Variable-speed variable-pitch wind generating set torque control method and system and computer readable medium

Abstract

The invention provides a variable-speed variable-pitch wind generating set torque control method and system and a computer readable medium, and the method comprises the steps: designing a rolling optimization torque and variable pitch instruction based on model predictive control (MPC) and sliding mode surface constraint on the basis of constructing a discretization model comprising a pneumatic mechanism, a transmission chain and a variable pitch mechanism; the robustness is enhanced in combination with the self-adaptive sliding mode surface design, the control gain is dynamically adjusted through fuzzy logic, and the power tracking precision and the mechanical loss suppression are further considered; meanwhile, the delay is reduced with the assistance of wind speed prediction feed-forward compensation, the method has high robustness and adaptive capacity, wind speed fluctuation and parameter time variation can be effectively handled, equipment is protected while the power generation efficiency is guaranteed, and an efficient and reliable solution is provided for wind power generation control.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation control, especially the torque control of a generating set, and particularly to a torque control method, system and computer-readable medium for a variable-speed variable-pitch wind turbine generator set. Background Art

[0002] The pitch control system of a wind turbine generator set is the core part of the wind turbine control system, which plays a key role in the safe, stable and efficient operation of the unit. Stable pitch control is a difficult point in the research of control technology for large wind turbine generator sets. Variable-speed variable-pitch regulation combines the advantages of variable-speed fixed-pitch control and fixed-speed variable-pitch control, and is the mainstream pitch control strategy. The main goal of a variable-speed variable-pitch unit is to maximize energy capture at different wind speeds, make the unit operate on the ideal power curve of the wind speed model, and ensure the safe and stable operation of the unit at the same time. Currently, typical control strategies include control methods such as PID-based control, model predictive control (MPC), sliding mode control (SMC), fuzzy logic or neural network.

[0003] Traditional PID control is the first proposed control strategy. The control system has a simple structure and is easy to implement. It adjusts the torque error through proportional, integral and derivative links to maintain the power tracking target. However, it is difficult for the PID parameters to remain optimal under large wind speed changes / change rates, resulting in insufficient dynamic response and problems such as overshoot or oscillation. Moreover, the adaptability of PID to nonlinear systems is insufficient, especially the control effect is poor in the case of sudden wind speed changes or turbulence. In traditional model predictive control (MPC) strategies, an accurate model is used to predict future states and optimize control inputs. There are often uncertainties in the models of wind turbine generator sets, such as changes in aerodynamic parameters, resulting in a decline in the performance of the accurate model of MPC, and defects in the accuracy and robustness of predictive control. Sliding mode control (SMC) has strong adaptability and robustness. By designing a sliding mode surface, the system trajectory is forced to converge to the desired state, and it has a good suppression ability for parameter changes and external disturbances. However, sliding mode control will introduce chattering phenomena, especially in actual systems, which will affect the service life of mechanical components, such as increased wear of the pitch mechanism. At the same time, designing appropriate sliding mode surfaces and control laws requires a lot of experience, is not easy to implement and adjust, and affects the control effect. Summary of the Invention

[0004] In view of the defects and deficiencies existing in the prior art, the purpose of the present invention is to provide a torque control method and system for a variable-speed variable-pitch wind turbine generator set. Through a hierarchical control architecture and multi-technology integration, a full closed-loop system from data acquisition, model identification to intelligent control is constructed to achieve the dual goals of efficient and stable power generation and equipment protection.

[0005] According to the first aspect of the object of the present invention, a torque control method for a variable-speed variable-pitch wind turbine generator set is proposed, and its specific implementation includes the following steps:

[0006] Step 1: Obtain the state sampling signals of the wind turbine generator set, including the wind speed signal, the unit speed, the torque, and the pitch angle, align the unit speed, the torque, and the pitch angle with the wind speed signal, and calculate the tip speed ratio λ;

[0007] Step 2: Construct an improved blade element momentum model Pmech = f(C p , β, λ) to characterize the dynamic correlation relationship between the wind energy capture efficiency C p , the tip speed ratio λ, and the pitch angle β, and update the parameter estimation values in real time based on the recursive least squares method to realize online identification and update of C p ;

[0008] Step 3: Construct a multi-objective prediction model based on the improved blade element momentum model, the dynamic model of the generator set drive train, and the pitch actuator model, and perform discretization processing using the forward Euler method to obtain a discretized state space model;

[0009] Step 4: Design an adaptive sliding mode surface with an integral term based on the power tracking error, eliminate the steady-state error through the integral term, and realize the fuzzy logic dynamic adjustment through the time-varying gain;

[0010] Step 5: Take the dynamically adjusted sliding mode surface as the core constraint, optimize the objective function of the discretized state space model and transform it into a convex optimization for solution, and design based on the continuous reaching law, use the hyperbolic tangent function to approximate the sign function to realize continuous chattering suppression, and output the optimal control sequence for the next N steps; and

[0011] Step 6: According to the optimal control sequence, perform coordinated control of the pitch torque and the torque.

[0012] According to the second aspect of the object of the present invention, a computer system is proposed, including:

[0013] One or more processors;

[0014] A memory that stores operable instructions, and the instructions, when executed by the one or more processors, cause the one or more processors to perform operations, and the operations include the process of executing the aforementioned torque control method for the variable-speed variable-pitch wind turbine generator set.

[0015] According to the third aspect of the object of the present invention, a computer-readable medium storing software is proposed, and the software includes instructions that can be executed by one or more computers, and the instructions, when executed by the one or more computers, perform the process of executing the aforementioned torque control method for the variable-speed variable-pitch wind turbine generator set.

[0016] Based on the torque control method of the variable-speed and variable-pitch wind turbine generator set according to the embodiments of the present invention above, the coordinated operation of torque and pitch angle is realized based on the predictive control strategy of adaptive sliding mode. The rolling optimization of MPC is used to handle multivariable constraints, improve the dynamic response speed, and reduce overshoot and oscillation. At the same time, adaptive sliding mode is adopted to compensate for model uncertainty and disturbances. Through integral term compensation, fuzzy logic dynamic gain adjustment, and continuous reaching law optimization, a robust control core with both steady-state accuracy and dynamic adaptability is constructed to suppress the jitter effects of random wind speed fluctuations, time-varying characteristics of blade aerodynamic parameters (such as power coefficient attenuation caused by aging and icing), and uncertainties of drive train damping on the control algorithm. Furthermore, wind speed prediction information is introduced to drive feedforward control, and the control strategy is adjusted in advance. When the wind speed change is predicted, the torque set value is adjusted in advance to compensate for disturbances in advance and smooth the power output and mechanical load, thereby improving the control accuracy.

[0017] It should be understood that all combinations of the foregoing concepts and additional concepts described in greater detail below can be regarded as part of the inventive subject matter of the present disclosure as long as such concepts do not contradict each other. In addition, all combinations of the claimed subject matter are regarded as part of the inventive subject matter of the present disclosure.

[0018] The foregoing and other aspects, embodiments, and features of the teachings of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of exemplary embodiments, will be apparent from the following description or will be learned through practice of the specific embodiments according to the teachings of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in various figures may be represented by the same reference numeral. For clarity, not every component is labeled in each figure. Now, embodiments of various aspects of the present invention will be described by way of example and with reference to the drawings.

[0020] Figure 1 is a schematic flow chart of the torque control method of the variable-speed and variable-pitch wind turbine generator set according to the embodiments of the present invention.

[0021] Figure 2 is a schematic diagram of the principle of the torque control method of the variable-speed and variable-pitch wind turbine generator set according to the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] In order to better understand the technical content of the present invention, specific embodiments are hereby given and described in conjunction with the accompanying drawings as follows.

[0023] Aspects of the present invention are described with reference to the accompanying drawings, in which numerous illustrative embodiments are shown. Embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of a number of ways, because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone, or in any suitable combination with other aspects disclosed by the present invention.

[0024] {Embodiment 1}

[0025] According to the torque control method of a variable-speed variable-pitch wind turbine generator set according to an embodiment of the present invention, on the basis of constructing a discretized model including an aerodynamic, a drive train, and a pitch mechanism, a model predictive control (MPC) and a sliding mode surface constraint rolling optimization are designed to optimize the torque and pitch commands, and an adaptive sliding mode surface is designed to enhance the robustness, and the control gain is dynamically adjusted by fuzzy logic to further balance the power tracking accuracy and mechanical loss suppression; at the same time, a wind speed prediction feedforward compensation is supplemented to reduce the delay, which has high robustness and adaptive ability, can effectively cope with wind speed fluctuations and parameter time-variation, protect the equipment while ensuring the power generation efficiency, and provide an efficient and reliable solution for wind power generation control.

[0026] Combined with the attached Figure 1 , Figure 2 The torque control method of the variable-speed variable-pitch wind turbine generator set shown in the embodiment, its specific implementation includes the following steps:

[0027] Step 1: Obtain the state sampling signals of the wind turbine generator set, including the wind speed signal, the unit speed, the torque, and the pitch angle, align the unit speed, the torque, and the pitch angle with the wind speed signal, and calculate the tip speed ratio λ;

[0028] Step 2: Construct an improved blade element momentum model Pmech = f(C p , β, λ) to characterize the dynamic correlation relationship between the wind energy capture efficiency (also known as the power coefficient) C p and the tip speed ratio λ and the pitch angle β, and update the parameter estimation value in real time based on the recursive least squares method to realize the online identification and update of C p ;

[0029] Step 3: Construct a multi-objective prediction model based on the improved blade element momentum model, the dynamic model of the generator drive train, and the pitch actuator model, and perform discretization processing using the forward Euler method to obtain a discretized state space model;

[0030] Step 4: Design an adaptive sliding mode surface with an integral term based on the power tracking error, eliminate the steady-state error through the integral term, and realize the fuzzy logic dynamic adjustment through the time-varying gain;

[0031] Step 5: Taking the dynamically adjusted sliding mode surface as the core constraint, optimizing the objective function of the discretized state space model and transforming it into convex optimization for solution, and based on the continuous reaching law design, using the hyperbolic tangent function to approximate the sign function to achieve continuous chattering suppression, and outputting the optimal control sequence for the next N steps; and

[0032] Step 6: According to the optimal control sequence, perform coordinated control of pitch moment and torque.

[0033] As an optional implementation manner, in Step 1, aligning the unit speed, torque, and pitch angle with the wind speed signal, and calculating the tip speed ratio λ, including:

[0034] Taking the time stamp of the wind speed signal as the reference, and according to the sampling frequency of 100 Hz, that is, the time step Δt = 0.01 s, linearly interpolating the unit speed ω(t), torque τ(t), and pitch angle β(t) so that the collected signals are at the same time point t i = i·Δt aligned; ω(t), τ(t), and β(t) respectively represent the sampled values at the corresponding time point t;

[0035] Calculating the tip speed ratio λ, λ = ω(t)*R / v(t), where R represents the radius of the wind turbine blade, in m; v(t) represents the wind speed at the i-th moment, in m / s.

[0036] It should be understood that in the optional embodiments of the present invention, on the basis of collecting data, further preprocessing of the original sensor signals (such as ω(t), τ(t), β(t), v(t)) can be performed, especially denoising processing. For example, using the wavelet threshold denoising (sym4 wavelet basis, 5-layer decomposition, with approximate symmetry, suitable for transient signals) algorithm to perform 5-layer wavelet decomposition on the original signal to obtain high-frequency coefficients (noise is mainly concentrated in the high frequency), and then applying the soft threshold function (threshold set to 3 times the noise standard deviation) to the high-frequency coefficients, and retaining the low-frequency trend component; then reconstructing the signal to obtain a smooth signal sequence. Thus, separating noise and effective signals through multi-resolution analysis, combining hard threshold processing to retain low-frequency components, achieving noise removal and signal detail retention (true wind speed trend).

[0037] On the basis of the denoised data, further perform data synchronization and resampling, that is, synchronize according to the sampling frequency of 100 Hz, and linearly interpolate ω(t), τ(t), and β(t) respectively with the time stamp of the wind speed signal as the reference to ensure that all signals are aligned at the same time point.

[0038] Thus, a unified data set is obtained, which is beneficial for subsequent modeling and data processing.

[0039] As an alternative implementation, in step 2, an improved blade element momentum model is constructed:

[0040] Pmech = f(C p , β, λ);

[0041] used to characterize the dynamic correlation between the wind energy capture efficiency C p and the tip speed ratio λ and the pitch angle β, and based on the recursive least squares method, the parameter estimation value is updated in real time to achieve online identification and update of C p , including:

[0042] Step 2-1, construct an improved blade element momentum model Pmech = f(Cp, β, λ), establish a direct mapping between the wind energy capture efficiency Cp and the input variables λ and β, and the specific expression is as follows:

[0043] Pmech(t) = ρπR 2 v(t) 3 Cp

[0044] where ρ represents the air density, ρ = 1.225 kg / m 3 , taking the standard air density (which can be corrected according to temperature and altitude);

[0045] Step 2-2, approximate the wind energy capture efficiency C p as a polynomial form:

[0046] Cp = k1λ + k2β + k3λ 2 + k4β 2 + k5λβ

[0047] In the formula, k1, k2, k3, k4, k5 are the coefficients of the polynomial respectively, reflecting the coupling effect of the tip speed ratio λ and the pitch angle β on the aerodynamic efficiency. Thus, the complex Cp(λ, β) curve is simplified into a linear combination that can be identified online;

[0048] Step 2-3, according to the polynomial form expression of C p and the measured power Pmech(t) = ρπR 2 v(t) 3 C p , construct a linear regression model: Y = Φ(t)θ + ∈(t);

[0049] The output Y(t) = Pmech(t), representing the measured mechanical power, is calculated based on the measured τ(t)*ω(t) (based on the core formula of the BEM theory, substituting Cp calculated according to the measured values in real time);

[0050] The special vector Φ(t) = [λ(t), β(t), λ(t) 2 , β(t), λ(t)β(t)];

[0051] The parameter vector θ = [k1, k2, k3, k4, k5] T ;

[0052] where ∈(t) represents the modeling error, which follows a normal distribution;

[0053] Step 2-4: Use the recursive least squares method to perform real-time updates at a update frequency of 10 Hz, obtain the real-time updated parameter vector θ, and thereby dynamically update C p Model parameters to avoid characteristic changes caused by factors such as blade contamination and aging, and preferentially track the latest operating conditions.

[0054] As an optional implementation, in Step 3, construct a multi-objective prediction model based on the improved blade element momentum model, the dynamic model of the generator drive train, and the pitch actuator model, and perform discretization using the forward Euler method to obtain the discretized state space model, including:

[0055] Step 3-1: Describe the mechanical dynamics based on the Newton-Euler equations and construct the dynamic model of the generator drive train, expressed as follows:

[0056]

[0057] where T aero = P mech / ω, representing the aerodynamic torque, with the unit of N·m; T gen = K τ u T , representing the electromagnetic torque, K τ is the generator torque coefficient, and u T is the torque control input; J is the moment of inertia (including the blade, gearbox, and generator), with the unit of kg·m 2 ; B is the viscous damping coefficient, with the unit of N·m·s / rad;

[0058] Step 3-2: Establish a pitch actuator model based on a first-order inertia link, expressed as follows:

[0059]

[0060] where T β represents the pitch time constant, used to characterize the mechanical response speed; K β represents the gain, used to characterize the input-output ratio; u β is the pitch control input; in this example, the optional pitch time constant T β = 0.2 s, and the gain K β = 1.2, used to control the input-output ratio;

[0061] Step 3-3: According to the time step Δt (adopting 0.01 s, Nyquist frequency ≥ 50 Hz, taking into account both calculation accuracy and real-time performance), the forward Euler method is used for discretization to obtain the discretized state space model, including:

[0062] Define the state variables as follows:

[0063]

[0064] Obtain the discrete state equation:

[0065] The rotational speed is updated as:

[0066] The pitch angle is updated as:

[0067] Where,

[0068] Thus, through explicit discretization, the calculation is simplified and the time-consuming iteration of the implicit method is avoided.

[0069] As an optional implementation manner, in Step 4, the adaptive sliding mode surface introducing an integral term is designed based on the power tracking error, and the steady-state error is eliminated through the integral term, and the fuzzy logic dynamic adjustment is realized through the time-varying gain, including:

[0070] Step 4-1: Define the power tracking error e(k):

[0071] e(k) = Pref(k) - Pmech(k)

[0072] Where, Pref(k) represents the reference power, which is determined according to the maximum power tracking curve preset according to the wind speed or the grid command;

[0073] It should be understood that in the embodiments of the present invention, the power generation control target is to make the actual power generation power close to the ideal value. If there is an error, it is controlled and adjusted through the core constraint of the sliding mode surface;

[0074] Step 4-2: Design the adaptive sliding mode surface σ(k) containing an integral term:

[0075]

[0076] In the formula, the integral term ∑e(i)Δt is used to eliminate the steady-state error, represents the error change rate;

[0077] Where, k(t) represents the time-varying gain: k(t) = k0 + Δk(t), k0 represents the initial value of the time-varying gain, and Δk(t) represents the gain adjustment amount;

[0078] According to the power tracking error e(k) and the error change rate Dynamically adjust Δk(t), and further dynamically update the time-varying gain k(t) to achieve a dynamically adjusted sliding mode surface σ(k), enhancing the adaptability to different working conditions.

[0079] Thus, by designing the time-varying gain, the dynamically updated positive adjustment amount is directly superimposed on the initial gain, and is dynamically adjusted through fuzzy logic to adapt to different wind speed conditions (such as increasing the gain during strong gusts to accelerate the response);

[0080] Among them, in step 4-2, the process of dynamically updating the time-varying gain k(t) includes:

[0081] Input the power tracking error e(k) and the error change rate Normalize to [-1, 1];

[0082] According to the design of the fuzzy rule base, output the gain adjustment amount Δk(t), with the universe of discourse [-0.5, 0.5];

[0083] Among them, the triangular function is divided into {negative large, negative small, zero, positive small, positive large}, and the fuzzy rule base is designed as: if e(k) is positive large and positive large, then the gain adjustment amount Δk(t) is positive large, enhancing the integral effect; if e(k) is negative small and zero, then the gain adjustment amount Δk(t) is negative small, suppressing overshoot; and

[0084] Dynamically update the time-varying gain k(t) according to k(t) = k0 + Δk(t), where the initial value k0 of the time-varying gain is set to 0.1, and dynamically adjust the sliding mode surface σ(k).

[0085] Thus, through the rule base, an adaptive dynamic update logic of "strong integration for large errors and weak integration for small errors" is realized, and the response speed and overshoot are balanced through the dynamic adaptive change of the time-varying gain.

[0086] As an optional implementation method, in step 5, with the dynamically adjusted sliding mode surface as the core constraint, optimize the objective function of the discretized state space model and transform it into a convex optimization for solution, so as to predict the control input sequence {u T (k), u β (k)} within the prediction time domain, and based on the design of the continuous reaching law, use the hyperbolic tangent function to approximate the sign function to achieve continuous chattering suppression, and output the optimal control sequence for the next N steps, including:

[0087] Step 5-1: With the dynamically adjusted sliding mode surface as the core constraint, determine the optimization objective function as:

[0088]

[0089] Among them, \(Q\) represents the power tracking error weight matrix, and its value is \(diag(1,\cdots,1)\); \(P\) represents the control input smoothness weight matrix, and its value is \(diag(r\) T ,\cdots,r β ); The torque control input weight \(r\) T = 10 -6 , to avoid drastic changes in current; the pitch rate weight \(r\) β = 10 -3 , aiming to protect the servo motor;

[0090] represents the torque limit constraint condition, which is determined by the generator capacity (rated torque), such as \([-3000N\cdot m, 3000N\cdot m]\);

[0091] \(\vert u_{\beta}(k + i)\vert\leq15^{\circ} / s\) represents the pitch rate limit constraint condition, which determines the mechanical safety constraint (usually the maximum rate is constrained at \(15^{\circ} / s\) to avoid the blades rotating too fast);

[0092] Thus, through the power tracking term, \(P_{mech}(t)\) is forced to approach \(P_{ref}\), the steady-state deviation is eliminated through the integral characteristic of the sliding mode surface, and at the same time, the frequent torque adjustment (reducing the gearbox impact) and fast pitch (reducing mechanical wear) are punished through the control smoothness term;

[0093] Step 5 - 2: The optimization objective function is quadratic and the constraints are linear, which is transformed into a quadratic programming QP problem and solved using the interior point method. In this example, IPOPT is used to solve it, and the open-source library qpoases is called. The input matrix dimension is \(2N\times2N\). When the step size \(N = 10\), the input matrix dimension is \(20\times20\), and the single - solution time \(<3ms\) (meeting the \(100Hz\) control period);

[0094] Moreover, based on the continuous reaching law, the sign function is approximated by the hyperbolic tangent function to suppress the chattering of the sliding mode control (i.e., the mechanical chattering caused by high - frequency switching):

[0095] \(u(k)=-\kappa\tanh(\sigma(k) / F)\)

[0096] Among them, \(\kappa\) represents the control strength, and its value in this example is \(500\), which is used to control the response strength to the sliding mode surface error; \(F\) represents the boundary layer thickness, and its value is \(0.1\), which is used to control the reduction of high - frequency switching, that is, to control the reduction of high - frequency oscillation through smooth switching.

[0097] Thus, through the solution, the optimal control input sequence \(\{u\) T (k),u β (k)\} for the next \(N\) steps is output. For example, when \(N = 10\), it represents the optimal control input sequence within the predicted \(100ms\) time domain.

[0098] It should be understood that in other embodiments, iteration can also be performed through other convex optimization solution methods, such as solving by the gradient descent method, which will not be elaborated here.

[0099] In an alternative embodiment, implicit constraint processing can be further performed by restricting the pitch angle range and rotational speed safety interval in the aerodynamic model. For example, the pitch angle range is [0°, 90°], and the rotational speed safety interval is [0, ωrated]. State boundary constraints (saturation constraints) are added to the prediction model to prevent state overshoot.

[0100] Thus, the adaptive sliding mode surface design forces the system state to converge along the preset sliding mode surface through the integral term and fuzzy logic, ensuring that the power tracking error remains within ±5% of the rated power even when there are large perturbations in the blade aerodynamic parameters (the error of traditional PID control expands to ±10% - ±15%).

[0101] For external disturbances of sudden gusts (wind speed mutation rate > 5 m / s), the system response delay < 0.1 second. Chattering is suppressed through the hyperbolic tangent function, and the torque fluctuation amplitude is controlled within ±10% of the rated torque, effectively protecting the gearbox and generator.

[0102] As an alternative implementation, in step 6, according to the optimal control sequence, pitch torque and torque coordinated control is performed, including the coordinated control of the generator converter and the pitch servo motor, as follows:

[0103] Step 6-1, Torque command conversion, tracking T through the current loop of the generator converter cmd :

[0104] T cmd = K τ u T (k)

[0105] Step 6-2, Based on the predicted wind speed value v pred (k + 1) drive feedforward compensation to obtain the feedforward torque T ff :

[0106] Feedforward torque T ff = 1 / 2 * ρπR 2 v pred (k + 1) 3 Cp(λpred, β(k)) / λpred;

[0107] where λpred = ω(k)R / v pred (k + 1), used to compensate for the change in aerodynamic torque to reduce the impact of feedback delay (about 50 ms);

[0108] Step 6-3, Conversion of the pitch angle control command β cmd Conversion:

[0109] β cmd = β trim + K β u β (k)

[0110] β trim represents the pitch angle reference angle, which is preset according to the average wind speed (for example, β = 0° at low wind speeds and increases at high wind speeds to unload power); trim K β u β (k) represents the real-time adjustment amount of the pitch angle, which is calculated from the optimization result; K β represents the pitch mechanism gain, which can be calibrated through actual measurement;

[0111] Step 6 - 4: Send the torque control command T cmd + T ff to the converter for torque control, and send the pitch angle control command β cmd to the pitch angle servo motor for pitch angle control, for example, send it to the actuators (converter, pitch angle servo motor) through the CAN bus to achieve the coordinated action of the actual torque and the pitch angle.

[0112] In the example of the present invention, ARIMA or LSTM can be used to predict the future 1 step (0.01 s). Taking ARIMA as an example, the wind speed in the previous 10 seconds is input to predict the future 1 step, with an accuracy of ±0.3 m / s.

[0113] In the embodiment of the present invention, only the optimized first-step control input {u T (k), u β (k)} is executed, and the subsequent sequence is discarded, that is, only the first term is applied to the actual control system; at the next sampling moment k + 1, the optimization problem is re-solved based on the new measured state (x(k + 1)) (rolling update).

[0114] Therefore, in the torque control strategy of the variable-speed and variable-pitch wind turbine of the present invention, an aerodynamic model (calculating the power generated by the wind blowing the blades), a drive train model (simulating the mechanical drive inside the generator, speed and torque transmission), and a pitch model (simulating the actuator for adjusting the blade angle, focusing on the blade rotation speed and response time) are used to construct the operation of the wind turbine. By designing an error regulator, i.e., a sliding surface, to stabilize the generated power, an integral term control is creatively designed and introduced into the sliding surface, which can not only quickly respond to errors but also eliminate long-term steady-state errors. And the parameters of the sliding mode surface are dynamically adjusted using fuzzy logic to achieve adaptive adjustment, so as to achieve a fast response of dynamic adaptability, suppress the jitter influence of random wind speed fluctuations, the time-varying characteristics of blade aerodynamic parameters, and the uncertainty of drive train damping on the control. Finally, through model predictive control (MPC), the next control commands (torque and blade angle) are optimized by rolling, so as to achieve both accurate power tracking (less error) and reduced mechanical losses (such as avoiding frequent blade angle changes and sudden torque changes that damage gears), and the optimal solution is calculated through the "convex optimization" algorithm to ensure the stable convergence of the system.

[0115] Through the coordinated control of pitch torque and torque of the present invention: for torque control, the calculated optimal torque command is sent to the converter to adjust the current of the generator to make the torque follow the target; for pitch control, the blade angle is adjusted in two parts, "reference angle" (preset in advance according to the average wind speed) + "real-time adjustment amount" (according to the current error), and the torque and blade angle are adjusted in advance according to the predicted wind speed for feedforward compensation to reduce the fluctuations caused by delay.

[0116] {Embodiment 2}

[0117] In this embodiment, combined with the torque control method of the variable-speed and variable-pitch wind turbine in the above embodiment, real-time performance data is input, including power error e(t), torque fluctuation Δτ(t), and pitch action times Npitch, and performance evaluation is carried out based on the root mean square error RMSE and the mean absolute value MAE:

[0118]

[0119] M represents the data points (36,000) in the most recent 1 hour.

[0120] Then, the fuzzy rules are adjusted by particle swarm optimization (PSO), and the central values of the fuzzy logic membership functions are optimized (a total of 10 parameters, 5 input membership centers + 5 output membership centers). The membership function used is: F = 0.7·RMSE + 0.3·MAE;

[0121] The PSO parameter settings are as follows:

[0122] Population size = 50, number of iterations = 100;

[0123] The inertial weight w linearly decreases from 0.9 to 0.4, and the learning factors \(c1 = c2 = 2\).

[0124] Execute PSO optimization and verify the new parameters at 2 am every day (low wind speed period, small system load) (for example, run for 10 minutes, if the RMSE drops > 5%, update it, otherwise retain the original parameters), write the optimal parameters into the fuzzy rule base, and restart the controller. Thus, through the tuning of PSO, it adapts to the wind speed distribution characteristics in different seasons (for example, when the turbulence intensity is high in winter, automatically enhance the weight of the error change rate), and improves the power tracking accuracy during long-term operation.

[0125] Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Those with ordinary knowledge in the technical field to which the present invention pertains can make various modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the scope defined in the claims.

Claims

1. A method for controlling torque of a variable speed and variable pitch wind turbine generator set, characterized in that: The following steps are involved: Step 1, obtaining the state sampling signal of the wind turbine generator set, including the wind speed signal, the unit speed, the torque and the pitch angle, and aligning the unit speed, the torque and the pitch angle with the wind speed signal to calculate the tip speed ratio λ; Step 2: Construct an improved blade element momentum model Pmech = f(C p ,β,λ), used to characterize the wind energy capture efficiency C p The dynamic relationship between the tip speed ratio λ and the pitch angle β is calculated, and the parameter estimation value is updated in real time based on the recursive least squares method to realize online identification and update C p ; Step 3: construct a multi-objective prediction model based on the improved blade element momentum model, the generator set transmission chain dynamic model and the pitch actuator model, and use the forward Euler method for discretization to obtain a discretized state space model; Step 4: Based on the power tracking error, an adaptive sliding surface with an integral term is designed to eliminate the steady-state error through the integral term, and the fuzzy logic dynamic adjustment is realized through the time-varying gain; Step 5: Taking the dynamically adjusted sliding surface as the core constraint, optimizing the objective function of the discretized state space model and converting it into convex optimization for solving, and based on the continuous reaching law design, using the hyperbolic tangent function to approximate the sign function, achieving continuous chattering suppression, and outputting the optimal control sequence for the next N steps; and Step 6: According to the optimal control sequence, the pitch and torque are coordinated and controlled.

2. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: The step of aligning the unit rotation speed, torque and pitch angle with the wind speed signal and calculating the tip speed ratio λ includes: Based on the wind speed signal timestamp, the sampling frequency is unified to 100 Hz, that is, the time step Δt = 0.01s, and the unit speed ω(t), torque τ(t) and pitch angle β(t) are linearly interpolated so that the collected signals are at the same time point t i =i·Δt alignment; ω(t), τ(t), β(t) represent the sampling values ​​at the corresponding time t respectively; Calculate the tip speed ratio λ, λ = ω(t)*R / v(t), where R represents the radius of the fan blade in m; v(t) represents the wind speed at time i in m / s.

3. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: In step 2, an improved blade element momentum model Pmech=f(C p ,β,λ), used to characterize the wind energy capture efficiency C p The dynamic relationship between the tip speed ratio λ and the pitch angle β is calculated, and the parameter estimation value is updated in real time based on the recursive least squares method to realize online identification and update C p ,include: Step 2-1: Construct an improved blade element momentum model Pmech=f(Cp,β,λ) to establish a direct mapping between the wind energy capture efficiency Cp and the input variables λ and β. The specific expression is as follows: Pmech(t)=ρπR 2 v(t) 3 Cp Where ρ represents the air density; Step 2-2: Wind energy capture efficiency C p Approximately in polynomial form: Cp=k1λ+k2β+k3λ 2 +k4β 2 +k5λβ In the formula, k1, k2, k3, k4, k5 are the coefficients of the polynomial respectively; Step 2-3, according to C p The polynomial expression and measured power Pmech(t)=ρπR 2 v(t) 3 C p , construct a linear regression model: Y = Φ(t)θ+∈(t); Output Y(t)=Pmech(t), represents the measured mechanical power, which is calculated based on the measured τ(t)*ω(t); Special vector Φ(t) = [λ(t), β(t), λ(t) 2 , β(t), λ(t)β(t)]; Parameter vector θ = [k1, k2, k3, k4, k5] T ; In the formula, ∈(t) represents the modeling error, which obeys the normal distribution; Step 2-4: Use the recursive least squares method to perform real-time update calculations at an update frequency of 10 Hz to obtain a real-time updated parameter vector θ.

4. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: In step 3, a multi-objective prediction model is constructed based on the improved blade element momentum model, the generator set transmission chain dynamic model and the pitch actuator model, and the forward Euler method is used for discretization processing to obtain a discretized state space model, including: Step 3-1: Construct the dynamic model of the generator transmission chain, which is expressed as follows: Among them, T aero =P mech / ω, represents the aerodynamic torque, the unit is N·m; T gen =K τ u T , represents the electromagnetic torque, K τ is the generator torque coefficient, u T is the torque control input; J is the moment of inertia, in kg·m 2 ; B is the viscous damping coefficient, unit is N·m·s / rad; Step 3-2: Establish a variable pitch actuator model based on the first-order inertia link, expressed as follows: Among them, T β Represents the pitch time constant, which is used to characterize the mechanical response speed; K β Indicates gain, which is used to characterize the input-output ratio; u β It is the pitch control input; Step 3-3: According to the time step Δt, the forward Euler method is used for discretization to obtain a discretized state space model, including: Define the state variables as follows: Obtain the discrete state equation: The speed is updated to: The pitch angle is updated as: in, 5. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: In step 4, the adaptive sliding surface with an integral term introduced based on the power tracking error design eliminates the steady-state error through the integral term, and realizes the dynamic adjustment of fuzzy logic through the time-varying gain, including: Step 4-1, define the power tracking error e(k): e(k)=Pref(k)-Pmech(k) Wherein, Pref(k) represents the reference power, which is determined according to the preset maximum power tracking curve of wind speed or the grid instruction; Step 4-2: Design the adaptive sliding surface σ(k) with integral term: In the formula, the integral term ∑e(i)Δt is used to eliminate the steady-state error, represents the error change rate; Wherein, k(t) represents the time-varying gain: k(t)=k0+Δk(t), k0 represents the initial value of the time-varying gain, and Δk(t) represents the gain adjustment amount; According to the power tracking error e(k) and the error change rate Dynamically adjust Δk(t), and further dynamically update the time-varying gain k(t) to realize the dynamically adjusted sliding surface σ(k) and enhance the adaptability to different working conditions.

6. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 5, characterized in that: In step 4-2, the process of dynamically updating the time-varying gain k(t) includes: Input power tracking error e(k) and error change rate Normalized to [-1,1]; According to the fuzzy rule base design, the output gain adjustment value Δk(t) is in the domain [-0.5, 0.5]; Among them, the triangular function is divided into {negative large, negative small, zero, positive small, positive large}, and the fuzzy rule base is designed as follows: if e(k) is positive and large and If the positive value is large, the gain adjustment value Δk(t) is large, which enhances the integral effect; if e(k) is small and negative If Δk(t) is zero, the gain adjustment amount Δk(t) is small, which suppresses overshoot; and The time-varying gain k(t) is dynamically updated according to k(t)=k0+Δk(t), where the initial value k0 of the time-varying gain is set to 0.1, and the sliding surface σ(k) is dynamically adjusted.

7. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: In step 5, the dynamically adjusted sliding surface is used as the core constraint, the objective function of the discretized state space model is optimized and converted into a convex optimization for solving, and based on the continuous reaching law design, the hyperbolic tangent function is used to approximate the sign function to achieve continuous chattering suppression, and the optimal control sequence for the next N steps is output, including: Step 5-1: Taking the dynamically adjusted sliding surface as the core constraint, determine the optimization objective function as: Where Q represents the power tracking error weight matrix, and its value is diag(1,…,1); P represents the control input smoothness weight matrix, and its value is diag(r T ,…,r β ), r T =10 -6 , r β =10 -3 ; represents the torque limit constraint, which is determined by the generator capacity; |u β (k+i)≤15° / s represents the pitch rate limit constraint, which determines the mechanical safety constraint; Step 5-2: The optimization objective function is quadratic, the constraint is linear, and it is converted into a quadratic programming QP problem, which is solved using the interior point method or gradient descent method; Moreover, based on the continuous reaching law, the hyperbolic tangent function is used to approximate the sign function to suppress the chattering of the sliding mode control: u(k)=-κtanh(σ(k) / F) Among them, κ represents the control intensity, and F represents the boundary layer thickness, which is used to control and reduce high-frequency switching; By solving, the optimal control sequence {u T (k),u β (k)}.

8. The torque control method of a variable speed and variable pitch wind turbine generator set according to claim 1, characterized in that: In step 6, according to the optimal control sequence, the pitch moment and torque coordinated control is performed, including the coordinated control of the generator converter and the pitch servo motor, as follows: Step 6-1: Torque command conversion, tracking T through the current loop of the generator converter cmd : T cmd =K τ u T (k) Step 6-2: Based on the wind speed prediction value v pred (k+1) drive feedforward compensation, obtain feedforward torque T ff : Feedforward torque T ff =1 / 2*ρπR 2 v pred (k+1) 3 Cp(λpred,β(k)) / λpred; Where λpred=ω(k)R / v pred (k+1), used to compensate for the change of aerodynamic torque; Step 6-3: Pitch angle control command β cmd Conversion: b cmd =b trim +K β you β (k) β trim Indicates the pitch angle reference angle, which is preset according to the average wind speed; K β u β (k) represents the real-time adjustment amount of the pitch angle; Step 6-4: Send torque control command T cmd +T ff To the converter for torque control and to send pitch angle control instructions β cmd The pitch angle servo motor is used to control the pitch angle, thereby achieving coordinated action between actual torque and variable pitch angle.

9. A computer system, characterized in that: include: one or more processors; A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors are caused to perform operations, wherein the operations include the process of executing the method described in any one of claims 1 to 8.

10. A computer-readable medium storing software, characterized in that: The software includes instructions that can be executed by one or more computers, and when the instructions are executed by the one or more computers, the process of the method according to any one of claims 1 to 8 is performed.

Images