An excitation control method and system for a front-end speed-regulating wind turbine generator
By establishing an excitation winding model and an expanded state observer based on fractional calculus theory, combined with sliding mode control, the problems of model mismatch and multi-machine coupling disturbance in the excitation control of front-end speed-regulating wind turbine units were solved, achieving higher dynamic performance and stability, and enhancing adaptability to complex environments.
Patent Information
- Application Number
- CN202611192975.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-07
- Publication Date
- 2026-11-10
AI Technical Summary
In the current excitation control of front-end speed-regulating wind turbine units, integer-order models cannot accurately describe the non-local physical characteristics of the excitation winding, resulting in a structural mismatch between the controller and the real system. It is difficult to balance the estimation accuracy of slow-varying wind disturbances and the noise suppression capability, and low-frequency oscillations are difficult to suppress under multi-machine coupling conditions.
A fractional-order calculus theory is used to establish an excitation winding model. A fractional-order augmented state equation system is constructed by combining a fractional-order extended state observer and sliding mode control to obtain a comprehensive disturbance estimate. The disturbance is then compensated by a fractional-order sliding mode surface to obtain the final excitation control quantity, thereby realizing real-time estimation and compensation of mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance.
It improves the dynamic performance and system stability of wind turbine units, enhances their ability to withstand complex operating environments and multi-source disturbances, and ensures the safe and stable operation of the system under multi-unit grid connection conditions.
Smart Images

Figure CN123074332A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of wind turbine control technology, and in particular to an excitation control method and system for a front-end speed-regulating wind turbine. Background Technology
[0002] Front-end variable-speed wind turbines use hydraulic torque converters to regulate speed and are directly coupled to the grid using electrically excited synchronous generators, structurally overcoming the limitations of traditional units that rely on converters for grid connection. However, the actual operation of this type of unit faces severe challenges, as the random fluctuations in input wind energy, perturbations in system parameters, grid faults, and multi-unit coupling continuously affect the unit's dynamics.
[0003] Currently, the excitation control strategies widely adopted in industry are mainly PI / PID control and power system stabilizers based on linearized models. While existing fractional-order control research has been explored in the wind power field—for example, fractional-order sliding mode control is used for current control and power quality improvement of permanent magnet synchronous motors, and fractional-order active disturbance rejection control is used for parameter optimization of doubly-fed induction generators—these studies generally face the following technical shortcomings: Traditional excitation models use integer-order ordinary differential equations to describe the excitation windings, neglecting non-local physical characteristics such as core hysteresis memory effect and eddy current diffusion, leading to a structural mismatch between the controller's internal model and the actual system. Existing extended state observers are designed based on an integer-order framework, but their disturbance reconstruction channels suffer from phase margin degradation in the low-frequency band, making it difficult to balance the estimation accuracy and noise suppression capability of slowly varying wind disturbances. Existing fractional-order sliding mode control is mostly aimed at single-machine infinite bus systems, lacking a systematic compensation scheme for the strong coupling disturbances of power angle and power flow under multi-machine grid-connected conditions, easily inducing low-frequency oscillations under multi-machine cooperative operation conditions.
[0004] Therefore, there is an urgent need for an excitation control method that can match the fractional-order characteristics of the excitation winding from the physical mechanism level and take into account both local disturbances and cross-machine coupling disturbances, so as to improve the dynamic performance and system stability of the front-end speed-regulating wind turbine. Summary of the Invention
[0005] To address the aforementioned technical problems, this application provides an excitation control method and system for a front-end speed-regulating wind turbine, which solves problems such as the mismatch of integer-order model mechanism in existing excitation control, insufficient accuracy of disturbance observer for reconstructing slow-varying wind disturbances, difficulty in suppressing low-frequency oscillations under multi-machine coupling conditions, and inherent chattering in sliding mode control.
[0006] In a first aspect, this application provides an excitation control method for a front-end speed-regulating wind turbine, comprising: Step S1: Obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, establish a fractional excitation model of the excitation winding, and construct a fractional augmented state equation system in combination with the operating state parameters. Step S2: Obtain mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Construct a fractional extended state observer based on the fractional augmented state equations, and perform online estimation of the extended state variables to obtain the initial comprehensive disturbance estimate. Step S3: Based on the generator terminal voltage and the generator power angle, construct a fractional sliding mode surface using a fractional integral operator to obtain an equivalent control term. Based on the initial comprehensive disturbance estimate, construct a disturbance feedforward compensation term and perform disturbance compensation on the equivalent control term to obtain the excitation control quantity after disturbance. Step S4: Obtain the operating status deviation signal of the neighboring units in the multi-unit grid-connected environment, and calculate the equivalent network coupling disturbance as an additional extended state variable by combining the sensitivity coefficient of the power grid network Jacobian matrix, and obtain the preliminary updated comprehensive disturbance estimate. Step S5: Construct a comprehensive objective function for the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
[0007] Secondly, this application provides an excitation control system for a front-end speed-regulating wind turbine, comprising: The model building module is used to obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, a fractional excitation model of the excitation winding is established, and a fractional augmented state equation system is constructed in combination with the operating state parameters. The disturbance observation module is used to acquire mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Based on the fractional augmented state equations, a fractional extended state observer is constructed to estimate the extended state variables online and obtain the initial comprehensive disturbance estimate. The sliding mode control module is used to construct a fractional sliding mode surface using a fractional integral operator based on the generator terminal voltage and the generator power angle, obtain an equivalent control term, construct a disturbance feedforward compensation term based on the initial comprehensive disturbance estimate, perform disturbance compensation on the equivalent control term, and obtain the excitation control quantity after disturbance. The collaborative compensation module is used to acquire the operating status deviation signal of neighboring units in a multi-unit grid-connected environment, and calculate the equivalent network coupling disturbance as an additional extended state variable by combining the sensitivity coefficient of the power grid network Jacobian matrix to obtain the preliminary updated comprehensive disturbance estimate. The closed-loop optimization module is used to construct the comprehensive objective function of the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
[0008] Compared with the prior art, the beneficial effects of this application are at least as follows: This application constructs a fractional-order excitation model for the excitation winding by integrating fractional-order calculus theory. It fully considers the memory effects of magnetic circuit saturation and hysteresis, as well as complex electromagnetic characteristics such as high-frequency skin effect and eddy current diffusion effect. This overcomes the shortcomings of traditional integer-order models in providing insufficient accuracy for wideband dynamic descriptions, thus improving the fidelity of the full-condition model. A fractional-order extended state observer is constructed to perform real-time online estimation of mechanical torque disturbances, excitation parameter perturbations, and network coupling disturbances. Combined with a fractional-order sliding surface and disturbance compensation mechanism, it effectively suppresses system parameter uncertainties and the influence of external disturbances, enhancing the dynamic adjustment quality and robust stability of voltage and power angle.
[0009] In a multi-unit grid-connected environment, this application acquires the operating state deviation signals of neighboring units and calculates equivalent coupling disturbances using the network Jacobian matrix sensitivity. These disturbances are then superimposed onto the observer as additional extended state variables, effectively decoupling and compensating for multi-unit coupling effects. This overcomes the limitations of traditional distributed control in terms of coordination and reduces the risk of oscillation coupling between units. By constructing a multi-objective comprehensive optimization function and combining frequency domain and pole stability constraints for global optimization, adaptive tuning and online updating of control parameters are achieved, ensuring the system's continuous adaptability to changes in operating conditions and parameter drift.
[0010] This application forms a complete technical closed loop from excitation system modeling, disturbance observation and compensation, multi-machine collaborative control to global parameter optimization. While improving the transient power angle stability and terminal voltage regulation accuracy of the front-end speed-regulating wind turbine, it effectively enhances the comprehensive resistance capability of the excitation control system to complex operating environments and multi-source disturbances, providing reliable technical support for the safe and stable operation of wind power under large-scale grid connection conditions, and has practical application value and promotion prospects. Attached Figure Description
[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 This is a flowchart of an excitation control method for a front-end speed-regulating wind turbine in an embodiment of this application; Figure 2 This is a three-layer architecture diagram of the excitation control method according to an embodiment of this application; Figure 3 This is a schematic diagram of the multi-machine cooperative disturbance compensation mechanism architecture in an embodiment of this application; Figure 4 This is a schematic diagram of the signal flow for the final excitation control quantity synthesis in an embodiment of this application; Figure 5 This is a comparison curve of the generator terminal voltage response in an embodiment of this application; Figure 6 This is a diagram showing the excitation voltage control output and chattering performance evaluation of an embodiment of this application; Figure 7 This is a schematic diagram of generator power angle deviation under cross-machine disturbance in multi-machine grid connection according to an embodiment of this application; Figure 8 This is a schematic diagram of the integrated perturbation reconstruction error of the fractional-order extended state observer in an embodiment of this application; Figure 9 This is a schematic diagram of the excitation control system of a front-end speed-regulating wind turbine according to an embodiment of this application. Detailed Implementation
[0013] This application provides an excitation control method and system for a front-end speed-regulating wind turbine. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0014] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1The excitation control method for a front-end speed-regulating wind turbine in this application embodiment includes: Step S1: Obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, establish a fractional excitation model of the excitation winding, and construct a fractional augmented state equation system in combination with the operating state parameters.
[0015] Step S1 involves constructing a fractional-order augmented state equation set, including: establishing a fractional-order excitation model based on fractional-order calculus theory, which includes the memory effect of magnetic circuit saturation and hysteresis, the high-frequency skin effect, and the eddy current diffusion effect; the fractional-order excitation model is as follows: ,in, for Caputo fractional differential operator (0 < <1), For fractional order, The excitation winding was identified through experiments. The generalized fractional inductance coefficient, This is the equivalent resistance of the excitation winding. For excitation current, Here, is the excitation voltage, and f is the generator's excitation winding; the generator power angle is selected based on the operating state parameters. Speed deviation q-axis transient electromotive force and excitation current As state variables, combined with the fractional excitation model, an augmented set of state equations containing fractional differential terms is constructed: ,in, The first derivative of the generator's power angle. This is the generator's rated synchronous angular frequency. For speed deviation, Let H be the first derivative of the speed deviation, and H be the unit's inertial time constant. The mechanical torque of the prime mover. Let be the electromagnetic torque of the generator, and D be the damping coefficient. Let d be the open-circuit transient time constant. , These are the d-axis synchronous reactance and transient reactance, respectively. The stator current is the d-axis current. The magnetizing electromotive force. The mutual inductance coefficient between the stator and rotor is... denoted as d-axis flux linkage.
[0016] Specifically, since the excitation winding of the front-end speed-regulating wind turbine is not an ideal integer-order inductor in nature, its magnetic circuit saturation and hysteresis effect give the core magnetization process a dependence on the historical trajectory of the excitation current. At the same time, the eddy current diffusion and magnetic field penetration behavior generated in the excitation winding and rotor shaft under high-frequency switching excitation follow the physical laws of fractional diffusion equations. Therefore, it is necessary to establish an excitation winding model based on fractional calculus theory. First, the operating parameters of the front-end speed-regulating wind turbine are acquired during operation, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage, and generator power angle. The excitation voltage is acquired in real-time by a voltage Hall sensor connected in parallel to the excitation winding end. The excitation current is acquired in real-time by a current Hall sensor connected in series in the excitation winding power supply circuit. The generator terminal voltage is obtained through a three-phase voltage transformer located on the generator output side. The actual mechanical speed is acquired by a speed sensor installed at the end of the generator rotor shaft, and the speed deviation is obtained by subtracting it from the rated speed. The rated speed refers to the standard mechanical rotational speed that the rotor should maintain when the wind turbine is operating at rated wind speed and rated power output. The q-axis transient electromotive force and generator power angle are both state quantities that cannot be directly measured. The q-axis transient electromotive force is calculated based on the stator voltage and current, transformed from the abc stationary coordinate system to the dq coordinate system where the rotor is synchronously rotating, after Parker transformation, and combined with the generator's d-axis transient reactance. The generator power angle is calculated and estimated online by substituting active power, reactive power, and terminal voltage parameters into the power angle equation. All the above operating state parameters are synchronously collected and stored in the controller using a unified sampling clock to ensure a consistent time base among the state variables.
[0017] A fractional-order excitation model for the excitation winding is established based on fractional-order calculus theory. This theory incorporates two physical mechanisms: the memory effect of magnetic circuit saturation and hysteresis, and the physical mechanisms of high-frequency skin effect and eddy current diffusion effect. The memory effect means that the current magnetic flux state of the core depends not only on the current excitation current but also on the past current history. The integral kernel function inherent in the fractional-order calculus operator has the ability to describe this long-term historical memory. The diffusion effect occurs during dynamic adjustment, generating complex eddy currents within the excitation winding and rotor shaft. The diffusion, heat transfer, and magnetic field penetration of these eddy currents in the metallic medium essentially conform to the fractional-order diffusion equation. Based on the aforementioned physical mechanisms of the memory effect of magnetic circuit saturation and hysteresis, as well as the high-frequency skin effect and eddy current diffusion effect, a fractional-order excitation model is established. This model uses fractional-order differential equations to characterize the dynamic voltage balance relationship of the excitation winding. The expression of this model is: To make the fractional order It can accurately characterize the hysteresis memory effect and electromagnetic diffusion characteristics of the excitation winding. Before establishing the fractional-order excitation model, the excitation winding is experimentally identified to determine the fractional order.
[0018] in, The excitation winding was identified through experiments, including: applying a wideband excitation signal covering the main operating frequency band to the excitation winding using a logarithmic sinusoidal sweep frequency method under generator shutdown or low load conditions. This included sinusoidal sweep signals, multi-frequency sinusoidal superposition signals, pseudo-random binary sequence signals, or step excitation signals. Simultaneously, excitation voltage and current response signals were acquired, the actual response frequency of the excitation winding was calculated, and experimental data on the actual amplitude-frequency and phase-frequency characteristics were obtained. A fractional-order frequency domain transfer function model of the excitation winding was constructed, the model response frequency of the excitation winding was calculated, and theoretical data on the amplitude-frequency and phase-frequency characteristics under the model framework were obtained. Parameter identification was performed using a nonlinear optimization algorithm, with the objective function being the minimization of amplitude and phase errors between the actual frequency response and the model frequency response. By performing joint optimization, the optimal fractional-order model parameters are obtained, thus completing the fractional-order process. The calibration.
[0019] Specifically, fractional order The value of the excitation winding is determined through experimental identification. This identification process is performed when the generator is in a shutdown state or under low load to avoid unnecessary interference from mechanical rotating parts to the electromagnetic measurement signal. At the start of the identification, a wideband excitation signal is applied to the generator's excitation winding through the excitation power supply. This excitation signal includes a sinusoidal sweep signal, a multi-frequency sinusoidal superposition signal, a pseudo-random binary sequence signal, or a step excitation signal. In this application, a logarithmic sinusoidal sweep method is used, so that the frequency of the excitation signal changes logarithmically with the scanning time. This achieves higher frequency resolution in the low-frequency band to ensure the identification accuracy of slowly changing dynamic characteristics, while maintaining sufficient scanning speed in the high-frequency band to shorten the overall identification time. The starting and ending frequencies of the logarithmic sinusoidal sweep need to cover the main operating frequency band of the excitation system so that the dynamic characteristics of the excitation winding are fully excited throughout the entire operating frequency band.
[0020] While applying the excitation signal, excitation voltage data are simultaneously acquired by voltage Hall sensors and current Hall sensors, respectively. and excitation current data The acquisition of both signals is triggered uniformly by the same sampling clock, and the sampling frequency is... Set to no less than the highest frequency of the excitation signal Ten times, that is, satisfying This ensures that at least ten sample points are collected in each cycle at the highest frequency component, which is sufficient to reconstruct the waveform characteristics of the signal without distortion. Multiple complete excitation cycles are collected continuously, and the data from multiple cycles are then averaged to reduce the impact of random noise on the identification results and improve the accuracy of parameter identification.
[0021] After continuously acquiring data for multiple excitation cycles, the excitation voltage and excitation current data are transformed from the time domain to the frequency domain, and the Fourier transform of the excitation voltage is calculated respectively. Fourier transform of excitation current The actual response frequency of the excitation winding is calculated based on the definition of the frequency response function. The frequency response function is the ratio of the Fourier transform of the excitation voltage to the Fourier transform of the excitation current, i.e. Where j is the imaginary unit, Let be the angular frequency variable. This calculation yields the actual amplitude-frequency characteristics of the excitation winding at various frequency points throughout the entire swept frequency band, i.e., the magnitude of the frequency response function. | and the actual phase-frequency characteristics, i.e., the argument of the frequency response function. The experimental data mentioned above formed the basis for subsequent model fitting.
[0022] While acquiring experimental response frequency data, a fractional-order frequency domain transfer function model of the excitation winding is constructed. This model is based on the Laplace domain form of the fractional-order excitation model, and its expression is: Where s is the Laplace operator, for this transfer function model, a sequence of frequency points exactly the same as the actual frequency sweep excitation is selected on the frequency axis, let s = The theoretical amplitude-frequency characteristics and theoretical phase-frequency characteristics of the model at each frequency point are calculated to obtain the theoretical response frequency data under the model framework.
[0023] Subsequently, a nonlinear optimization algorithm was used to identify the parameters of the fractional-order excitation model. The Levenberg-Marquardt algorithm was preferred, as it combines the advantages of gradient descent and Gauss-Newton methods, exhibiting good global convergence when far from the optimal solution and fast local convergence when close to the optimal solution. The objective function for parameter identification was defined as the weighted sum of the squared amplitude and phase errors between the experimentally measured response frequency and the theoretical response frequency calculated by the model. The optimization process was synchronously adjusted through iteration. , The values of the three parameters, α, etc., are chosen to gradually decrease the objective function value until it converges to the global minimum. During the iteration process, the search space of the fractional order α is constrained to the interval between 0 and 1 to ensure its physical rationality. The iteration stops when the objective function converges to below a preset threshold. The parameter combination obtained at this point is the optimal fractional order model parameter, and the specific value of the corresponding fractional order α is calibrated. The preset threshold is usually set to 1% to 5% of the initial value of the objective function, or determined based on the identifiable error lower limit calculated from the accuracy of the measurement sensor and the number of bits of the A / D conversion. The specific value is set comprehensively based on the signal-to-noise ratio of the actual measurement system and the model accuracy requirements required by the engineering.
[0024] Experimental calibration shows that the fractional order α of the excitation winding of a front-end speed-regulating wind turbine typically ranges from 0.8 to 1.0, with its specific value determined by the degree of magnetic circuit saturation, the excitation operating frequency, and the electromagnetic characteristics of the winding. Under conditions of weak magnetic circuit saturation and low excitation operating frequency, the historical memory effect of the excitation winding is weak, and the value of α is close to 1. In this case, the fractional order model can approximately degenerate into a traditional integer-order RL model. As the degree of magnetic circuit saturation increases, the excitation operating frequency rises, and the skin effect and eddy current diffusion effect gradually strengthen, the non-local memory characteristics of the excitation winding increase accordingly, and the corresponding fractional order α gradually decreases. To further establish a quantitative relationship between α and the operating state of the excitation winding, an empirical expression can be used: Where S is the magnetic saturation index and f is the excitation frequency. and These are proportionality coefficients obtained through experiments. This expression indicates that the higher the degree of magnetic saturation or the higher the excitation frequency, the more... The smaller the value, the more pronounced the fractional-order memory characteristics exhibited by the excitation winding. The fractional-order calibration is completed through the above experimental calibration procedure. The precise calibration enables the fractional-order excitation model to accurately match the real physical dynamic characteristics of a specific generator excitation winding, providing accurate model parameter support for the subsequent construction of the fractional-order augmented state equation system.
[0025] Based on the completion of the fractional-order excitation model, the generator power angle is selected from the acquired operating state parameters. Speed deviation q-axis transient electromotive force and excitation current As state variables, an augmented set of state equations containing fractional differential terms is constructed by combining the fractional excitation model: The third and fourth equations in the above augmented state equations both adopt... The use of the Caputo fractional differential operator instead of the traditional first-order differential allows the memory dynamic characteristics of the excitation circuit to be directly embedded into the state-space description of the system.
[0026] The fractional excitation model established in this step is integrated into the generator augmented state equations. This fractional model breaks through the limitations of traditional integer-order models in describing the slow-varying electromagnetic memory effect from a mechanistic perspective, providing a control internal model foundation that matches the real physical system for the subsequent design of fractional-order extended state observers and sliding mode controllers.
[0027] Step S2: Obtain mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Construct a fractional extended state observer based on the fractional augmented state equations, and perform online estimation of the extended state variables to obtain the initial comprehensive disturbance estimate.
[0028] Step S2 includes: acquiring the mechanical torque disturbance, excitation parameter perturbation, and network coupling disturbance, integrating them into a comprehensive disturbance, and defining it as an extended state variable; based on the fractional-order augmented state equations, introducing an observation error feedback correction term, and constructing a fractional-order extended state observer, which includes a state estimation channel and a disturbance reconstruction channel; the state estimation channel is used to estimate the q-axis transient electromotive force online. The perturbation reconstruction channel employs a fractional-order calculus structure for online reconstruction of the extended state variables. Numerical solutions are obtained using fractional integral operators to obtain initial estimates of the overall disturbance. ,in, for Caputo fractional differential operator of order 1. For fractional order, This is the observed estimate of the q-axis transient electromotive force. Let d be the open-circuit transient time constant. , These are the d-axis synchronous reactance and transient reactance, respectively. The magnetizing electromotive force. This represents the initial combined disturbance estimate for the expanded state variables. For observation error feedback correction term, To account for observation errors, the observer gain is selected based on the constraints. and ,in, For state error gain, This is the perturbation reconstruction gain.
[0029] Specifically, in actual operation, the front-end speed-regulating wind power generation system faces multiple uncertainties, including mechanical torque disturbances caused by random wind speed fluctuations, excitation parameter perturbations caused by temperature variations in generator excitation winding resistance and inductance variations due to magnetic circuit saturation, and network coupling disturbances transmitted through the transmission network from neighboring units under multi-unit grid-connected conditions. These disturbances collectively affect the dynamic path of the q-axis transient electromotive force (EMF), resulting in an unknown disturbance term in the mathematical model of this path. Since this disturbance term cannot be directly obtained through sensor measurements, an extended state observer needs to be constructed for its online estimation. Firstly, the aforementioned mechanical torque disturbances, excitation parameter perturbations, and network coupling disturbances are integrated into a comprehensive disturbance. This comprehensive disturbance reflects the equivalent impact of all unknown uncertainties on the dynamics of the q-axis transient EMF and is defined as an extended state variable.
[0030] Based on the fractional augmented state equations established in step S1, a fractional extended state observer is constructed. This observer includes two parallel channels: a state estimation channel and a disturbance reconstruction channel. The state estimation channel is used to estimate the q-axis transient electromotive force online. Its construction method involves introducing the extended state variables as known estimated values into the equations based on the fractional dynamic equations of the q-axis transient electromotive force in step S1, and superimposing an observation error feedback correction term to drive the estimated values to converge to the true values. Specifically, it is expressed as follows: When the estimated value deviates from the true value, the correction term will generate a correction signal proportional to the deviation, which is injected into the observer's state estimation channel, pushing the estimated value closer to the true value. The perturbation reconstruction channel uses a fractional-order calculus structure to reconstruct the extended state variables online, specifically expressed as: This channel uses fractional integral operators for numerical solutions to obtain initial estimates of the overall disturbance. Both of the above channels use The Caputo fractional differential operator is consistent with the differential order of the fractional augmented state equations in step S1, ensuring the matching of the observer's dynamic characteristics with the system model in the frequency domain.
[0031] Among them, the observer gain is selected based on the constraints. and Includes: defining the observer state error vector The state error vector includes the observation error of the q-axis transient electromotive force. Reconstruction error of combined disturbances Using the rate of change of the derivative of the extended state variable h(t) as the driving input, the error dynamic matrix form of the fractional-order extended state observer is obtained: ,in, The observer error state matrix, The perturbation input matrix; According to the stability theory of fractional linear systems, to ensure the asymptotic convergence of the observer's dynamic error, the eigenvalues of the state matrix... It must satisfy the fractional asymptotic stability condition, i.e. The eigenvalues are derived from the characteristic equation corresponding to the error state matrix: Determine; under the constraint that all eigenvalues satisfy the fractional asymptotic stability condition, select the observer gain according to the frequency domain tuning criterion. and .
[0032] Specifically, observer gain and The selection of the observer must be based on strict constraints to ensure the stability and estimation accuracy of the observer. First, the error dynamic equation of the observer is established, and the state error vector of the observer is defined as follows: The state error vector includes the observation error of the q-axis transient electromotive force. Reconstruction error of combined disturbances Using the rate of change of the derivative of the combined disturbance, h(t), as the driving input, the matrix form of the observer error dynamics is derived as follows: According to the Mittag-Leffler stability theory of fractional linear systems, to ensure that the dynamic error of the observer asymptotically converges to zero, the state matrix... All eigenvalues It must lie entirely within the fractional asymptotically stable region of the complex plane, which is determined by the constraint condition. The eigenvalues are derived from the characteristic equation corresponding to the error state matrix: It is determined that all eigenvalues must lie entirely within the fractional asymptotically stable region of the complex plane, and this stability criterion is consistent with that specified in step S1. The values are directly related. The closer the value is to 1, the closer the stability region is to the left half of the open plane of a traditional integer-order system; The smaller the value, the more the stability region shrinks towards the negative real axis, and the stricter the constraint on the observer gain.
[0033] Under the constraint that all the above eigenvalues satisfy the fractional asymptotic stability condition, the observer gain is selected according to the frequency domain tuning criterion during frequency domain tuning. and The frequency domain tuning criterion is the phase margin criterion, which requires that the phase margin of the observer estimation loop at the cutoff frequency be no less than 45°. Preferred at 45 60 Between these two ranges, this interval achieves the best trade-off between robust stability and dynamic response speed. It can withstand model perturbations and parameter drift in actual operating conditions while ensuring that the observer has sufficient phase following capability in the turbulent disturbance frequency band of the wind farm, providing a timely and consistent disturbance estimation signal for the subsequent feedforward compensation of the sliding mode control law. This is achieved by selecting a gain that satisfies the above eigenvalue constraints and frequency domain tuning criteria. and It can guarantee the finite-time convergence of observation errors under the Mittag-Leffler stability principle.
[0034] This step constructs a fractional-order extended state observer that includes state estimation and disturbance reconstruction channels based on the fractional-order augmented state equations, obtaining an initial comprehensive disturbance estimate, which provides real-time and high-quality disturbance information input for the feedforward compensation of the fractional-order sliding mode controller in subsequent steps.
[0035] Step S3: Based on the generator terminal voltage and generator power angle, construct a fractional sliding mode surface using a fractional integral operator to obtain the equivalent control term. Based on the initial comprehensive disturbance estimate, construct a disturbance feedforward compensation term and perform disturbance compensation on the equivalent control term to obtain the excitation control quantity after disturbance.
[0036] Step S3 includes: obtaining the generator terminal voltage deviation. and generator power angle deviation ,in, The target generator terminal voltage, This is the actual generator terminal voltage. For the target generator power angle, The actual generator power angle is used; based on the generator terminal voltage deviation and generator power angle deviation, a fractional integral sliding surface is constructed using fractional integral calculation: ,in, The sliding mode coefficient for generator terminal voltage error. This is the sliding mode coefficient for the generator power angle error. Let be a fractional integral operator; by letting The equivalent control term is derived from the conditions. : ,in, This is the derivative of the generator terminal voltage deviation. This is the derivative of the generator power angle deviation. The generalized fractional inductance coefficient, This is the equivalent resistance of the excitation winding. For excitation current, The mutual inductance coefficient between the stator and rotor is... For the d-axis flux linkage, Let d be the open-circuit transient time constant. for Caputo fractional differential operator of order 1. Fractional order; based on the initial comprehensive perturbation estimate. Constructing disturbance feedforward compensation terms : , For robust gain switching, This is the disturbance compensation coefficient. The sign function is based on the perturbation feedforward compensation term. For equivalent control items Perform disturbance compensation to obtain the excitation control quantity after disturbance compensation. , .
[0037] Specifically, the generator terminal voltage has been obtained in step S1. and generator power angle Furthermore, the initial comprehensive disturbance estimate has already been obtained in step S2. To achieve highly robust excitation regulation, a fractional-order sliding mode controller is designed. The core of fractional-order sliding mode control lies in constructing a fractional-order sliding surface, allowing the system state to slide towards the equilibrium point along this surface. The structure of the fractional-order sliding surface directly determines the robustness and chattering characteristics of the controller. The generator terminal voltage deviation is then obtained. This deviation is equal to the target generator terminal voltage. Subtract the actual generator terminal voltage The generator terminal voltage deviation reflects the degree of deviation between the generator terminal voltage and the target value, and is the primary adjustment target for excitation control. Obtaining the generator power angle deviation... This deviation reflects the degree of deviation between the generator rotor angle and the target operating point. The stability of the power angle is the key to suppressing low-frequency oscillations. The generator terminal voltage deviation and the generator power angle deviation together constitute the two input error signals of the fractional sliding surface.
[0038] Based on generator terminal voltage deviation and generator power angle deviation A fractional sliding surface is constructed using a fractional integral operator. The expression for this sliding surface is as follows: ,in This is a fractional integral operator with order α⁻¹, where α takes values between 0 and 1. The negative value actually represents fractional integral operation. The difference between this fractional integral sliding surface and the classic integer integral sliding surface is that the fractional integral operator contains a Volterra integrator kernel, which can weight and accumulate the historical trajectory of the error, enhancing the controller's ability to sense the accumulated deviation. At the same time, the fractional integrator is equivalent to a low-pass filter in the frequency domain, with a slope of -20(1-α)dB / dec. This characteristic makes the sliding variable s much less sensitive to high-frequency noise than the integer sliding surface.
[0039] By making fractional-order sliding surfaces The equivalent control term is derived from the condition that equals 0. Its function is to maintain the system state on the fractional sliding surface. The derivation of the equivalent control term is as follows: Take the α-th order fractional differential of the fractional sliding surface s, and let... When s equals 0, the fractional dynamic equation of the excitation current in the fractional augmented state equation set established in step S1 is substituted into the equation, and the equivalent control term is obtained after simplification. The expression is : The equivalent control term is derived based on the fractional-order model in step S1, which includes fractional-order differential operators, so that the controller's action matches the dynamic characteristics of the fractional-order model.
[0040] Based on the initial comprehensive disturbance estimate obtained in step S2 Constructing disturbance feedforward compensation terms The purpose of this compensation term is to offset the adverse effects of the combined disturbance on the system, thereby reducing the burden on the sliding mode switching term. The expression for the disturbance feedforward compensation term is: In the disturbance feedforward compensation term, The term utilizes the comprehensive disturbance estimate reconstructed in real time by the observer in step S2 to perform feedforward compensation for the disturbance. When the disturbance estimate... When the actual integrated disturbance is reflected, this term directly cancels out the impact of the disturbance, enabling the equivalent control term to maintain sliding mode motion. This term is a robust switching term used to compensate for observation residuals and unmodeled dynamics. It is based on a perturbation feedforward compensation term. For equivalent control items Perform disturbance compensation to obtain the excitation control quantity after disturbance compensation. : The control quantity is equal to and sum.
[0041] This step utilizes the sliding surface constructed by the fractional integral operator to compress the equivalent gain bandwidth in the frequency domain, thereby reducing the sensitivity to high-frequency noise. At the same time, it uses the comprehensive disturbance estimate provided by the fractional extended state observer in step S2 to construct a feedforward compensation term, which greatly reduces the requirement for the value of the switching gain. Finally, under the premise of ensuring the strong robustness of the system, low-chattering excitation control is achieved.
[0042] Figure 2 This is a three-layer architecture diagram of the excitation control method of this application. The random wind speed input is transmitted to the excitation generator FO-EESG model via a front-end speed regulating device (hydraulic / differential gear train). The fractional-order electrically excited synchronous generator model established in step S1 is used to describe the fractional-order dynamic characteristics of the generator. The fractional-order extended state observer constructed in step S2 uses the generator model to perform online estimation of the comprehensive disturbance. Simultaneously, the network coupling disturbance of the multi-machine power grid, after being calculated by the power grid network Jacobian matrix sensitivity matrix, is injected into the extended channel of the observer as additional disturbance information. The comprehensive disturbance estimate output by the observer... Entering the disturbance compensation feedforward channel in step S3, the fractional sliding surface constructed in step S3 is used with a fractional integral operator. The voltage deviation and power angle deviation at the opposite end are processed, and the equivalent control term is derived from the sliding mode surface. The disturbance compensation feedforward channel generates the compensation term, and the two are superimposed to obtain the excitation control quantity after disturbance compensation.
[0043] Step S4: Obtain the operating status deviation signal of neighboring units in a multi-unit grid-connected environment, and calculate the equivalent network coupling disturbance as an additional extended state variable by combining the sensitivity coefficient of the power grid network Jacobian matrix, and obtain the preliminary updated comprehensive disturbance estimate.
[0044] Step S4 includes: In a multi-unit grid-connected environment, for the i-th unit, obtaining the operating status deviation signal of the neighboring j-th unit at the current time t, including the predicted generator power angle deviation of the neighboring j-th unit. and predicted generator terminal voltage deviation By combining the active power angle sensitivity and reactive power voltage sensitivity in the Jacobian matrix of the power grid network, the equivalent network coupling disturbance of the i-th generating unit is calculated. : ,in, The generator power angle of the adjacent j-th unit. Let be the generator terminal voltage of the adjacent j-th unit. The active power of the i-th unit Power angle of the j-th unit The sensitivity coefficient, The reactive power of the i-th unit For the terminal voltage of the j-th unit The sensitivity coefficient is obtained; the equivalent network coupled disturbance is superimposed on the disturbance reconstruction channel of the fractional-order extended state observer, and the updated comprehensive disturbance estimate is output.
[0045] Specifically, in scenarios involving multiple wind turbine units operating in a grid-connected configuration, the power angle-power flow coupling effect between units cannot be ignored. Power fluctuations in one unit can be transmitted to neighboring units through the transmission network, forming a cross-unit propagation chain of disturbances. Therefore, a multi-unit collaborative disturbance compensation mechanism needs to be introduced on top of the single-unit control framework. First, for the i-th unit, the operating state deviation signal of the j-th neighboring unit at time t is obtained. This operating state deviation signal includes the predicted generator power angle deviation of the j-th neighboring unit. and predicted generator terminal voltage deviation .
[0046] Among them, obtaining the operating status deviation signal of the j-th neighboring unit at time t includes: there is a communication delay when the operating status deviation signal of the j-th neighboring unit is transmitted to the i-th unit through the communication network. The i-th generating unit receives the lagging generator power angle deviation Deviation from the lagging generator terminal voltage ; Calculate the first-order rate of change of the lagging generator power angle deviation and the lagging generator terminal voltage deviation; Utilizing the lagging generator power angle deviation, lagging generator terminal voltage deviation, and their corresponding first-order rate of change, combined with communication delay... The Taylor series forward prediction method is used to calculate the predicted generator power angle deviation at the current time t. and the predicted generator terminal voltage at the current time t The predicted generator power angle deviation and the predicted generator terminal voltage deviation are used as the operating state deviation signal of the j-th unit adjacent to the current time t.
[0047] Specifically, due to the unavoidable communication delay when the measured deviation signal from the j-th neighboring unit is transmitted to the i-th unit via the communication network... The i-th generating unit actually receives a lagging signal, including a lagging generator power angle deviation. Deviation from the lagging generator terminal voltage Calculate the first-order rate of change of the power angle deviation of the lagging generator. The first-order rate of change of the generator terminal voltage deviation By utilizing the lag generator power angle deviation and its corresponding first-order rate of change in conjunction with communication delay The Taylor series forward prediction method is used to linearly extrapolate and predict the state at the current real time t, that is, through the formula... Calculate the predicted generator power angle deviation at the current time t using the formula. Calculate the predicted generator terminal voltage deviation at the current time t. Through this forward prediction, the predicted generator power angle deviation and generator terminal voltage deviation at the current time t are used as the operating status deviation signal of the j-th neighboring unit at the current time, thus offsetting the phase attenuation caused by communication delay in terms of timing.
[0048] Because the network topology and ground-state power flow remain relatively stable over short periods, the sensitivity coefficients in the Jacobian matrix of the power grid exhibit slow time-varying characteristics. Therefore, online solutions are not required within each microsecond-level excitation control cycle. Instead, the upper-level energy management system or wide-area measurement system performs asynchronous calculations on a slow timescale. The calculated sensitivity coefficient matrix is then transmitted via the communication network to the local controller of the i-th generating unit and stored as a constant. Based on this, the active power-power angle sensitivity coefficients in the Jacobian matrix of the power grid are combined... With reactive power-voltage sensitivity coefficient Calculate the equivalent network coupling disturbance experienced by the i-th unit. Its expression is Summation is performed on all neighboring j-th generating units. Figure 3 This is a schematic diagram of the multi-machine cooperative disturbance compensation mechanism architecture of this application. Combined with... Figure 3 Taking a grid-connected system containing three front-end speed-regulating wind turbine generators as an example, each unit (unit 1, unit 2, and unit 3) is equipped with an independent local controller. Each local controller integrates the fractional-order electrically excited synchronous generator model established in step S1 and the fractional-order extended state observer constructed in step S2.
[0049] While performing local state estimation and control law calculation, each local controller uploads key operating status information, such as generator power angle and generator terminal voltage, to the network-coupled Jacobian calculation module of the wide-area measurement system or energy management system via the communication network. Figure 3 In the network Jacobian coupled computation module, based on the current network topology and ground-state power flow information, the active power-power angle sensitivity coefficients in the power grid network Jacobian matrix are called. With reactive power-voltage sensitivity coefficient And combined with the power angle deviation from each neighboring unit Deviation from terminal voltage According to the expression Calculate the equivalent network coupling disturbance for each unit. , Figure 3 The dashed arrows in the diagram illustrate the path of the coupled disturbance to the feedforward injection for disturbance compensation of each unit, and the calculated... As additional extended state variables, they are injected into the disturbance reconstruction channel of the fractional extended state observer in the local controller of each unit, and the updated comprehensive disturbance estimate is output. This updated estimate replaces the initial comprehensive disturbance estimate in step S2 and is used to construct the disturbance feedforward compensation term in step S3, so that the controller of each unit can generate the feedforward compensation amount in advance to offset the effect before the disturbance of the neighboring unit actually affects the dynamics of the unit.
[0050] This step effectively shortens the equivalent time delay between disturbance perception and compensation action through the aforementioned multi-machine collaborative disturbance compensation mechanism, combined with a dual compensation mechanism of forward prediction and fractional-order extended state observer residual adaptive elimination.
[0051] Step S5: Construct the comprehensive objective function of the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
[0052] Step S5 includes: constructing a multi-objective optimization model and defining the control parameter variables as follows. ,in, The sliding mode coefficient for generator terminal voltage error. This is the sliding mode coefficient for the generator power angle error. For robust gain switching, This is the disturbance compensation coefficient. For state error gain, To reconstruct the perturbation gain; construct the comprehensive objective function for the control parameters of the fractional-order extended state observer and the fractional-order sliding surface: ,in, Let be the generator terminal voltage deviation at time t. The target generator terminal voltage, Let be the actual generator terminal voltage at time t. Let t be the generator power angle deviation. For the target generator power angle, Let t be the actual generator power angle. This is the disturbance feedforward compensation term, where T is the total time step. , and These are the weighting coefficients, and Set the search space boundary for the control parameter variables and use frequency domain stability and pole stability as hard constraints for the optimization process; solve the comprehensive objective function under hard constraints based on the global optimization algorithm to obtain the optimal parameter combination; update the fractional-order extended state observer and fractional-order sliding surface based on the optimal parameter combination, and repeat steps S2 to S3 to obtain the comprehensive disturbance estimate after model update, update the excitation control quantity, obtain the final excitation control quantity, and feed it back to the generator excitation winding.
[0053] Specifically, since the selection of each control parameter in steps S1 to S4 directly affects the dynamic response speed, steady-state accuracy, and robustness of the system, and since these parameters are interdependent, it is difficult to obtain the globally optimal configuration through manual trial and error. Therefore, a multi-objective optimization model needs to be established, and the decision variable vector to be optimized is defined as follows: A comprehensive objective function J(x) is constructed to evaluate the performance of the fractional-order extended state observer and the fractional-order sliding surface control. This objective function simultaneously considers the fast tracking capability of the generator terminal voltage, the suppression effect of power angle oscillation, and the smoothness of the sliding mode control law. Its expression is as follows: ,in, , and The weighting coefficients and Weighting coefficient The weights characterizing the dynamic performance of the terminal voltage are assigned to the excitation system, which is tasked with maintaining the stability of the terminal voltage. Therefore, the excitation system is assigned the highest weight to ensure that the system has a fast voltage response speed and a very small steady-state tracking error. Characterizes the power angle suppression and damping weight, which is used to improve the dynamic damping characteristics in multi-machine grid-connected scenarios and suppress low-frequency power angle oscillations caused by cross-machine coupling; The weights characterize chattering and energy suppression, and penalize the high-frequency compensation terms in the control law to prevent power device overheating and control chattering caused by excessive switching gain or observer gain.
[0054] While constructing the comprehensive objective function, it is necessary to set the search space boundaries for each decision variable to ensure that the control loop has a clear physical meaning and can be safely implemented in the digital controller. Generator terminal voltage error sliding mode coefficient. The search boundary is limited to 0.1 to 50.0; the generator power angle error sliding mode coefficient The search boundary is limited to 0.1 to 30.0; robust switching gain. The search boundary is limited to 0.01 to 2.00, with its upper bound compressed to 2 because of the feedforward decoupling capability of the fractional-order extended state observer, which ensures robustness without excessive switching gain; the perturbation compensation coefficient The search boundary is limited to 0.10 to 1.50; state error gain The search boundary is limited to 1.0 to 100.0; perturbation reconstruction gain The search boundary is limited to 1.0 to 200.0.
[0055] Furthermore, the system eigenvalues and frequency domain characteristics during the optimization process are used as hard penalty constraints to ensure that the tuned closed-loop control system meets the stability requirements. Let the open-loop transfer function of the closed-loop system be... It must satisfy phase margin constraints, gain margin constraints, and pole stability constraints. The phase margin constraint requires... ,in, For phase margin, The cutoff frequency; gain margin constraint requirements. ,in, For gain margin, The phase crossover frequency, i.e., satisfying The frequency range; the pole stability constraint requires all roots of the closed-loop characteristic equation to be based on the Mittag-Leffler stability criterion. It must lie entirely within the asymptotically stable region of the complex plane, i.e., satisfy... The aforementioned frequency domain constraints are embedded as hard penalty terms in the optimization process. When a candidate solution does not satisfy any of the constraints, the corresponding objective function value is given a maximum penalty value.
[0056] Under the aforementioned hard constraints, a global optimization algorithm is used to solve the comprehensive objective function and obtain the optimal parameter combination. The preferred global optimization algorithm is Particle Swarm Optimization (PSO). PSO initializes a swarm of random particles, each representing a set of candidate decision variables. During iteration, each particle updates its velocity and position based on its own historical best position and the swarm's historical best position, gradually approaching the global optimum. When the iteration converges, the parameter combination corresponding to the globally optimal particle is output as the optimal parameter combination. The state error gain in the fractional-order extended state observer is then updated based on this optimal parameter combination. and perturbation reconstruction gain And the generator terminal voltage error sliding mode coefficient in the fractional-order sliding mode surface And generator power angle error sliding mode coefficient While keeping the fractional-order excitation model parameters unchanged, the updated observer and controller parameters are substituted into the state estimation channel and disturbance reconstruction channel of the fractional-order extended state observer in step S2 to obtain the comprehensive disturbance estimate after model update. The updated comprehensive disturbance estimate is then substituted into the disturbance feedforward compensation term in step S3 to calculate the updated excitation control quantity. This final excitation control quantity is then fed back to the generator excitation winding to achieve closed-loop optimization control of the front-end speed-regulating wind turbine excitation system.
[0057] To verify the feasibility of the above multi-objective tuning process, a typical composite working condition of a 6.25MVA, two-pole (4-pole), 50Hz electrically excited synchronous wind turbine generator set under rated wind speed turbulence disturbance and 20% sudden load was used as an example. The optimal control parameter tuning results after global optimization convergence are shown in Table 1.
[0058] Table 1 Optimal Control Parameter Tuning
[0059] The optimal tuning values in Table 1 verify the convergence effectiveness of the multi-objective optimization tuning method proposed in this application. Among them, the generator terminal voltage error sliding mode coefficient... The setting is 15.42 and the generator power angle error sliding mode coefficient. Set to 8.15. Value greater than This reflects the priority relationship in step S3: generator terminal voltage deviation is the primary adjustment target for excitation control, while generator power angle deviation is the secondary suppression target. Robust switching gain. The optimal setpoint is only 0.25, far lower than the value usually required in traditional sliding mode control. This is due to the disturbance feedforward compensation capability of the fractional-order extended state observer in step S2, and the fact that the observation residual is much smaller than the original disturbance in step S3, thus reducing the time required for control. This aligns with the technical logic, and the disturbance compensation coefficient... A setting of 1.00 indicates that the FO-ESO's disturbance estimate is accurate and reliable in both amplitude and phase, and can be fully feedforwarded proportionally. This setting indirectly verifies the observer's reconstruction accuracy and phase following capability in the low-frequency, slowly varying disturbance band, as well as the state error gain. The perturbation reconstruction gain is 42.50. The value is 112.30, and both satisfy the condition. > Typical extended state observer gain ratios, where A larger value is chosen to ensure that the perturbation reconstruction channel has a sufficiently fast convergence speed. The values are chosen to be relatively moderate to ensure the smoothness of the state estimation channel and avoid excessive amplification of measurement noise. Regarding frequency domain constraint verification, phase margin... It is 52.4 fall into to The preferred range, gain margin The value is 8.7dB, which is greater than 6dB. Both satisfy the hard frequency domain stability constraint set in step S5, verifying that the parameter combination obtained by tuning has sufficient robust stability boundary in actual engineering.
[0060] Figure 4The diagram shows the signal flow of the final excitation control quantity synthesis in this application, clearly illustrating the complete generation path of the control quantity. Figure 4 On the left, the input error signal of the sliding surface is composed of the deviation between the target generator terminal voltage and the actual terminal voltage, and the deviation between the target power angle and the actual power angle. The two are weighted by fractional integral operators to form a fractional sliding surface, where the sliding coefficient adopts the optimal tuning values listed in Table 1. Figure 4 The middle section is divided into two branches: the upper branch consists of the equivalent control term and the switching term. The equivalent control term is derived based on the sliding surface approach condition. This term is used to maintain the ideal motion of the system state on the sliding surface. The gain in the switching term has been compressed to a smaller value through global optimization in step S5, thereby effectively suppressing chattering. The lower branch is the disturbance feedforward compensation term. The comprehensive disturbance estimate on which it depends has undergone initial observation in step S2, multi-machine coupling update in step S4, and re-observation after parameter optimization in step S5, ensuring the estimation accuracy. Finally, the three signals of the equivalent control term, the switching term, and the disturbance feedforward compensation term are superimposed by an adder to generate the final excitation control quantity, which is fed back to the generator excitation winding to achieve closed-loop control.
[0061] This step achieves global optimization by combining multi-objective optimization with hard constraints on frequency domain stability. It realizes the systematic and synchronous tuning of the fractional-order extended state observer and the fractional-order sliding surface control parameters, avoiding manual trial and error. This improves the accuracy of disturbance estimation and the robustness, dynamic response, and smoothness of excitation control.
[0062] To quantitatively verify the superiority of the front-end speed-regulating wind turbine excitation control method and model based on fractional calculus in this application, a grid-connected dynamic model of a wind farm with multiple units densely connected was built in the Matlab / Simulink simulation platform. The tested unit adopted the aforementioned standard parameters, namely, an electrically excited synchronous generator with a single unit capacity of 6.25 MVA, two pole pairs, and a rated frequency of 50 Hz. To comprehensively simulate the multivariate uncertainties and time-varying disturbances in actual operation, the test scenario was set as follows: Input wind disturbance: a random turbulent wind speed with a mean of 12 m / s and a turbulence intensity of 15% was superimposed to simulate severe nonlinear mechanical torque disturbances; Severe grid disturbance: at t=2.0s, a sudden load change of 20% occurred near the generator grid bus; at t=5.0s, the operation of the j-th adjacent unit triggered a strong coupling impact between the power angle and the power flow, which was used to verify the vibration damping performance of the system. The control scheme proposed in this application is compared with the traditional industrial general scheme under the same conditions and the dynamic simulation of the whole time. By sampling and monitoring the output response of the two control schemes under the compound working condition, the core indicators such as the dynamic response of the terminal voltage, the steady-state tracking accuracy, the chattering performance of the control law and the damping ratio of the multi-machine system are extracted. The quantitative numerical comparison of the differences is shown in Table 2.
[0063] Table 2. Numerical Differences Between Traditional Integer-Order Control Schemes and the Scheme in this Application
[0064] As shown in Table 2, regarding the dynamic deviation of the terminal voltage, the proposed solution exhibits a maximum dynamic deviation of only 3.82% under a 20% sudden load surge, a significant reduction of 69.32% compared to the 12.45% of the traditional integer-order solution. This demonstrates that the fractional-order sliding mode controller combined with the disturbance feedforward compensation mechanism can quickly suppress voltage drops caused by sudden load changes. In terms of steady-state accuracy, the proposed solution achieves a terminal voltage steady-state tracking error of only 0.04%, a reduction of 95.29% compared to the 0.85% of the traditional solution. This is attributed to the low-frequency, high-gain characteristics of the fractional-order integral sliding mode surface and the zero-error compensation of the extended state observer for static disturbances. Regarding chattering control, the proposed solution achieves a chattering amplitude of only 0.02 pu in the excitation voltage control output, a reduction of 95.83% compared to the 0.48 pu of the traditional solution. This is attributed to… The high-precision feedforward compensation of the integrated disturbance by the fractional-order extended state observer significantly reduces the sliding mode switching gain, fundamentally weakening the chattering excitation source. Regarding system damping, the proposed scheme increases the low-frequency oscillation damping ratio of the multi-machine grid-connected system from 0.042 in the traditional scheme to 0.125, an increase of 197.62%, enabling the system to transition from the weakly damped region to the safe damped region, demonstrating excellent cross-machine oscillation suppression capability. In terms of dynamic response speed, the integrated disturbance reconstruction convergence time is shortened from 0.35s to 0.08s, a reduction of 77.14%, and the low-frequency oscillation residual settling time is shortened by 83.06%, from 4.25s to 0.72s, verifying the dual speed advantages of the fractional-order extended state observer in capturing fast-changing disturbances and fractional-order sliding mode control in calming electromechanical oscillations.
[0065] To visually demonstrate the waveform differences of the above quantitative indicators, Figures 5 to 8 The time-domain response curves for terminal voltage, excitation voltage, power angle deviation, and disturbance reconstruction error are presented respectively. Figure 5 To compare the generator terminal voltage response curves under the two control schemes, by Figure 5 It is evident that after a sudden load is applied at t=2.0s, the traditional solution exhibits a large voltage drop, numerous oscillations, and slow recovery. In contrast, the solution proposed in this application has a shallower voltage drop depth, shorter recovery time, and superior dynamic disturbance rejection performance. Figure 6 The graph shows the excitation voltage control output and chattering performance evaluation. After applying a disturbance at t=2s, the traditional solution has large output ripple and obvious chattering, while the solution in this application has smooth and stable output, which can effectively suppress chattering, thus proving the effectiveness of chattering suppression. Figure 7The diagram shows the generator power angle deviation under multi-generator grid-connected cross-generator disturbance. After the disturbance is applied at t=5.0s, the traditional scheme exhibits continuous oscillation. The peak power angle deviation under the scheme of this application is about 0.02rad, which effectively reduces the impact amplitude and converges quickly after the disturbance. In contrast, the traditional scheme still has residual low-frequency fluctuations at the end of the 8s simulation, which further confirms the quantitative difference between damping ratio and adjustment time in Table 2. Figure 8 The diagram shows the integrated perturbation reconstruction error of the fractional-order extended state observer. The error exhibits instantaneous spikes at perturbation times of t=2s and t=5s, with peak values of approximately 0.39pu and 0.23pu, respectively. Subsequently, the error rapidly converges to zero, verifying that the fractional-order extended state observer in step S2 has excellent perturbation tracking capability under rapidly changing and strong uncertain conditions, and providing a reliable information basis for perturbation feedforward compensation in step S3.
[0066] The above describes an excitation control method for a front-end speed-regulating wind turbine in an embodiment of this application. The following describes an excitation control system for a front-end speed-regulating wind turbine in an embodiment of this application. Please refer to [link to relevant documentation]. Figure 9 The excitation control system of a front-end speed-regulating wind turbine in this application embodiment includes: The model building module is used to obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, a fractional excitation model of the excitation winding is established, and a fractional augmented state equation system is constructed in combination with the operating state parameters.
[0067] The disturbance observation module is used to acquire mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Based on the fractional augmented state equations, a fractional extended state observer is constructed to estimate the extended state variables online and obtain the initial comprehensive disturbance estimate.
[0068] The sliding mode control module is used to construct a fractional sliding mode surface using a fractional integral operator based on the generator terminal voltage and generator power angle, obtain the equivalent control term, construct a disturbance feedforward compensation term based on the initial comprehensive disturbance estimate, perform disturbance compensation on the equivalent control term, and obtain the excitation control quantity after disturbance.
[0069] The collaborative compensation module is used to acquire the operating status deviation signal of neighboring units in a multi-unit grid-connected environment. Combined with the sensitivity coefficient of the Jacobian matrix of the power grid network, it calculates the equivalent network coupling disturbance as an additional extended state variable to obtain the preliminary updated comprehensive disturbance estimate.
[0070] The closed-loop optimization module is used to construct the comprehensive objective function of the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after the model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
[0071] In summary, this application addresses the challenges of memory effect modeling, multi-source disturbance suppression, and sliding mode chattering in the excitation system of a front-end speed-regulating wind turbine under strong uncertainty and multi-machine coupling. It proposes a systematic control scheme based on fractional-order calculus theory. By establishing a fractional-order excitation model that considers magnetic circuit saturation memory and high-frequency eddy current diffusion effects, the system's internal model accurately matches the actual physical dynamics. A fractional-order extended state observer is constructed to achieve rapid online reconstruction of mechanical torque, excitation parameters, and network coupling disturbances. Furthermore, the synergistic effect of fractional-order integral sliding surface and feedforward compensation effectively reduces chattering while maintaining strong robustness. A multi-machine collaborative disturbance compensation mechanism is introduced to shorten the cross-machine coupling sensing delay, and the control parameters are systematically tuned through multi-objective optimization. Simulation results show that compared with traditional integer-order control, the proposed scheme improves upon traditional methods in terms of terminal voltage dynamic deviation, steady-state accuracy, chattering suppression, and multi-machine low-frequency oscillation damping. The disturbance reconstruction and oscillation adjustment times are significantly shortened, fully verifying its superiority and engineering practicality in improving excitation dynamic quality, suppressing electromechanical oscillations, and enhancing grid-connected stability.
[0072] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0073] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. An excitation control method for a front-end speed-regulating wind turbine generator, characterized in that, include: Step S1: Obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, establish a fractional excitation model of the excitation winding, and construct a fractional augmented state equation system in combination with the operating state parameters. Step S2: Obtain mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Construct a fractional extended state observer based on the fractional augmented state equations, and perform online estimation of the extended state variables to obtain the initial comprehensive disturbance estimate. Step S3: Based on the generator terminal voltage and the generator power angle, construct a fractional sliding mode surface using a fractional integral operator to obtain an equivalent control term. Based on the initial comprehensive disturbance estimate, construct a disturbance feedforward compensation term and perform disturbance compensation on the equivalent control term to obtain the excitation control quantity after disturbance. Step S4: Obtain the operating status deviation signal of the neighboring units in the multi-unit grid-connected environment, and calculate the equivalent network coupling disturbance as an additional extended state variable by combining the sensitivity coefficient of the power grid network Jacobian matrix, and obtain the preliminary updated comprehensive disturbance estimate. Step S5: Construct a comprehensive objective function for the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
2. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 1, characterized in that, Step S1 involves constructing a fractional augmented state equation system, including: Based on fractional calculus theory, a fractional excitation model is established. The fractional calculus theory includes the memory effect of magnetic circuit saturation and hysteresis, the high-frequency skin effect, and the eddy current diffusion effect. The fractional-order excitation model is as follows: ,in, for Caputo fractional differential operator (0 < <1), For fractional order, The excitation winding was identified through experiments. The generalized fractional inductance coefficient, This is the equivalent resistance of the excitation winding. For excitation current, is the excitation voltage, and f is the excitation winding of the generator; Select the generator power angle of the operating state parameters. Speed deviation q-axis transient electromotive force and excitation current As state variables, combined with the aforementioned fractional-order excitation model, a set of augmented state equations containing fractional-order states is constructed: ,in, The first derivative of the generator's power angle. This is the generator's rated synchronous angular frequency. For speed deviation, Let H be the first derivative of the speed deviation, and H be the unit's inertial time constant. The mechanical torque of the prime mover. Let be the electromagnetic torque of the generator, and D be the damping coefficient. Let d be the open-circuit transient time constant. , These are the d-axis synchronous reactance and transient reactance, respectively. This refers to the d-axis stator current. This is the magnetizing electromotive force. The mutual inductance coefficient between the stator and rotor is... denoted as d-axis flux linkage.
3. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 2, characterized in that, The excitation winding was identified through experiments, including: When the generator is stopped or under low load, a wideband excitation signal covering the main operating frequency band is applied to the excitation winding using a logarithmic sinusoidal frequency sweep method. This includes sinusoidal frequency sweep signals, multi-frequency sinusoidal superposition signals, pseudo-random binary sequence signals, or step excitation signals. Simultaneously, the excitation voltage and excitation current response signals are collected, the actual response frequency of the excitation winding is calculated, and experimental data of the actual amplitude-frequency characteristics and phase-frequency characteristics are obtained. Construct a fractional-order frequency domain transfer function model of the excitation winding, calculate the model response frequency of the excitation winding, and obtain theoretical data on the amplitude-frequency characteristics and phase-frequency characteristics under the model framework; A nonlinear optimization algorithm is used for parameter identification, with the objective function being to minimize the amplitude and phase errors between the actual response frequency and the model response frequency. This algorithm is applied to the fractional-order parameters. By performing joint optimization, the optimal fractional-order model parameters are obtained, thus completing the fractional-order process. The calibration.
4. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 1, characterized in that, Step S2 includes: Mechanical torque disturbance, excitation parameter perturbation, and network coupling disturbance are acquired and integrated into a comprehensive disturbance, which is defined as an extended state variable. Based on the aforementioned fractional augmented state equations, an observation error feedback correction term is introduced to construct a fractional extended state observer, which includes a state estimation channel and a disturbance reconstruction channel. The state estimation channel is used for online estimation of the q-axis transient electromotive force: The perturbation reconstruction channel employs a fractional-order calculus structure for online reconstruction of the extended state variables. Numerical solutions are obtained using fractional integral operators to obtain initial estimates of the overall disturbance. ,in, for Caputo fractional differential operator of order 1. For fractional order, This is the observed estimate of the q-axis transient electromotive force. Let d be the open-circuit transient time constant. , These are the d-axis synchronous reactance and transient reactance, respectively. This is the magnetizing electromotive force. This is the initial comprehensive disturbance estimate of the extended state variables. For observation error feedback correction term, To account for observation errors, the observer gain is selected based on the constraints. and ,in, For state error gain, This is the perturbation reconstruction gain.
5. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 4, characterized in that, Select the observer gain based on the constraints. and include: Define the observer state error vector The state error vector includes the observation error of the q-axis transient electromotive force. Reconstruction error of combined disturbances ; Using the rate of change of the derivative of the extended state variable h(t) as the driving input, the error dynamic matrix form of the fractional-order extended state observer is obtained: ,in, The observer error state matrix, The perturbation input matrix; According to the stability theory of fractional linear systems, to ensure the asymptotic convergence of the observer's dynamic error, the eigenvalues of the state matrix... It must satisfy the fractional asymptotic stability condition, i.e. The eigenvalues are derived from the characteristic equation corresponding to the error state matrix: Sure; Under the constraint that all the eigenvalues satisfy the fractional asymptotic stability condition, the observer gain is selected according to the frequency domain tuning criterion. and .
6. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 1, characterized in that, Step S3 includes: Obtain generator terminal voltage deviation and generator power angle deviation ,in, The target generator terminal voltage, This is the actual generator terminal voltage. For the target generator power angle, This represents the actual generator power angle. Based on the generator terminal voltage deviation and the generator power angle deviation, a fractional integral sliding surface is constructed using fractional integral calculation: ,in, This is the sliding mode coefficient for the generator terminal voltage error. This is the sliding mode coefficient for the generator power angle error. It is a fractional integral operator; By order The equivalent control term is derived from the conditions. : ,in, This is the derivative of the generator terminal voltage deviation. This is the derivative of the generator power angle deviation. The generalized fractional inductance coefficient, This is the equivalent resistance of the excitation winding. For excitation current, The mutual inductance coefficient between the stator and rotor is... For the d-axis flux linkage, Let d be the open-circuit transient time constant. for Caputo fractional differential operator of order 1. It is a fractional order; Based on the initial comprehensive disturbance estimate Constructing a disturbance feedforward compensation term : , For robust gain switching, This is the disturbance compensation coefficient. For symbolic functions, Based on the disturbance feedforward compensation term For the equivalent control item Perform disturbance compensation to obtain the excitation control quantity after disturbance compensation. , .
7. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 1, characterized in that, Step S4 includes: In a multi-unit grid-connected environment, for the i-th unit, obtain the operating status deviation signal of the neighboring j-th unit at time t, including the predicted generator power angle deviation of the neighboring j-th unit. And predicted generator terminal voltage deviation ; By combining the active power angle sensitivity and reactive power voltage sensitivity in the Jacobian matrix of the power grid network, calculate the equivalent network coupling disturbance of the i-th generating unit. : ,in, The generator power angle of the adjacent j-th unit, Let be the generator terminal voltage of the adjacent j-th unit. The active power of the i-th unit Power angle of the j-th unit The sensitivity coefficient, The reactive power of the i-th unit For the terminal voltage of the j-th unit The sensitivity coefficient; Couple the perturbation amount of the equivalent network The updated comprehensive perturbation estimate is superimposed onto the perturbation reconstruction channel of the fractional-order extended state observer and output.
8. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 7, characterized in that, The operating status deviation signal of the j-th neighboring unit at time t includes: There is a communication delay when the operating status deviation signal of the neighboring j-th unit is transmitted to the i-th unit via the communication network. The i-th generating unit receives the lagging generator power angle deviation Deviation from the lagging generator terminal voltage ; Calculate the first-order rate of change of the lagging generator power angle deviation and the lagging generator terminal voltage deviation; Using the lagging generator power angle deviation, the lagging generator terminal voltage deviation, and the corresponding first-order rate of change, combined with the communication delay... The Taylor series forward prediction method is used to calculate the predicted generator power angle deviation at the current time t. and the predicted generator terminal voltage at the current time t ; The predicted generator power angle deviation and the predicted generator terminal voltage deviation are used as the operating state deviation signal of the j-th unit adjacent to the current time t.
9. The excitation control method for a front-end speed-regulating wind turbine generator according to claim 1, characterized in that, Step S5 includes: Construct a multi-objective optimization model and define the control parameter variables as follows: ,in, This is the sliding mode coefficient for the generator terminal voltage error. This is the sliding mode coefficient for the generator power angle error. For robust gain switching, This is the disturbance compensation coefficient. For state error gain, For perturbation reconstruction gain; Construct the combined objective function for the fractional-order extended state observer and the control parameters of the fractional-order sliding surface: ,in, Let be the generator terminal voltage deviation at time t. The target generator terminal voltage, Let be the actual generator terminal voltage at time t. Let t be the generator power angle deviation. For the target generator power angle, Let t be the actual generator power angle. This is the disturbance feedforward compensation term, where T is the total time step. , and These are the weighting coefficients, and ; The search space boundary of the control parameter variables is set, and frequency domain stability and pole stability are used as hard constraints in the optimization process. The optimal parameter combination is obtained by solving the comprehensive objective function under the hard constraints based on the global optimization algorithm. Based on the optimal parameter combination, update the fractional-order extended state observer and the fractional-order sliding surface, and repeat steps S2 to S3 to obtain the comprehensive disturbance estimate after model update, update the excitation control quantity, obtain the final excitation control quantity, and feed it back to the generator excitation winding.
10. An excitation control system for a front-end speed-regulating wind turbine, used to implement the excitation control method for a front-end speed-regulating wind turbine as described in any one of claims 1-9, characterized in that, The system includes: The model building module is used to obtain the operating state parameters of the front-end speed-regulating wind turbine, including excitation voltage, excitation current, speed deviation, q-axis transient electromotive force, generator terminal voltage and generator power angle. Based on fractional calculus theory, a fractional excitation model of the excitation winding is established, and a fractional augmented state equation system is constructed in combination with the operating state parameters. The disturbance observation module is used to acquire mechanical torque disturbance, excitation parameter perturbation and network coupling disturbance, and generate extended state variables. Based on the fractional augmented state equations, a fractional extended state observer is constructed to estimate the extended state variables online and obtain the initial comprehensive disturbance estimate. The sliding mode control module is used to construct a fractional sliding mode surface using a fractional integral operator based on the generator terminal voltage and the generator power angle, obtain an equivalent control term, construct a disturbance feedforward compensation term based on the initial comprehensive disturbance estimate, perform disturbance compensation on the equivalent control term, and obtain the excitation control quantity after disturbance. The collaborative compensation module is used to acquire the operating status deviation signal of neighboring units in a multi-unit grid-connected environment, and calculate the equivalent network coupling disturbance as an additional extended state variable by combining the sensitivity coefficient of the power grid network Jacobian matrix to obtain the preliminary updated comprehensive disturbance estimate. The closed-loop optimization module is used to construct the comprehensive objective function of the control parameters of the fractional-order extended state observer and the fractional-order sliding surface, update the fractional-order extended state observer and the fractional-order sliding surface, obtain the comprehensive disturbance estimate after model update, and recalculate the final excitation control quantity, which is then fed back to the generator excitation winding.
