Wind turbine generator impedance parameter identification method based on machine learning

Abstract

The embodiment of the invention provides a wind turbine generator impedance parameter identification method based on machine learning, and belongs to the field of doubly-fed wind turbine generators. The method comprises the following steps: setting electrical parameters, partial control parameters and working point parameters of the doubly-fed wind turbine generator set as known conditions, taking PI control parameters in a rotor side inverter controller as unknown quantities, and constructing a grey box impedance model of the doubly-fed wind turbine generator set; identifying PI control parameters in the model by using a machine learning algorithm; and substituting the identified PI control parameters into a grey box impedance model of the doubly-fed wind turbine generator, outputting a return ratio matrix, and judging the stability of the doubly-fed wind turbine generator system by adopting a generalized Nyquist stability criterion based on the return ratio matrix. According to the method, the impedance characteristic of the system can be accurately obtained without depending on the structure and the internal parameters of the wind power plant, so that the grid-connected characteristic and the dynamic response of the wind power plant are effectively evaluated, and the defect of impedance obtaining under the black / grey box characteristic in a traditional method is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of doubly-fed wind turbine generator sets, and in particular to a method for identifying impedance parameters of a wind turbine generator set based on machine learning. Background Art

[0002] As a clean and renewable energy source, wind energy is of great significance in replacing traditional fossil fuel power generation. As one of the most mature renewable energy technologies, wind power has been widely commercialized around the world. The wind power industry has developed rapidly in recent years, and the installed capacity has increased year by year, which not only helps to increase the diversity of energy supply, but also is an important measure to achieve low-carbon transformation of energy structure. Large-scale wind farms are composed of hundreds or even thousands of wind turbines, which are of various types, with different control parameters and different operating modes. Relevant studies have shown that when wind farms are connected to weak AC systems and series compensation systems, subsynchronous oscillations may occur, which brings severe challenges to the safe and stable operation of large-scale wind farm grid-connected systems. The impedance method is an important means of analyzing wind farms at this stage. The impedance modeling of wind farms is carried out, and after the impedance model is obtained, the stability and other aspects of the wind farm can be analyzed.

[0003] At present, impedance acquisition can be divided into two categories: one is the sweep frequency method based on harmonic injection, which can be further extended to an impedance identification method, and the other is an analytical method based on wind turbine model derivation. The sweep frequency method can obtain the accurate impedance value of the wind farm at a specific frequency point, but it requires harmonic injection equipment and the impedance acquisition speed is slow. The analytical method based on formula derivation can obtain a detailed impedance model.

[0004] However, in actual engineering applications, wind farm models usually present "black / gray box" characteristics due to the strict confidentiality of data by wind turbine manufacturers and operators. This confidentiality makes it difficult to obtain an accurate impedance model through theoretical derivation, which makes the theoretical derivation of wind turbine impedance face many challenges and limitations in practical applications. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a method for identifying impedance parameters of a wind turbine generator system based on machine learning, so as to fully or at least partially solve the technical problems existing in the above-mentioned prior art.

[0006] In order to achieve the above object, an embodiment of the present invention provides a method for identifying impedance parameters of a wind turbine generator set based on machine learning, comprising: The electrical parameters, some control parameters and working point parameters of the doubly-fed wind turbine are set as known conditions, and the PI control parameters in the rotor-side inverter controller are taken as unknown quantities to construct a gray box impedance model of the doubly-fed wind turbine. Using a machine learning algorithm to identify PI control parameters in the gray box impedance model of the doubly-fed wind turbine generator set; wherein the PI control parameters include proportional and integral parameters of the inner loop of the current control and proportional and integral parameters of the outer loop of the power control; The identified PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator system, and a back-ratio matrix is ​​output. Based on the back-ratio matrix, the generalized Nyquist stability criterion is used to judge the stability of the doubly-fed wind turbine generator system.

[0007] Optionally, using a machine learning algorithm to identify PI control parameters in a grey box impedance model of the doubly-fed wind turbine generator system includes: Construct a sample data set and generate a regression prediction model through machine learning training; Obtaining an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substituting the impedance scanning data set into the regression prediction model, and outputting predicted PI control parameters; The predicted PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator set, and the transient response of the active power under the conditions of actual PI control parameters and predicted PI control parameters is compared to verify the accuracy and reliability of the predicted PI control parameters.

[0008] Optionally, the sample data set includes a random matrix of control parameters and a frequency scanning sampling point data matrix thereof, wherein the frequency scanning sampling point data matrix is ​​characterized by sampling the voltage and current of the sampling points using a sweeping frequency method with harmonic injection, and obtaining frequency scanning data through Fourier transform.

[0009] Optionally, the back-ratio matrix is ​​as follows:

[0010] In the formula, represents the grid impedance matrix, represents the impedance characteristics of the doubly-fed wind turbine generator set, , , , They respectively represent the amplitude relationships from d-axis to d-axis, d-axis to q-axis, q-axis to d-axis, and q-axis to q-axis in the rotating coordinate system.

[0011] Optionally, a generalized Nyquist stability criterion is used to substitute the identified PI control parameters into the gray box impedance model of the doubly-fed wind turbine generator system, output a back-ratio matrix, and based on the back-ratio matrix, a generalized Nyquist stability criterion is used to judge the stability of the doubly-fed wind turbine generator system, including: It is determined whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane. If not, it is determined that the doubly-fed wind turbine system is stable. Otherwise, it is determined that the doubly-fed wind turbine system is unstable.

[0012] On the other hand, the present invention also provides a wind turbine impedance parameter identification system based on machine learning, comprising: A construction unit is used to set the electrical parameters, some control parameters and working point parameters of the doubly-fed wind turbine generator set as known conditions, take the PI control parameters in the rotor-side inverter controller as unknown quantities, and construct a gray box impedance model of the doubly-fed wind turbine generator set; An identification unit, used to identify PI control parameters in the gray box impedance model of the doubly-fed wind turbine generator set using a machine learning algorithm; wherein the PI control parameters include proportional and integral parameters of the inner loop of the current control and proportional and integral parameters of the outer loop of the power control; The judging unit is used to substitute the identified PI control parameters into the grey box impedance model of the doubly-fed wind turbine generator system, output a back-ratio matrix, and judge the stability of the doubly-fed wind turbine generator system based on the back-ratio matrix and the generalized Nyquist stability criterion.

[0013] Optionally, the identification unit is specifically used for: Construct a sample data set and generate a regression prediction model through machine learning training; Obtaining an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substituting the impedance scanning data set into the regression prediction model, and outputting predicted PI control parameters; The predicted PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator set, and the transient response of the active power under the conditions of actual PI control parameters and predicted PI control parameters is compared to verify the accuracy and reliability of the predicted PI control parameters.

[0014] Optionally, the judging unit is specifically configured to: It is determined whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane. If not, it is determined that the doubly-fed wind turbine system is stable. Otherwise, it is determined that the doubly-fed wind turbine system is unstable.

[0015] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for identifying wind turbine impedance parameters based on machine learning when executing the program.

[0016] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for identifying wind turbine impedance parameters based on machine learning are implemented.

[0017] This application proposes an impedance parameter identification method based on a machine learning algorithm, including the establishment of a parameter identification model and the basic process of machine learning parameter identification. Through impedance modeling driven by measured data, the impedance characteristics of the system can be accurately obtained without relying on the structure and internal parameters of the wind farm, so as to effectively evaluate the grid-connected characteristics and dynamic response of the wind farm. It can not only improve the accuracy of the simulation model, but also ensure the accurate evaluation of the stability and reliability of the system, solve the shortcomings of the traditional method in impedance acquisition under the "black / gray box" characteristics, and provide new ideas and methods for the modeling and research of the grid-connected impedance characteristics of large wind farms under the "black / gray box" background, and lay a model foundation for the stability analysis of large-scale renewable energy grid connection.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings: Figure 1 It is an implementation flow chart of a method for identifying wind turbine impedance parameters based on machine learning provided by an embodiment of the present invention; Figure 2 is a structural diagram of a doubly-fed wind turbine system provided by an embodiment of the present invention; Figure 3 This is a frequency scanning result diagram of a doubly-fed wind turbine generator system provided by an embodiment of the present invention; Figure 4 This is a machine learning regression fitting sample comparison diagram provided by an embodiment of the present invention; Figure 5 It is a BP neural network training sample error percentage diagram provided by an embodiment of the present invention; Figure 6 An active step transient response diagram of actual parameters and predicted parameters provided by an embodiment of the present invention; Figure 7 It is a characteristic trajectory diagram of a doubly-fed wind turbine system provided by an embodiment of the present invention; Figure 8 It is a time domain waveform of the voltage on the grid side of a doubly-fed wind turbine generator under a gray box characteristic provided by an embodiment of the present invention; Fig. 9 This is a system characteristic trajectory diagram under a 15% parameter error provided by an embodiment of the present invention; Fig.10 This is a voltage time domain waveform diagram when a parameter error is 15% provided by an embodiment of the present invention; Fig.11is a detailed implementation flow chart of a method for identifying wind turbine impedance parameters based on machine learning provided by an embodiment of the present invention; Fig.12 It is a structural schematic diagram of a wind turbine impedance parameter identification system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0021] See also Figure 1 As shown, it is an implementation flow chart of a method for identifying wind turbine impedance parameters based on machine learning provided by an embodiment of the present invention, comprising the following execution steps: Step 100: The electrical parameters, some control parameters and operating point parameters of the doubly-fed wind turbine are set as known conditions, the PI control parameters in the rotor-side inverter controller are taken as unknown quantities, and a grey box impedance model of the doubly-fed wind turbine is constructed.

[0022] In some embodiments, see Figure 2 As shown, it is a structural diagram of a doubly-fed wind turbine system provided by an embodiment of the present invention, in which the gray box part is unknown and the white box part is known. Considering the power outer loop, the current inner loop and the phase-locked loop, the specific functions of each part are as follows: (1) Doubly-fed induction generator (DFIG), which is used to receive mechanical energy and convert it into electrical energy; (2) The dq decoupling control part is responsible for converting the three-phase voltage and current into the direct axis (d axis) and quadrature axis (q axis) components in the dq coordinate system. (3) The RSC (rotor-side converter) is connected between the rotor and the DC bus of the DFIG to control the rotor current. The GSC (grid-side converter) is connected between the DC bus and the grid to control the current flowing to the grid. Both converters include filters to reduce harmonics and electromagnetic interference. (4) PI controller, which is used to adjust voltage and current. The PI controller reduces errors through proportional (P) and integral (I) actions to improve the stability and response speed of the system. The meanings of the parameters in the white box part are as follows: and : Represents the voltage components of the d-axis and q-axis, which are the voltage components in the rotating coordinate system and are used to control the magnetic field and torque of the motor. and : represents the current components of the d-axis and q-axis, which are also the current components in the rotating coordinate system and are used to control the magnetic field and torque of the motor. and : These are reference values ​​of voltage and current, usually set by an external controller or control algorithm to guide the operation of the motor. : represents the product of synchronous inductance and angular velocity, used to calculate the back electromotive force of the motor. PI: represents PI controller. : This represents the DC bus voltage, which is the input voltage to the power electronic converter. : This represents the transformation from the three-phase stationary coordinate system (abc) to the rotating coordinate system (dq), which is used to convert the three-phase AC signal into two orthogonal DC signals.

[0023] In some implementations, the DFIG main circuit parameters are shown in Table 1: Table 1: DFIG main circuit parameters

[0024] Step 101: Using a machine learning algorithm to identify PI control parameters in the gray box impedance model of the doubly fed wind turbine generator system; wherein the PI control parameters include proportional and integral parameters of the inner loop of current control and proportional and integral parameters of the outer loop of power control.

[0025] It should be noted that machine learning is the mainstream of artificial intelligence and is widely used in power grid fault identification, intelligent decision-making assistance, human-computer interaction, load forecasting and other fields. With the help of previous application ideas, machine learning mainly realizes regression and classification functions. Machine learning includes traditional machine learning, deep learning, reinforcement learning and deep reinforcement learning. The machine learning impedance identification method introduced in this paper uses four different algorithms, namely back propagation (BP) neural network, radial basis function (RBF) neural network, generalized regression (GR) neural network and support vector regression (SVR). Under supervised learning, this paper uses these algorithms to perform regression prediction and error comparison on the parameters of the inner and outer loops, and selects an algorithm with fast calculation and high accuracy for application.

[0026] Specifically, when executing step 101, the following steps may be specifically performed: S1010: Build a sample data set and generate a regression prediction model through machine learning training.

[0027] Preferably, the sample data set includes a random matrix of control parameters and a frequency scanning sampling point data matrix thereof, wherein the frequency scanning sampling point data matrix is ​​characterized by a frequency scanning method using harmonic injection to sample the voltage and current of the sampling points and obtain frequency scanning data through Fourier transform.

[0028] In some implementations, after building a doubly-fed wind turbine system model, a harmonic injection frequency sweep method is used to analyze the correctness of the theoretical model. Simulink is used to build a simulation model, and the voltage and current at the sampling point are sampled. The frequency sweep data is obtained by Fourier transform, and the frequency sweep data is compared with the Bode plot of the theoretical model, such as Figure 3 As shown, the upper and lower figures in the upper left corner represent the Bode diagrams of the theoretical model of amplitude and phase angle comparison under the Zdd coordinate, the upper and lower figures in the lower left corner represent the Bode diagrams of the theoretical model of amplitude and phase angle comparison under the Zqd coordinate, the upper and lower figures in the upper right corner represent the Bode diagrams of the theoretical model of amplitude and phase angle comparison under the Zdq coordinate, and the upper and lower figures in the lower right corner represent the Bode diagrams of the theoretical model of amplitude and phase angle comparison under the Zqq coordinate. The theoretical model fits the sweep frequency points well, proving that the sweep frequency data is correct and usable.

[0029] In the prediction of the gray box, the operating point of the theoretical model is obtained by actual measurement. In the theoretical model, the inner loop parameters to be identified are set as random numbers to generate a random number matrix. With the help of the theoretical model, the random number matrix will generate a frequency scanning sampling matrix under different control parameters. The parameter random number matrix and the frequency scanning sampling matrix generate sample data accordingly. In the theoretical model, the measured parameters are generated as a random number matrix, and the corresponding frequency scanning sampling matrix is ​​calculated. The frequency range of the scanning sampling matrix is ​​10Hz-200Hz. Through the above process, model data generation is realized.

[0030] S1011: Acquire an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substitute the impedance scanning data set into the regression prediction model, and output predicted PI control parameters.

[0031] S1012: Substitute the predicted PI control parameters into the grey box impedance model of the doubly-fed wind turbine generator system, compare the transient response of the active power under the conditions of the actual PI control parameters and the predicted PI control parameters, and verify the accuracy and reliability of the predicted PI control parameters.

[0032] In some embodiments, the gray box model parameter identification of the doubly fed wind turbine generator set based on machine learning includes: in the "gray box" model simulation, the electrical parameters, some control parameters and working point parameters of the doubly fed wind turbine generator set are set as known conditions, and the PI control parameters in the rotor side inverter controller are regarded as unknown quantities. Each has its own advantages in terms of the accuracy of the regression prediction of the training sample and the training time. In order to simplify the complexity of the model, it is assumed that the control parameters of the d-axis and the q-axis are equal, and the four parameters that need to be identified are the proportional and integral parameters of the inner loop of the current control and the proportional and integral parameters of the outer loop of the control power control. The constructed sample data is trained by machine learning to generate a regression prediction model. After the regression prediction model is constructed, the impedance scanning data set is obtained by using a frequency sweeper or other methods for the gray box system of the doubly fed wind turbine generator set in actual operation. Substitute the acquired impedance scanning data set into the prediction model, and the prediction model will output the predicted PI control parameters. Then, the inner loop proportional (kp1) integral (ki1) coefficient and the outer loop proportional (kp2) integral (ki2) coefficient were regressed and predicted under different algorithms, and four machine learning algorithms were trained to generate model prediction data using parameter random matrices and corresponding frequency scan sampling matrices as sample data. Figure 4 and Figure 5 As shown in the figure, the BP algorithm has the smallest error in fitting the four parameters, and some samples of RBF have large errors. The GRNN and SVR algorithms have good fitting effects on the two proportional parameters, but the fitting errors on the two integral parameters are large, and the trained model is unusable. Finally, the model trained by the BP neural network is used for parameter identification of the doubly fed wind turbine, and the results are shown in Table 2.

[0033] Table 2: Performance comparison between actual and predicted values

[0034] After substituting the predicted PI control parameters into the simulation model, the active power transient response under the actual parameters and predicted parameters is compared. The results are as follows: Figure 6 As shown in the figure, it can be observed that the two are highly fitted in the entire transient response process, and the prediction error is small. This shows that the predicted parameter values ​​have high accuracy and reliability and can effectively reflect the dynamic characteristics of the system. Therefore, the predicted parameter values ​​can be applied to the dynamic characteristics analysis of the system, providing a solid foundation for further research and application. Step 102: Substitute the identified PI control parameters into the grey box impedance model of the doubly-fed wind turbine generator system, output a back-ratio matrix, and based on the back-ratio matrix, use the generalized Nyquist stability criterion to determine the stability of the doubly-fed wind turbine generator system.

[0035] Specifically, the back ratio matrix is ​​as follows:

[0036] In the formula, represents the grid impedance matrix, represents the impedance characteristics of the doubly-fed wind turbine generator set, , , , They respectively represent the amplitude relationships from d-axis to d-axis, d-axis to q-axis, q-axis to d-axis, and q-axis to q-axis in the rotating coordinate system.

[0037] In some embodiments, when executing step 102, the following steps may be specifically performed: determining whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane; if not, determining that the doubly-fed wind turbine system is stable; otherwise, determining that the doubly-fed wind turbine system is unstable.

[0038] In some embodiments, the stability analysis process of the gray box doubly fed wind turbine model is as follows: The Nyquist stability criterion is to determine whether the system meets the necessary and sufficient conditions for stability based on the amplitude angle principle in the complex function. Substituting the identified parameters into the doubly fed wind turbine model, the system changes from a "gray box" model to a "white box" model with a clear mechanism. The generalized Nyquist stability criterion can be used to implement the stability analysis of the system to analyze whether the back-ratio matrix can meet the criterion. If the two eigenvalues ​​of the back-ratio matrix (λ1(s), λ2(s)) do not surround the point (-1, j0) on the complex plane, the system is stable, otherwise the system is unstable. The results of the system Nyquist analysis are as follows: Figure 7 As shown in Figure 2, the two characteristic roots of the system are not surrounded by the point (-1, j0), which proves that the system is stable. The time domain simulation diagram of the corresponding system is shown in Figure 2. Figure 8 As shown, the waveform has no divergence trend, indicating that the system can operate stably.

[0039] In some embodiments, the influence of impedance identification error on stability analysis is also analyzed. In the parameter identification of the system, when the proportional and integral errors of the inner loop are 15%, the characteristic trajectory of the system is as follows Fig. 9 As shown, the simulation time domain waveform under the corresponding parameters is as follows Fig.10 As shown. Fig.10 It can be seen that when the parameter identification error is large, the stability judgment result will change. Therefore, accurate parameter identification in the gray box system is of great significance.

[0040] In some embodiments, see Fig.11As shown, a detailed implementation flow chart of a wind turbine impedance parameter identification method based on machine learning provided by an embodiment of the present invention is mainly composed of three parts: model data generation, data prediction model and stability verification. First, by deeply analyzing the operating characteristics of the doubly fed wind turbine system, a "gray box" impedance model (gray box theoretical model and simulation model) of the doubly fed wind turbine is constructed. If it is verified by the sweep frequency diagram, the control parameter random number matrix and its sweep frequency sampling point data matrix are output, and sample data is constructed based on it. Subsequently, a machine learning algorithm is used for regression fitting, and a prediction model is generated through error judgment. The sweep frequency data of the gray box model is brought into the prediction model, and the control parameters in the model are accurately identified. Finally, based on the generalized Nyquist criterion, the stability of the doubly fed wind turbine system is analyzed.

[0041] The technical effects achieved by this application are as follows: 1. Compared with the sweep frequency method and the analytical method, this patent can overcome the "black / gray box" problem of wind farm structure and control parameters at this stage, and can still effectively perform modeling and analysis in the absence of detailed internal parameters. 2. The proposed algorithm can perform regression prediction and error comparison on the parameters of the inner loop and the outer loop, and can automatically select an algorithm with fast calculation and high accuracy for application. 3. It can be characterized by the relationship between the voltage and current at the device or system port, and can directly characterize the dynamic characteristics and stability of the system in the frequency domain without relying on detailed internal model information. 4. The modeling and research of the grid-connected impedance characteristics of large wind farms under the background of "black / gray box" provides new ideas and methods, and lays a model foundation for the stability analysis of large-scale renewable energy grid connection.

[0042] See also Fig.12 FIG. 1 is a schematic diagram of a wind turbine impedance parameter identification system based on machine learning provided by an embodiment of the present invention, comprising: A construction unit 120 is used to set the electrical parameters, some control parameters and working point parameters of the doubly-fed wind turbine generator set as known conditions, take the PI control parameters in the rotor-side inverter controller as unknown quantities, and construct a gray box impedance model of the doubly-fed wind turbine generator set; An identification unit 121 is used to identify the PI control parameters in the gray box impedance model of the doubly-fed wind turbine generator using a machine learning algorithm; wherein the PI control parameters include the proportional and integral parameters of the current control inner loop and the proportional and integral parameters of the control power control outer loop; The judgment unit 122 is used to substitute the identified PI control parameters into the gray box impedance model of the doubly-fed wind turbine generator system, output a back-ratio matrix, and judge the stability of the doubly-fed wind turbine generator system based on the back-ratio matrix and the generalized Nyquist stability criterion.

[0043] In some implementations, the identification unit 121 is specifically configured to: Construct a sample data set and generate a regression prediction model through machine learning training; Obtaining an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substituting the impedance scanning data set into the regression prediction model, and outputting predicted PI control parameters; The predicted PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator set, and the transient response of the active power under the conditions of actual PI control parameters and predicted PI control parameters is compared to verify the accuracy and reliability of the predicted PI control parameters.

[0044] In some implementations, the determining unit 122 is specifically configured to: It is determined whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane. If not, it is determined that the doubly-fed wind turbine system is stable. Otherwise, it is determined that the doubly-fed wind turbine system is unstable.

[0045] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the wind turbine impedance parameter identification method based on machine learning described in any one of the above embodiments are implemented.

[0046] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the wind turbine impedance parameter identification method based on machine learning described in any of the above embodiments are implemented.

[0047] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0048] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0049] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0050] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0051] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0052] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0053] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0054] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0055] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A wind turbine impedance parameter identification method based on machine learning, characterized in that: include: The electrical parameters, some control parameters and working point parameters of the doubly-fed wind turbine are set as known conditions, and the PI control parameters in the rotor-side inverter controller are taken as unknown quantities to construct a gray box impedance model of the doubly-fed wind turbine. Using a machine learning algorithm to identify PI control parameters in the gray box impedance model of the doubly-fed wind turbine generator set; wherein the PI control parameters include proportional and integral parameters of the inner loop of the current control and proportional and integral parameters of the outer loop of the power control; The identified PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator system, and a back-ratio matrix is ​​output. Based on the back-ratio matrix, the generalized Nyquist stability criterion is used to judge the stability of the doubly-fed wind turbine generator system.

2. The wind turbine impedance parameter identification method based on machine learning according to claim 1 is characterized in that: The PI control parameters in the grey box impedance model of the doubly-fed wind turbine generator are identified by using a machine learning algorithm, including: Construct a sample data set and generate a regression prediction model through machine learning training; Obtaining an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substituting the impedance scanning data set into the regression prediction model, and outputting predicted PI control parameters; The predicted PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator set, and the transient response of the active power under the conditions of actual PI control parameters and predicted PI control parameters is compared to verify the accuracy and reliability of the predicted PI control parameters.

3. The wind turbine impedance parameter identification method based on machine learning according to claim 2 is characterized in that: The sample data set includes a control parameter random matrix and a frequency scanning sampling point data matrix thereof, wherein the frequency scanning sampling point data matrix is ​​characterized by sampling the voltage and current of the sampling points by a frequency scanning method using harmonic injection, and obtaining frequency scanning data by Fourier transform.

4. The wind turbine impedance parameter identification method based on machine learning according to claim 1 is characterized in that: The back ratio matrix is ​​as follows: In the formula, represents the grid impedance matrix, represents the impedance characteristics of the doubly-fed wind turbine generator set, , , , They respectively represent the amplitude relationships from d-axis to d-axis, d-axis to q-axis, q-axis to d-axis, and q-axis to q-axis in the rotating coordinate system.

5. The wind turbine impedance parameter identification method based on machine learning according to claim 4 is characterized in that: The generalized Nyquist stability criterion is used to substitute the identified PI control parameters into the gray box impedance model of the doubly-fed wind turbine generator system, and the back-ratio matrix is ​​output. Based on the back-ratio matrix, the generalized Nyquist stability criterion is used to judge the stability of the doubly-fed wind turbine generator system, including: It is determined whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane. If not, it is determined that the doubly-fed wind turbine system is stable. Otherwise, it is determined that the doubly-fed wind turbine system is unstable.

6. A wind turbine impedance parameter identification system based on machine learning, characterized in that: include: A construction unit is used to set the electrical parameters, some control parameters and working point parameters of the doubly-fed wind turbine generator set as known conditions, take the PI control parameters in the rotor-side inverter controller as unknown quantities, and construct a gray box impedance model of the doubly-fed wind turbine generator set; An identification unit, used to identify PI control parameters in the gray box impedance model of the doubly-fed wind turbine generator set using a machine learning algorithm; wherein the PI control parameters include proportional and integral parameters of the inner loop of the current control and proportional and integral parameters of the outer loop of the power control; The judging unit is used to substitute the identified PI control parameters into the grey box impedance model of the doubly-fed wind turbine generator system, output a back-ratio matrix, and judge the stability of the doubly-fed wind turbine generator system based on the back-ratio matrix and the generalized Nyquist stability criterion.

7. The wind turbine impedance parameter identification system based on machine learning according to claim 6, characterized in that: The identification unit is specifically used for: Construct a sample data set and generate a regression prediction model through machine learning training; Obtaining an impedance scanning data set of a gray box system of a doubly-fed wind turbine generator set, substituting the impedance scanning data set into the regression prediction model, and outputting predicted PI control parameters; The predicted PI control parameters are substituted into the grey box impedance model of the doubly-fed wind turbine generator set, and the transient response of the active power under the conditions of actual PI control parameters and predicted PI control parameters is compared to verify the accuracy and reliability of the predicted PI control parameters.

8. The wind turbine impedance parameter identification system based on machine learning according to claim 6, characterized in that: The judging unit is specifically used for: It is determined whether the eigenvalues ​​of the back-ratio matrix surround the target point on the complex plane. If not, it is determined that the doubly-fed wind turbine system is stable. Otherwise, it is determined that the doubly-fed wind turbine system is unstable.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the wind turbine impedance parameter identification method based on machine learning as described in any one of claims 1 to 5 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind turbine impedance parameter identification method based on machine learning as described in any one of claims 1 to 5 are implemented.

Images