Wind turbine generator transmission chain torsional vibration suppression method based on adaptive model predictive control

Abstract

The invention discloses a wind turbine generator transmission chain torsional vibration suppression method based on adaptive model predictive control. In order to solve the problem that most of existing wind turbine generator transmission chain torsional vibration suppression methods are difficult to adapt to transmission chain parameter changes, the wind turbine generator transmission chain torsional vibration suppression method based on self-adaptive model predictive control is designed by combining parameter identification and model predictive control and comprises the following steps that firstly, a transmission chain double-mass-block flexible shaft model is established; and establishing a transmission chain discrete state space equation and a parameter identification model thereof. Then, identifying a transmission chain stiffness coefficient based on an extended Kalman filter; adaptively updating prediction model parameters based on an identification result; and finally, solving a damping torque instruction based on model predictive control, and adding the damping torque instruction to an original electromagnetic torque instruction so as to suppress torsional vibration of the transmission chain. Compared with a torsional vibration suppression method based on a band-pass filter and conventional model predictive control, the method disclosed by the invention can effectively suppress the torsional vibration of the transmission chain under the condition that the parameters of the transmission chain are uncertain, reduces the fatigue load of the transmission chain, and shows relatively high robustness.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation technology, and in particular to a method for suppressing torsional vibration of the transmission chain of wind turbines based on adaptive model predictive control. Background Technology

[0002] As wind turbines develop towards higher power outputs, the operational safety of the drivetrain becomes increasingly prominent. As a typical underdamped system, the wind turbine drivetrain is susceptible to imbalances in mechanical and electromagnetic torques due to instantaneous and alternating loads such as wind speed fluctuations, leading to torsional vibration (TVR). TVR not only increases the load on drivetrain components (such as the main shaft, gearbox, and couplings) and other coupled components (such as blades and towers), causing damage to turbine parts, but also causes fluctuations in generator output power, reducing power quality. Therefore, researching strategies to suppress torsional vibration in wind turbine drivetrains is of significant theoretical and engineering importance for ensuring the safe and stable operation of the turbine, reducing maintenance costs, and extending its service life.

[0003] Domestic and international scholars have conducted considerable research on methods for suppressing torsional vibration in wind turbine drivetrains. Some studies focus on controlling mechanical torque through pitch control to dampen drivetrain oscillations; however, some literature indicates that controlling electromagnetic torque commands is more effective than controlling mechanical torque in increasing drivetrain damping and suppressing torsional vibration. A commonly used method for suppressing torsional vibration involves using a bandpass filter (BPF) to extract the natural oscillation frequency component of the drivetrain at generator speed, amplifying it, and then superimposing it onto the generator torque reference value to provide characteristic frequency electrical damping for the drivetrain; however, as a fixed-parameter controller, it relies heavily on accurate model parameters. Some researchers employ advanced controllers for torsional vibration suppression, such as sliding mode control (which may exhibit chattering) and H... ∞ Control (whose weighting function design depends on experience); in addition, Model Predictive Control (MPC) has also been introduced into torsional vibration suppression research.

[0004] Most of the methods described above design fixed-parameter controllers or controllers based on precise mechanistic models, whose control performance depends on the accuracy of the transmission chain model. However, in actual operation, due to the effects of gearbox aging and wear, as well as unmodeled dynamics, the parameters of the wind turbine transmission chain model exhibit time-varying characteristics, making them difficult to obtain accurately. Large-scale changes in the wind turbine's operating points and alterations in shaft parameters can lead to changes in torsional vibration frequency, causing a reduction in the performance of the torsional vibration suppressor or even its failure. Therefore, it is necessary to study torsional vibration suppression methods that can adapt to changes in transmission chain parameters. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the existing technology by providing a method for suppressing torsional vibration in the transmission chain of wind turbines based on Adaptive Model Predictive Control (AMPC).

[0006] The technical solution to achieve the objective of this invention is: a method for suppressing torsional vibration in the transmission chain of a wind turbine based on adaptive model predictive control, the method comprising the following steps:

[0007] Step 1: Establish a flexible shaft model with two mass blocks in the wind turbine transmission chain, and derive its discrete state space equations, i.e., the prediction model and parameter identification model.

[0008] Step 2: Identify the stiffness coefficient of the transmission chain based on the Extended Kalman Filter (EKF);

[0009] Step 3: Use the identified stiffness coefficients for adaptive updates of the prediction model;

[0010] Step 4: Based on the prediction model and model predictive control method, calculate the additional damping torque command for torsional vibration suppression;

[0011] Step 5: The additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression.

[0012] Furthermore, step 1 specifically includes:

[0013] Step 1-1: Convert the high-speed side variables of the wind turbine drive train to the low-speed side;

[0014] Steps 1-2: Establish the dynamic equations of the two mass blocks in the transmission chain and the differential equations of motion for the twist angle;

[0015] Steps 1-3: Based on steps 1-1 to 1-2, derive the discrete state-space equations and parameter identification models of the transmission chain;

[0016] In steps 1-3, the discrete state-space equation of the transmission chain, i.e., the prediction model, is as follows:

[0017]

[0018] In the formula, the state variable x = [Δω θ] s ] T System output y = [Δω]; Control input u = [T] d ]; Disturbance input v d =[T g T a ] T T a It is the aerodynamic torque of the wind turbine, Tg It refers to the electromagnetic torque of the wind turbine; x [k+1] x [k] These are the state variables at time k+1 and time k, respectively; u [k] v is the control input at time k; d[k] The perturbation input at time k; y [k] Let θ be the system output at time k; A, B, D, and C are the system coefficient matrices; T is the system offset length; Δω is the difference between the wind turbine speed and the generator speed, i.e., the torsional velocity; θ s For twist angle; T d For additional damping torque; J r J is the moment of inertia of the wind turbine. g K represents the generator's moment of inertia. s and D s These are the equivalent stiffness coefficient and damping coefficient of the transmission chain, respectively.

[0019] The parameter identification model is as follows:

[0020]

[0021] In the formula, x [k] x [k-1] The state variables at time k and time k-1 are respectively; x 1[k] x 2[k] x 3[k] x [k] The 1st, 2nd, and 3rd elements; x 1[k-1] x 2[k-1] x 3[k-1] x [k-1] The 1st, 2nd, and 3rd elements; u [k-1] This is the control input at time k-1; z [k] Let z be the observation vector at time k; 1[k] z 2[k] z [k] The first and second elements; f(*) and h(*) are the nonlinear state transition function and observation function, respectively; m [k-1] The process white noise at time k-1 follows an expectation of 0 and a covariance of Q. c The normal distribution; n [k] Let K be the measurement white noise at time k, which follows a function with expectation of 0 and covariance of R. c It follows a normal distribution.

[0022] Furthermore, step 2 specifically includes two stages: prediction and correction update.

[0023] The prediction equation is:

[0024]

[0025] The correction and update equation is:

[0026]

[0027] in,

[0028]

[0029] In the formula, The prior state estimate at time k; These are the posterior state estimates at time k and time k-1, respectively. Let P be the prior state estimation error covariance matrix at time k; [k] Let K be the posterior state estimation error covariance matrix at time k; [k] Let A be the Kalman gain at time k; l[k] Let f be the Jacobian matrix of the partial derivatives of f with respect to x, and let k be the state transition matrix at time k. For A l[k] transpose of M; [k] Let f be the Jacobian matrix of the partial derivatives of f with respect to m. For M [k] Transpose of; Let be the Jacobian matrix of the partial derivatives of h with respect to m, and represent the observation matrix at time k. for transpose of N; [k] Let h be the Jacobian matrix of the partial derivatives of h with respect to n. For N [k] The transpose of ; I is a 3×3 identity matrix; f1, f2, f3 represent the 1st, 2nd, and 3rd nonlinear equations of the nonlinear state transition function f(), respectively; h1, h2 represent the 1st and 2nd nonlinear equations of the nonlinear observation function h(), respectively; x1, x2, x3 represent the 1st, 2nd, and 3rd elements of the state variable x, respectively; m1, m2, m3 represent the 1st, 2nd, and 3rd elements of the process white noise m, respectively; n1, n2 represent the 1st and 2nd elements of the measurement white noise n, respectively. They are respectively The 1st, 2nd, and 3rd elements.

[0030] Furthermore, step 2 also includes: introducing a first-order low-pass filter to smooth the identification results of the extended Kalman filter.

[0031] Furthermore, in step 3, the identified stiffness coefficients are used for adaptive updating of the prediction model, expressed as:

[0032]

[0033] In the formula, A1 is the posterior estimate of the transmission chain stiffness coefficients identified by EKF.new A1 represents the prediction model after updating the model parameters, while A2 represents the prediction model before the update.

[0034] Furthermore, step 4 specifically includes:

[0035] Step 4-1: Predict the system output at p future time points based on the updated prediction model from Step 3.

[0036] Step 4-2: Perform feedback correction on the predicted output of Step 4-1;

[0037] Step 4-3: Perform rolling optimization based on the value function and constraints to obtain the additional damping torque command value.

[0038] Furthermore, in step 4-1, the predicted system output for the next p time steps is:

[0039]

[0040] In the formula, p is the prediction time domain; m is the control time domain; Y is the predicted system output vector for the next p time steps; y(k+1), y(k+2), and y(k+p) are the system outputs at time steps k+1, k+2, and k+p, respectively; x [k] v is the state vector at time k; d[k] Let U be the disturbance input vector at time k; U is the predicted system control input vector for the next p times; u(k), u(k+1), and u(k+m-1) are the system control inputs at times k, k+1, and k+m-1, respectively; F, G, and H are coefficient matrices.

[0041] Furthermore, the feedback correction mechanism in step 4-2 is as follows:

[0042]

[0043] In the formula, e(k) is the prediction error at time k; Δω a (k) and Δω p (k) represents the actual and predicted torsional velocities at time k, respectively; Δω p (k+i) represents the predicted torsional velocity at time k+i; Δω c (k+i) represents the corrected predicted torsional velocity at time k+i; α i This is the feedback coefficient.

[0044] Furthermore, the value function and constraints in step 4-3 are as follows:

[0045]

[0046] In the formula, J is the value function; q and r are the weighting coefficients of the torsional load and damping torque amplitude of the transmission chain, respectively; Δω(k+i) is the torsional velocity at time k+i; T d T d max T d min These are the additional damping torque and its upper and lower limits, respectively; ΔT d ΔT d max ΔT d min These represent the rate of change of the additional damping torque and its upper and lower limits, respectively.

[0047] For the perturbation input v in the prediction model d[k] After compensation, the value function J is equivalent to:

[0048]

[0049] In the formula, y [k] Let k be the system output vector at time k. For y [k] The transpose of ; Q is the weight coefficient matrix consisting of q elements; R is the weight coefficient matrix consisting of r elements;

[0050] Let Y = Fx [k] +GU+Hv d[k] Substituting into the above equation and simplifying, we obtain the equivalent quadratic programming form of the value function as follows:

[0051]

[0052] Furthermore, in step 5, the additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression, as shown below:

[0053] T e * =T g * +T d

[0054] In the formula, T g * The electromagnetic torque command for the original torque controller; T d Additional damping torque command calculated for the adaptive model predictive controller; T e * This is the electromagnetic torque command after adding damping torque.

[0055] Compared with the prior art, the significant advantages of this invention are:

[0056] (1) This invention combines the online parameter identification capability of the extended Kalman filter with the constraint optimization and predictive control capability of model predictive control to form an adaptive closed-loop optimization control. Through online real-time parameter identification and dynamic updating of the predictive model, the controller achieves the adaptation of the time-varying characteristics of key parameters of the transmission chain, reduces the dependence on the accurate model, and solves the problem of performance degradation of existing fixed parameter controllers when facing transmission chain aging, wear, and changes in operating conditions. Compared with the traditional torsional vibration suppression method, it has higher robustness.

[0057] (2) The present invention suppresses torsional vibration of the transmission chain based on adaptive model predictive control. It has multi-objective optimization capability and multi-constraint processing capability, and can optimize and solve for the actual working conditions of wind turbine units.

[0058] (3) The present invention performs first-order low-pass filtering on the identification results of the extended Kalman filter, effectively filtering out high-frequency noise and disturbances that may exist in the identification process, ensuring that the updated prediction model parameters are smooth and reliable, and improving the robustness and engineering practicality of the control system.

[0059] (4) This invention is based on the existence of a perturbation input term v in the prediction model. d[k] The present invention, based on the equivalent derivation of the value function, performs feedforward compensation for the disturbance term during the rolling optimization process. Existing model predictive control-based transmission chain torsional vibration suppression methods do not consider compensating for the disturbance term. Therefore, compared to existing torsional vibration suppression methods, the transmission chain torsional vibration suppression method based on adaptive model predictive control provided by this invention improves the prediction accuracy of the prediction model and enhances the torsional vibration suppression performance through model updates and disturbance term compensation.

[0060] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0061] Figure 1 This is a control block diagram of the wind turbine drive train torsional vibration suppression method based on adaptive model predictive control according to the present invention.

[0062] Figure 2 Here is a block diagram of a dual-mass flexible shaft model of the transmission chain and its transfer function in one embodiment.

[0063] Figure 3 This is a schematic diagram of the transmission chain stiffness coefficient identification principle based on an extended Kalman filter in one embodiment.

[0064] Figure 4 This is a schematic diagram of a model prediction controller in one embodiment.

[0065] Figure 5 This is a schematic diagram of turbulent wind speed used for simulation calculation in one embodiment.

[0066] Figure 6 This is a diagram showing the identification results of the transmission chain stiffness coefficient in one embodiment of the present invention.

[0067] Figure 7 This is a comparison graph showing the effect of the present invention and the conventional method in reducing the torsional speed fluctuation of the transmission chain in one embodiment.

[0068] Figure 8 This is a comparison diagram showing the effect of the present invention and the conventional method in reducing torque ripple of the transmission chain spindle in one embodiment.

[0069] Figure 9 This is a comparison graph showing the results of the present invention and the conventional method in reducing the amplitude of the spindle torque spectral density in one embodiment. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0071] It should be noted that if the embodiments of the present invention involve descriptions such as "first" and "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include at least one of those features. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0072] To address the problem that most current methods for suppressing torsional vibration in wind turbine drive trains are unable to adapt to changes in drive train parameters, this invention combines parameter identification and model predictive control to design a wind turbine drive train torsional vibration suppression method based on adaptive model predictive control.

[0073] In one embodiment, combined Figure 1 This paper provides a method for suppressing torsional vibration in the drive train of a wind turbine based on adaptive model predictive control. The method includes the following steps:

[0074] Step 1: Establish a flexible shaft model with two mass blocks in the wind turbine transmission chain and derive its discrete state space equations, namely the prediction model and parameter identification model.

[0075] Step 2: Identify the stiffness coefficient of the transmission chain based on the extended Kalman filter;

[0076] Step 3: Use the identified stiffness coefficients for adaptive updates of the prediction model;

[0077] Step 4: Based on the prediction model and model predictive control method, calculate the additional damping torque command for torsional vibration suppression;

[0078] Step 5: The additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression.

[0079] Furthermore, step 1 specifically includes:

[0080] Step 1-1: Convert the high-speed side variables of the wind turbine drive train to the low-speed side;

[0081] Steps 1-2: Establish the dynamic equations of the two mass blocks in the transmission chain and the differential equations of motion for the twist angle;

[0082] Steps 1-3: Based on steps 1-1 to 1-2, derive the discrete state-space equations and parameter identification models of the transmission chain;

[0083] Here, in steps 1-3, combined Figure 2 The discrete state-space equation of the transmission chain, i.e., the prediction model, is as follows:

[0084]

[0085] In the formula, the state variable x = [Δω θ] s ] T System output y = [Δω]; Control input u = [T] d ]; Disturbance input v d =[T g T a ] T T a It is the aerodynamic torque of the wind turbine, T g It refers to the electromagnetic torque of the wind turbine; x [k+1] x [k] These are the state variables at time k+1 and time k, respectively; u [k] v is the control input at time k; d[k] The perturbation input at time k; y [k] Let θ be the system output at time k; A, B, D, and C are the system coefficient matrices; T is the system offset length; Δω is the difference between the wind turbine speed and the generator speed, i.e., the torsional velocity; θ s For twist angle; T d For additional damping torque; J r J is the moment of inertia of the wind turbine. g K represents the generator's moment of inertia. s and D s These are the equivalent stiffness coefficient and damping coefficient of the transmission chain, respectively.

[0086] The parameter identification model is as follows:

[0087]

[0088] In the formula, x [k] x [k-1] The state variables at time k and time k-1 are respectively; x 1[k] x 2[k] x 3[k] x [k] The 1st, 2nd, and 3rd elements; x 1[k-1] x 2[k-1] x 3[k-1] x [k-1] The 1st, 2nd, and 3rd elements; u [k-1] This is the control input at time k-1; z [k] Let z be the observation vector at time k; 1[k] z 2[k] z [k] The first and second elements; f(*) and h(*) are the nonlinear state transition function and observation function, respectively; m [k-1] The process white noise at time k-1 follows an expectation of 0 and a covariance of Q. c The normal distribution; n [k] Let K be the measurement white noise at time k, which follows a function with expectation of 0 and covariance of R. c It follows a normal distribution.

[0089] Furthermore, in one embodiment, combined with Figure 3 Step 2 specifically includes two stages: prediction and correction / update.

[0090] The prediction equation is:

[0091]

[0092] The correction and update equation is:

[0093]

[0094] in,

[0095]

[0096] In the formula, The prior state estimate at time k; These are the posterior state estimates at time k and time k-1, respectively. Let P be the prior state estimation error covariance matrix at time k; [k] Let K be the posterior state estimation error covariance matrix at time k; [k]Let A be the Kalman gain at time k; l[k] Let f be the Jacobian matrix of the partial derivatives of f with respect to x, and let k be the state transition matrix at time k. For A l[k] transpose of M; [k] Let f be the Jacobian matrix of the partial derivatives of f with respect to m. For M [k] transpose; Let be the Jacobian matrix of the partial derivatives of h with respect to m, and represent the observation matrix at time k. for transpose of N; [k] Let h be the Jacobian matrix of the partial derivatives of h with respect to n. For N [k] The transpose of ; I is a 3×3 identity matrix; f1, f2, f3 represent the 1st, 2nd, and 3rd nonlinear equations of the nonlinear state transition function f(), respectively; h1, h2 represent the 1st and 2nd nonlinear equations of the nonlinear observation function h(), respectively; x1, x2, x3 represent the 1st, 2nd, and 3rd elements of the state variable x, respectively; m1, m2, m3 represent the 1st, 2nd, and 3rd elements of the process white noise m, respectively; n1, n2 represent the 1st and 2nd elements of the measurement white noise n, respectively. They are respectively The 1st, 2nd, and 3rd elements.

[0097] Furthermore, in one embodiment, in order to reduce the interference of high-frequency noise, step 2 further includes: introducing a first-order low-pass filter to smooth the identification results of the extended Kalman filter.

[0098] Preferably, the low-pass filter can be expressed as:

[0099]

[0100] In the formula, G LPF (s) is the transfer function of the low-pass filter; τ is the time constant; s is the Laplace variable.

[0101] Furthermore, in one embodiment, in step 3, the identified stiffness coefficients are used for adaptive updating of the prediction model, expressed as:

[0102]

[0103] In the formula, A1 is the posterior estimate of the transmission chain stiffness coefficients identified by EKF. new A1 represents the prediction model after updating the model parameters, while A2 represents the prediction model before the update.

[0104] Furthermore, in one embodiment, combined with Figure 4 Step 4 specifically includes:

[0105] Step 4-1: Predict the system output at p future time points based on the updated prediction model from Step 3.

[0106] Step 4-2: Perform feedback correction on the predicted output of Step 4-1;

[0107] Step 4-3: Perform rolling optimization based on the value function and constraints to obtain the additional damping torque command value.

[0108] Preferably, in some embodiments, in step 4-1, the predicted system output for the next p time periods is:

[0109]

[0110] In the formula, p is the prediction time domain; m is the control time domain; Y is the predicted system output vector for the next p time steps; y(k+1), y(k+2), and y(k+p) are the system outputs at time steps k+1, k+2, and k+p, respectively; x [k] v is the state vector at time k; d[k] Let U be the disturbance input vector at time k; U is the predicted system control input vector for the next p times; u(k), u(k+1), and u(k+m-1) are the system control inputs at times k, k+1, and k+m-1, respectively; F, G, and H are coefficient matrices.

[0111] Preferably, in some embodiments, the feedback correction mechanism in step 4-2 is as follows:

[0112]

[0113] In the formula, e(k) is the prediction error at time k; Δω a (k) and Δω p (k) represents the actual and predicted torsional velocities at time k, respectively; Δω p (k+i) represents the predicted torsional velocity at time k+i; Δω c (k+i) represents the corrected predicted torsional velocity at time k+i; α i This is the feedback coefficient.

[0114] Preferably, in some embodiments, the value function and constraints in step 4-3 are as follows:

[0115]

[0116] In the formula, J is the value function; q and r are the weighting coefficients of the torsional load and damping torque amplitude of the transmission chain, respectively; Δω(k+i) is the torsional velocity at time k+i; T d T dmax T d min These are the additional damping torque and its upper and lower limits, respectively; ΔT d ΔT d max ΔT d min These represent the rate of change of the additional damping torque and its upper and lower limits, respectively.

[0117] The rolling optimization process is as follows:

[0118] Because the prediction model contains a perturbation input v d[k] Therefore, this disturbance input term needs to be compensated. The value function J is then equivalent to:

[0119]

[0120] In the formula, y [k] Let k be the system output vector at time k. For y [k] The transpose of ; Q is the weight coefficient matrix consisting of q elements; R is the weight coefficient matrix consisting of r elements;

[0121] Let Y = Fx [k] +GU+Hv d[k] Substituting into the above equation and simplifying, we obtain the equivalent quadratic programming form of the value function as follows:

[0122]

[0123] Thus, the value function has been transformed into a standard quadratic programming form, which can be solved using the quadprog function in MATLAB; moreover, it is evident that this quadratic programming form includes consideration of the perturbation input term v. d[k] During the rolling optimization solution process, feedforward compensation can be applied to this term to ensure the prediction accuracy of the prediction model and improve control performance.

[0124] Furthermore, in step 5, the additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression, as shown below:

[0125] T e * =T g * +T d

[0126] In the formula, T g * The electromagnetic torque command for the original torque controller; T d Additional damping torque command calculated for the adaptive model predictive controller; T e* This is the electromagnetic torque command after adding damping torque.

[0127] In one embodiment, a wind turbine drive train torsional vibration suppression system based on adaptive model predictive control is provided, the system comprising:

[0128] The first module is used to establish a flexible shaft model of a dual-mass block in the wind turbine drive train and derive its discrete state-space equations, i.e., the prediction model and parameter identification model; it is also used to initialize the extended Kalman filter and set the initial values ​​of the stiffness coefficients and the state estimation error covariance matrix used for stiffness coefficient identification.

[0129] The second module is used to identify the stiffness coefficient of the transmission chain based on the extended Kalman filter;

[0130] The third module is used to adaptively update the prediction model by using the identified stiffness coefficients.

[0131] The fourth module is used to calculate the additional damping torque command for torsional vibration suppression based on the predictive model and model predictive control method.

[0132] The fifth module is used to transmit the additional damping torque command to the wind turbine torque controller for torsional vibration suppression.

[0133] Specific limitations regarding the torsional vibration suppression system for wind turbine drivetrain based on adaptive model predictive control can be found in the limitations of the torsional vibration suppression method for wind turbine drivetrain based on adaptive model predictive control mentioned above, and will not be repeated here. Each module in the aforementioned torsional vibration suppression system for wind turbine drivetrain based on adaptive model predictive control can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0134] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements:

[0135] Step 1: Establish a flexible shaft model with two mass blocks in the wind turbine transmission chain, and derive its discrete state space equations, i.e., the prediction model and parameter identification model.

[0136] Step 2: Identify the stiffness coefficient of the transmission chain based on the extended Kalman filter;

[0137] Step 3: Use the identified stiffness coefficients for adaptive updates of the prediction model;

[0138] Step 4: Based on the prediction model and model predictive control method, calculate the additional damping torque command for torsional vibration suppression;

[0139] Step 5: The additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression.

[0140] For specific limitations on each step, please refer to the limitations on the torsional vibration suppression method of wind turbine drive train based on adaptive model predictive control mentioned above, which will not be repeated here.

[0141] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being implemented when executed by a processor:

[0142] Step 1: Establish a flexible shaft model with two mass blocks in the wind turbine transmission chain, and derive its discrete state space equations, i.e., the prediction model and parameter identification model.

[0143] Step 2: Identify the stiffness coefficient of the transmission chain based on the extended Kalman filter;

[0144] Step 3: Use the identified stiffness coefficients for adaptive updates of the prediction model;

[0145] Step 4: Based on the prediction model and model predictive control method, calculate the additional damping torque command for torsional vibration suppression;

[0146] Step 5: The additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression.

[0147] For specific limitations on each step, please refer to the limitations on the wind turbine drive train torsional vibration suppression method based on adaptive model predictive control mentioned above, which will not be repeated here.

[0148] As a specific example, the invention will be further verified and illustrated in one embodiment.

[0149] A wind turbine simulation model was built in the OpenFAST-MATLAB / Simulink joint platform simulation environment, including a wind turbine with a rated power of 5MW. Specific parameters are shown in Table 1. The model was then configured as follows: Figure 5 The turbulent wind speed shown is used to induce torsional vibration in the wind turbine drive train; at the same time, the drive train stiffness coefficient is reduced by 50% to verify the adaptability of the method described in this invention in dealing with parameter uncertainties.

[0150] Table 1. Basic parameters of the wind turbine used in the simulation.

[0151]

[0152]

[0153] Depend on Figure 6 As can be seen, the parameter identifier based on extended Kalman filtering according to the method of this invention can accurately identify the stiffness coefficient of the transmission chain (with an error within 2%) and update the prediction model parameters; Figure 7 and Figure 8 It is evident that, compared to torsional vibration suppression methods based on bandpass filters without additional damping and torsional vibration suppression based on conventional model predictive control, the torsional vibration suppression method based on adaptive model predictive control described in this invention can effectively reduce the fluctuations in torsional speed and spindle torque, and reduce the torsional load on the transmission chain. Figure 9 This further verifies that the method described in this invention can effectively reduce the spectral density amplitude of the spindle torque at the torsional vibration frequency.

[0154] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention without departing from its spirit and scope should be included within the protection scope of the present invention.

Claims

1. A method for suppressing torsional vibration in the transmission chain of a wind turbine based on adaptive model predictive control, characterized in that, The method includes the following steps: Step 1: Establish a flexible shaft model with two mass blocks in the wind turbine transmission chain and derive its discrete state space equations, namely the prediction model and parameter identification model. Step 2: Identify the stiffness coefficient of the transmission chain based on the extended Kalman filter; Step 3: Use the identified stiffness coefficients for adaptive updates of the prediction model; Step 4: Based on the prediction model and model predictive control method, calculate the additional damping torque command for torsional vibration suppression; Step 5: The additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression.

2. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 1, characterized in that, Step 1 specifically includes: Step 1-1: Convert the high-speed side variables of the wind turbine drive train to the low-speed side; Steps 1-2: Establish the dynamic equations of the two mass blocks in the transmission chain and the differential equations of motion for the twist angle; Steps 1-3: Based on steps 1-1 to 1-2, derive the discrete state-space equations and parameter identification models of the transmission chain; In steps 1-3, the discrete state-space equation of the transmission chain, i.e., the prediction model, is as follows: In the formula, the state variable x = [Δω θ] s ] T System output y = [Δω]; Control input u = [T] d ]; Disturbance input v d =[T g T a ] T T a It is the aerodynamic torque of the wind turbine, T g It refers to the electromagnetic torque of the wind turbine; x [k+1] x [k] These are the state variables at time k+1 and time k, respectively; u [k] v is the control input at time k; d[k] The perturbation input at time k; y [k] Let θ be the system output at time k; A, B, D, and C are the system coefficient matrices; T is the system offset length; Δω is the difference between the wind turbine speed and the generator speed, i.e., the torsional velocity; θ s For twist angle; T d For additional damping torque; J r J is the moment of inertia of the wind turbine. g K represents the generator's moment of inertia. s and D s These are the equivalent stiffness coefficient and damping coefficient of the transmission chain, respectively. The parameter identification model is as follows: In the formula, x [k] x [k-1] The state variables at time k and time k-1 are respectively; x 1[k] x 2[k] x 3[k] x [k] The 1st, 2nd, and 3rd elements; x 1[k-1] x 2[k-1] x 3[k-1] x [k-1] The 1st, 2nd, and 3rd elements; u [k-1] This is the control input at time k-1; z [k] Let z be the observation vector at time k; 1[k] z 2[k] z [k] The first and second elements; f(*) and h(*) are the nonlinear state transition function and observation function, respectively; m [k-1] The process white noise at time k-1 follows an expectation of 0 and a covariance of Q. c The normal distribution of n; [k] Let K be the measurement white noise at time k, which follows a function with expectation of 0 and covariance of R. c It follows a normal distribution.

3. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 1, characterized in that, Step 2 specifically includes two stages: prediction and correction / update. The prediction equation is: The correction and update equation is: in, In the formula, The prior state estimate at time k; These are the posterior state estimates at time k and time k-1, respectively. Let P be the prior state estimation error covariance matrix at time k; [k] Let K be the posterior state estimation error covariance matrix at time k; [k] Let A be the Kalman gain at time k; l[k] Let f be the Jacobian matrix of the partial derivatives of f with respect to x, and let k be the state transition matrix at time k. For A l[k] transpose of M; [k] Let f be the Jacobian matrix of the partial derivatives of f with respect to m. For M [k] Transpose of; Let be the Jacobian matrix of the partial derivatives of h with respect to m, and represent the observation matrix at time k. for transpose of N; [k] Let h be the Jacobian matrix of the partial derivatives of h with respect to n. For N [k] The transpose of ; I is a 3×3 identity matrix; f1, f2, f3 represent the 1st, 2nd, and 3rd nonlinear equations of the nonlinear state transition function f(), respectively; h1, h2 represent the 1st and 2nd nonlinear equations of the nonlinear observation function h(), respectively; x1, x2, x3 represent the 1st, 2nd, and 3rd elements of the state variable x, respectively; m1, m2, m3 represent the 1st, 2nd, and 3rd elements of the process white noise m, respectively; n1, n2 represent the 1st and 2nd elements of the measurement white noise n, respectively. They are respectively The 1st, 2nd, and 3rd elements.

4. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 3, characterized in that, Step 2 also includes: introducing a first-order low-pass filter to smooth the identification results of the extended Kalman filter.

5. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 1, characterized in that, In step 3, the identified stiffness coefficients are used for adaptive updates of the prediction model, expressed as: In the formula, A1 is the posterior estimate of the transmission chain stiffness coefficients identified by EKF. new A1 represents the prediction model after updating the model parameters, while A2 represents the prediction model before the update.

6. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 1, characterized in that, Step 4 specifically includes: Step 4-1: Predict the system output at p future time points based on the updated prediction model from Step 3. Step 4-2: Perform feedback correction on the predicted output of Step 4-1; Step 4-3: Perform rolling optimization based on the value function and constraints to obtain the additional damping torque command value.

7. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 2 or 6, characterized in that, In step 4-1, the predicted system output for the next p time steps is: In the formula, p is the prediction time domain; m is the control time domain; Y is the predicted system output vector for the next p time steps; y(k+1), y(k+2), and y(k+p) are the system outputs at time steps k+1, k+2, and k+p, respectively; x [k] v is the state vector at time k; d[k] Let U be the disturbance input vector at time k; U is the predicted system control input vector for the next p times; u(k), u(k+1), and u(k+m-1) are the system control inputs at times k, k+1, and k+m-1, respectively; F, G, and H are coefficient matrices.

8. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 6, characterized in that, The feedback correction mechanism in step 4-2 is as follows: In the formula, e(k) is the prediction error at time k; Δω a (k) and Δω p (k) represents the actual and predicted torsional velocities at time k, respectively; Δω p (k+i) represents the predicted torsional velocity at time k+i; Δω c (k+i) represents the corrected predicted torsional velocity at time k+i; α i This is the feedback coefficient.

9. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 7, characterized in that, The value function and constraints in step 4-3 are as follows: In the formula, J is the value function; q and r are the weighting coefficients of the torsional load and damping torque amplitude of the transmission chain, respectively; Δω(k+i) is the torsional velocity at time k+i; T d T d max T d min These are the additional damping torque and its upper and lower limits, respectively; ΔT d ΔT d max ΔT d min These represent the rate of change of the additional damping torque and its upper and lower limits, respectively. For the perturbation input v in the prediction model d[k] After compensation, the value function J is equivalent to: In the formula, y [k] Let k be the system output vector at time k. For y [k] The transpose of ; Q is the weight coefficient matrix consisting of q elements; R is the weight coefficient matrix consisting of r elements; Let Y = Fx [k] +GU+Hv d[k] Substituting into the above equation and simplifying, we obtain the equivalent quadratic programming form of the value function as follows:

10. The method for suppressing torsional vibration of wind turbine drive train based on adaptive model predictive control according to claim 1, characterized in that, In step 5, the additional damping torque command is transmitted to the wind turbine torque controller for torsional vibration suppression, as shown below: T e * =T g * +T d In the formula, T g * The electromagnetic torque command for the original torque controller; T d Additional damping torque command calculated for the adaptive model predictive controller; T e * This is the electromagnetic torque command after adding damping torque.

Images