A wind turbine operating condition screening and parameter identification method and device

Abstract

This invention relates to the field of wind power equipment monitoring and parameter identification technology, and particularly to a method and apparatus for screening and identifying operating conditions and parameters of wind turbine generators. During grid-connected operation, voltage and current at the grid connection point are collected, and a sliding data window is constructed according to the sampling period. A nonlinear state-space reference model containing the control parameters to be identified is established, and the current response within the window is calculated. Based on the model, the dynamic sensitivity of each parameter to be identified is calculated, and a Fisher information matrix is ​​constructed and accumulated. Completeness indicators such as the condition number are calculated on the Fisher information matrix and compared with a threshold, retaining valid data segments and discarding invalid ones. A nonlinear least-squares identification problem is constructed using only the valid data segments, and the proportional and integral parameters of the phase-locked loop and each control link of the converter are output using a damped Gauss-Newton iterative method. This method eliminates the need for shutdown frequency sweeping or manual fault injection, and utilizes daily random disturbances to achieve automatic screening of operating conditions, improving identification accuracy and reducing testing and data processing costs.

Description

Technical Field

[0001] This invention relates to the field of wind power equipment monitoring and parameter identification technology, and in particular to a method and apparatus for screening wind turbine operating conditions and identifying parameters. Background Technology

[0002] As the proportion of wind power connected to the grid continues to increase, the dynamic characteristics of the power system become more complex, and stability problems such as wideband oscillations are more likely to occur. To conduct stability analysis, fault diagnosis, and optimize operation control strategies, dispatching and maintenance departments often need to master the key control parameters of the wind turbine converter control system, such as the phase-locked loop, the ratio of the rotor-side / grid-side converter current loop to the power / voltage outer loop, and integral parameters. However, in actual engineering practice, these control parameters are usually kept in a "black box" state due to manufacturer confidentiality restrictions, making them difficult to obtain externally, resulting in insufficient model accuracy and limited reliability of analysis conclusions.

[0003] Existing parameter acquisition and identification technologies mainly include two categories: offline testing and online identification.

[0004] Offline testing methods (such as frequency domain sweeping and artificial short circuits) typically require the wind turbine to be shut down and disconnected from the grid, and rely on specialized signal injection devices or artificial faults to obtain excitation and response data that can be used for identification. These methods are costly, result in significant downtime losses, are complex to organize and implement, and depend on specialized equipment and site conditions, making them difficult to conduct routinely during operation and maintenance.

[0005] Online identification methods typically utilize daily operational data, combined with extended Kalman filtering or intelligent optimization algorithms (such as particle swarm optimization and whale algorithms) to achieve parameter estimation. However, wind turbines typically experience relatively stable input and output during most normal operating periods, resulting in insufficient data excitation and a tendency for problems such as unidentifiable parameters, multiple solutions, or estimation drift. Some studies empirically select periods with larger disturbances for identification, but the lack of a unified and repeatable theoretical basis for operating condition selection may lead to insufficient excitation and identification failure, or redundant excitation increasing data processing and computation costs. Other methods attempt to judge the sufficiency of data information using indicators such as positive definiteness of signal spectral density, but it is still difficult to reliably guarantee that the selected data provides sufficient and decoupled identification information for all parameters to be identified, and may miss high-frequency, short-term but information-rich disturbances and response segments.

[0006] Therefore, there is an urgent need for a method that can automatically filter high-value data segments with sufficient information and complete and identifiable parameters from massive historical or real-time operating data without affecting grid-connected operation, and on this basis, achieve reliable identification of control parameters, so as to balance engineering feasibility, identification accuracy and calculation efficiency. Summary of the Invention

[0007] This invention provides a method and apparatus for screening operating conditions and identifying parameters of wind turbine generators, which can effectively solve the problems in the background art.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] A method for screening operating conditions and identifying parameters of wind turbine generators includes the following steps:

[0010] During the grid-connected operation of wind turbine units, voltage and current data at the grid connection point are collected in real time, and a sliding data window containing continuous sampling points is constructed according to the sampling period.

[0011] A nonlinear state-space reference model containing the control parameters to be identified is established based on the sliding data window, and the calculated value of the current response within the sliding data window is obtained from the reference model.

[0012] The dynamic sensitivity of the control parameter to be identified to the current response is calculated based on the reference model, and the Fisher information matrix corresponding to the sliding data window is constructed.

[0013] The completeness index representing the completeness and identifiability of parameters is calculated based on the Fisher information matrix, and the completeness index is compared with a preset threshold to filter the operating conditions corresponding to the sliding data window, retaining the valid operating condition data segments that meet the parameter completeness and identifiability and discarding the invalid operating condition data segments.

[0014] The nonlinear least squares identification problem is constructed using the effective data segment, and the identification result of the control parameter to be identified is output through iterative solution.

[0015] Furthermore, before constructing the sliding data window, the inherent parameters of the wind turbine are obtained, the control parameter vector to be identified is determined, and its initial value is set.

[0016] The inherent parameters of the wind turbine generator include: generator stator resistance R. s Rotor resistance R r Stator inductance L s Rotor inductance L r Magnetizing inductance L m ; Grid-side converter filter resistor R f Filter inductor L f DC bus capacitor C; synchronous speed ω s Rotor speed ω r Stator active power reference value Stator reactive power reference value DC bus voltage reference value Speed ​​reference value AC voltage reference value Sampling period T s Sliding data window length N, T s The time interval between two consecutive samples is N, and the number of sampling points within the sliding window is N.

[0017] The control parameter vector to be identified is defined as follows: ,in , These are the proportional and integral coefficients of the phase-locked loop controller to be identified. , These are the proportional and integral coefficients of the inner loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the inner loop controller of the grid-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the grid-side converter to be identified, respectively. Before the identification process begins... initial value Set as the empirical nominal value.

[0018] Furthermore, the collected three-phase instantaneous voltages and three-phase instantaneous current The Park transformation is performed to obtain the voltage input vector and the measured current output vector in the synchronous rotating coordinate system.

[0019] ;

[0020] ;

[0021] in, , These represent the d-axis and q-axis components of the grid connection point voltage in a synchronous rotating coordinate system, respectively. , These are the d-axis and q-axis components of the grid-connected point current in the synchronous rotating coordinate system; the data passes through the sliding data window for each sampling period T. s The window is updated by adding the latest sampling point to the end of the window and discarding the oldest sampling point, so that the window always contains the latest N samples.

[0022] Furthermore, the nonlinear state-space reference model is represented in a synchronously rotating coordinate system as follows:

[0023] ;

[0024] in, Let x(t) be the first derivative of the system state vector with respect to time t, where x(t) is the system state vector, u(t) is the system input vector, u(t) includes at least the d-axis and q-axis components of the grid-connected point voltage in the synchronous rotating coordinate system, y(t) is the output vector of the d-axis and q-axis components of the grid-connected current, f() is the nonlinear vector function of the state equation, g() is the nonlinear vector function of the output equation, and θ is the vector of control parameters to be identified.

[0025] Furthermore, the reference model is modeled using the synchronous rotating coordinate system of the doubly-fed wind turbine grid-connected system, and its state variable x is defined as:

[0026] ;

[0027] in, , These are the d and q components of the stator flux linkage in the global dq coordinate system, respectively. , These represent the dq components of the rotor current in the global dq coordinate system. , These represent the d and q components of the grid-side converter output current in the global dq coordinate system. This is the DC bus voltage. The phase-locked loop output angle. , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the rotor-side converter, respectively. , These are the state variables of the d-axis and q-axis integral elements of the outer loop controller of the rotor-side converter, respectively. For the state variables of the integral element of the phase-locked loop controller, , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the grid-side converter, respectively. The state variables are the integral elements of the d-axis outer loop controller of the grid-side converter.

[0028] Furthermore, the state equations of the reference model are composed of the following differential equations:

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] Where σ is the leakage magnetic coefficient and S slip slip ratio and ; , These are the d and q components of the rotor voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the inner current loop on the rotor side converter, respectively. , , These are stator active power, stator reactive power, and grid-side active power, respectively. , These are the d-axis and q-axis components of the grid-side converter output voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the grid-side converter's inner current loop, respectively.

[0046] Furthermore, in the reference model, the stator current component is represented by state variables as follows:

[0047] ;

[0048] ;

[0049] in, , These are the d-axis and q-axis components of the stator current in the global dq coordinate system, respectively.

[0050] Stator active power and reactive power are expressed as follows:

[0051] ;

[0052] ;

[0053] ;

[0054] ;

[0055] ;

[0056] ;

[0057] ;

[0058] Output equation It consists of the following equations:

[0059] ;

[0060] .

[0061] Furthermore, the construction of the dynamic sensitivity and Fisher information matrix includes: parameters to be identified The parameters θ in i At each sampling time t k Compute the Jacobian matrix and parametric partial derivative matrix of the system:

[0062] ,

[0063] ;

[0064] State sensitivity is calculated using discrete recursion:

[0065] ;

[0066] Calculate t k Output sensitivity at any time:

[0067] ,

[0068] The output sensitivities of each parameter are concatenated to form a sensitivity matrix:

[0069] ,

[0070] By summing the sensitivity matrices at N sampling times within the sliding window, we obtain the Fisher information matrix:

[0071] ;

[0072] in, Let be the partial derivative matrix of the state equation with respect to the state. The state equation with respect to parameter θ i The partial derivative or partial derivative matrix, For the state with respect to parameter θ i The sensitivity vector, To output the parameter θ i The sensitivity vector is R, where R is the measurement noise covariance matrix.

[0073] Furthermore, the completeness index is expressed as the condition number of the Fisher information matrix M, specifically by performing eigenvalue decomposition on M to obtain the largest eigenvalue λ. max With the smallest eigenvalue λ min Calculate the condition number:

[0074] ;

[0075] Compared with the completeness threshold γ, when The input and output data of the sliding window are saved as valid data segments when they are in use, and discarded otherwise; the least squares objective function is constructed using only the valid data segments.

[0076] ;

[0077] Calculate the residual vector in the j-th iteration:

[0078] ;

[0079] Calculate the gradient approximation of the objective function:

[0080] ;

[0081] Damped Gauss-Newton update:

[0082] ;

[0083] When satisfied or Terminate iteration and output ;in, Let γ be the condition number and γ be the completeness threshold. Let the least squares objective function be... Let μ be the residual vector, and μ be the damping factor. Let be the identity matrix, and ε be the iterative convergence threshold.

[0084] A wind turbine operating condition screening and parameter identification device includes:

[0085] The data acquisition and sliding window construction module is used to acquire voltage and current data at the grid connection point and construct a sliding data window;

[0086] The reference model module is used to establish a nonlinear state-space reference model containing the control parameters to be identified and output the calculated current response value.

[0087] The sensitivity and Fisher information matrix construction module is used to calculate dynamic sensitivity and construct the Fisher information matrix corresponding to the sliding window;

[0088] The completeness index judgment and data segment filtering module is used to calculate the completeness index and compare it with the completeness threshold to filter valid data segments.

[0089] The parameter identification module is used to construct a nonlinear least squares identification problem using only the effective data segment and iteratively solve and output the control parameters to be identified.

[0090] The technical solution of this invention can achieve the following technical effects:

[0091] This invention eliminates the need for offline testing operations such as frequency sweep signal injection and artificial short circuits. It directly utilizes natural excitations such as voltage fluctuations and random wind speed disturbances during grid-connected operation to complete the identification, avoiding economic losses caused by downtime and grid disconnection, and significantly reducing engineering implementation costs. By constructing a Fisher information matrix and calculating the condition number or equivalent completeness index and comparing it with a threshold, the sufficiency of data segment information is quantitatively evaluated, overcoming the problem of traditional online identification relying on experience to select disturbance conditions, making the screening process repeatable and interpretable.

[0092] Only data segments that meet the completeness requirements are retained for the solution, which mathematically enhances the ability to decouple and identify all parameters to be identified, reduces the risks of unidentifiable data, multiple solutions, and parameter drift, and improves the reliability and stability of the identification results. Under the sliding window framework, massive amounts of data are quickly identified and invalid segments are discarded, while only high-value segments are subjected to nonlinear least squares iteration, reducing the computational overhead caused by redundant data. Attached Figure Description

[0093] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0094] Figure 1 Flowchart of wind turbine operating condition screening and parameter identification method;

[0095] Figure 2 This is a topological structure diagram of an embodiment of the present invention. Detailed Implementation

[0096] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0097] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0098] This invention provides a method for screening operating conditions and identifying parameters of wind turbine generators, comprising the following steps:

[0099] like Figure 1 The diagram shows a flowchart of the invention's method for screening operating conditions and identifying parameters of wind turbine generators. This method comprises five stages: data acquisition and sliding window construction, reference model establishment and current response calculation, dynamic sensitivity calculation and Fisher information matrix construction, operating condition screening based on completeness indicators, and parameter identification based on valid operating condition data segments.

[0100] During the grid-connected operation of wind turbine units, voltage and current data at the grid connection point are collected in real time, and a sliding data window containing continuous sampling points is constructed according to the sampling period. Based on the sliding data window, a nonlinear state-space reference model containing the control parameters to be identified is established, and the current response within the sliding data window is calculated from the reference model. Based on the reference model, the dynamic sensitivity of the control parameters to be identified to the current response is calculated, and the Fisher information matrix corresponding to the sliding data window is constructed. Based on the Fisher information matrix, a completeness index characterizing the completeness and identifiability of the parameters is calculated, and the completeness index is compared with a preset threshold to filter the operating conditions corresponding to the sliding data window. Valid operating condition data segments that meet the parameter completeness and identifiability are retained, while invalid operating condition data segments are discarded. A nonlinear least squares identification problem is constructed using the valid data segments, and the identification result of the control parameters to be identified is output through iterative solution.

[0101] like Figure 2The diagram shows a topology of a wind turbine grid-connected system. This invention further provides a wind turbine operating condition screening and parameter identification device, comprising: a data acquisition and sliding window construction module for acquiring grid-connected point voltage and current data and constructing a sliding data window; a reference model module for establishing a nonlinear state-space reference model containing the control parameters to be identified and outputting calculated current response values; a sensitivity and Fisher information matrix construction module for calculating dynamic sensitivity and constructing the Fisher information matrix corresponding to the sliding window; a completeness index judgment and data segment screening module for calculating a completeness index and comparing it with a completeness threshold to screen valid data segments; and a parameter identification module for constructing a nonlinear least-squares identification problem using only valid data segments and iteratively solving to output the control parameters to be identified. Each module can be implemented by a processor-executed software program or by an FPGA / DSP in conjunction with a host computer.

[0102] The implementation platform can be a site-side industrial control computer / edge computing box (CPU / GPU optional), a high-precision sampling device (PMU or high sampling rate fault waveform recorder / protection measurement channel), and time synchronization (GPS / BeiDou or IEEE1588). The preferred data acquisition point is the grid connection point (PCC) or the low-voltage side grid connection point of the wind turbine transformer; if it is the site aggregation point, it is necessary to ensure that the voltage and current measurements can reflect the dynamic response of a single unit or equivalent unit.

[0103] The above methods and devices can be used to embed online screening and parameter identification into existing scheduling / maintenance platforms, forming a sustainable "black box parameter virtual measurement" capability, avoiding the downtime and dedicated equipment costs of traditional offline frequency scanning / short circuit tests.

[0104] In this embodiment, the inherent parameters of the wind turbine are obtained before constructing the sliding data window, the vector of control parameters to be identified is determined, and its initial value is set.

[0105] The inherent parameters of a wind turbine include: generator stator resistance R. s Rotor resistance R r Stator inductance L s Rotor inductance L r Magnetizing inductance L m ; Grid-side converter filter resistor R f Filter inductor L f DC bus capacitor C; synchronous speed ω s Rotor speed ω r Stator active power reference value Stator reactive power reference value DC bus voltage reference value Speed ​​reference value AC voltage reference value Sampling period T sSliding data window length N, T s The time interval between two consecutive samples is N, and the number of sampling points within the sliding window is N.

[0106] The vector of control parameters to be identified is defined as follows: ,in , These are the proportional and integral coefficients of the phase-locked loop controller to be identified. , These are the proportional and integral coefficients of the inner loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the inner loop controller of the grid-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the grid-side converter to be identified, respectively. Before the identification process begins... initial value Set as the empirical nominal value.

[0107] The inherent parameters are derived from manufacturer data, nameplates / commissioning records, site model libraries, or historical identification results. For some parameters that are difficult to obtain accurately, such as filter equivalent parameters that drift with temperature, a two-stage identification method can be used as an "optional" step in the instruction manual: first, fix the control parameters to estimate the electrical parameters, or first, fix the electrical parameters to estimate the control parameters. Initial values ​​can be manufacturer settings, empirical settings, or the results of the previous identification. To avoid iterative divergence, upper and lower bounds can be set for the parameters, such as making the proportional and integral coefficients non-negative and limiting their engineering range, and using projection or truncation during iteration.

[0108] In this embodiment, the collected three-phase instantaneous voltage and three-phase instantaneous current The Park transformation is performed to obtain the voltage input vector and the measured current output vector in the synchronous rotating coordinate system.

[0109] ;

[0110] ;

[0111] in, , These represent the d-axis and q-axis components of the grid connection point voltage in a synchronous rotating coordinate system, respectively. , These represent the d-axis and q-axis components of the grid-connected point current in the synchronous rotating coordinate system; the data is processed through a sliding data window over each sampling period T. sThe window is updated by adding the latest sampling point to the end of the window and discarding the oldest sampling point, so that the window always contains the latest N samples.

[0112] Specifically, real-time measurement of three-phase instantaneous voltage and current The voltage vector in the synchronous rotating coordinate system is obtained by performing the Park transformation on it. and measured current vector A sliding window is formed by collecting data at N time points. Each sampling period T... s New data points are collected and added to the end of the window data sequence, while the first data point is discarded, so that the window always stores the latest N data samples.

[0113] The sliding window mechanism allows the system to run online without storing massive amounts of historical data; data preprocessing and time alignment reduce the impact of noise and phase errors on sensitivity calculation and information matrix construction, thereby improving the reliability of the screening criteria.

[0114] Furthermore, the nonlinear state-space reference model is represented in a synchronously rotating coordinate system as follows:

[0115] ;

[0116] in, Let x(t) be the first derivative of the system state vector with respect to time t, where x(t) is the system state vector, u(t) is the system input vector, u(t) includes at least the d-axis and q-axis components of the grid-connected point voltage in the synchronous rotating coordinate system, y(t) is the output vector of the d-axis and q-axis components of the grid-connected current, f() is the nonlinear vector function of the state equation, g() is the nonlinear vector function of the output equation, and θ is the vector of control parameters to be identified.

[0117] The reference model can be either a mechanistic model or a data-driven model, such as a pre-trained neural network / grey box model, which can provide gradient information of the output with respect to the parameters to support dynamic sensitivity calculations: mechanistic models can explicitly derive the Jacobian and parameter partial derivatives; data-driven models can use automatic differentiation to obtain the gradient.

[0118] The reference model is modeled using a synchronous rotating coordinate system of a doubly-fed wind turbine grid-connected system, and its state variable x is defined as:

[0119] ;

[0120] in, , These are the d and q components of the stator flux linkage in the global dq coordinate system, respectively. , These represent the dq components of the rotor current in the global dq coordinate system. , These represent the d and q components of the grid-side converter output current in the global dq coordinate system. This is the DC bus voltage. The phase-locked loop output angle. , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the rotor-side converter, respectively. , These are the state variables of the d-axis and q-axis integral elements of the outer loop controller of the rotor-side converter, respectively. For the state variables of the integral element of the phase-locked loop controller, , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the grid-side converter, respectively. The state variables are the integral elements of the d-axis outer loop controller of the grid-side converter.

[0121] The state equations of the reference model consist of the following differential equations:

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] ;

[0129] ;

[0130] ;

[0131] ;

[0132] ;

[0133] ;

[0134] ;

[0135] ;

[0136] ;

[0137] ;

[0138] Wherein, the left side of the above formula represents the first derivative of the parameter with respect to time, and σ is the leakage magnetic coefficient. S slip slip ratio and ; , These are the d and q components of the rotor voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the inner current loop on the rotor side converter, respectively. , , These are stator active power, stator reactive power, and grid-side active power, respectively. , These are the d-axis and q-axis components of the grid-side converter output voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the grid-side converter's inner current loop, respectively.

[0139] in:

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] In the reference model, the stator current component is represented by the state variables as follows:

[0145] ;

[0146] ;

[0147] in, , These are the d-axis and q-axis components of the stator current in the global dq coordinate system, respectively.

[0148] Stator active power and reactive power are expressed as follows:

[0149] ;

[0150] ;

[0151] ;

[0152] ;

[0153] ;

[0154] ;

[0155] ;

[0156] Output equation It consists of the following equations:

[0157] ;

[0158] .

[0159] In this embodiment, the construction of dynamic sensitivity and Fisher information matrix includes: parameters to be identified The parameters θ in i At each sampling time t k Using the current state vector and input vector Calculate the Jacobian matrix and parameter partial derivative matrices of the system:

[0160] ,

[0161] ;

[0162] State sensitivity is calculated using discrete recursion:

[0163] ;

[0164] Calculate t k Output sensitivity at any time:

[0165] ,

[0166] The output sensitivities of each parameter are concatenated to form a sensitivity matrix:

[0167] ,

[0168] By summing the sensitivity matrices at N sampling times within the sliding window, we obtain the Fisher information matrix:

[0169] ;

[0170] in, Let be the partial derivative matrix of the state equation with respect to the state. Let be the partial derivative or partial derivative matrix of the state equation with respect to the parameter θi. Let be an n-dimensional state sensitivity vector. For an n×n dimensional system, the Jacobian matrix is... For n-dimensional parametric biased inputs, The output is the sensitivity vector for parameter θi, and R is the measurement noise covariance matrix.

[0171] The completeness index is the condition number of the Fisher information matrix M, specifically obtained by performing eigenvalue decomposition on M to obtain the largest eigenvalue λ. max With the smallest eigenvalue λ min Calculate the condition number:

[0172] ;

[0173] Compared with the completeness threshold γ, when The input and output data of the sliding window are saved as valid data segments when they are in use, otherwise they are discarded; the least squares objective function is constructed using only the valid data segments.

[0174] ;

[0175] In the measured output vector at time k, To utilize the current parameters at time k Calculate the output vector obtained from the state-space model; For L2 norm operations;

[0176] Calculate the residual vector in the j-th iteration:

[0177] ;

[0178] Calculate the gradient approximation of the objective function:

[0179] ;

[0180] Damped Gauss-Newton update:

[0181] ;

[0182] When satisfied or Terminate iteration and output ;in, Let γ be the condition number and γ be the completeness threshold. Let the least squares objective function be... Let μ be the residual vector, and μ be the damping factor. Let be the identity matrix, and ε be the iterative convergence threshold.

[0183] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for screening operating conditions and identifying parameters of wind turbine generators, characterized in that, Includes the following steps: During the grid-connected operation of wind turbine units, voltage and current data at the grid connection point are collected in real time, and a sliding data window containing continuous sampling points is constructed according to the sampling period. A nonlinear state-space reference model containing the control parameters to be identified is established based on the sliding data window, and the calculated value of the current response within the sliding data window is obtained from the reference model. The dynamic sensitivity of the control parameter to be identified to the current response is calculated based on the reference model, and the Fisher information matrix corresponding to the sliding data window is constructed. The completeness index representing the completeness and identifiability of parameters is calculated based on the Fisher information matrix, and the completeness index is compared with a preset threshold to filter the operating conditions corresponding to the sliding data window, retaining the valid operating condition data segments that meet the parameter completeness and identifiability and discarding the invalid operating condition data segments. A nonlinear least squares identification problem is constructed using the effective data segment, and the identification result of the control parameter to be identified is output through iterative solution. Before constructing the sliding data window, the inherent parameters of the wind turbine are obtained, the vector of control parameters to be identified is determined, and its initial value is set. The inherent parameters of the wind turbine generator include: generator stator resistance R. s Rotor resistance R r Stator inductance L s Rotor inductance L r Magnetizing inductance L m ; Grid-side converter filter resistor R f Filter inductor L f DC bus capacitor C; synchronous speed ω s Rotor speed ω r Stator active power reference value Stator reactive power reference value DC bus voltage reference value Speed ​​reference value AC voltage reference value Sampling period T s Sliding data window length N, T s The time interval between two consecutive samples is N, and the number of sampling points within the sliding window is N. The control parameter vector to be identified is defined as follows: ,in , These are the proportional and integral coefficients of the phase-locked loop controller to be identified. , These are the proportional and integral coefficients of the inner loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the rotor-side converter to be identified. , These are the proportional and integral coefficients of the inner loop controller of the grid-side converter to be identified. , These are the proportional and integral coefficients of the outer loop controller of the grid-side converter to be identified, respectively. Before the identification process begins... initial value Set as an empirical nominal value; The collected three-phase instantaneous voltage and three-phase instantaneous current The Park transformation is performed to obtain the voltage input vector and the measured current output vector in the synchronous rotating coordinate system. ； ； in, , These represent the d-axis and q-axis components of the grid connection point voltage in a synchronous rotating coordinate system, respectively. , These are the d-axis and q-axis components of the grid-connected point current in the synchronous rotating coordinate system; the data passes through the sliding data window for each sampling period T. s The window is updated by adding the latest sampling point to the end of the window and discarding the oldest sampling point, so that the window always contains the latest N samples. The construction of the dynamic sensitivity and Fisher information matrix includes: parameters to be identified The parameters θ in i At each sampling time t k Compute the Jacobian matrix and parametric partial derivative matrix of the system: ， ； State sensitivity is calculated using discrete recursion: ； Computing t k Output sensitivity at time: ， The output sensitivities of each parameter are concatenated to form a sensitivity matrix: ， By summing the sensitivity matrices at N sampling times within the sliding window, we obtain the Fisher information matrix: ； in, Let be the partial derivative matrix of the state equation with respect to the state. The state equation with respect to parameter θ i The partial derivative or partial derivative matrix, For the state with respect to parameter θ i The sensitivity vector, To output the parameter θ i The sensitivity vector, where R is the measurement noise covariance matrix; The completeness index is the condition number of the Fisher information matrix M, specifically obtained by performing eigenvalue decomposition on M to obtain the largest eigenvalue λ. max With the smallest eigenvalue λ min Calculate the condition number: ； Compared with the completeness threshold γ, when The input and output data of the sliding window are saved as valid data segments when they are in use, and discarded otherwise; the least squares objective function is constructed using only the valid data segments. ； Calculate the residual vector in the j-th iteration: ； Calculate the gradient approximation of the objective function: ； Damped Gauss-Newton update: ； When satisfied or Terminate iteration and output ;in, Let γ be the condition number and γ be the completeness threshold. Let the least squares objective function be... Let μ be the residual vector, and μ be the damping factor. Let be the identity matrix, and ε be the iterative convergence threshold.

2. The method for screening operating conditions and identifying parameters of wind turbine generators according to claim 1, characterized in that, The nonlinear state-space reference model is represented in a synchronously rotating coordinate system as follows: ； in, Let x(t) be the first derivative of the system state vector with respect to time t, where x(t) is the system state vector, u(t) is the system input vector, u(t) includes at least the d-axis and q-axis components of the grid-connected point voltage in the synchronous rotating coordinate system, y(t) is the output vector of the d-axis and q-axis components of the grid-connected current, f() is the nonlinear vector function of the state equation, g() is the nonlinear vector function of the output equation, and θ is the vector of control parameters to be identified.

3. The method for screening operating conditions and identifying parameters of wind turbine generators according to claim 2, characterized in that, The reference model is modeled using a synchronous rotating coordinate system of a doubly-fed wind turbine grid-connected system, and its state variable x is defined as: ； in, , These are the d and q components of the stator flux linkage in the global dq coordinate system, respectively. , These represent the dq components of the rotor current in the global dq coordinate system. , These represent the d and q components of the grid-side converter output current in the global dq coordinate system. This is the DC bus voltage. The phase-locked loop output angle. , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the rotor-side converter, respectively. , These are the state variables of the d-axis and q-axis integral elements of the outer loop controller of the rotor-side converter, respectively. For the state variables of the integral element of the phase-locked loop controller, , These are the state variables of the d-axis and q-axis integral elements of the inner loop controller of the grid-side converter, respectively. The state variables are the integral elements of the d-axis outer loop controller of the grid-side converter.

4. The method for screening operating conditions and identifying parameters of wind turbine generators according to claim 3, characterized in that, The state equations of the reference model consist of the following differential equations: ； ； ； ； ； ； ； ； ； ； ； ； ； ； ； ； Where σ is the leakage magnetic coefficient and S slip slip ratio and ; , These are the d and q components of the rotor voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the inner current loop on the rotor side converter, respectively. , , These are stator active power, stator reactive power, and grid-side active power, respectively. , These are the d-axis and q-axis components of the grid-side converter output voltage in the global dq coordinate system, respectively. , These are the reference values ​​for the d-axis and q-axis currents of the grid-side converter's inner current loop, respectively.

5. The method for screening operating conditions and identifying parameters of wind turbine generators according to claim 4, characterized in that, In the reference model, the stator current component is represented by state variables as follows: ； ； in, , These are the d-axis and q-axis components of the stator current in the global dq coordinate system, respectively. Stator active power and reactive power are expressed as follows: ； ； ； ； ； ； ； Output equation It consists of the following equations: ； 。 6. A device for screening operating conditions and identifying parameters of a wind turbine generator set, characterized in that, The wind turbine operating condition screening and parameter identification method applicable to any one of claims 1 to 5 includes: The data acquisition and sliding window construction module is used to acquire voltage and current data at the grid connection point and construct a sliding data window; The reference model module is used to establish a nonlinear state-space reference model containing the control parameters to be identified and output the calculated current response value. The sensitivity and Fisher information matrix construction module is used to calculate dynamic sensitivity and construct the Fisher information matrix corresponding to the sliding window; The completeness index judgment and data segment filtering module is used to calculate the completeness index and compare it with the completeness threshold to filter valid data segments. The parameter identification module is used to construct a nonlinear least squares identification problem using only the effective data segment and iteratively solve and output the control parameters to be identified.

Images