A subsynchronous damping control method based on model-free adaptive control

Abstract

This invention discloses a subsynchronous damping control method based on model-free adaptive control. The subsynchronous damping controller structure is first designed based on the model-free adaptive control method. A virtual equivalent time-varying linearized data model of the power system is dynamically established using system input and output data to adaptively control the controlled system, thereby suppressing subsynchronous oscillations. The method achieves consistent ultimate boundedness of the closed-loop system's tracking error and bounded-input-bounded-output stability. Relying on dynamic linearization technology to extract dynamic features from system input and output data, the invention eliminates the control method's reliance on modeling accuracy, enhances controller adaptability, requires minimal data, reduces control algorithm complexity, and achieves excellent real-time performance.

Description

Technical Field

[0001] The present invention relates to the technical field of power system control, and in particular to a subsynchronous damping control method based on model-free adaptive control. Background Art

[0002] Currently, in the research on suppressing subsynchronous oscillations in wind farms, the addition of subsynchronous damping controllers in doubly-fed wind turbine converters has attracted widespread attention due to their low cost and low control difficulty. However, during wind farm operation, there are many uncertainties and the operating conditions of wind turbines vary greatly. Existing subsynchronous damping controllers only consider a single or a few operating points when designing their control parameters, which may not be suitable for all operating conditions of the system. Furthermore, the parameter design of existing subsynchronous damping controllers relies on an accurate dynamic model of the wind power system, which is a form of model-based control (MBC). Due to the complex and highly nonlinear control process of wind power transmission systems, MBC-based control methods inevitably contain unmodeled dynamic links and other uncertainties, which affect the control effectiveness of the subsynchronous damping controllers. Summary of the Invention

[0003] The present invention aims to provide a subsynchronous damping control method based on model-free adaptive control. Based on this method, a subsynchronous damping controller structure is designed. A virtual equivalent time-varying linearized data model of the power system is dynamically established using system input and output data to adaptively control the controlled system, thereby suppressing subsynchronous oscillations. Subsequently, the method theoretically demonstrates the consistent ultimate boundedness of the tracking error and bounded input-bounded output stability of the closed-loop system. The present invention leverages dynamic linearization technology to extract dynamic features from system input and output data, thereby eliminating the control method's reliance on modeling accuracy and enhancing controller adaptability. This method requires minimal data, reduces control algorithm complexity, and provides excellent real-time performance.

[0004] The object of the present invention can be achieved by the following technical solution: A subsynchronous damping control method based on model-free adaptive control, comprising the following steps:

[0005] S1: Select the signal with significant subsynchronous component when subsynchronous oscillation occurs as the input signal of the subsynchronous damping controller;

[0006] S2: collect the selected input signal;

[0007] S3: Optimize the initial parameters of the subsynchronous damping controller using genetic algorithm;

[0008] S4: The subsynchronous damping controller adopts a tight linearization method and uses the input signal collected in S2 to establish a virtual time-varying equivalent linearized mathematical model of the power system;

[0009] S5: The subsynchronous damping controller uses a one-step-forward weighted prediction algorithm suitable for damping control to generate a control signal according to the linearized mathematical model constructed in S4;

[0010] S6: Add the control signal generated in S4 to the wind turbine converter.

[0011] As a further solution of the present invention, the input signal in S2 is a system state variable deviation signal that remains constant when the system is in steady state but changes significantly when subsynchronous oscillation occurs.

[0012] As a further solution of the present invention, in S4, the power system virtual equivalent time-varying linearized data model includes:

[0013] The system output estimation algorithm is:

[0014]

[0015] Where, is the estimated value of the system output signal y(k); c (k) is the estimated value of the pseudo partial derivative of the wind power transmission system; Δu(k) = u(k) - u(k-1); u(k) represents the input signal of the system at time k; γ is the error gain located at (0,1); is the estimated error of the output;

[0016] The adaptive iterative algorithm for the corresponding pseudo partial derivative estimate is:

[0017]

[0018] Where Γ(k)=η / (Δu(k) 2 +μ), which represents the weight coefficient of the change in the pseudo partial derivative step size; η is a positive constant representing the step size factor of the pseudo partial derivative estimation algorithm; μ is the penalty coefficient representing the change in the pseudo partial derivative; F = 1-γ.

[0019] As a further solution of the present invention, the one-step-forward predictive control algorithm suitable for damping control in S4 is:

[0020]

[0021] Where ρ0, ρ1 are step size factors; λ represents the penalty coefficient of the system input signal; y * Represents the expected output of the system.

[0022] As a further solution of the present invention, the initial parameters of the subsynchronous damping controller in S3 include a penalty coefficient, a step size factor and an error gain; the parameter optimization selection establishes an objective function based on the time and error square product integral index, and uses a genetic algorithm to optimize the initial parameters of the subsynchronous damping controller.

[0023] As a further solution of the present invention, the parameter optimization method comprises the following steps:

[0024] S31: Establishing an objective function based on the time and square error product integral index;

[0025] S32: Set the initial values of the parameters of the subsynchronous damping controller [λ, ρ0, ρ1, μ, η, γ]. The selection of these initial values has a certain impact on the optimization effect and the required time, and needs to be reasonably set according to the parameter design principles; where λ is the penalty coefficient of the control criterion function, ρ0, ρ1 are the step factors of the one-step-forward predictor, μ is the penalty coefficient of the pseudo partial derivative estimator, η is the step factor of the pseudo partial derivative estimator, and γ is the error gain;

[0026] S33: Optimize the parameters of the subsynchronous damping controller using a genetic algorithm, set the population size and the maximum number of iterations;

[0027] S34: running a simulation program related to the objective function to calculate the objective function value;

[0028] S35: Determine whether the optimized parameter value meets the constraint condition of the stability proof. If so, go to S36; if not, add a penalty term to the objective function and go to S36;

[0029] S36: Determine whether the number of iterations reaches the maximum number of iterations. If so, select the set of [λ, ρ0, ρ1, μ, η, γ] values with the smallest objective function value as the final optimal solution; if not, go to S33 for the next optimization iteration;

[0030] As a further solution of the present invention, parameter design:

[0031] Step 1: The smaller λ is, the faster the system responds, but overshoot may occur; the larger λ is, the slower the system responds;

[0032] Step 2: The step size inductors ρ0 and ρ1 affect the learning speed of the controller. The larger their values, the greater the amplitude of the controller output change per step.

[0033] Step 3: μ is the weight coefficient of the pseudo-partial derivative performance index function (a positive value), which is used to determine the constraint strength of the pseudo-partial derivative. If μ is too small, the possibility of system instability will increase; if μ is too large, the oscillation amplitude of the controlled system output will increase and the convergence speed will slow down.

[0034] Step 4: η is used to adjust the pseudo partial derivative estimation speed, η∈(0,2]. The larger η is, the smaller the oscillation amplitude of the controlled system output;

[0035] Step 5: γ is the gain of the output estimation error, γ∈(0,1). It affects the convergence speed of the system. The larger γ is, the faster the output of the controlled system will converge. However, if γ is too large, the system tracking error may not converge effectively.

[0036] Beneficial effects of the present invention:

[0037] 1. The subsynchronous damping controller of the present invention does not require modeling based on internal mechanisms and is applicable to nonlinear complex systems such as power systems. It has low algorithm complexity and is not affected by the accuracy of modeling.

[0038] 2. The subsynchronous damping controller of the present invention continuously updates the pseudo-partial derivative value by acquiring input and output (I / O) data online, has strong adaptability, and is suitable for wind power systems with large changes in operating conditions.

[0039] 3. Compared with control methods based on neural network control, the subsynchronous damping controller of the present invention does not require any external test signals or training processes, requires less real-time data and calculations, is low in cost, and has higher engineering practical value.

[0040] 4. The subsynchronous damping controller of the present invention has a relatively complete theoretical proof under some practical assumptions, which can ensure the monotonic convergence of the closed-loop system tracking error and the bounded input-bounded output stability;

[0041] 5. The subsynchronous damping control method based on model-free adaptive control proposed in the present invention can effectively suppress the subsynchronous oscillation phenomenon of the wind farm. The method can adapt to various operating points that may exist during the operation of the wind farm. It only needs to be installed in the wind turbine converter. There is no need to change the operating mode of the wind power transmission system, and the control cost is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] Figure 1 Schematic diagram of the subsynchronous damping controller structure of the present invention;

[0044] Figure 2 Flowchart of the parameter optimization method of the present invention;

[0045] Figure 3 This is a diagram of the installation position of the subsynchronous oscillation damping controller of the present invention in a wind power system;

[0046] Figure 4The suppression effect of the subsynchronous damping controller of the present invention when the system operating state changes at different times;

[0047] Figure 5 The wind power transmission system of the present invention includes wind turbine groups 1 and 2;

[0048] Figure 6 This is the time domain waveform of the active output of the wind turbine group 1 when a large disturbance occurs in the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] A subsynchronous damping control method based on model-free adaptive control includes the following steps:

[0051] S1: Select the signal with significant subsynchronous component when subsynchronous oscillation occurs as the input signal of the subsynchronous damping controller;

[0052] S2: collect the selected input signal;

[0053] S3: Optimize the initial parameters of the subsynchronous damping controller using genetic algorithm;

[0054] S4: The subsynchronous damping controller adopts a tight linearization method and uses the input signal collected in S2 to establish a virtual time-varying equivalent linearized mathematical model of the power system;

[0055] S5: The subsynchronous damping controller uses a one-step-forward weighted prediction algorithm suitable for damping control to generate a control signal according to the linearized mathematical model constructed in S4;

[0056] S6: Add the control signal generated in S4 to the wind turbine converter.

[0057] The wind power transmission system can be regarded as a single-input single-output discrete-time nonlinear system, which can be described by the following model:

[0058] y(k+1)=f(y(k),...,y(kN y ),u(k),...,u(kN u ))+d(k) (1)

[0059] Where u(k) and y(k) represent the input and output of the system at time k respectively; N u , Ny are the dimensions of input and output respectively; f(·) is an unknown nonlinear function; d(k) represents the finite unknown disturbance to the system at time k.

[0060] When the system satisfies the following two assumptions, the equation can be linearized equivalently:

[0061] Assumption 1: Except for a finite number of time points, the partial derivatives of the function f(·) with respect to the input u(k) are continuous.

[0062] Assumption 2: Except for a finite number of time points, the formula satisfies the generalized Lipschitz condition, that is:

[0063] |y(k+1)-y(k)|≤b|u(k)-u(k-1)|

[0064] Where b is a positive constant; |u(k)-u(k-1)|≠0.

[0065] Assumption 1 is a typical constraint for general nonlinear systems, while Assumption 2 states that bounded changes in input energy should produce bounded changes in system output energy. These assumptions are satisfied by most practical systems. Since the power system is a real physical system, energy is conserved, satisfying Assumptions 1 and 2. The finite time points represent the occurrence of certain unexpected events (e.g., disturbances such as power outages).

[0066] Using the compact linearization method, the formula can be written as:

[0067] y(k+1)=y(k)+φ c (k)Δu(k)+Δd(k) (2)

[0068] Where, φ c (k) represents the pseudo partial derivative of equation (1); |φ c (k)| is bounded; Δu(k) = u(k) - u(k - 1); d(k) = d(k) - d(k - 1), and |Δd(k)| < Ω, Ω is a positive constant representing the limit of Δd(k).

[0069] In order to obtain the iterative algorithm for estimating pseudo partial derivatives, the following system output estimation algorithm based on the tight format linearization method is established:

[0070]

[0071] Where, That is the estimated value of the system output signal y(k); is the estimated error of the output; γ is the error gain located in (0,1).

[0072] The adaptive iterative algorithm for the corresponding pseudo partial derivative estimate is:

[0073]

[0074] Where Γ(k)=η / (Δu(k) 2 +μ), which represents the weight coefficient of the pseudo partial derivative step change. The larger Γ(k), The faster the convergence speed, but if Γ(k) is too large, the system is prone to The convergence speed is too fast and the algorithm becomes unstable. η is a positive constant representing the step size factor of the pseudo partial derivative estimation algorithm. μ is a positive constant representing the penalty coefficient of the change of the pseudo partial derivative to determine the constraint strength of the pseudo partial derivative. F = 1-γ, which is in the interval (0,1).

[0075] In order to improve the ability to track time-varying parameters during the iteration process, it is necessary to build the following algorithm reset mechanism:

[0076]

[0077] Where ε is a sufficiently small positive number, and Proven to be bounded.

[0078] The above equations constitute the pseudo partial derivative estimator control algorithm.

[0079] Furthermore, the one-step-forward predictor obtains the control signal of the subsynchronous damping controller based on the established data model and according to a one-step-forward weighted prediction algorithm suitable for damping control, namely:

[0080] Considering that the subsynchronous damping controller is an additional control, after the system recovers stability, the control signal u(k) of the subsynchronous damping controller should be close to 0 to avoid affecting other normal control functions of the system. Therefore, the following new control criterion function is established:

[0081]

[0082] Among them, y * (k+1) represents the expected output of the system at time (k+1). For the problem of suppressing subsynchronous oscillations, the expected output is 0. λ is a positive number, which represents the penalty coefficient of the system input signal.

[0083] Substituting into the function, we get:

[0084]

[0085] Taking the partial derivative of the formula with respect to u(k) and setting the partial derivative to 0, we get:

[0086]

[0087] The formula is the control algorithm of the one-step-ahead predictor.

[0088] Furthermore, the stability proof mathematically proves that the tracking error of the closed-loop system after adopting the subsynchronous damping controller is uniformly bounded and bounded input-bounded output stability, that is:

[0089] When y * =const., there exists a positive real number λ min , when λ>λ min The following conclusions hold:

[0090] 1) Tracking error of wind power system output e(k) = y * -y(k) is uniformly eventually bounded (UUB), and

[0091] 2) The system input and output sequences {u(k)}, {y(k)} are bounded, that is, the system is BIBO stable.

[0092] In order to prove 1), the system estimation tracking error is introduced Then the tracking error e(k) can be expressed as:

[0093] e(k)=e′(k)-e0(k) (9)

[0094] in,

[0095] Therefore, if we can prove that e0(k) and e′(k) are uniformly eventually bounded, we can prove that e(k) is uniformly eventually bounded and the system as a whole is BIBO stable.

[0096] For the uniform eventual boundedness of e0(k):

[0097] According to y(k+1) and The iterative formula of , subtracting the two can get the iterative formula of e0(k+1):

[0098]

[0099] Construct the Lyapunov function V1(k)=|e0(k)| to prove the boundedness of e0(k), then:

[0100]

[0101] Where a1 = γ and a2 = |(φ c (k)-φ c (k))Δu(k)|+|Δd(k)|.

[0102] According to the above definition, a1=γ∈(0,1) and all terms in a2 are bounded, then it can be proved that at any time k, e0(k) is uniformly bounded, and

[0103] For the uniform eventual boundedness of e′(k):

[0104] Will Substitution The iterative formula of e′(k) can be obtained:

[0105]

[0106] Introducing vectors

[0107]

[0108] U(k)=[u(k),u(k-1)] T

[0109] Then e(k)=e′(k)-e0(k) can be expressed as:

[0110]

[0111] Similarly, the algorithm model of u(k) can be further expressed as:

[0112]

[0113] Substituting the above formula into vector U(k), we have:

[0114]

[0115] in,

[0116] Because |φ c (k)| is bounded, let its upper bound be Then there exists λ min >0 and positive constants M1, M2, M3, M4, so that when λ>λ min When , the following conclusions are satisfied:

[0117] (a)

[0118] (b)

[0119] (c)

[0120] (d)M2+M4<1

[0121] According to conclusion (b) and 0<ρ1≤1, we can see that the compatibility norm of A(k) satisfies ||A(k)||v ≤ρ1M3<1. Let d1=ρ1M3, and we know that ||U(0)|| v , taking the norm on both sides of the equation, we have:

[0122]

[0123] Substituting the formula into the formula, we can get:

[0124]

[0125] According to conclusion (b), select an appropriate ρ0 in (0,1] so that

[0126]

[0127] Let d2=1-ρ0M2, take the norm on both sides of the equation, and we can get

[0128]

[0129] Let d3 = ρ0M4. According to conclusion (c), the above formula can be changed to:

[0130]

[0131] Let the right side of the inequality be equal to g(k+1), that is

[0132]

[0133] then, The problem of proving the boundedness of e′(k) is transformed into proving the boundedness of the function g(k).

[0134] First, we obtain the iterative relationship of g(k):

[0135]

[0136] Among them,

[0137]

[0138] According to conclusion (d), since M2+M4<1, we can get:

[0139] d2=1-ρ0M2>ρ0(M2+M4)-ρ0M2=d3 (22)

[0140] Using this property, the following proof process can be established:

[0141]

[0142] Substituting the formula into the formula, we can get the iterative relationship of g(k):

[0143] g(k + 1) < (d1 + d2)g(k) + γd2|e0(k)| (24)

[0144] Secondly, construct the Lyapunov function V2(k) = g(k) to prove the boundedness of g(k):

[0145]

[0146] where a3 = 1 - d1 - d2, a4 = γd2|e0(k)|.

[0147] Select an appropriate ρ1 such that ρ1M3 ≤ ρ0M2. At this time, 0 ≤ a3 = 1 - d1 - d2 < 1. Therefore, it can be proved that g(k) is uniformly ultimately bounded, and

[0148] Since it can be proved that e′(k) has uniform ultimate boundedness at any time k, and

[0149] According to the uniform boundedness of e0(k) and e′(k) obtained by the derivation, it can be proved that the tracking error e(k) of the system is uniformly ultimately bounded, and

[0150]

[0151] For the BIBO stability of the system input-output sequences {u(k)}, {y(k)}:

[0152] Since e(k) is bounded and y * = 0, it can be proved that the output sequence {y(k)} has bounded stability; for the input sequence {u(k)}, using the formula, the Lyapunov function can also be constructed to prove its boundedness:

[0153]

[0154] where a5 = 1 - d1, a6 = ρ0M1(|e′(k)| + γ|e0(k)|).

[0155] Since 0 < a5 < 1 and a6 is bounded, it can be proved that the input sequence {u(k)} is also bounded and stable.

[0156] Furthermore, the parameter optimization method optimizes the initial parameters of subsynchronous damping, that is:

[0157] The control parameters to be optimized are [φ c (1), λ, ρ0, ρ1, μ, η, γ′], where since the pseudo partial derivative can be adaptively updated according to different operating conditions, the value of φ c (1) can be set relatively roughly.

[0158] The optimization method steps are as follows:

[0159] S1: Establish the objective function. The function established based on the time and square error product integral index is:

[0160]

[0161] Where e(t) represents the tracking error of the subsynchronous damping controller e(k) = y * -y(k).

[0162] Step 2: Set the initial values of [λ, ρ0, ρ1, μ, η, γ′]. The selection of the initial values of these parameters has a certain impact on the optimization effect and the required time, so they need to be set reasonably.

[0163] The design principles are as follows:

[0164] 1) The smaller λ is, the faster the system responds, but overshoot may occur; the larger λ is, the slower the system responds.

[0165] 2) The step size factor affects the learning speed of the controller. The larger its value, the greater the amplitude of each step change of the controller output.

[0166] 3) μ is the weight coefficient (positive value) of the pseudo-partial derivative performance index function, which is used to determine the constraint strength of the pseudo-partial derivative. If μ is too small, the possibility of system instability will increase; if μ is too large, the oscillation amplitude of the controlled system output will increase, and the convergence speed will slow down.

[0167] 4) η is used to adjust the pseudo partial derivative estimation speed, η∈(0,2]. The larger η is, the smaller the oscillation amplitude of the controlled system output is.

[0168] 5) γ is the gain of the output estimation error e0(k), γ∈(0,1). It affects the convergence speed of the system. The larger γ is, the faster the output of the controlled system will converge. However, if γ is too large, the system tracking error may not converge effectively.

[0169] S3: Genetic algorithm (GA) was used to optimize the control parameters. The GA toolbox in MATLAB was used, with the population size set to 50 and the maximum number of iterations set to 100.

[0170] S4: Calculate the objective function value.

[0171] S5: Determine whether the optimized values of [λ,ρ0,ρ1] satisfy the following constraints:

[0172] 1) Choose a suitable λ so that when λ>λ min hour,

[0173]

[0174] Where, Indicates |φ c (k)|. Since the update direction of the pseudo partial derivative is generally getting closer to 0, and according to the reset mechanism, the pseudo partial derivative has the sign preservation property, so It can generally be considered as |φ c (1)|.

[0175] 2) Choose a suitable ρ1 so that

[0176]

[0177] If [λ,ρ0,ρ1] does not satisfy the constraints, a penalty term is added to the objective function shown in the formula.

[0178] S6: Determine whether the number of iterations has reached the maximum number of iterations. If not, go to Step 3 for the next optimization iteration; if so, select the set of [λ, ρ0, ρ1, μ, η, γ] values with the smallest objective function value as the final optimal solution.

[0179] like Figure 1 The figure shows the structure diagram of the sub-synchronous damping controller of the present invention, which includes a system state quantity deviation signal acquisition device, a pseudo-partial derivative estimator that uses the system state quantity deviation signal at the current moment to establish a virtual equivalent time-varying linearization data model of the system, and a one-step-forward predictor that obtains the control signal of the sub-synchronous damping controller based on the virtual equivalent time-varying linearization model.

[0180] Figure 2 The parameter optimization flow chart of the subsynchronous damping controller of the present invention includes the following steps:

[0181] S1: Establish the objective function. The function established based on the time and square error product integral index is:

[0182]

[0183] Where e(t) represents the tracking error of the subsynchronous damping controller e(k) = y * -y(k).

[0184] S2: Set the initial values of [λ, ρ0, ρ1, μ, η, γ]. The selection of these initial values has a certain impact on the optimization effect and the required time, so they need to be set reasonably.

[0185] The design principles are as follows:

[0186] 1) The smaller λ is, the faster the system responds, but overshoot may occur; the larger λ is, the slower the system responds.

[0187] 2) The step size factor affects the learning speed of the controller. The larger its value is, the greater the amplitude of each step change of the controller output.

[0188] 3) μ is the weight coefficient (positive value) of the pseudo-partial derivative performance index function, which is used to determine the constraint strength of the pseudo-partial derivative. If μ is too small, the possibility of system instability will increase; if μ is too large, the oscillation amplitude of the controlled system output will increase, and the convergence speed will slow down.

[0189] 4) η is used to adjust the pseudo partial derivative estimation speed, η∈(0,2]. The larger η is, the smaller the oscillation amplitude of the controlled system output is.

[0190] 5) γ is the gain of the output estimation error e0(k), γ∈(0,1). It affects the convergence speed of the system. The larger γ is, the faster the output of the controlled system will converge. However, if γ is too large, the system tracking error may not converge effectively.

[0191] S3: Genetic algorithm (GA) was used to optimize the control parameters. The GA toolbox in MATLAB was used, with the population size set to 50 and the maximum number of iterations set to 100.

[0192] S4: Calculate the objective function value.

[0193] S5: Determine whether the optimized values of [λ,ρ0,ρ1] satisfy the following constraints:

[0194] 1) Choose a suitable λ so that when λ>λ min hour,

[0195]

[0196] Where, Indicates |φ c (k)|. Since the update direction of the pseudo partial derivative is generally getting closer to 0, and according to the reset mechanism, the pseudo partial derivative has the sign preservation property, so It can generally be considered as |φ c (1)|.

[0197] 2) Choose a suitable ρ1 so that

[0198]

[0199] If [λ,ρ0,ρ1] does not satisfy the constraints, a penalty term is added to the objective function shown in the formula.

[0200] S6: Determine whether the number of iterations has reached the maximum number of iterations. If not, go to Step 3 for the next optimization iteration; if so, select the set of [λ, ρ0, ρ1, μ, η, γ] values with the smallest objective function value as the final optimal solution.

[0201] A specific embodiment of the present invention is described below.

[0202] The verification example used in the present invention is a wind power transmission system including a doubly fed wind farm and series capacitor compensation. The system structure diagram after adding the subsynchronous damping controller of the present invention is shown in FIG. Figure 3 As shown, the system state quantity deviation signal acquisition device here acquires the deviation signal Δu of the d-axis component of the fan terminal voltage ds The doubly-fed wind farm is composed of several identical 1.5MW doubly-fed wind turbines. Each doubly-fed wind turbine is connected to the same busbar through a 0.69 / 35kV on-site step-up transformer T1 for grid-connected power generation. The entire doubly-fed wind farm is simulated using a single-machine equivalent model. The entire wind farm is then connected to the 220kV line through a 35 / 220kV transformer T2, and finally connected to the 500kV line through a 220 / 500kV step-up transformer T3 for long-distance power transmission. Series compensation capacitors are installed in the 500kV line for compensation. At wind speed v wind =8m / s, fan active output P output =0.2885pu, fan reactive output Q output = 0p.u., number of wind turbines in operation n = 1000, series compensation degree K c =12% working condition according to Figure 2 The flow chart of the present invention is used to solve the parameters of the subsynchronous damping controller, and the solution is: c (1) = -15, λ = 400.13, ρ0 = 0.98, ρ1 = 0.61, μ = 20.04, η = 1.95, γ = 0.92, and considering the subsynchronous oscillation frequency range, the sampling period is 0.001s.

[0203] Example 1:

[0204] Building in MATLAB / SIMULINK Figure 3 The transient simulation model shown in the figure sets the initial operating state of the system to be v wind =9m / s,P output =0.4108pu,Q output = 0p.u., n = 1200, the system starts fixed series compensation at t = 6s (K c = 20%), change Q at t = 9s output = -0.2pu, change Q at t = 11s output = -0.3pu. Then the wind farm active output P output and pseudo partial derivative φ c The time domain waveform that changes with time is as follows Figure 4 shown.

[0205] It can be found that at t = 6s, 9s, and 11s, due to sudden changes in the operating state, the pseudo-partial derivatives of the subsynchronous damping controller of the present invention begin to self-update in real time based on the fluctuations of the control input and output, effectively suppressing SSCI. This demonstrates that the subsynchronous damping control method based on model-free adaptive control of the present invention has good dynamic tracking control capabilities. In addition, the adaptive update capability of the proposed controller enables it to successfully suppress SSCI triggered by various disturbances even when the system operating point is constantly changing.

[0206] Example 2:

[0207] During the operation of the system, large disturbances such as machine trips and short circuits are inevitable. Therefore, it is necessary to verify the controller's ability to suppress subsynchronous oscillations when large disturbances occur. In Experimental Example 2, two aggregated wind turbines are used to represent wind turbine groups 1 and 2 under different operating conditions. Figure 5 As shown in the figure, the initial speed of wind turbine group 1 is 7m / s, and the speed of wind turbine group 2 is 9m / s. The reactive output of the two wind turbine groups is Q output = 0p.u., number of wind turbines in operation n = 1000. Series compensation capacitor (K c =15%) was put into operation at startup. When t = 6s, wind turbine group 2 tripped; and when t = 9s, a three-phase ground short circuit fault occurred on the high-voltage side of T2, and the fault was cleared after 50ms. Under the control of the subsynchronous damping control method based on model-free adaptive control for wind turbine converters of the present invention, the dynamic response waveform of wind turbine group 1 is as follows: Figure 6 shown.

[0208] It can be seen that without additional control, the system waveform begins to oscillate at t=6s and continues to diverge. However, when the subsynchronous damping control method based on model-free adaptive control of the present invention is added to the control, the system subsynchronous oscillations caused by the two large disturbances of machine tripping and short-circuit fault can still be successfully suppressed, indicating that the subsynchronous damping control method based on model-free adaptive control for wind turbine converters of the present invention is still effective in suppressing oscillations under large disturbances.

[0209] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0210] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention as claimed.

Claims

1. A subsynchronous damping control method based on model-free adaptive control, characterized in that: The steps include: S1: Select the signal with significant subsynchronous component when subsynchronous oscillation occurs as the input signal of the subsynchronous damping controller; S2: collect the selected input signal; S3: Optimize the initial parameters of the subsynchronous damping controller using genetic algorithm; S4: The subsynchronous damping controller adopts a tight linearization method and uses the input signal collected in S2 to establish a virtual equivalent time-varying linear mathematical model of the power system; S5: The subsynchronous damping controller uses a one-step-forward weighted prediction algorithm suitable for damping control to generate a control signal according to the linearized mathematical model constructed in S4; S6: Add the control signal generated in S5 to the wind turbine converter; In S4, the virtual equivalent time-varying linearized mathematical model of the power system includes: The system output estimation algorithm is: Where, is the estimated value of the system output signal y(k); c (k) represents the estimated value of the pseudo partial derivative of the wind power transmission system; Δu(k) = u(k) - u(k-1), where u(k) represents the input signal of the system at time k; γ is the error gain located in (0,1); is the estimated error of the output; The adaptive iterative algorithm for the corresponding pseudo partial derivative estimate is: Where Γ(k)=η / (Δu(k) 2 +μ), represents the weight coefficient of the pseudo partial derivative step size change; η represents the positive constant of the step size factor of the pseudo partial derivative estimation algorithm; μ represents the penalty coefficient of the pseudo partial derivative change; F = 1-γ; The one-step-forward weighted prediction algorithm suitable for damping control in S5 is: Where ρ0, ρ1 are step size factors; λ represents the penalty coefficient of the system input signal; y * Represents the expected output of the system.

2. The subsynchronous damping control method based on model-free adaptive control according to claim 1, characterized in that: The input signal in S2 is a system state variable deviation signal that remains constant when the system is in steady state but changes significantly when subsynchronous oscillation occurs.

3. The subsynchronous damping control method based on model-free adaptive control according to claim 1, characterized in that: The initial parameters of the subsynchronous damping controller in S3 include a penalty coefficient, a step size factor, and an error gain; the parameter optimization selection establishes an objective function based on the time and error square product integral index, and uses a genetic algorithm to optimize the initial parameters of the subsynchronous damping controller.

4. The subsynchronous damping control method based on model-free adaptive control according to claim 3, characterized in that: The parameter optimization method comprises the following steps: S31: Establishing an objective function based on the time and square error product integral index; S32: Set the initial parameter values [λ, ρ0, ρ1, μ, η, γ] of the subsynchronous damping controller described in claim 1. The selection of these initial parameter values has a certain impact on the optimization effect and the required time, and needs to be reasonably set according to the parameter design principle; wherein λ is the penalty coefficient of the control criterion function, ρ0, ρ1 are the step factors of the one-step-forward predictor, μ is the penalty coefficient of the pseudo partial derivative estimator, η is the step factor of the pseudo partial derivative estimator, and γ is the error gain; S33: Optimizing the parameters of the subsynchronous damping controller using a genetic algorithm, setting a population size and a maximum number of iterations; S34: running a simulation program related to the objective function to calculate the objective function value; S35: Determine whether the optimized parameter value meets the constraint conditions of the stability proof. If so, go to S36; if not, add a penalty term to the objective function and go to S36; S36: Determine whether the number of iterations has reached the maximum number of iterations. If so, select the set of [λ, ρ0, ρ1, μ, η, γ] values with the smallest objective function value as the final optimal solution; if not, go to S33 for the next optimization iteration.

5. The subsynchronous damping control method based on model-free adaptive control according to claim 4, characterized in that: The initial parameter design steps of the subsynchronous damping controller in S3 are as follows: Step 1: The smaller λ is, the faster the system responds, but overshoot may occur; the larger λ is, the slower the system responds; Step 2: The step size inductors ρ0 and ρ1 affect the learning speed of the controller. The larger the value, the greater the amplitude of the controller output change per step. Step 3: μ is the weight coefficient of the pseudo partial derivative performance index function (the value is positive), which is used to determine the constraint strength of the pseudo partial derivative. If μ is too small, the possibility of system instability will increase; if μ is too large, the oscillation amplitude of the controlled system output will increase and the convergence speed will slow down. Step 4: η is used to adjust the pseudo partial derivative estimation speed, η∈(0,2], the larger η is, the smaller the oscillation amplitude of the controlled system output; Step 5: γ is the gain of the output estimation error, γ∈(0,1), which affects the convergence speed of the system. The larger γ is, the faster the convergence speed of the controlled system output will be; however, if γ is too large, the system tracking error may not converge effectively.

Images