, achieves efficient bidirectional energy flow between the internal DC bus and the external power grid of the test system, significantly improving energy feedback efficiency and system energy utilization. Specifically, in a first aspect, this invention provides a dynamic loading test system for wind turbine gearboxes based on energy feedback, used for conducting multi-condition dynamic simulation loading tests on the gearbox under test. The system includes:

[0006] The main drive module is connected to the input shaft of the gearbox;

[0007] The composite loading module includes a multi-degree-of-freedom loading mechanism and a load motor. The multi-degree-of-freedom loading mechanism is connected to the outer wall of the gearbox and is used to apply a first force to the outside of the gearbox. The load motor is connected to the output shaft of the gearbox and is used to apply a second force to the output shaft of the gearbox. The first force is a non-rotational force, and the second force is a rotational force.

[0008] The intelligent energy management unit includes an electrically connected supercapacitor module and a four-quadrant PWM converter. The supercapacitor module is connected to the internal DC power of the test system and the AC power of the external power grid through the four-quadrant PWM converter. The four-quadrant PWM converter is used to convert the internal DC power and the AC power of the external power grid to each other. The supercapacitor module is used to output or absorb the rapidly changing current of the system, or to smooth out the DC bus voltage fluctuations.

[0009] A multi-dimensional sensor array, comprising multiple sets of sensors, is respectively set at various loading positions on the input shaft, output shaft, and side of the gearbox;

[0010] The central coordinating controller is communicatively connected to the main drive module, the multi-degree-of-freedom loading mechanism, the load mechanism, the intelligent energy management unit, and the multi-dimensional sensor array. It is used to obtain the real-time loading parameters of the gearbox through the multi-dimensional sensor array and the power status of each power module through the intelligent energy management unit, so as to send loading control commands to the main drive module, the multi-degree-of-freedom loading mechanism, and the load motor to load the gearbox.

[0011] In the first aspect, the main drive module includes: a permanent magnet synchronous main motor and a vector frequency converter electrically connected, wherein the output shaft of the permanent magnet synchronous main motor is connected to the input shaft of the gearbox for simulating the aerodynamic torque input of the wind turbine, and the vector frequency converter is also communicatively connected to the central coordinating controller for controlling the output frequency of the permanent magnet synchronous main motor.

[0012] In the first aspect, the multi-degree-of-freedom loading mechanism includes:

[0013] Several magnetorheological fluid actuators connected in parallel are respectively connected between the loading interface plate on the gearbox housing and the foundation platform via ball joints. Each magnetorheological fluid actuator is equipped with a position and force sensor.

[0014] Secondly, the present invention provides a dynamic loading test control method for wind turbine gearboxes based on energy feedback, the control method comprising:

[0015] Step S1: Call the pre-stored target load data, construct the inverse dynamics model, and predict the output parameters of the gearbox based on the inverse dynamics model; the target load data is the pre-stored gearbox normal operating condition load parameter data.

[0016] Step S2: Obtain the current system state and initial parameters, and predict the behavior of the tested system in a finite time domain in the future using a pre-established inverse dynamics model to obtain a predicted output sequence.

[0017] Step S3: Based on the behavior and predicted output sequence of the tested system, construct a unified, constrained multi-objective optimization model, and generate loading instructions for the gearbox based on the multi-objective optimization model.

[0018] Step S4: Control the test system to load the gearbox according to the generated loading command;

[0019] Step S5: At the next sampling time, repeat steps S1-S4 based on the new measurement value.

[0020] In the second aspect, step S1, constructing the inverse dynamics model includes:

[0021] Based on the target load data, inverse dynamics is used to solve the problem using the parameters of the gearbox, loading mechanism, and load motor under test. This allows for the calculation of the target speed and torque of the main drive motor, the target torque of the load motor, and the target force and displacement of several magnetorheological actuators.

[0022] In the second aspect, the inverse dynamics model includes:

[0023]

[0024] Where I = diag([I1, I2, ..., I n ]), is a diagonal matrix of moments of inertia, θ=[θ1,θ2,...,θ n ] T Let be an n-dimensional angular displacement vector containing the input shaft, each gear pair, and the output shaft; C be the damping matrix; K be the stiffness matrix; and τ be the angular displacement vector. in To apply the torque vector, τ out For load torque vector;

[0025] F = J -T Formula 2

[0026] Where F represents the six-dimensional force / torque at the gearbox interface, derived from the target load data, i.e., the load applied to the gearbox; F = [F x ,F y ,F z M x M y M z ] T J is the Jacobian matrix, which maps the small motions on the moving platform to the rate of change of the actuator rod length. τ is the force / torque vector of the six magnetorheological actuators, τ = [τ1, τ2, τ3, τ4, τ5, τ6]. T ;

[0027]

[0028] Where, τ m Let p be the electromagnetic torque generated by the motor, and p be the number of pole pairs of the motor. L is the flux linkage amplitude generated by the permanent magnet. d L q For direct-axis and quadrature-axis inductors. d i q These are the direct-axis and quadrature-axis currents.

[0029] In the second aspect, obtaining the current system state and initial parameters includes:

[0030] Real-time data from the system's main drive module, composite loading module, and intelligent energy management unit are acquired through multi-dimensional sensors.

[0031] In the second aspect, in step S2, the predicted output sequence includes:

[0032] in, Includes all target variables that need to be tracked, such as the main drive input torque τ_in_pred, the rotational speed ω_in_pred, and the six-dimensional force / torque at the gearbox interface [F]. x _pred,F_y_pred,F_z_pred,M x _pred,M_y_pred,M_z_pred].

[0033] In the second aspect, the multi-objective optimization model in step S3 includes:

[0034]

[0035] Where k is the current control time, k+i is the future prediction time (i=1,2,…,p), X d (k+i) is the target load spectrum vector at time k+i in the future, including input torque, speed and six-dimensional force and moment), X(k+i|k) is the predicted state based on the system dynamics model, Δu(k+j|k) is the control increment sequence, Q and R are weight matrices, and p and m are the prediction time domain and control time domain, respectively.

[0036] Thirdly, the present invention provides a test system that can be applied to a dynamic loading test control method for wind turbine gearboxes based on energy feedback.

[0037] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0038] 1. The wind turbine gearbox dynamic loading test system based on energy feedback of the present invention, by introducing an intelligent energy management unit and integrating a supercapacitor module and a four-quadrant PWM converter, realizes efficient bidirectional energy flow between the DC bus inside the test system and the external power grid, significantly improves energy feedback efficiency and system energy utilization, effectively suppresses power fluctuations on the grid side and reduces the impact on the grid, while reducing dependence on large-capacity grid-side converters, reducing equipment costs and multi-stage energy conversion losses.

[0039] 2. By adopting a composite loading module that combines a parallel multi-degree-of-freedom magnetorheological fluid actuation mechanism with a high dynamic load motor, and supplemented by real-time monitoring with a multi-dimensional sensor array, the system achieves high-precision, low-coupling independent loading of six-dimensional force / torque on the wind turbine gearbox. This significantly improves the dynamic response speed and simulation realism of multi-directional composite loads under extreme transient conditions, effectively reproducing the high-frequency impact and complex stress state in actual wind loads.

[0040] 3. By leveraging the central coordinating controller and model predictive control algorithms to uniformly schedule and optimize the main drive module, composite loading module, and energy management unit, the system achieves high-precision coordinated control of actuators and intelligent dynamic energy allocation under multivariable strongly coupled operating conditions. This not only significantly improves the control synchronization and result reliability of the test process, but also enhances the system's adaptability to different gearbox models and varying operating conditions. Attached Figure Description

[0041] Figure 1 This is a module connection diagram of the gearbox dynamic loading test system based on energy feedback in this embodiment;

[0042] Figure 2 This is a topology diagram of the intelligent energy management system in this embodiment;

[0043] Figure 3This is a schematic diagram of the composite loading module in this embodiment;

[0044] Figure 4 This is a three-dimensional topology diagram of the gearbox dynamic loading test system based on energy feedback in this embodiment;

[0045] Figure 5 This is the control flowchart of the dynamic loading test control method for wind turbine gearbox based on energy feedback in this embodiment;

[0046] Figure 6 This is a flowchart of the dynamic loading test control method for wind turbine gearbox based on energy feedback in this embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0048] Example 1:

[0049] Please see Figure 1-6 This embodiment provides a wind turbine gearbox dynamic loading test system based on energy feedback, which is used to conduct multi-condition dynamic simulation loading tests on the gearbox under test. The system includes: a main drive module, a composite loading module, an intelligent energy management unit, a multi-dimensional sensor array, and a central collaborative controller.

[0050] The main drive module is connected to the input shaft of the gearbox; the composite loading module includes a multi-degree-of-freedom loading mechanism and a load motor. The multi-degree-of-freedom loading mechanism is connected to the outer wall of the gearbox and is used to apply a first force to the outside of the gearbox. The load motor is connected to the output shaft of the gearbox and is used to apply a second force to the output shaft of the gearbox. The first force is a non-rotational force, and the second force is a rotational force; the intelligent energy management unit includes an electrically connected supercapacitor module and a four-quadrant PWM converter. The supercapacitor module is connected to the internal DC power of the test system and the AC power of the external power grid through the four-quadrant PWM converter. The four-quadrant PWM converter is used to convert the internal DC power into AC power. The system converts AC power from the external power grid to AC power. The supercapacitor module is used to output or absorb rapidly changing current in the system, or to smooth DC bus voltage fluctuations. The multi-dimensional sensor array includes multiple sets of sensors, which are respectively set at various loading positions on the input shaft, output shaft, and side of the gearbox. The central coordinating controller is communicatively connected to the main drive module, the multi-degree-of-freedom loading mechanism, the load mechanism, the intelligent energy management unit, and the multi-dimensional sensor array. It is used to obtain real-time loading parameters of the gearbox through the multi-dimensional sensor array and to obtain the power status of each power-consuming module through the intelligent energy management unit, so as to send loading control commands to the main drive module, the multi-degree-of-freedom loading mechanism, and the load motor to load the gearbox.

[0051] Specifically, the wind turbine gearbox dynamic loading test system based on energy feedback in this embodiment, by introducing an intelligent energy management unit and integrating a supercapacitor module and a four-quadrant PWM converter, achieves efficient bidirectional energy flow between the DC bus inside the test system and the external power grid. This significantly improves energy feedback efficiency and system energy utilization, effectively suppresses power fluctuations on the grid side and reduces the impact on the grid. Simultaneously, it reduces dependence on large-capacity grid-side converters, lowering equipment costs and multi-stage energy conversion losses. By employing a composite loading module combining a parallel multi-degree-of-freedom magnetorheological fluid actuation mechanism and a high-dynamic load motor, supplemented by real-time monitoring with a multi-dimensional sensor array, the system achieves high-precision, low-coupling independent loading of the wind turbine gearbox's six-dimensional force / torque. This significantly improves the dynamic response speed under extreme transient conditions and the simulation realism of multi-directional composite loads, effectively reproducing the high-frequency impacts and complex stress states in actual wind loads. By leveraging the central coordinating controller and model predictive control algorithms to uniformly schedule and optimize the main drive module, composite loading module, and energy management unit, the system achieves high-precision coordinated control of actuators and intelligent dynamic energy allocation under multivariable strongly coupled operating conditions. This not only significantly improves the control synchronization and result reliability of the test process but also enhances the system's adaptability to different gearbox models and varying operating conditions.

[0052] For the intelligent energy management unit, it is responsible for the efficient scheduling and feedback of the system energy, with a response time ≤ 10 ms. This unit mainly consists of a four-quadrant PWM converter (bidirectional DC / AC converter, such as a common two-level voltage source type PWM converter), a supercapacitor energy storage module, and an energy coordination controller. The four-quadrant PWM converter realizes the bidirectional energy flow between the AC bus and the DC bus. The supercapacitor energy storage module acts as an instantaneous power fluctuation buffer unit, providing / absorbing the rapidly changing power demand of the system and suppressing the voltage fluctuation of the DC bus. The energy coordination controller collects the following data in real time through the CAN bus: the voltage / current of the main drive bus, calculates the required power P_drive of the drive motor; the power on the DC side of the load motor controller, calculates the generated power P_load_gen of the load motor; the SOC and the maximum charge / discharge power (P_cap_max) of the supercapacitor BMS (Battery Management System); the power on the AC side of the four-quadrant converter (P_grid), and dynamically adjusts the energy flow direction according to the monitored data.

[0053] The energy scheduling strategy used in this part is as follows: If P_load_gen > P_drive: then (P_load_gen - P_drive) is preferentially stored in the supercapacitor. When the supercapacitor reaches the storage limit, the excess part is fed back to the power grid. If P_load_gen < P_drive: then the supercapacitor releases (P_drive - P_load_gen). When the energy of the supercapacitor is insufficient, the insufficient part is supplemented by the power grid. Among them, the generated power of the load motor is P_load_gen, and the required power of the drive motor is P_driv. If the system needs to absorb regenerative energy, such as braking, it is preferentially absorbed by the supercapacitor. When the supercapacitor is full, the excess energy can be optionally fed back to the power grid.

[0054] For the composite loading module, the multi-degree-of-freedom loading mechanism adopts a parallel magnetorheological fluid (MRF) (MagnetorheologicalFluid) loading technology. Based on the rheological effect of MRF (the viscosity changes reversibly and rapidly with the magnetic field strength), it realizes the continuous, stepless, and rapid (millisecond-level response) adjustment of the output force. Six MRF actuators are respectively connected to the loading interface plate on the gearbox housing and the foundation platform through spherical hinges, such as Figure 3As shown, the configuration parameters satisfy: actuator working stroke ≥ ±50mm; maximum output force ≥ 200kN; excitation coil response time ≤ 5ms. It possesses advantages such as high structural rigidity, low inertia, and good decoupling of motion for each degree of freedom. It can independently load three translational degrees of freedom (radial x, y, axial z) and two rotational degrees of freedom (pitch, yaw) or equivalent force and torque, totaling five non-rotational degrees of freedom (torque), onto a wind turbine gearbox. Each magnetorheological actuator incorporates a displacement / force sensor, forming an independent force / position servo control loop. Through decoupling algorithms such as real-time feedforward compensation based on the kinematic / dynamic model of a parallel mechanism or cross-coupling suppression algorithms based on sensor feedback, the mutual interference between loading commands for each degree of freedom is minimized, improving loading accuracy.

[0055] For multi-dimensional sensor arrays, they are distributed in key parts of the gearbox such as the input shaft, output shaft, and housing to collect parameters such as speed, torque, vibration, temperature, and strain. The sensor array signals are directly connected to the central coordinating controller to realize closed-loop feedback control.

[0056] As the sole and highest decision-making unit of the system, the central coordinating controller interacts with each module via industrial Ethernet, coordinates and controls the operation of the main drive module, composite loading module, and intelligent energy management unit, and processes multi-dimensional sensor array data to achieve coordinated control of the entire system.

[0057] In one specific implementation, the main drive module includes: a permanent magnet synchronous main motor and a vector frequency converter electrically connected, wherein the output shaft of the permanent magnet synchronous main motor is connected to the input shaft of the gearbox for simulating the aerodynamic torque input of the wind turbine, and the vector frequency converter is also communicatively connected to the central coordinating controller for controlling the output frequency of the permanent magnet synchronous main motor.

[0058] In one specific embodiment, the multi-degree-of-freedom loading mechanism includes: a plurality of parallel magnetorheological actuators, which are respectively connected between the loading interface plate 1 on the gearbox housing and the foundation platform 3 via ball joints 2, and each magnetorheological actuator is equipped with a position and force sensor.

[0059] The load motor is a high-dynamic-response permanent magnet synchronous motor and its matching driver, connected in series with the parallel multi-degree-of-freedom loading mechanism. Together, they act on the output shaft of the gearbox under test, primarily responsible for loading the rotational degree of freedom (torque), and can absorb or provide rotational kinetic energy when necessary. Under the unified command of the central coordinating controller, the load motor and the multi-degree-of-freedom parallel loading mechanism coordinate their actions to accurately reproduce the six-dimensional force / torque (radial Fx, Fy, axial Fz, bending moment Mx, My, Mz) experienced by the gearbox under actual wind conditions.

[0060] Example 2:

[0061] This embodiment provides a dynamic loading test control method for wind turbine gearboxes based on energy feedback. The control method includes: Step S1, calling pre-stored target load data, constructing an inverse dynamics model, and predicting the gearbox output parameters based on the inverse dynamics model; the target load data is pre-stored gearbox conventional operating condition load parameter data, specifically including: Xd(t)=[τin,ωin,Fx,Fy,Fz,Mx,My,Mz]; Step S2, obtaining the current system state and initial parameters, predicting the behavior of the tested system in a finite time domain in the future through the pre-established inverse dynamics model, and obtaining a predicted output sequence; Step S3, constructing a unified, constrained multi-objective optimization model based on the behavior of the tested system and the predicted output sequence, and generating loading instructions for the gearbox based on the multi-objective optimization model; Step S4, controlling the tested system to load the gearbox according to the generated loading instructions; Step S5, repeating steps S1-S4 based on the new measurement values ​​at the next sampling time.

[0062] Furthermore, for step S1, constructing the inverse dynamics model includes:

[0063] Based on the target load data, inverse dynamics is used to solve the problem using the parameters of the gearbox, loading mechanism, and load motor under test. This allows for the calculation of the target speed and torque of the main drive motor, the target torque of the load motor, and the target force and displacement of several magnetorheological actuators.

[0064] In the second aspect, the inverse dynamics model includes:

[0065]

[0066] Where I = diag([I1, I2, ..., I n ]), is a diagonal matrix of moments of inertia, θ=[θ1,θ2,...,θ n ] T Let be an n-dimensional angular displacement vector containing the input shaft, each gear pair, and the output shaft; C be the damping matrix; K be the stiffness matrix; and τ be the angular displacement vector. in To apply the torque vector, τ out For load torque vector;

[0067] F = J -T Formula 2

[0068] Where F represents the six-dimensional force / torque at the gearbox interface, derived from the target load data, i.e., the load applied to the gearbox; F = [F x ,F y ,F z M x M y Mz ] T J is the Jacobian matrix, which maps the small motions on the moving platform to the rate of change of the actuator rod length. τ is the force / torque vector of the six magnetorheological actuators, τ = [τ1, τ2, τ3, τ4, τ5, τ6]. T ;

[0069]

[0070] Where, τ m Let p be the electromagnetic torque generated by the motor, and p be the number of pole pairs of the motor. L is the flux linkage amplitude generated by the permanent magnet. d L q For direct-axis and quadrature-axis inductors. d i q These are the direct-axis and quadrature-axis currents.

[0071] In some specific implementations, obtaining the current system status and initial parameters includes: acquiring real-time data from the system's main drive module, composite loading module, and intelligent energy management unit through multi-dimensional sensors.

[0072] In some specific implementations, the initial parameters acquired by the multi-dimensional sensor array include: actual rotational speed ω_act, current I_act, actuator excitation voltage / current u_mrf_act, output force F_mrf_act, displacement s_mrf_act, torque τ_act, temperature T, vibration Vib, etc., as well as the status of the system's intelligent energy management unit, including: supercapacitor SOC, maximum charge / discharge power P_cap_max, DC bus voltage V_dc, AC side power P_grid, etc.

[0073] In some specific implementations, step S2, predicting the output sequence includes: Specifically, it predicts the system behavior within a finite time domain of p steps (i.e., from k+1 to k+p) based on the current system state x(k) and the established system dynamics model (including the gearbox model, Stewart platform inverse dynamics model, motor model, etc.).

[0074] in, Includes all target variables that need to be tracked, such as the main drive input torque τ_in_pred, the rotational speed ω_in_pred, and the six-dimensional force / torque at the gearbox interface [F]. x _pred,F_y_pred,F_z_pred,M x _pred,M_y_pred,M_z_pred].

[0075] In some specific implementations, the multi-objective optimization model in step S3 includes:

[0076]

[0077] Where k is the current control time, k+i is the future prediction time (i=1,2,…,p), X d (k+i) is the target load spectrum vector at time k+i in the future, including input torque, speed and six-dimensional force and moment), X(k+i|k) is the predicted state based on the system dynamics model, Δu(k+j|k) is the control increment sequence, Q and R are weight matrices, and p and m are the prediction time domain and control time domain, respectively.

[0078] In rolling optimization control, this invention considers the energy management state in a coordinated manner. By explicitly incorporating the state of the intelligent energy management unit (such as supercapacitor SOC, available power (P_cap_max)) and grid interaction power limits into the optimization objective function or constraints, the power allocation strategy of each actuator is dynamically adjusted. Under the premise of ensuring loading accuracy, internal circulating energy is used preferentially, and grid interaction is minimized.

[0079] The controller incorporates a machine learning module, such as reinforcement learning or online parameter identification algorithms. This module can automatically adjust MPC model parameters, weight coefficients, or inverse dynamics model parameters based on historical test data (such as the deviation between actual response and target, and control effect), thereby adapting to the differences in dynamic characteristics of different wind turbine gearbox models and continuously improving control accuracy and system adaptability. For example, the online parameter identification method employs recursive least squares (RLS) with a forgetting factor, updating the equivalent moment of inertia I, damping coefficient C, or stiffness coefficient K in the gearbox dynamics model online in real time based on the deviation between the actual applied force / torque, displacement / velocity sensor feedback, and target command. Another example is the use of reinforcement learning to adjust weights. A policy search-based reinforcement learning algorithm dynamically adjusts the diagonal elements of the weight matrices Q and R in the MPC objective function based on the statistics of tracking errors (such as root mean square error RMSE) and energy consumption indicators within historical test cycles to find the optimal balance point under specific operating conditions.

[0080] In summary, compared with the prior art, the embodiments have the following technical effects:

[0081] 1. High Energy Efficiency: Compared with traditional closed-loop power test benches that use unidirectional rectification to feed back to the grid, the intelligent energy management unit achieves closed-loop energy circulation within the test system through bidirectional energy flow design and supercapacitor energy storage, improving energy feedback efficiency by more than 30%, as shown in Table 1. This is mainly due to: 1) the closed-loop energy circulation reduces the number of AC-DC-AC conversions; 2) the supercapacitor efficiently absorbs / releases instantaneous fluctuation energy, reducing energy throughput losses at the grid interface; and 3) the high-efficiency operation of the bidirectional converter.

[0082] Table 1 Comparison of Indicators

[0083] index Traditional system This invention Test conditions Grid power fluctuation 15.2％ 4.8％ 5MW sudden unloading test AC-DC-AC conversion times 3 times / cycle ≤1 time / cycle Gust simulation conditions Coupling interference 9.7％ 2.1％ Multi-directional step load superposition

[0084] 2. Improved accuracy in working condition simulation: The parallel magnetorheological fluid (MRF) loading technology of the composite loading module effectively eliminates coupling interference between degrees of freedom. The static loading force error is ≤ ±0.5%FS, and the coupling interference is ≤3%. The static and quasi-static loading accuracy of each degree of freedom (including torque and non-torque) is significantly improved. This is attributed to the high stiffness and low coupling characteristics of the parallel mechanism combined with the precise force control and decoupling algorithm of the magnetorheological fluid (MRF) actuator.

[0085] With a dynamic response time (90% rise time of step load) ≤ 40ms, the millisecond-level rheological response characteristics of magnetorheological fluid (MRF) are the core guarantee, which can accurately reproduce the multi-directional composite loads experienced by wind turbine gearboxes in actual operation.

[0086] For extreme conditions such as sudden changes in wind speed, it can accurately simulate the amplitude, waveform and duration of impact loads. The experimental data such as the correlation coefficient or root mean square error of the time-domain waveform of key load channels (such as the equivalent load at key bearings) are more than 90% consistent with the actual operating data, providing a more realistic test basis for the reliability assessment of wind turbine gearboxes.

[0087] 3. Intelligent system operation: The application of the central collaborative control strategy enables seamless collaborative work among the various modules of the test system. The MPC algorithm optimizes decision-making in a unified manner, ensuring that the main drive, multi-dimensional loading, and energy management are strictly synchronized in terms of time and objectives, avoiding control conflicts or response delays caused by traditional separate control.

[0088] The self-learning function of the control algorithm enables the system to adapt to the test requirements of different models and power wind turbine gearboxes. For new gearbox models, the system's adaptive adjustment time is reduced by more than 70%, eliminating the need for extensive parameter adjustments.

[0089] Multi-dimensional sensor arrays and real-time data processing capabilities enable comprehensive monitoring and fault early warning throughout the testing process, improving testing efficiency by over 40%. This is attributed to: 1) high-precision rapid loading reducing the number of iterations per test; 2) automated and intelligent control reducing manual intervention; and 3) comprehensive monitoring and early warning reducing downtime due to faults. This significantly shortens the testing cycle of wind turbine gearboxes, providing efficient technical support for new product development and quality inspection.

[0090] Example 3:

[0091] This invention provides a test system that can be applied to a dynamic loading test control method for wind turbine gearboxes based on energy feedback.

[0092] This method enables efficient bidirectional energy flow between the internal DC bus and the external power grid of the test system, significantly improving energy feedback efficiency and system energy utilization. It effectively suppresses power fluctuations on the grid side and reduces the impact on the grid, while reducing dependence on large-capacity grid-side converters, thus lowering equipment costs and multi-stage energy conversion losses.

[0093] Those skilled in the art will readily understand that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic loading test control method for wind turbine gearboxes based on energy feedback, characterized in that, The control method includes: Step S1: Retrieve pre-stored target load data to construct an inverse dynamics model, and predict gearbox output parameters based on the inverse dynamics model; the target load data is pre-stored gearbox normal operating condition load parameter data; the inverse dynamics model includes: Formula 1 in, I = diag([I1, I2, ..., I n Let θ be the diagonal matrix of moments of inertia, θ = [θ1, θ2, ..., θ]. n ]ᵀ is an n-dimensional angular displacement vector containing the input shaft, each gear pair, and the output shaft. C Here is the damping matrix. K Here is the stiffness matrix. To apply torque vector, For load torque vector; Formula 2 in, F The six-dimensional force / torque at the gearbox interface is derived from the target load data, i.e., the load applied to the gearbox; F = [F x , F y , F z M x M y M z ]ᵀ, where J is the Jacobian matrix, which maps small motions on the moving platform to the rate of change of the actuator rod length. τ Let τ be the force / torque vector of the six magnetorheological actuators, τ = [τ1, τ2, τ3, τ4, τ5, τ6]ᵀ; Formula 3 in, The electromagnetic torque generated by the motor, p This represents the number of pole pairs of the motor. The magnitude of the magnetic flux generated by the permanent magnet. For direct-axis and quadrature-axis inductors, For direct-axis and quadrature-axis currents; Step S2: Obtain the current system state and initial parameters, and predict the behavior of the tested system in a finite time domain in the future using a pre-established inverse dynamics model to obtain a predicted output sequence. Step S3: Based on the behavior and predicted output sequence of the tested system, construct a unified, constrained multi-objective optimization model, and generate loading instructions for the gearbox based on the multi-objective optimization model. Step S4: Control the test system to load the gearbox according to the generated loading command; Step S5: At the next sampling time, repeat steps S1-S4 based on the new measurement value.

2. The dynamic loading test control method for wind turbine gearbox based on energy feedback according to claim 1, characterized in that, In step S1, constructing the inverse dynamics model includes: Based on the target load data, inverse dynamics is used to solve the problem using the parameters of the gearbox, loading mechanism, and load motor under test. This allows for the calculation of the target speed and torque of the main drive motor, the target torque of the load motor, and the target force and displacement of several magnetorheological actuators.

3. The dynamic loading test control method for wind turbine gearbox based on energy feedback according to claim 1, characterized in that, Obtaining the current system status and initial parameters includes: Real-time data from the system's main drive module, composite loading module, and intelligent energy management unit are acquired through multi-dimensional sensors.

4. The dynamic loading test control method for wind turbine gearbox based on energy feedback according to claim 3, characterized in that: In step S2, the predicted output sequence includes: Y_pred(k) = [ŷ(k+1|k)ᵀ, ŷ(k+2|k)ᵀ, ..., ŷ(k+p|k)ᵀ]ᵀ; Where ŷ(k+i|k) contains all the target variables that need to be tracked, such as the main drive input torque τ_in_pred, the rotational speed ω_in_pred, and the six-dimensional force / torque [F] at the gearbox interface. x _pred, F_y_pred, F_z_pred, M x _pred, M_y_pred, M_z_pred].

5. The dynamic loading test control method for wind turbine gearbox based on energy feedback according to claim 1, characterized in that, The multi-objective optimization model in step S3 includes: Formula 4 Where k is the current control time, k+i is the future prediction time (i=1,2,…,p), X d (k+i) is the target load spectrum vector at time k+i in the future, which includes input torque, rotational speed and six-dimensional force and torque. X(k+i|k) is the predicted state based on the system dynamics model. Δu(k+j|k) is the control increment sequence. Q and R are weight matrices. p and m are the prediction time domain and control time domain, respectively.

6. A gearbox dynamic loading test system based on energy feedback, applicable to the wind turbine gearbox dynamic loading test control method based on energy feedback as described in any one of claims 1-4, characterized in that, The system is used to perform multi-condition dynamic simulation loading tests on the gearbox under test, and includes: The main drive module is connected to the input shaft of the gearbox; The composite loading module includes a multi-degree-of-freedom loading mechanism and a load motor. The multi-degree-of-freedom loading mechanism is connected to the outer wall of the gearbox and is used to apply a first force to the outside of the gearbox. The load motor is connected to the output shaft of the gearbox and is used to apply a second force to the output shaft of the gearbox. The first force is a non-rotational force, and the second force is a rotational force. The intelligent energy management unit includes an electrically connected supercapacitor module and a four-quadrant PWM converter. The supercapacitor module is connected to the internal DC power of the test system and the AC power of the external power grid through the four-quadrant PWM converter. The four-quadrant PWM converter is used to convert the internal DC power and the AC power of the external power grid to each other. The supercapacitor module is used to output or absorb the rapidly changing current of the system, or to smooth out the DC bus voltage fluctuations. A multi-dimensional sensor array, comprising multiple sets of sensors, is respectively set at various loading positions on the input shaft, output shaft, and side of the gearbox; The central coordinating controller is communicatively connected to the main drive module, the multi-degree-of-freedom loading mechanism, the load mechanism, the intelligent energy management unit, and the multi-dimensional sensor array. It is used to obtain the real-time loading parameters of the gearbox through the multi-dimensional sensor array and the power status of each power module through the intelligent energy management unit, so as to send loading control commands to the main drive module, the multi-degree-of-freedom loading mechanism, and the load motor to load the gearbox.

7. The gearbox dynamic loading test system based on energy feedback according to claim 6, characterized in that, The main drive module includes: The permanent magnet synchronous main motor and the vector frequency converter are electrically connected. The output shaft of the permanent magnet synchronous main motor is connected to the input shaft of the gearbox to simulate the aerodynamic torque input of the wind turbine. The vector frequency converter is also communicatively connected to the central coordinating controller to control the output frequency of the permanent magnet synchronous main motor.

8. The gearbox dynamic loading test system based on energy feedback according to claim 6, characterized in that, The multi-degree-of-freedom loading mechanism includes: Several magnetorheological fluid actuators connected in parallel are respectively connected between the loading interface plate on the gearbox housing and the foundation platform via ball joints. Each magnetorheological fluid actuator is equipped with a position and force sensor.

Images