A Distributed Model Predictive Control Method for Offshore Front-End Regulated Wind Turbines

Abstract

This invention discloses a distributed model predictive control method for offshore front-end speed-regulating wind turbines, belonging to the field of new energy power generation technology. The method decomposes the overall control task into three parallel subsystems: pitch control, hydraulic torque regulation, and excitation, and establishes local discrete predictive models for each. In each sampling period, rolling optimization of each subsystem is performed in parallel, and the future predicted trajectories of generator speed and terminal voltage are introduced as globally consistent variables. By constructing an augmented Lagrangian cost function based on the alternating direction multiplier method and iteratively solving it, the local predicted trajectories are driven to converge to the globally consistent optimal trajectory. Finally, the first instruction of the optimal control sequence obtained after iterative convergence is output to each actuator, achieving coordinated control of the three subsystems. This allows for rapid and smooth simultaneous satisfaction of frequency and voltage synchronization conditions before grid connection, effectively suppressing coupling interference and wind speed fluctuations, and improving grid connection quality and operational reliability.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation technology, and more specifically to a distributed model predictive control method for offshore front-end speed-regulating wind turbine units. Background Technology

[0002] Offshore wind power, as an important renewable energy source, is rapidly developing towards large-scale and mass production due to its abundant wind energy resources, high stability, and suitability for large-scale development. As the capacity of individual wind turbine units continues to increase, and the penetration rate of wind power and other new energy sources in the power system rises, the inertia support capacity of the power system is weakened, placing unprecedentedly high demands on the control performance and grid connection quality of the wind turbine units themselves.

[0003] Against this backdrop, a front-end speed-regulating wind turbine (FSCWT) integrating a hydraulic torque converter has attracted attention due to its unique mechanical configuration. This unit achieves "flexible decoupling" between the variable-speed operation of the turbine and the constant-speed operation of the generator through a hydraulic torque-regulating speed control system. This not only broadens the wind energy capture range but also effectively buffers aerodynamic load impacts using hydraulic damping, significantly reducing the frequency of mechanical wear on the pitch system. This characteristic has great potential to improve the long-term reliability and economy of offshore wind farms, which face high operation and maintenance costs and harsh operating environments.

[0004] However, the mechanical advantages of the FSCWT come at the cost of a dramatic increase in the complexity of its control system. Its operation control involves three subsystems with vastly different and tightly coupled dynamic characteristics: the pitch system, the hydraulic torque converter, and the excitation system. This forms a typical multivariable, nonlinear, and strongly coupled complex control object. This strong coupling characteristic is particularly prominent during the quasi-synchronous phase before grid connection: grid connection requires the generator's output voltage to be strictly matched with the grid in terms of frequency, amplitude, and phase, but control actions on any subsystem will interfere with the state variables of other subsystems through coupling relationships.

[0005] Currently, conventional control for such units often employs a distributed PID strategy. Each subsystem controller is independently designed and operates in isolation, lacking a global coordination mechanism and thus unable to effectively handle the aforementioned strong coupling conflicts. This leads to overshoot and continuous oscillations in generator speed and terminal voltage during grid synchronization, making it difficult to quickly and smoothly meet the synchronization conditions for both frequency and voltage. For offshore wind turbines connected to the grid via long-distance submarine cables, connecting units exhibiting such oscillations to the grid will generate enormous inrush currents and power fluctuations, not only violating grid connection guidelines but also potentially threatening the stable operation of the power grid.

[0006] Therefore, how to provide a distributed model predictive control method for offshore front-end speed-regulating wind turbines is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a distributed model predictive control method for offshore front-end speed-regulating wind turbines, which can collaboratively optimize the dynamic behavior of each subsystem. While suppressing internal coupling interference and external wind speed disturbances, it can achieve fast, accurate, and overshoot-free synchronous control of generator frequency and voltage, thereby ensuring high-quality grid connection of the unit, improving the dynamic performance of offshore wind power grid connection and grid acceptance capacity, and solving the multivariable, strongly coupled, and nonlinear control problems of front-end speed-regulating wind turbines.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A distributed model predictive control method for offshore front-end speed-regulating wind turbines includes the following steps:

[0010] S1. Decompose the control task of the front-end speed-regulating wind turbine into a pitch subsystem, a hydraulic torque-regulating speed-regulating subsystem, and an excitation subsystem, and establish local discrete prediction models for the pitch subsystem, the hydraulic torque-regulating speed-regulating subsystem, and the excitation subsystem, respectively.

[0011] S2. At each sampling moment, the current state of the pitch subsystem, hydraulic torque regulation subsystem and excitation subsystem is collected, and based on the local discrete prediction model, rolling optimization is performed to predict the state trajectory of each pitch subsystem, hydraulic torque regulation subsystem and excitation subsystem in the future prediction time domain.

[0012] S3. Taking the generator speed and terminal voltage required for grid connection synchronization as the synchronization target, the future predicted trajectory of the generator speed and terminal voltage is used as a global consistency variable.

[0013] S4. Construct an augmented Lagrangian cost function that includes the global consistency variables, and use the alternating direction multiplier method to iteratively solve the problem so that the local predicted trajectories of the pitch subsystem, the hydraulic torque regulation subsystem, and the excitation subsystem converge to the globally consistent optimal trajectory.

[0014] S5. After iterative convergence, the first control command in the optimal control sequence corresponding to the pitch subsystem, hydraulic torque variable speed control subsystem, and excitation subsystem is output to the corresponding pitch system, hydraulic torque variable speed control system, and excitation system to achieve coordinated control before grid connection.

[0015] Furthermore, in S1, the structure of the front-end speed-regulating wind turbine includes a wind turbine, a gearbox, a hydraulic torque converter, and a brushless electrically excited synchronous generator.

[0016] Furthermore, S1 includes:

[0017] A local discrete prediction model for the pitch subsystem is established based on the linearized aero-mechanical dynamic model of the wind turbine and its drive train.

[0018] A local discrete prediction model for the hydraulic torque converter speed regulation subsystem is established based on the linearized dynamic equations of the hydraulic torque converter and the generator rotor motion.

[0019] A local discrete prediction model for the excitation subsystem is established based on the linearized electromagnetic dynamic equations of the generator.

[0020] Furthermore, in S2, the rolling optimization is based on a quadratic objective function, which includes system state tracking error and control input cost.

[0021] Furthermore, in the rolling optimization objective function of the hydraulic torque variable speed control subsystem and the excitation subsystem, an augmented Lagrange term is added to penalize the deviation between the globally consistent variable and the globally consensus value in its local predictions.

[0022] Furthermore, in S3, the synchronization target of the generator speed is the grid synchronization speed, and the synchronization target of the terminal voltage is the grid voltage.

[0023] Furthermore, in S4, the iterative solution process using the alternating direction multiplier method includes the following sub-steps:

[0024] S41, the global consistency variables and dual variables of the pitch subsystem, hydraulic torque variable speed control subsystem and excitation subsystem are inherited or initialized based on the current state;

[0025] S42. Solve their respective local optimization problems in parallel to obtain the local optimal control sequence and the corresponding local prediction sequence of the global consistency variable;

[0026] S43, the pitch subsystem, hydraulic torque variable speed control subsystem and excitation subsystem exchange the local prediction sequences and calculate the updated global consensus variables through a distributed average consensus algorithm;

[0027] S44, the pitch subsystem, the hydraulic torque converter speed control subsystem, and the excitation subsystem update their dual variables locally;

[0028] S45. Determine whether the iteration has converged. If it has converged, proceed to S46. If it has not converged, return to step S41.

[0029] S46. After iterative convergence, each subsystem applies the first element of its optimal control sequence to the physical actuator.

[0030] Furthermore, in S42, the updated global consistency variable value is the average value of the local prediction sequences corresponding to the pitch subsystem, hydraulic torque regulation subsystem, and excitation subsystem.

[0031] Furthermore, it also includes: using a state observer to perform feedback correction on the initial state of the corresponding prediction model based on the actual measured values ​​of the pitch subsystem, hydraulic torque regulation subsystem, and excitation subsystem.

[0032] Furthermore, in S5, the objective of the coordinated control is to ensure that the frequency and amplitude of the generator output voltage meet the quasi-synchronous grid connection conditions.

[0033] As can be seen from the above technical solution, compared with the prior art, this invention provides a distributed model predictive control method for offshore front-end speed-regulating wind turbines. To achieve grid-connected coordinated control of offshore front-end speed-regulating wind turbines, the pitch system, hydraulic torque-regulating speed control system, and excitation system are taken as the core objects of coordinated control, replacing the traditional decentralized control strategy. Specific beneficial technical effects are as follows:

[0034] By coordinating and optimizing the dynamic processes of the three major subsystems, precise synchronization of generator frequency and terminal voltage is achieved. At the same time, the model prediction is used to suppress turbulent wind fluctuations, thereby reducing grid connection impact and mechanical stress on the pitch system, and improving the dynamic performance and quality of the unit's grid connection.

[0035] Meanwhile, to ensure the safe and reliable operation of the system before grid connection and to minimize the harmful effects of the multivariable, strongly coupled, and nonlinear characteristics of the unit, distributed model predictive control (DMM) and the alternating direction multiplier method (ADMM) are applied to the coordinated control system of the front-end speed-regulating wind turbine. This control method establishes local discrete predictive models for the pitch, hydraulic speed regulation, and excitation subsystems based on the unit's structure. Then, it predicts the future dynamics of the system through online rolling optimization. Generator speed and terminal voltage are introduced as consistency variables, and an augmented Lagrangian cost function is constructed using the ADMM algorithm and iteratively solved to obtain the globally optimal control sequence, achieving coordinated control of the operating states of each subsystem controller. Furthermore, by combining the predictive information from each subsystem, turbulent wind speed disturbances are suppressed, improving the system's dynamic performance and ensuring rapid and smooth synchronization of frequency and voltage. This is beneficial for the stable grid connection of the unit and enhances the grid connection quality and operational reliability of the system under complex operating conditions. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0037] Figure 1 This is a structural diagram of the front-end speed regulating unit provided in an embodiment of the present invention;

[0038] Figure 2 This is a diagram illustrating the coordinated control structure of the front-end speed regulating unit provided in an embodiment of the present invention.

[0039] Figure 3 This is a diagram illustrating the operating modes of the front-end speed regulating unit provided in an embodiment of the present invention.

[0040] Figure 4 This is a structural diagram of the no-load grid-connected model of the front-end speed regulating unit provided in an embodiment of the present invention;

[0041] Figure 5 This is a diagram of the DMPC coordination control structure provided in an embodiment of the present invention;

[0042] Figure 6 The ADMM-DMPC coordination control flowchart provided in this embodiment of the invention;

[0043] Figure 7 This is a flowchart illustrating a distributed model predictive control method for offshore front-end speed-regulating wind turbines provided by the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] See Figure 7 This invention discloses a distributed model predictive control method for offshore front-end speed-regulating wind turbines, comprising:

[0046] S1. Considering the multivariable and strongly coupled characteristics of offshore front-end speed-regulating wind turbines during grid connection, the system control task is decomposed into three parallel local subsystems: pitch regulation, hydraulic torque regulation, and excitation, and local discrete prediction models are constructed for each of them.

[0047] S2. Collect the actual operating status of each subsystem at the current moment, and based on the local discrete prediction model, perform rolling optimization in parallel at each sampling moment to predict the state trajectory of each subsystem in the future prediction time domain.

[0048] S3. To meet the standards of frequency and voltage consistency simultaneously according to grid synchronization requirements, the generator speed ω is introduced. g and terminal voltage V tThe future predicted trajectory is used as a globally consistent variable;

[0049] S4. Establish an augmented Lagrangian cost function based on the alternating direction multiplier method (ADMM), and evaluate and optimize each local predicted trajectory by iteratively solving the cost function, so that it converges to a consistent global optimal trajectory.

[0050] S5. Based on the optimal control sequence evaluated after iterative convergence as described above, extract the first control command. The output is sent to the corresponding pitch control system, hydraulic torque converter, and excitation system to achieve smooth and synchronous control of frequency and voltage before grid connection.

[0051] Specifically, this invention decomposes the overall control task into three parallel subsystems: pitch control, hydraulic torque regulation, and excitation, and establishes local discrete prediction models for each. During each sampling period, rolling optimization of each subsystem is performed in parallel, and the future predicted trajectories of generator speed and terminal voltage are introduced as globally consistent variables. By constructing an augmented Lagrangian cost function based on the alternating direction multiplier method and iteratively solving it, each local predicted trajectory is driven to converge to a globally consistent optimal trajectory. Finally, the first instruction of the optimal control sequence obtained after iterative convergence is output to each actuator, achieving coordinated control of the three subsystems. This allows for rapid and smooth simultaneous satisfaction of frequency and voltage synchronization conditions before grid connection, effectively suppressing coupling interference and wind speed fluctuations, and improving grid connection quality and operational reliability.

[0052] Specifically, the structure of a front-end speed-regulating wind turbine is as follows: Figure 1 As shown, its core components include a wind turbine, gearbox, hydraulic torque converter, and brushless electrically excited synchronous generator. The hydraulic torque converter consists of a pump impeller, turbine, and guide wheel, with the guide wheel guiding the fluid flow direction. By adjusting the opening angle of the guide wheel, the pump impeller torque can be changed, thereby regulating the output torque and speed, ultimately achieving "front-end speed regulation" and converting the variable-speed input of the wind turbine into a constant-speed output of the generator. See [link to front-end speed regulation unit coordination control structure] for details. Figure 2 .

[0053] In one specific embodiment, the mathematical model of each part of the front-end speed regulating unit includes:

[0054] (1) Wind turbine:

[0055] The aerodynamic torque and power generated by the wind turbine capturing wind energy are calculated using the following formulas:

[0056] (1)

[0057] (2)

[0058] In the formula: Tw The torque of the wind turbine is N·m; n w The rotor speed is r / min; P w The output power of the wind turbine is expressed in kW. v Wind speed, m / s; R Let be the radius of the wind turbine, in meters. ρ air density, kg / m³ 3 C p Wind energy utilization coefficient; β The pitch angle; λ The tip speed ratio; ω is the angular velocity of the wind turbine.

[0059] (2) Transmission chain:

[0060] The derivation and calculation are based on the following assumption: the input shaft of the synchronous generator moves counterclockwise. According to the right-hand rule, the positive direction of the angular velocity is determined, and the positive and negative directions of the torque are correspondingly defined as: counterclockwise is positive, and clockwise is negative.

[0061] ① Gearbox:

[0062] The output speed, power, and torque of the low speed ratio speed-increasing gearbox are expressed as shown in equation (3).

[0063] (3)

[0064] In the formula, T w This refers to the wind turbine torque; P w This refers to the output power of the wind turbine. n w This refers to the wind turbine rotation speed; i 1 represents the growth rate ratio; η 1 represents transmission efficiency. η 1 = 0.975; n 1 represents the output speed; P 1 represents output power; T 1 represents the output torque.

[0065] ②Hydraulic torque converter:

[0066] The key performance parameters of the hydraulic torque converter, including speed, speed ratio, torque and power, are shown in Equation (4).

[0067] (4)

[0068] In the formula: T w This refers to the wind turbine torque; P w This refers to the output power of the wind turbine.n w This refers to the wind turbine rotation speed; v Wind speed; λ The tip speed ratio; n 0 represents the output speed of the hydraulic torque converter; α 1 and α 2 represents the structural parameters of the first and second row planetary arrays; i TB The ratio of pump impeller speed to turbine speed. i TB = n T / n B ; K TB This is the ratio of pump impeller torque to turbine torque. K TB = T T / T B ; T G This refers to the output torque of the hydraulic torque converter. P G This refers to the output power of the hydraulic torque converter; η m =0.97, which is the mechanical friction correction coefficient for the hydraulic torque converter.

[0069] (3) EESG model:

[0070] The EESG excitation system model is as follows:

[0071] (5)

[0072] In the formula: The rate of change of the excitation voltage; E fd This is the excitation voltage, measured in volts (V). The time constant of the excitation system; Sampling time; For PWM gain; For direct-axis armature reactive reactance; The resistance of the excitation winding; To control the voltage.

[0073] When establishing the EESG model, the effects of damping winding, speed variation, and stator winding transients are ignored, but the role of excitation winding is taken into account. Based on this, the third-order model is derived by combining equation (5): (6)

[0074] In the formula: K is the armature reaction coupling coefficient, c is the excitation coupling coefficient. ; , Let d and q be the transient potentials; The rate of change of the transient potential along the q-axis; , These are the subtransient potentials along the d-axis and q-axis; The rates of change of the subtransient potential along the d-axis and q-axis; , , These are the d-axis reactance, transient reactance, and q-axis reactance. This refers to the d-axis winding current. The d-axis subtransient reactance; T d0 The d-axis open-circuit time constant; The open-circuit time during the transient state of the d-axis; , For the open-circuit time during the subtransient states of the d-axis and q-axis; Uc To control the voltage.

[0075] In one specific embodiment, the division of the operating modes of the front-end speed-regulating wind turbine includes:

[0076] Based on changes in wind speed and turbine operating conditions, this embodiment divides the operation of the front-end speed-regulating wind turbine into six modes, such as... Figure 3 As shown, the modes include standby, startup, maximum wind energy capture, constant speed, constant power, and shutdown protection. For each of these modes and its different control objectives, a corresponding coordinated control strategy needs to be determined.

[0077] This study investigates the coordinated control of various subsystems during grid connection of front-end variable speed wind turbine units operating in maximum wind energy capture mode.

[0078] In one specific embodiment, the grid connection synchronization standard for front-end speed-regulating wind turbine units includes:

[0079] Employing a brushless electrically excited synchronous generator, the front-end speed-regulating wind turbine unit can achieve grid connection through self-synchronization or quasi-synchronization, similar to conventional hydropower and thermal power units. Its no-load grid connection model is as follows: Figure 4 As shown.

[0080] The instantaneous voltage value on the grid side is:

[0081] (7)

[0082] In the formula, U ms Voltage amplitude; For grid voltage u s angular frequency; This represents the initial phase angle of the system voltage.

[0083] Let the output voltage of EESG be:

[0084] (8)

[0085] In the formula: U mg Voltage amplitude; Output voltage u g angular frequency; The initial phase angle of the output voltage.

[0086] The voltage difference between the generator and the system before grid connection can be obtained from equations (7) and (8). u e Phase angle difference and angular frequency difference They are respectively:

[0087] (9)

[0088] To achieve ideal grid connection, the generator needs to ensure the performance of equation (9). u e , and The phase sequence and voltage waveform must be equal to zero, and simultaneously, they must be consistent with the power grid. Given that this requirement is difficult to fully meet in engineering practice, a certain deviation is permissible during grid connection operation. This standard is called the quasi-synchronization condition, which specifically includes the following three conditions:

[0089] (1) Permissible deviation range of phase angle: .

[0090] (2) Voltage allowable deviation range: ±5% to ±10% of rated voltage.

[0091] (3) Frequency allowable deviation range: ±0.2% to ±0.5% of the rated frequency.

[0092] To facilitate rapid synchronization of the EESG after grid connection, in practice, while meeting the aforementioned quasi-synchronization conditions, the generator output voltage should be intentionally set to be slightly ahead in phase and slightly higher in frequency than the grid. In this way, the generator can absorb active load immediately upon closing, and the resulting braking torque will accelerate its entry into synchronous operation.

[0093] In one specific embodiment, the DMPC coordination control structure proposed in this invention includes:

[0094] To address the challenges of controlling multi-variable, strongly coupled systems before grid connection of front-end variable-speed wind turbine generators, this invention designs a coordinated control framework based on DMPC, the structure of which is as follows: Figure 5As shown, this architecture decomposes the overall coordination and control task of the unit into three parallel local MPC controllers, corresponding to the propeller, hydraulic speed regulation and excitation subsystems respectively.

[0095] Specifically, unlike traditional distributed control, the core of this invention's framework lies in introducing consistency variables and designing an iterative optimization mechanism based on the alternating direction multiplier method (ADMM) to achieve close coordination among local controllers.

[0096] During grid connection, the generator speed ω g and terminal voltage V t These are key coupled variables characterizing the overall system state and directly affecting grid connection quality. Therefore, we treat the future predicted trajectories of these two variables as consistency variables. Each local controller i Each will generate its own prediction sequence for these variables. ω g，i (k) and V t，i (k), and through information interaction and iteration, ensure that all local prediction sequences eventually converge to a common, globally consistent trajectory, i.e. ω g，i → ω g，c and V t，i → V t，c Here, the subscript 'c' represents the consensus variable after consensus is reached. This method transforms the complex global optimization problem into a series of local optimization subproblems that can be solved in parallel on each subsystem controller, greatly reducing the complexity of online computation while ensuring control performance.

[0097] Specifically, the ADMM algorithm includes:

[0098] The standard form of ADMM is:

[0099] (10)

[0100] In the formula: ; , It is a convex function with respect to x and z.

[0101] Its augmented Lagrangian function is shown in equation (11), and the corresponding ADMM iteration steps are given by equations (12)–(14).

[0102] (11)

[0103] (12)

[0104] (13)

[0105] (14)

[0106] In the formula: λ 1 represents the augmented Lagrange multiplier vector; As a penalty factor, k is the number of iterations.

[0107] In the standard ADMM iteration, each subproblem is solved sequentially in a predetermined order. The output of the preceding subproblem is directly used as the input of the subsequent subproblem, and the Lagrange multipliers are updated after all subproblems have been iterated once.

[0108] The ADMM algorithm is used to augment the standard MPC optimization problem for each subsystem. ADMM incorporates consistency constraints into the objective function of each subsystem by introducing augmented Lagrangian terms.

[0109] Specifically, the pitch subsystem MPC design:

[0110] Based on equations (1)-(3), a steady-state operating point is selected, and the above nonlinear dynamic equations are linearized using a first-order Taylor series expansion to obtain a continuous-time state-space expression. The state vector is defined as... Input is The disturbance is The linearized model is as follows:

[0111] (15)

[0112] In the formula: variables marked with "~" represent their deviation from the operating point value. Matrix Let be the Jacobian matrix of the system at the operating point.

[0113] (1) Prediction Model:

[0114] The continuous model described above is discretized using a zero-order hold (ZOH), with a sampling period of [missing information]. T s The discrete-time incremental state-space prediction model required for the MPC controller is obtained as follows:

[0115] (16)

[0116] In the formula: , This represents the control increment at the current moment.

[0117] (2) Rolling optimization:

[0118] MPC determines the current control input by solving a finite-time optimal control problem at each sampling time. The objective function Jp of the pitch subsystem is designed as a quadratic cost function, aiming to minimize the deviation between the predicted output and the reference trajectory in the prediction time Np, and the severity of the control action in the control time Nc.

[0119] Objective function:

[0120] (17)

[0121] In the formula: Np For prediction in the time domain; Nc To control the time domain; This represents the increment of the propeller pitch angle; Let be the predicted value of the wind turbine rotation speed at time k+j from time k. This is a reference value for the wind turbine speed at the future time k+j, i.e. . Let be the rate of change of the pitch angle at time k+j, calculated at time k. Q p and R p It is a positive definite weight matrix.

[0122] Constraints:

[0123] 1) Pitch angular position constraints:

[0124] (18)

[0125] 2) Pitch angle change rate constraint:

[0126] (19)

[0127] (3) Feedback correction:

[0128] The core of the feedback correction mechanism is a state observer that uses a Kalman filter to obtain the actual pitch angle of the pitch system at the beginning of each control cycle k. Wind turbine speed The measured value.

[0129] Measured values Compared with the predicted value at the previous time point The Kalman filter uses this prediction error, combined with the noise statistics of the system and measurements, to generate the optimal estimate of the current state. .

[0130] (20)

[0131] In the formula: K p Kalman gain; This is the current optimal estimate. This is the estimated value from the previous moment; For measurement; C y This is the output matrix of the state-space model.

[0132] This optimal estimate This serves as the initial state for the current moment's rolling optimization.

[0133] Specifically, the MPC design of the hydraulic torque converter speed control subsystem:

[0134] The core objective of the hydraulic speed control subsystem is to increase the generator speed. The rotational speed is precisely adjusted and stabilized at 1500 r / min to maintain synchronization with the 50Hz power grid, thus meeting the grid connection requirements for frequency synchronization. This decision directly impacts the consistency variables. .

[0135] The dynamic model of the subsystem can be obtained from equation (4) and the rotor motion equation of the generator:

[0136] (twenty one)

[0137] In the formula: J g The moment of inertia of the generator. T em It is electromagnetic torque. T G Pump impeller speed Generator speed and guide vane opening angle The nonlinear function is the output torque of the hydraulic torque converter, where the turbine speed = generator speed = 1500.

[0138] (1) Prediction Model:

[0139] The model is linearized near the synchronous speed operating point. The state vector is defined as follows. The input is The predicted pump impeller speed transmitted from upstream. As a measurable disturbance Incorporate into the model. After linearization and discretization, the prediction model is obtained in the following form:

[0140] (twenty two)

[0141] (2) Rolling optimization:

[0142] The objective function, based on the standard quadratic form, adds an augmented Lagrangian term based on ADMM to penalize the locally predicted generator speed. With globally consistent variables The deviation between them.

[0143] (twenty three)

[0144] In the formula: Np For prediction in the time domain; Nc To control the time domain; This represents the increment of the guide vane opening. and The globally consistent rotational speed and its corresponding dual variable obtained in the previous iteration; As a penalty factor, Let k be the predicted value at time k+j in the future. This represents the control increment at time k for the future time k+j.

[0145] Constraints:

[0146] 1) Guide vane opening angle position constraint:

[0147] (twenty four)

[0148] 2) Guide vane opening angle change rate constraint:

[0149] (25)

[0150] 3) Frequency deviation constraint:

[0151] (26)

[0152] (3) Feedback correction:

[0153] At time k, the actual generator speed is measured. Using a Kalman filter, based on The difference between the original and predicted values ​​is used to update the optimal estimate of the dynamic state of the transmission chain. .Will As initial conditions for optimization.

[0154] Specifically, the excitation subsystem MPC design:

[0155] The linear differential equations for the dynamics of the generator's dq-axis flux linkage and current:

[0156] (27)

[0157] (28)

[0158] In the formula: For stator d-axis flux linkage; For stator q-axis flux linkage; It is the excitation current; The electrical system state matrix; The electric angular velocity of the generator; For input control matrix; To control input variables; V t This refers to the generator terminal voltage amplitude. V d The stator d-axis voltage component; V q This refers to the q-axis voltage component of the stator. i d , i q These are the stator d-axis and q-axis currents.

[0159] Within the DMPC framework, receive future speed prediction sequences from the speed regulation MPC. Using a time-varying system matrix in the prediction time domain greatly improves the accuracy of the prediction model.

[0160] (1) Prediction Model:

[0161] Discretizing the above model and linearizing the output equation yields the standard discrete-time state-space prediction model:

[0162] (29)

[0163] In the formula: x e It is a state vector containing magnetic flux and current.

[0164] (2) Rolling optimization:

[0165] objective function J e It also includes ADMM augmentations to ensure its locally predicted terminal voltage. V t ( k ) and globally consistent variables V t，c ( k (To maintain consistency)

[0166] (30)

[0167] In the formula: Np For prediction in the time domain; Nc To control the time domain; This represents the increment of the excitation voltage. and This represents the globally consistent voltage and its corresponding dual variable obtained in the previous iteration. The weight matrix Q... e A large value will be assigned to strongly constrain voltage tracking performance and minimize the deviation between the predicted terminal voltage and the grid voltage reference value.

[0168] Constraints:

[0169] 1) Excitation voltage constraint:

[0170] (31)

[0171] 2) Voltage deviation constraint:

[0172] (32)

[0173] (3) Feedback correction:

[0174] The actual terminal voltage of the generator is measured at time k. State observer utilizes To correct the electrical state estimation of the generator .use As the initial state.

[0175] For details, see Figure 6 The coordinated control based on the ADMM-DMPC algorithm provided by this invention is as follows:

[0176] At each sampling period k, the three local controllers execute a fast internal iterative process in parallel until the predictions for the consistency variables converge.

[0177] 1) Each controller measures the current state. x i ( k The iteration counter m=0. The initial values ​​of the consistency and dual variables are inherited from the previous time step k-1.

[0178] 2) Parallel optimization. Pitch MPC solves its optimization problem. J p The optimal control sequence is obtained. And the predicted wind turbine speed trajectory. Speed ​​regulation MPC uses the consistency variables from the previous round. and dual variables Solve its augmented optimization problem J s To obtain local optimal control and new local speed prediction Excitation MPC usage As model parameters, and used and Solve its augmented optimization problem J e To obtain local optimal control and new local voltage prediction .

[0179] 3) Information exchange and consistency variable updates. Each subsystem shares its newly calculated prediction sequences through a communication network. and The updated value of the globally consistent variable is calculated using a distributed average consensus algorithm, typically the average of all local predictions.

[0180] (33)

[0181] (34)

[0182] 4) Dual variable update. Each subsystem updates its dual variables locally:

[0183] (35)

[0184] (36)

[0185] 5) Convergence check. Check if the residual is less than the threshold. If convergence is achieved, the iteration ends; otherwise, set m = m + 1 and return to step 2.

[0186] 6) Control Implementation. After iterative convergence, each subsystem will assign the first element of its optimal control sequence. It acts on the physical actuator. The system enters the next sampling time k+1.

[0187] Through the above process, the DMPC strategy proposed in this invention tightly couples the independent decisions of each subsystem with the global objective. By coordinating the optimization of consistency variables, it ensures that the synchronization conditions of frequency and voltage can be met quickly and smoothly before grid connection.

[0188] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0189] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A distributed model predictive control method for offshore front-end speed-regulating wind turbine units, characterized in that, Includes the following steps: S1. Decompose the control task of the front-end speed-regulating wind turbine into a pitch subsystem, a hydraulic torque-regulating speed-regulating subsystem, and an excitation subsystem, and establish local discrete prediction models for the pitch subsystem, the hydraulic torque-regulating speed-regulating subsystem, and the excitation subsystem, respectively. S2. At each sampling moment, the current state of the pitch subsystem, hydraulic torque regulation subsystem and excitation subsystem is collected, and based on the local discrete prediction model, rolling optimization is performed to predict the state trajectory of each pitch subsystem, hydraulic torque regulation subsystem and excitation subsystem in the future prediction time domain. S3. Taking the generator speed and terminal voltage required for grid connection synchronization as the synchronization target, the future predicted trajectory of the generator speed and terminal voltage is used as a global consistency variable. S4. Construct an augmented Lagrangian cost function that includes the global consistency variables, and use the alternating direction multiplier method to iteratively solve the problem so that the local predicted trajectories of the pitch subsystem, the hydraulic torque regulation subsystem, and the excitation subsystem converge to the globally consistent optimal trajectory. S5. After iterative convergence, the first control command in the optimal control sequence corresponding to the pitch subsystem, hydraulic torque variable speed control subsystem, and excitation subsystem is output to the corresponding pitch system, hydraulic torque variable speed control system, and excitation system to achieve coordinated control before grid connection.

2. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, In S1, the structure of the front-end speed-regulating wind turbine includes a wind turbine, a gearbox, a hydraulic torque converter, and a brushless electrically excited synchronous generator.

3. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, S1 includes: A local discrete prediction model for the pitch subsystem is established based on the linearized aero-mechanical dynamic model of the wind turbine and its drive train. A local discrete prediction model for the hydraulic torque converter speed regulation subsystem is established based on the linearized dynamic equations of the hydraulic torque converter and the generator rotor motion. A local discrete prediction model for the excitation subsystem is established based on the linearized electromagnetic dynamic equations of the generator.

4. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, In S2, the rolling optimization is based on a quadratic objective function, which includes the system state tracking error and the control input cost.

5. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 4, characterized in that, In the rolling optimization objective function of the hydraulic torque variable speed control subsystem and the excitation subsystem, an augmented Lagrange term is added to penalize the deviation between the global consensus variable and the global consensus value in its local predictions.

6. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, In S3, the synchronization target of the generator speed is the grid synchronization speed, and the synchronization target of the terminal voltage is the grid voltage.

7. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, In S4, the iterative solution process using the alternating direction multiplier method includes the following sub-steps: S41, the pitch subsystem, hydraulic torque-regulating speed control subsystem, and excitation subsystem solve their respective local optimization problems in parallel based on the current state, inheriting or initializing the globally consistent variables and dual variables; S42. Based on the local optimization problem, obtain the locally optimal control sequence and the corresponding local prediction sequence of the global consistency variable; S43, the pitch subsystem, hydraulic torque variable speed control subsystem and excitation subsystem exchange the local prediction sequences and calculate the updated global consensus variables through a distributed average consensus algorithm; S44, the pitch subsystem, the hydraulic torque converter speed control subsystem, and the excitation subsystem update their dual variables locally; S45. Determine whether the iteration has converged. If it has converged, proceed to S46. If it has not converged, return to step S41. S46. After iterative convergence, each subsystem applies the first element of its optimal control sequence to the physical actuator.

8. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 7, characterized in that, In S43, the updated global consistency variable value is the average value of the local prediction sequence corresponding to the pitch subsystem, hydraulic torque regulation subsystem, and excitation subsystem.

9. The distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, It also includes: using a state observer to perform feedback correction on the initial state of the corresponding prediction model based on the actual measured values ​​of the pitch subsystem, hydraulic torque regulation subsystem, and excitation subsystem.

10. A distributed model predictive control method for offshore front-end speed-regulating wind turbines according to claim 1, characterized in that, In S5, the goal of the coordinated control is to make the frequency and amplitude of the generator output voltage meet the quasi-synchronous grid connection conditions.

Images