A finite time stabilization control method for a wind turbine random nonlinear system

Abstract

The present application relates to the technical field of wind power generation control, and particularly relates to a finite time stabilization control method for a random nonlinear system of a wind turbine generator set, comprising: dividing the wind turbine generator set into a two-order wind wheel transmission chain mechanical subsystem and a one-order generator electrical subsystem, constructing a bidirectional strong coupling correlation topology by introducing state nonlinear power terms and control coupling terms; then determining homogeneous degree parameters of the controller based on a homogeneity theory and a homogeneous degree general recursive calculation, combining control gains and scaling parameters to construct a decentralized state feedback controller; subsequently establishing a double closed loop signal flow network composed of the double subsystems, the double controllers and two random noise modules by using software; further performing iterative solution by using a discretized Euler-Maruyama numerical solver to output time domain evolution data; finally setting different control parameters to repeatedly solve, calculating system state fluctuation characteristic data and control characteristic data and classifying and tuning control scaling parameters according to the same.

Description

Technical Field

[0001] This invention relates to the field of wind power generation control technology, and in particular to a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator set. Background Technology

[0002] Wind turbine generators operate under variable field conditions for extended periods, facing multiple continuous random disturbances such as turbulent wind speed disturbances, mechanical shaft vibrations, grid voltage fluctuations, and high-frequency electrical noise from inverters. To accurately capture the operating status of the generator set, existing technologies typically deconstruct the wind power generation system into a front-end mechanical body and a rear-end generator electrical terminal. Because the system encompasses multiple physical multi-domain units, including the wind turbine, flexible drive shaft, and generator, the entire system exhibits extremely strong random nonlinearity and strong coupling dynamic characteristics across subsystems and terminals.

[0003] Most existing wind power control technologies employ traditional PID control, sliding mode control, or adaptive control strategies. These schemes are essentially asymptotically stable controls, and their state variables have slow convergence response speeds under random disturbances. Because they cannot suppress the system state to zero within a defined time, existing wind power control systems struggle to meet the urgent engineering requirements for rapid stabilization of large wind turbines under severe turbulence and highly random operating conditions. Summary of the Invention

[0004] To overcome the above deficiencies, this invention provides a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator set, aiming to improve the problem that existing wind power control systems are unable to meet the requirements for rapid stabilization of large wind turbine generator sets under severe conditions of strong turbulence and high randomness.

[0005] This invention provides the following technical solution: a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator, comprising:

[0006] S1. Divide the wind turbine generator set into a second-order wind turbine transmission chain mechanical subsystem and a first-order generator electrical subsystem. Establish the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems respectively. Introduce state nonlinear power terms and control coupling terms into the two subsystems to construct a bidirectional strongly coupled correlation topology.

[0007] S2. Based on homogeneity theory, construct a mechanical end controller and an electrical end controller that receive speed and torque state variables. Determine the homogeneity parameters of the controller through recursive calculation using the general formula for homogeneity. Calculate and output the control input by combining the control gain parameters and control scaling parameters.

[0008] S3. Use software to establish a dual closed-loop signal flow network consisting of the two subsystems, two controllers, and two random noise modules that generate random disturbance terms;

[0009] S4. The discretized Euler-Maruyama numerical solver is used to iteratively solve the interconnection model in the dual closed-loop signal flow network, and the time-domain evolution data of the state variables and control inputs are output.

[0010] S5. Set different control gain parameters and control scaling parameters and repeatedly perform iterative solutions to calculate the system state fluctuation characteristic data and control characteristic data derived from the time domain evolution data, and adjust the control scaling parameters accordingly.

[0011] Preferably, in S1, establishing the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems respectively includes:

[0012] The speed state variable and the torque state variable are established as the state variables of the second-order wind turbine transmission chain mechanical subsystem, and the torque state variable outputs mechanical coupling to the first-order generator electrical subsystem;

[0013] The rotational speed state variable is established as the state variable of the first-order generator electrical subsystem and transmitted to the two controllers as a feedback signal;

[0014] Establish the pitch angle control input as the control input of the second-order wind turbine drive chain mechanical subsystem, and establish the voltage control input as the control input of the first-order generator electrical subsystem.

[0015] Independent first-standard Brownian motion and second-standard Brownian motion are introduced as random disturbance terms and input to the second-order wind turbine transmission chain mechanical subsystem and the first-order generator electrical subsystem, respectively.

[0016] Preferably, in S1, the introduction of state nonlinear power terms and control coupling terms into the two subsystems to construct a bidirectional strongly coupled topology includes:

[0017] In the mechanical subsystem of the second-order wind turbine drive chain, the torque state variable is introduced into the drift term of the speed state variable in a nonlinear power form;

[0018] In the diffusion term of the second-order wind turbine drive chain mechanical subsystem, a voltage control input is introduced, and a control coupling term is formed by the nonlinear product of the voltage control input and the pitch angle control input.

[0019] A nonlinear power term for the torque state variable is introduced into the random disturbance term of the first-order generator electrical subsystem, and multiplied by the second standard Brownian motion through a sine function;

[0020] By controlling the dynamic interaction between coupling terms and random perturbation terms, a bidirectional strongly coupled correlation topology is constructed.

[0021] Preferably, in S2, the determination of the homogeneity parameters of the controller through recursive calculation using the homogeneity formula includes:

[0022] Based on the power exponent, preset initial homogeneity parameter and initial scaling parameter set by each of the two subsystems, the corresponding homogeneity parameter is selected within the negative value constraint range.

[0023] The subsequent homogeneity parameters are calculated using the recursive formula for homogeneity parameters;

[0024] The control scaling parameters for each of the two subsystems are calculated using the state scaling recursive formula.

[0025] Preferably, in S2, calculating and outputting the control input by combining the control gain parameter and the control scaling parameter includes:

[0026] The control scaling parameters are applied to the state variables of the second-order wind turbine drive chain mechanical subsystem by a preset fractional power to perform state transformation and power transformation.

[0027] By combining the preset control gain parameters, the transformed variables are cascaded with the speed state variables of the first-order generator electrical subsystem introduced across the terminal to calculate the multi-state collaborative sensing feedback variables.

[0028] The multi-state collaborative sensing feedback variables are subjected to nonlinear power transformation to output the pitch angle control input.

[0029] The speed state variables of the first-order generator electrical subsystem are subjected to square nonlinear power transformation to solve for the electrical terminal state feedback variables.

[0030] The electrical terminal state feedback variables are subjected to nonlinear power transformation, and the output voltage controls the input.

[0031] Preferably, in S3, establishing a dual closed-loop signal flow network includes:

[0032] The first noise module inputs a first random disturbance term consisting of a first standard Brownian motion into the second-order wind turbine transmission chain mechanical subsystem, and transmits its state variable to the mechanical end controller.

[0033] The mechanical end controller is used to input the pitch angle control input to the mechanical subsystem of the second-order wind turbine drive chain;

[0034] The second noise module inputs a second random disturbance term consisting of the second standard Brownian motion into the first-order generator electrical subsystem, and transmits its state variable back as a feedback signal.

[0035] The voltage control input is calculated using the electrical terminal controller and then input into the closed electrical main control loop within the first-order generator electrical subsystem.

[0036] Preferably, the establishment of the dual closed-loop signal flow network further includes:

[0037] The state variables of the first-order generator electrical subsystem are used as shared feedback signals and simultaneously input to both the mechanical and electrical controllers.

[0038] The calculated voltage control input is simultaneously introduced across the line into the mechanical subsystem of the second-order wind turbine drive chain. By utilizing the dynamic interaction between the voltage control input and the pitch angle control input, a cross-end collaborative control loop for the entire system's operating status is established between the two controllers.

[0039] Preferably, in S4, the iterative solution of the interconnection model in the dual closed-loop signal flow network using a discretized Euler-Maruyama numerical solver includes:

[0040] Set a fixed simulation step size and total simulation duration, and initialize the state variables at the start time;

[0041] Two random number seeds are fixed to reproduce discrete Brownian motion increments;

[0042] Using the state variables from the previous moment, the simulation step size, and the control input, the mechanical and electrical state variables at the current moment are updated using discrete difference expressions, and the time-domain evolution data is output.

[0043] Preferably, in S5, calculating the system state fluctuation characteristic data and control characteristic data derived from the time-domain evolution data includes:

[0044] Obtain the random fluctuation values ​​of state variables near the equilibrium point and the convergence time at each time point in the time-domain evolution data;

[0045] Calculate system state fluctuation characteristic data to characterize operational stability and response speed;

[0046] Extract the control input values ​​from the time-domain evolution data and calculate the control characteristic data used to characterize the energy consumption of regulation.

[0047] Preferably, in S5, adjusting the control scaling parameters accordingly includes:

[0048] Obtain the corresponding preset control priority tendency based on the wind farm operating conditions;

[0049] When the control priority is to prioritize high precision and steady state, reduce the value of the control scaling parameter;

[0050] When the control priority is to prioritize high response speed and real-time performance, increase the value of the control scaling parameter;

[0051] The adjusted control scaling parameters are fed back to the two controllers for categorized tuning.

[0052] The present invention has the following beneficial effects:

[0053] 1. In this invention, by introducing a product term of voltage and pitch angle into the diffusion term of the mechanical subsystem and a nonlinear power term of torque state into the disturbance term of the electrical subsystem, a comprehensive cascading and bidirectional coordinated stabilization of the front-end mechanical body and the back-end electrical device in the channel is realized; the problem of traditional wind power control implementing completely unidirectional decoupling control between the mechanical end and the electrical end, resulting in cross-end data discontinuity and causing severe impact of flexible torque on the unit's transmission chain shaft system is solved.

[0054] 2. In this invention, a distributed state feedback controller constructed based on homogeneity and finite-time stability theory is used to strictly recursively determine the controller parameters using the general formula of homogeneity and state scaling. This enables the rapid finite-time convergence of the state variables of each system to the zero equilibrium point within a preset time interval, and the amplitude of steady-state random fluctuations after convergence is precisely suppressed to a very small proportion. This solves the problem that traditional wind power control strategies are mostly asymptotically stable control with slow convergence response speed, which cannot meet the finite-time rapid stabilization requirements of large wind turbine units under strong turbulence conditions.

[0055] 3. In this invention, a quantifiable performance analysis mechanism is established, including indicators such as steady-state error, mean square error, maximum transient deviation, and convergence time. Based on different wind farm conditions and control priority tendencies, the values ​​of the control scaling parameters fed back to the two controllers are classified and tuned: when high precision and steady-state priority are required, the scaling parameters are reduced, achieving high steady-state precision and small transient impact fluctuations with conservative control output; when abrupt changes require high response speed, the scaling parameters are increased, significantly compressing the average convergence time of multiple state variables. This solves the problem that conventional single control configurations cannot accommodate and dynamically adapt to the complex operating scenarios of different wind farms, resulting in an inability to flexibly switch between system control precision, response speed, operational stability, and energy economy. Attached Figure Description

[0056] Figure 1 This is a flowchart of a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention;

[0057] Figure 2 The system state response curve of the finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention. Schematic diagram;

[0058] Figure 3 The system state response curve of the finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention. Schematic diagram;

[0059] Figure 4 The system state response curve of the finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention. Schematic diagram;

[0060] Figure 5 The control input response curve of the finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention. Schematic diagram;

[0061] Figure 6 The control input response curve of the finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator proposed in this invention. Schematic diagram;

[0062] Figure 7 This invention proposes a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator. Phase plane trajectory diagram;

[0063] Figure 8 This invention proposes a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator. Phase plane trajectory diagram. Detailed Implementation

[0064] The technical solutions in 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.

[0065] In embodiments of the present invention, a finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator set is provided, such as... Figure 1 As shown, it includes the following steps:

[0066] S1. Divide the wind turbine generator set into a second-order wind turbine transmission chain mechanical subsystem and a first-order generator electrical subsystem. Establish the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems respectively. Introduce state nonlinear power terms and control coupling terms into the two subsystems to construct a bidirectional strongly coupled correlation topology.

[0067] Furthermore, in S1, the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems are established respectively, including:

[0068] The speed state variable and the torque state variable are established as the state variables of the second-order wind turbine transmission chain mechanical subsystem, and the torque state variable outputs mechanical coupling to the first-order generator electrical subsystem.

[0069] The rotational speed state variable is established as the state variable of the first-order generator electrical subsystem and transmitted to the two controllers as a feedback signal.

[0070] The pitch angle control input is established as the control input of the second-order wind turbine drive chain mechanical subsystem, and the voltage control input is established as the control input of the first-order generator electrical subsystem.

[0071] Independent first-standard Brownian motion and second-standard Brownian motion are introduced as random disturbance terms and input to the second-order wind turbine drive chain mechanical subsystem and the first-order generator electrical subsystem, respectively.

[0072] Furthermore, in S1, state nonlinear power terms and control coupling terms are introduced into the two subsystems to construct a bidirectional strongly coupled topology, including:

[0073] In the mechanical subsystem of the second-order wind turbine drive chain, the torque state variable is introduced into the drift term of the speed state variable in a nonlinear power form;

[0074] In the diffusion term of the second-order wind turbine drive chain mechanical subsystem, a voltage control input is introduced, and a control coupling term is formed by the nonlinear product of the voltage control input and the pitch angle control input.

[0075] A nonlinear power term for the torque state variable is introduced into the random disturbance term of the first-order generator electrical subsystem, and multiplied by the second standard Brownian motion through a sine function;

[0076] By controlling the dynamic interaction between coupling terms and random perturbation terms, a bidirectional strongly coupled correlation topology is constructed.

[0077] Specifically, in the system physical partitioning and state establishment phase, the wind turbine generator is deconstructed into two interconnected dynamic subsystems. Specifically, the wind turbine drivetrain constitutes a second-order mechanical subsystem, whose internal state is characterized in real time by two continuously changing physical quantities: the wind turbine speed deviation state variable. Dynamic state variables of transmission chain torque The torque state variable The torque variation of the flexible shaft in the drive train, reflecting the key variable connecting the wind turbine and the generator, outputs mechanical coupling to the subsequent first-order generator electrical subsystem. The generator side then constitutes the first-order electrical subsystem, whose core state is determined by the generator speed deviation state variable. The variable is characterized and transmitted as a feedback signal to both the mechanical and electrical controllers. At the control input, the second-order wind turbine drive chain mechanical subsystem is controlled by the pitch angle. To implement regulation, the first-order generator electrical subsystem is controlled by the generator terminal voltage input. Adjustments are implemented. To address the uncontrollable wind field environment, the system introduces independent standard Brownian motions. and As random disturbance terms, they represent random disturbances such as turbulent wind speed disturbance, mechanical vibration, grid voltage fluctuation, and inverter noise.

[0078] In the dynamic construction and operation of the bidirectional strongly coupled topology, the state nonlinearity power term and the control coupling term achieve a deep reconstruction of the system's inherent nonlinearity. Within the second-order wind turbine drive train mechanical subsystem, the drive train state... This is a first-order pure control drive state, while the wind turbine state... By relying on the power-law nonlinear feedback of the transmission chain, the torque state variable is introduced into the drift term of the wind turbine speed deviation state variable in a nonlinear power-law form, forming an internal series cascade structure. Simultaneously, a voltage control input is introduced into the diffusion term of the mechanical subsystem. Through voltage control input and pitch angle control input The nonlinear product constitutes the control coupling term, representing the coupling of turbulent wind speed disturbance to the rotor speed in a real wind power system. A nonlinear power term for the torque state variable is introduced into the stochastic disturbance term of the first-order generator electrical subsystem, and then coupled with a sine function and the second standard Brownian motion. Multiplication represents the coupling of electrical noise to generator speed, reflecting the influence of wind turbine torque fluctuations on the generator. By controlling the dynamic interaction between the coupling term and the random disturbance term, a bidirectional strongly coupled topological network is ultimately established between the two subsystems.

[0079] To ensure the mathematical completeness and sufficiency of the above topology and to directly support the subsequent solution of the control law, the stochastic differential dynamics equations of this strongly coupled network are expressed as follows:

[0080] ;

[0081] The complete physical and mathematical meanings of the parameters and symbols in the above equations are defined as follows: This refers to the deviation in wind turbine speed; For the dynamic torque of the transmission chain; This refers to the generator speed deviation. This is a control value for the pitch angle; This is a generator terminal voltage control quantity; and These are mutually independent standard Brownian motions; The time variable is represented by the fractional power exponents 5 / 3 and 1, which correspond to the first and second level nonlinear power exponents established within the mechanical subsystem, respectively; the fractional power exponent 7 / 5 corresponds to the nonlinear shaft work power exponent introduced into the electrical subsystem at the cross-end. Furthermore, in the equation... To transform the representation of custom symbolic functions, for ,in It is a standard symbolic function for determining the direction of a variable.

[0082] In practical industrial applications, the operation of this topology network begins with the synchronous parallel acquisition of wind speed capture and sensor signals. Within each sampling cycle, the rotational speed deviation on the rotor side and the axial torque of the drive shaft are converted into digital signals. and Meanwhile, the electromagnetic speed deviation of the generator The data is fed back from the other end. The nonlinear power-law cascade terms in the state equation. The trajectory of the wind turbine speed drift term on the mechanical side was dynamically rewritten. When fluctuations occur in the external wind field, this was addressed through the random disturbance term channel. and The interaction between the mechanical side and the torque fluctuations are achieved through... The real-time mapping is applied to the generator speed diffusion channel on the electrical side, while the voltage term of the control strategy on the electrical side is implemented through... The cross-end reverse action acts on the wind turbine speed fluctuation channel on the mechanical side. By fully cascading physical control variables and state variables in the diffusion and drift channels, the finite-time stabilization target of the entire system in a probabilistic sense obtains support at the underlying physical topology level, laying a complete state-space foundation for the accurate solution of the subsequent homogeneous distributed control law.

[0083] S2. Based on homogeneity theory, construct mechanical end controller and electrical end controller that receive speed and torque state variables. Determine the homogeneity parameters of the controller by recursive calculation using the general formula for homogeneity. Calculate and output the control input by combining the control gain parameters and control scaling parameters.

[0084] Furthermore, in S2, the homogeneity parameters of the controller are determined by recursive calculation using the homogeneity formula, including:

[0085] Based on the power exponent, preset initial homogeneity parameter and initial scaling parameter set by each of the two subsystems, select the corresponding homogeneity parameter within the range of negative values.

[0086] The subsequent homogeneity parameters are calculated using the recursive formula for homogeneity parameters;

[0087] The control scaling parameters for each of the two subsystems are calculated using the state scaling recursive formula.

[0088] Furthermore, in S2, the control inputs are calculated and output by combining the control gain parameter and the control scaling parameter, including:

[0089] The control scaling parameters are applied to the state variables of the second-order wind turbine drive chain mechanical subsystem by a preset fractional power to perform state transformation and power transformation.

[0090] By combining the preset control gain parameters, the transformed variables are cascaded with the speed state variables of the first-order generator electrical subsystem introduced across the terminal to calculate the multi-state collaborative sensing feedback variables.

[0091] The multi-state collaborative sensing feedback variables are subjected to nonlinear power transformation to output the pitch angle control input.

[0092] The speed state variables of the first-order generator electrical subsystem are subjected to square nonlinear power transformation to solve for the electrical terminal state feedback variables.

[0093] The electrical terminal state feedback variables are subjected to nonlinear power transformation, and the output voltage controls the input.

[0094] Specifically, in the recursive calculation stage of the homogeneity parameter and scaling parameter, the core design parameters of the controller are all determined recursively based on homogeneity theory through rigorous algebraic formulas. First, the power exponent selected by the second-order wind turbine drive chain mechanical subsystem (subsystem 1) is... , And the power exponent selected for the first-order generator electrical subsystem (subsystem 2). Set initial homogeneity parameters (Among them, mechanical end) ,satisfy The range of negative numbers is constrained; electrical terminals ,satisfy (The range of negative values ​​is constrained). Simultaneously, initial state scaling parameters are set. With initial control scaling item .

[0095] Based on this, a general recursive algorithm for calculating the homogeneity parameter and state scaling is introduced. The general formula of this algorithm is:

[0096] ;

[0097] ;

[0098] Using the above general formula, the preset initial parameters and the power exponents of the corresponding subsystems are substituted into the formula for cascade calculation. For the mechanical subsystem (s=1), from... and The subsequent state variables are calculated recursively using homogeneous exponent parameters. ;Depend on The corresponding subsequent control scaling parameters are calculated recursively. And finally, the power exponent parameter of the forward control gain is calculated. For the electrical subsystem (s=2), since it is a first-order system, the corresponding control scaling parameters can be directly calculated based on the initial parameters and the power relationship. And the state homogeneous power exponent parameter The entire recursive process is driven entirely by deterministic algebraic expressions.

[0099] During the specific calculation and real-time output phase of the control input, the mechanical end controller (controller 1) and the electrical end controller (controller 2) receive the collected state variables and combine them with the preset control gain parameters. , , , Multi-level nonlinear state transformations are implemented. For the mechanical end controller, its operation process first utilizes the calculated control scaling parameters. The dynamic state variable of torque acting on the transmission chain in fractional power form Transform it into an intermediate scaling variable The wind turbine speed deviation state variable The corresponding scaling item is Subsequently, combined with preset control gain parameters... The transformed variables The multi-state collaborative sensing feedback variable is calculated by summing and transforming the generator speed deviation state variable introduced from the electrical subsystem across the terminal. Its operation process satisfies:

[0100] ;

[0101] Finally, this multi-state collaborative sensing feedback variable Perform a nonlinear power transformation and multiply by a factor containing control gain. and The forward coefficient term is used to calculate and output the pitch angle control input. The complete solution algorithm expression is as follows:

[0102] ;

[0103] For the electrical terminal controller, it receives the generator speed deviation state variable. First, a zero-scaling boundary transformation is performed to obtain... Subsequently, a squared nonlinear power transform is applied to it to calculate the electrical terminal state feedback variables. ,satisfy Finally, this is combined with the corresponding electrical feedback gain parameters. With control gain The feedback variable is subjected to a nonlinear power transformation to calculate and output the voltage control input. The complete solution algorithm expression is as follows:

[0104] ;

[0105] In practical industrial applications, the above algorithm is encapsulated within a dual-controller microprocessor unit and runs in a loop. At each sampling moment, the mechanical end controller reads the local... , and cross-end transmission According to containing The nonlinear homogeneous control law directly outputs a continuous and smooth pitch angle control signal. The electrical terminal controller reads synchronously. According to containing The control law outputs a generator terminal voltage regulation signal. Since all homogeneity and scaling parameters are rigorously derived from general formulas based on homogeneity theory, the repeated blind trial-and-error of gain parameters in traditional control strategies is completely avoided. This enables the calculated control input to drive the actuator to quickly suppress and stabilize the system state to near zero within a finite time when facing highly random and turbulent conditions, ensuring that the wind power closed-loop interconnected system exhibits extremely strong disturbance rejection performance and a deterministic convergence trajectory.

[0106] S3. Use software to establish a dual closed-loop signal flow network consisting of two subsystems, two controllers, and two random noise modules that generate random disturbance terms;

[0107] Furthermore, in S3, establishing a dual closed-loop signal flow network includes:

[0108] The first noise module inputs a first random disturbance term consisting of a first standard Brownian motion into the second-order wind turbine transmission chain mechanical subsystem, and transmits its state variable to the mechanical end controller.

[0109] The mechanical end controller is used to input the pitch angle control input to the mechanical subsystem of the second-order wind turbine drive chain;

[0110] The second noise module inputs a second random disturbance term consisting of the second standard Brownian motion into the first-order generator electrical subsystem, and transmits its state variable back as a feedback signal.

[0111] The voltage control input is calculated using the electrical terminal controller and then input into the closed electrical main control loop within the first-order generator electrical subsystem.

[0112] Furthermore, establishing a dual-closed-loop signal flow network also includes:

[0113] The state variables of the first-order generator electrical subsystem are used as shared feedback signals and simultaneously input to both the mechanical and electrical controllers.

[0114] The calculated voltage control input is simultaneously introduced across the line into the mechanical subsystem of the second-order wind turbine drive chain. By utilizing the dynamic interaction between the voltage control input and the pitch angle control input, a cross-end collaborative control loop for the entire system's operating status is established between the two controllers.

[0115] Specifically, in the modular construction phase of the closed-loop simulation system, the entire solution is fully mapped and reproduced using the open-source modeling software OpenModelica. The closed-loop control system built within this simulation platform mainly consists of a second-order wind turbine and transmission chain subsystem module, a first-order generator subsystem module, a dedicated mechanical controller module, an electrical regulating controller module, and two independent random noise modules working together. The two random noise modules generate independent and reproducible standard Brownian increments using fixed random number seeds (specifically, 12345 and 54321). and These underlying modules are tightly coupled through signal flow, forming a dual closed-loop network structure to cope with strong turbulence and highly random disturbances.

[0116] In the basic closed-loop construction of the network topology, the signal flow is mainly divided into two relatively independent main control loops: mechanical and electrical. In the mechanical basic network, the first noise module inputs a first random disturbance term consisting of the first standard Brownian motion into the second-order wind turbine and transmission chain subsystem. This mechanical subsystem receives input and outputs the wind turbine speed deviation. With drive train torque dynamics Two core state variables are directly transmitted to the mechanical controller, enabling multi-state collaborative perception of the mechanical dynamics. After performing state transformation and control law calculations on the received signals, the mechanical controller outputs the pitch angle control input in the reverse direction to the second-order wind turbine and drivetrain subsystem. This closes the deterministic feedback control main loop at the mechanical end. In the electrical foundation network, the second noise module synchronously inputs a second random disturbance term, consisting of the second standard Brownian motion, into the first-order generator subsystem. The noise and the nonlinear torque dynamics transmitted by the mechanical side are affected. After the dual interference, the generator subsystem generates and sends back a feedback signal, the physical meaning of which is the generator speed deviation. After receiving the feedback signal, the electrical controller calculates the voltage control input in real time using its built-in nonlinear square transformation algorithm. It is then input into the first-order generator subsystem to close the basic main control loop of the electrical terminal.

[0117] To overcome the limitations of traditional cross-domain decoupled control, this signal flow network further introduces a cross-end collaborative control loop through cascading interaction. Specifically, the operation process involves the feedback state variables of the first-order generator electrical subsystem. It not only transmits signals to its own electrical controller but also serves as a globally shared feedback signal, simultaneously transmitting them back and inputting them to both the mechanical and electrical controllers. This allows the dedicated mechanical controller to collect real-time speed fluctuations from the electrical side. In the reverse transmission path of the control command, the voltage control input calculated by the electrical controller... The system employs a dual-path parallel flow: one path acts on the generator subsystem, controlling its own electrical regulation, while the other path crosses the border and is introduced into the front-end second-order wind turbine and drivetrain mechanical subsystem. The coupling terms directly participate in the state evolution iteration on the mechanical side.

[0118] In the specific code-level reproduction of the OpenModelica software, the discrete iteration and interaction of the entire strongly coupled network are implemented through specific algorithmic expressions. The wind turbine and drive chain subsystem at each time step... Internally, through expressions as well as The mechanical end of the system undergoes state evolution. The generator subsystem, however, uses an expression... Run the electrical terminal state evolution. These expressions contain... The algorithm extracts the state variables from the previous simulation time step. The algorithm corresponds to a custom sign power function transformation.

[0119] In practical applications, the aforementioned dual-closed-loop signal flow network, when running on the software platform, is accompanied by fluctuations in the shaft power output of the downstream generator subsystem. The mechanical subsystem transmits torque through the transmission chain. It continuously provides mechanical coupling input to the back-end electrical subsystems. Meanwhile, voltage control input is introduced across the network's front-end diffusion path. In the wind turbine speed fluctuation channel, and with the pitch angle control input The logarithmic transformation term undergoes a high-order nonlinear product operation, forming the control coupling term. Through the dynamic interaction between the voltage control input and the pitch angle control input in the mechanical-side diffusion term, and the intertwining of torque state and electrical noise on the electrical side, the entire system is ultimately packaged into a unified information and energy correlation topology network with bidirectional physical flow characteristics, providing network architecture support for the high-precision finite-time stabilization tuning of the entire system.

[0120] S4. The discretized Euler-Maruyama numerical solver is used to iteratively solve the interconnection model in the dual closed-loop signal flow network, and output the time-domain evolution data of the state variables and control inputs.

[0121] Furthermore, in S4, the discretized Euler-Maruyama numerical solver is used to iteratively solve the interconnection model in the dual closed-loop signal flow network, including:

[0122] Set a fixed simulation step size and total simulation duration, and initialize the state variables at the start time;

[0123] Two random number seeds are fixed to reproduce discrete Brownian motion increments;

[0124] Using the state variables from the previous moment, the simulation step size, and the control input, the mechanical and electrical state variables at the current moment are updated using discrete difference expressions, and the time-domain evolution data is output.

[0125] Specifically, in the numerical simulation parameter configuration and initialization phase, to accurately reproduce the closed-loop model of the wind power stochastic nonlinear twin-system with Brownian perturbations in digital space, the solver explicitly defines the underlying parameters through the software environment. In experimental and engineering applications, the numerical solver selected is the fixed-step Euler-Maruyama algorithm, i.e., a discretized Euler-Maruyama numerical solver, to adapt to the numerical integration requirements of stochastic differential equations. The system is set to a fixed simulation step size. Total simulation time Before the iterative calculation begins, the algorithm initializes the flow network of the state variables at the initial time, precisely assigning the initial state value at time zero to the corresponding memory-level register. Specifically, the initial value is set as: the initial value of the wind turbine speed deviation. Initial dynamic value of transmission chain torque Initial value of generator speed deviation Meanwhile, to ensure strict experimental consistency under strong turbulence and highly random conditions, the noise module in the network was forced to use two fixed random number seeds: the first random seed 12345 and the second random seed 54321. This design can effectively reproduce and simulate real discrete Brownian motion increments. and Furthermore, it ensures that the discrete numerical solution results are fully reproducible, providing a deterministic benchmark disturbance scenario for subsequent performance evaluation of the control law.

[0126] In the recursive iterative stage of the multidimensional state-time domain evolution, the discretized Euler-Maruyama numerical solver updates the mechanical and electrical state variables at the current moment using specific discrete difference expressions, based on the state variables from the previous moment, the simulation step size, and the real-time output of the pitch angle control and generator terminal voltage control quantities from the controller, in a parallel and interleaved manner. For the mechanical state evolution, the discrete update algorithm for the second-order wind turbine and transmission chain subsystem consists of two levels of state chains: the first level uses the transmission torque state from the previous moment to update the wind turbine speed deviation at the current moment through a custom sign power transformation, satisfying the expression... The second stage, acting as a first-order pure control drive state, directly integrates the value from the previous moment with the current pitch angle control quantity using linear steps, satisfying the expression... Regarding the electrical terminal state evolution, the discrete update algorithm for the first-order generator subsystem, under the coupling effect of front-end nonlinear mechanical torque fluctuations, uses the generator terminal voltage control quantity calculated at the current moment and the Brownian increment generated by the second noise module to jointly update the generator speed deviation state at the current moment, satisfying the expression... In all the discrete difference calculation expressions mentioned above, The operator refers to the historical value of the variable at the previous sampling time. All levels of nonlinear fractional power exponents, such as five-thirds, seven-fifths, and fifteen-eighths, strictly follow the mathematical boundary values ​​determined by homogeneity theory.

[0127] In practical industrial applications and system closed-loop verification, each discrete loop update of the solver generates a continuous data trajectory in the time domain, thus outputting complete time-domain evolution data of state variables and control inputs. These time-domain response time series data are captured in real time and transformed into system state response curves and control input response curves, where the pitch angle control input... The time-domain response curve is as follows Figure 5 As shown, the generator terminal voltage control input The time-domain response curve is as follows Figure 6 As shown, this allows for intuitive and comprehensive verification of the stable trajectory stabilized over a finite time at the time domain level. Through quantitative analysis of this evolution data, the system can accurately obtain the rotor speed deviation of approximately [value missing]. (It behaves approximately under certain specific sets of smoothing parameters) (Internal value basically returns to zero) and rapidly approaches zero (e.g.) Figure 2 As shown), the dynamic torque of the transmission chain is approximately... The internal rapid approach to zero (e.g.) Figure 3 As shown), the generator speed deviation is approximately... The precise time required for rapid convergence to near zero (e.g.) Figure 4(As shown). Furthermore, this data supports further calculation of the probability characteristics of the converged state variables maintaining small fluctuations under continuous standard Brownian perturbations, ensuring that its steady-state fluctuation amplitude is much less than one percent of the initial deviation. This indicates that the proposed method exhibits advanced dynamic control characteristics in discrete simulation networks, characterized by non-divergence, no high-frequency local oscillations, and smooth, unsaturated control output, providing complete data support for subsequent quantitative evaluation under multiple sets of control parameters. Figure 7 The wind turbine speed deviation is given. Deviation from generator speed phase plane trajectory, Figure 8 The wind turbine speed deviation is given. With pitch angle control input The phase plane trajectory.

[0128] S5. Set different control gain parameters and control scaling parameters and repeatedly perform iterative solutions to calculate the system state fluctuation characteristic data and control characteristic data derived from the time domain evolution data, and adjust the control scaling parameters accordingly.

[0129] Furthermore, in S5, the calculation of system state fluctuation characteristic data and control characteristic data derived from time-domain evolution data includes:

[0130] Obtain the random fluctuation values ​​of state variables near the equilibrium point and the convergence time at each time point in the time-domain evolution data;

[0131] Calculate system state fluctuation characteristic data to characterize operational stability and response speed;

[0132] Extract the control input values ​​from the time-domain evolution data and calculate the control characteristic data used to characterize the energy consumption of regulation.

[0133] Furthermore, in S5, the adjustment of the control scaling parameters includes:

[0134] Obtain the corresponding preset control priority tendency based on the wind farm operating conditions;

[0135] When the control priority is to prioritize high precision and steady state, reduce the value of the control scaling parameter;

[0136] When the control priority is to prioritize high response speed and real-time performance, increase the value of the control scaling parameter;

[0137] The adjusted control scaling parameters are fed back to the two controllers for categorized tuning.

[0138] Specifically, in the multi-parameter comparison simulation and feature data quantization calculation stage, in order to comprehensively evaluate and optimize the system's control performance under strong turbulence and multi-coupling conditions, the solver repeatedly executes discretized numerical solutions according to a preset parameter configuration scheme. The system obtains diverse time-domain evolution data by comparing three specific parameter sets laterally. These three parameter sets are specifically defined as: a baseline set, whose control parameter configuration is the control gain. Mechanical end first-stage feedback gain Mechanical end second-stage feedback gain Electrical feedback gain The high-gain group implements aggressive control by multiplying the feedback gain of each subsystem; specific parameter configurations are as follows: ; and for large manufacturers, the feedback gain remains unchanged, but the global control gain, which is used as a control scaling parameter, is increased. The specific parameter configuration is as follows: .

[0139] For each repeated solution output in the time domain, the system extracts discrete features from the multidimensional state time series at fixed time intervals throughout the entire simulation time. First, the algorithm dynamically acquires three state variables at each sampling time point. The random fluctuation values ​​near the zero equilibrium point and the convergence time to reach the steady-state boundary. Based on this, the algorithm eliminates the random influence of local state anomalies by calculating the average value of multiple state variables, and calculates the system state fluctuation characteristic data used to characterize the stability and response speed. This characteristic data specifically includes three quantitative indicators: the first is the average steady-state error, which reflects the steady-state accuracy. The benchmark group's calculation results are The second item is the mean square error, which reflects the waveform accuracy. Among them, the calculation results of the large factory group are the best. The third item is the average maximum transient deviation, which reflects the ability to suppress transient shocks. The benchmark group yielded the best results. Simultaneously, the algorithm extracts the pitch angle control quantity from the time-domain evolution data. With voltage control quantity The control input time series values ​​are used to calculate control characteristic data to characterize the energy consumption and mechanical wear economy of regulation. This characteristic data includes two quantitative energy consumption indicators: the average effective value of the control input, which reflects the overall control output of the system. and the effective value of the average rate of change reflecting the smoothing performance of the actuator. The data comparison shows that the benchmark group achieved a significant advantage in conservative configuration in both energy consumption indicators, demonstrating lower system energy consumption and economy.

[0140] During the control scaling parameter feedback tuning phase based on operating condition preferences, the system dynamically adjusts the control scaling parameters within the two controllers based on quantified state fluctuation characteristic data and energy consumption characteristic data. Its operation process first obtains the corresponding preset control priority preference based on the real-time operating conditions of the wind farm.

[0141] When wind farm conditions require high precision and prioritize steady-state operation, a conservative control strategy is chosen, using control parameters. The system accordingly reduces the global control scaling parameter. The value of is determined and fed back to the mechanical and electrical controllers. At this point, although the average convergence time is relatively moderate... However, the system achieves extremely high steady-state accuracy and the strongest ability to suppress transient shocks, while significantly reducing control output and system energy consumption.

[0142] When the wind farm operates under highly variable and turbulent conditions, with extremely high requirements for response speed, low latency, and real-time performance, the system implements aggressive regulation by increasing the amplification factor of aerodynamic and excitation control commands. The system then adjusts the global control scaling parameters accordingly. The value of the control input (e.g., increasing it to the value of 0.1 corresponding to the large factory group). At this time, although the effective value of the control input is... It will increase accordingly. However, this scaling parameter is in fractional powers. and The adjustment, applied to the controllers at both ends, significantly accelerated the physical convergence process of the entire system's state evolution. The adjusted average convergence time was substantially reduced to the optimal value. This not only eliminates speed fluctuations and torque shocks that are prone to occur under turbulent conditions, but also improves and enhances the controller's smooth adjustment characteristics under high-frequency white noise due to the increased scaling gain. Through this classification tuning method, the system can flexibly select the optimal control parameter configuration scheme, achieving an optimal balance of control accuracy, stability, and energy economy under different wind field scenarios while ensuring the finite-time stabilization effectiveness of the stochastic system.

[0143] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator set, characterized in that, include: S1. Divide the wind turbine generator set into a second-order wind turbine transmission chain mechanical subsystem and a first-order generator electrical subsystem. Establish the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems respectively. Introduce state nonlinear power terms and control coupling terms into the two subsystems to construct a bidirectional strongly coupled correlation topology. S2. Based on homogeneity theory, construct a mechanical end controller and an electrical end controller that receive speed and torque state variables. Determine the homogeneity parameters of the controller through recursive calculation using the general formula for homogeneity. Calculate and output the control input by combining the control gain parameters and control scaling parameters. S3. Use software to establish a dual closed-loop signal flow network consisting of the two subsystems, two controllers, and two random noise modules that generate random disturbance terms; S4. The discretized Euler-Maruyama numerical solver is used to iteratively solve the interconnection model in the dual closed-loop signal flow network, and the time-domain evolution data of the state variables and control inputs are output. S5. Set different control gain parameters and control scaling parameters and repeatedly perform iterative solutions to calculate the system state fluctuation characteristic data and control characteristic data derived from the time domain evolution data, and adjust the control scaling parameters accordingly.

2. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S1, establishing the speed and torque state variables, pitch angle and voltage control inputs, and random disturbance terms corresponding to the two subsystems respectively includes: The speed state variable and the torque state variable are established as the state variables of the second-order wind turbine transmission chain mechanical subsystem, and the torque state variable outputs mechanical coupling to the first-order generator electrical subsystem; The rotational speed state variable is established as the state variable of the first-order generator electrical subsystem and transmitted to the two controllers as a feedback signal; Establish the pitch angle control input as the control input of the second-order wind turbine drive chain mechanical subsystem, and establish the voltage control input as the control input of the first-order generator electrical subsystem. Independent first-standard Brownian motion and second-standard Brownian motion are introduced as random disturbance terms and input to the second-order wind turbine transmission chain mechanical subsystem and the first-order generator electrical subsystem, respectively.

3. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S1, the introduction of state nonlinear power terms and control coupling terms into the two subsystems to construct a bidirectional strongly coupled topology includes: In the mechanical subsystem of the second-order wind turbine drive chain, the torque state variable is introduced into the drift term of the speed state variable in a nonlinear power form; In the diffusion term of the second-order wind turbine drive chain mechanical subsystem, a voltage control input is introduced, and a control coupling term is formed by the nonlinear product of the voltage control input and the pitch angle control input. A nonlinear power term for the torque state variable is introduced into the random disturbance term of the first-order generator electrical subsystem, and multiplied by the second standard Brownian motion through a sine function; By controlling the dynamic interaction between coupling terms and random perturbation terms, a bidirectional strongly coupled correlation topology is constructed.

4. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S2, the determination of the homogeneity parameters of the controller through recursive calculation using the homogeneity formula includes: Based on the power exponent, preset initial homogeneity parameter and initial scaling parameter set by each of the two subsystems, the corresponding homogeneity parameter is selected within the negative value constraint range. The subsequent homogeneity parameters are calculated using the recursive formula for homogeneity parameters; The control scaling parameters for each of the two subsystems are calculated using the state scaling recursive formula.

5. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S2, calculating and outputting the control input by combining the control gain parameter and the control scaling parameter includes: The control scaling parameters are applied to the state variables of the second-order wind turbine drive chain mechanical subsystem by a preset fractional power to perform state transformation and power transformation. By combining the preset control gain parameters, the transformed variables are cascaded with the speed state variables of the first-order generator electrical subsystem introduced across the terminal to calculate the multi-state collaborative sensing feedback variables. The multi-state collaborative sensing feedback variables are subjected to nonlinear power transformation to output the pitch angle control input. The speed state variables of the first-order generator electrical subsystem are subjected to square nonlinear power transformation to solve for the electrical terminal state feedback variables. The electrical terminal state feedback variables are subjected to nonlinear power transformation, and the output voltage controls the input.

6. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S3, establishing a dual closed-loop signal flow network includes: The first noise module inputs a first random disturbance term consisting of a first standard Brownian motion into the second-order wind turbine transmission chain mechanical subsystem, and transmits its state variable to the mechanical end controller. The mechanical end controller is used to input the pitch angle control input to the mechanical subsystem of the second-order wind turbine drive chain; The second noise module inputs a second random disturbance term consisting of the second standard Brownian motion into the first-order generator electrical subsystem, and transmits its state variable back as a feedback signal. The voltage control input is calculated using the electrical terminal controller and then input into the closed electrical main control loop within the first-order generator electrical subsystem.

7. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 6, characterized in that, The establishment of the dual closed-loop signal flow network also includes: The state variables of the first-order generator electrical subsystem are used as shared feedback signals and simultaneously input to both the mechanical and electrical controllers. The calculated voltage control input is simultaneously introduced across the line into the mechanical subsystem of the second-order wind turbine drive chain. By utilizing the dynamic interaction between the voltage control input and the pitch angle control input, a cross-end collaborative control loop for the entire system's operating status is established between the two controllers.

8. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S4, the iterative solution of the interconnection model in the dual closed-loop signal flow network using the discretized Euler-Maruyama numerical solver includes: Set a fixed simulation step size and total simulation duration, and initialize the state variables at the start time; Two random number seeds are fixed to reproduce discrete Brownian motion increments; Using the state variables from the previous moment, the simulation step size, and the control input, the mechanical and electrical state variables at the current moment are updated using discrete difference expressions, and the time-domain evolution data is output.

9. The finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S5, the calculation of system state fluctuation characteristic data and control characteristic data derived from the time-domain evolution data includes: Obtain the random fluctuation values ​​of state variables near the equilibrium point and the convergence time at each time point in the time-domain evolution data; Calculate system state fluctuation characteristic data to characterize operational stability and response speed; Extract the control input values ​​from the time-domain evolution data and calculate the control characteristic data used to characterize the energy consumption of regulation.

10. A finite-time stabilization control method for a stochastic nonlinear system of a wind turbine generator according to claim 1, characterized in that, In S5, adjusting the control scaling parameters according to them includes: Obtain the corresponding preset control priority tendency based on the wind farm operating conditions; When the control priority is to prioritize high precision and steady state, reduce the value of the control scaling parameter; When the control priority is to prioritize high response speed and real-time performance, increase the value of the control scaling parameter; The adjusted control scaling parameters are fed back to the two controllers for categorized tuning.

Images