Virtual synchronous machine control double-fed wind power plant small signal steady state discrimination method and system, and medium

Abstract

The invention relates to the technical field of new energy grid-connected stability control, in particular to a small-signal steady-state discrimination method and system for a double-fed wind power plant controlled by a virtual synchronous machine and a medium, and the method comprises the steps: building a linear state space model for a double-fed fan controlled by a single virtual synchronous machine; listing a characteristic polynomial, and deducing a single-machine stability criterion inequality by using a Routh criterion; the method comprises the following steps: constructing a full-order state space matrix for a doubly-fed wind power plant comprising M units; decomposing the matrix into M independent subsystems, and performing eigenvalue decomposition on the current collection network impedance matrix to obtain a maximum eigenvalue; the M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and an equivalent single-machine subsystem; and carrying out Routh criterion derivation on the equivalent single machine subsystem to obtain an explicit stability criterion inequality of the double-fed wind power plant grid-connected system and carrying out steady state discrimination. According to the invention, the problem that the stability boundary cannot be explicitly displayed under the complex power grid topology in the prior art is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of new energy grid-connected stability control technology, and in particular to a method, system and storage medium for small-signal steady-state discrimination of doubly fed wind farms controlled by virtual synchronous machines. Background Technology

[0002] In the field of wind power grid connection, Virtual Synchronous Generator (VSG) control technology significantly improves the grid support capability of wind power systems by simulating the inertia and damping characteristics of synchronous generators. However, in areas with weak grid structures, i.e., weak grids, the stable operation of wind farm clusters faces severe challenges. In such scenarios, there is a complex dynamic coupling relationship between the control parameters of wind turbines and the strength characteristics of the grid topology. If the linkage constraint boundary between these three cannot be accurately quantified, it will directly lead to system oscillations or even grid disconnection accidents.

[0003] Existing stability assessment techniques are mostly based on single-machine equivalent models, simplifying wind farms into a single virtual machine group for analysis. While this model simplifies calculations, it completely ignores the electrical coupling and power oscillation transfer effects between wind turbine groups. When multiple units operate in tandem, the stability boundary output by the single-machine model deviates significantly from the actual operating conditions, creating safety hazards. Existing technologies typically separate grid strength analysis from control parameter tuning, treating grid strength as a fixed background condition and failing to establish its dynamic mathematical relationship with virtual inertia and damping coefficients. This separation makes it impossible to establish a unified three-dimensional stability criterion, forcing engineers to rely on experience for local parameter adjustments and failing to anticipate the systemic impact of parameter combinations on global stability. Traditional numerical simulation methods require repeated solutions to high-order differential equations, resulting in excessive computation time and making it difficult to embed in real-time control systems. Frequency domain impedance methods rely on prior knowledge of specific network topologies and are insufficiently adaptable to the ever-changing structures of actual power grids. Neither method can generate closed-loop or explicit criteria suitable for online control, preventing wind farms from dynamically adjusting control parameters according to grid conditions.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method, system, and medium for small-signal steady-state discrimination in doubly-fed wind farms controlled by virtual synchronous machines, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for small-signal steady-state discrimination in a virtual synchronous machine-controlled doubly-fed wind farm, the method comprising: A linearized state-space model including virtual inertia, damping coefficient, and grid parameters is established for a doubly fed wind turbine controlled by a single virtual synchronous machine. Based on the linearized state-space model, the characteristic polynomial is written, and the Routh criterion is applied to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane. For a doubly fed wind farm with M generating units, construct a full-order state-space matrix that includes dynamic coupling between generating units and the impedance of the collector network; The full-order state space matrix is ​​decomposed into M independent subsystems according to the block diagonalization operation, and the impedance matrix of the collector network is decomposed into eigenvalues ​​to obtain the maximum eigenvalue as the equivalent network strength characterization. The M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid; The Routh criterion is derived by performing the equivalent single-machine subsystem on the equivalent single-machine subsystem. Combined with the maximum eigenvalue and the weak grid approximation condition, the explicit stability criterion inequality of the doubly fed wind farm grid-connected system is obtained. The real-time collected power grid strength characteristic value and wind turbine control parameters are input into the explicit stability criterion inequality for online discrimination; when the inequality is true, a system stability signal is output, triggering normal wind turbine operation; when the inequality is false, an instability warning signal is output, triggering adaptive adjustment of the virtual inertia or the damping coefficient.

[0007] Furthermore, the parameters of the linearized state-space model include: The virtual inertia parameters and damping coefficient parameters defined in the virtual synchronous machine control loop; The resistance and inductance characteristics of the stator winding of the doubly fed wind turbine; Equivalent grid reactance parameters from the wind turbine connection point to the grid common coupling point; The reactive power control loop of the virtual synchronous machine is used to adjust the reactive power regulation coefficient parameter for regulating voltage deviation.

[0008] Furthermore, the single-machine stability criterion inequality includes: The damping coefficient parameter must be greater than the sum of the linear reference term based on the virtual inertia parameter and the product adjustment term based on the reactive power adjustment coefficient parameter and the equivalent grid reactance parameter; The proportional relationship of the linear reference term is determined by the resistance characteristic parameter of the stator winding, and the weighting relationship of the product adjustment term is determined by the inductance characteristic parameter of the stator winding.

[0009] Furthermore, the impedance matrix of the collector network is decomposed into eigenvalues ​​to obtain the maximum eigenvalue, including: The collector network impedance matrix obtained by the block diagonalization operation is subjected to a symmetry transformation to obtain a symmetric matrix; The symmetric matrix is ​​subjected to spectral analysis to obtain all eigenvalues ​​and generate an eigenvalue set structure; The feature component with the largest value is selected from the feature value set structure as the power grid strength characterization.

[0010] Furthermore, the M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and one equivalent stand-alone subsystem, including: The reactive coupling path of the inter-unit interaction corresponding to the M-1 isolated subsystems is replaced with a static zero-impedance coupling model to achieve electrical decoupling; Based on the electrical decoupling state, the virtual inertia parameters and damping coefficient parameters of the original wind turbine system are kept unchanged for the equivalent single-machine subsystem. The maximum eigenvalue is coupled with the equivalent grid reactance parameter to perform network strength calculation, thereby generating the normalized equivalent reactance characterization parameter at the external grid port of the equivalent single-machine subsystem. The normalized equivalent reactance characterization parameters are used to replace the grid reactance parameters to complete the grid interface modeling of the equivalent single-machine subsystem.

[0011] Furthermore, the Routh-Hurwitz criterion is derived for the equivalent single-machine subsystem, including: The normalized equivalent reactance characterization parameters completely replace the original grid reactance parameter terms included in the single-machine stability criterion inequality; Based on the above, the single-machine stability criterion inequality is simplified by parameter merging based on the linear approximation constraint rule of the collector network impedance parameters under weak grid operation conditions. The maximum eigenvalue and the reference impedance component under weak grid conditions are imported into the simplified single-machine stability criterion inequality, and the explicit stability criterion inequality of the doubly-fed wind farm grid-connected system is output.

[0012] Furthermore, the linear approximation constraint rules under weak grid operating conditions include: Under the condition that the external grid connection impedance value is continuously greater than the total impedance of the wind farm collection network, the intrinsic reactance component of the transmission network included in the grid parameters is obtained as a basic reference quantity. Based on the topology of the wind farm's power collection network, a network topology constant factor is generated, and a linear proportional constraint relationship is established between the network topology constant factor and the basic reference quantity. The linear proportional constraint relationship is used as the calculation criterion for the linear approximation constraint rule to generate the reference impedance component.

[0013] Furthermore, the explicit stability criterion inequality includes: The maximum eigenvalue in the equivalent single-machine subsystem is correlated with the virtual inertia parameter and damping coefficient parameter to generate a power grid strength constraint relationship by performing a stability correlation mapping operation. Based on the aforementioned weak grid approximation conditions, the grid reactance parameters in the single-machine stability criterion inequality are transformed into network-dependent stability boundary variables. The power grid strength constraint relationship is integrated into the network-dependent stability boundary variable expression to form a three-dimensional stability criterion structure that includes a virtual inertia boundary control term, a damping coefficient adjustment term, and a network strength dominant term.

[0014] A virtual synchronous machine-controlled small-signal steady-state discrimination system for doubly-fed wind farms, the system comprising: The model building module establishes a linearized state-space model for a doubly fed wind turbine controlled by a single virtual synchronous machine, including virtual inertia, damping coefficient, and grid parameters. The Routh derivation module is based on the linear state-space model to write the characteristic polynomial and applies the Routh criterion to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane. The full-order matrix module constructs a full-order state-space model for a doubly fed wind farm with M units, including dynamic coupling between units and the impedance of the collector network. The diagonal operation module decomposes the full-order state space matrix into M independent subsystems according to the block diagonalization operation, and performs eigenvalue decomposition on the collector network impedance matrix to obtain the maximum eigenvalue as the equivalent network strength characterization. The equivalent system module transforms M independent subsystems into M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid. The criterion construction module performs Routh criterion derivation on the equivalent single-machine subsystem, and obtains the explicit stability criterion inequality of the doubly fed wind farm grid-connected system by combining the maximum eigenvalue and the weak grid approximation condition. The stability discrimination module inputs the real-time collected power grid strength characteristic values ​​and wind turbine control parameters into an explicit stability inequality for online discrimination. When the inequality is true, it outputs a system stability signal, triggering normal wind turbine operation; when the inequality is false, it outputs an instability warning signal, triggering adaptive adjustment of virtual inertia or damping coefficient.

[0015] A computer-readable storage medium is characterized in that the computer-readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, can implement the virtual synchronous machine controlled doubly-fed wind farm small-signal steady-state discrimination method.

[0016] The technical solution of this invention can achieve the following technical effects: By constructing a technical chain of single-machine Routh criterion, wind farm block diagonal decoupling, grid strength characteristic value quantification, and weak grid equivalent normalization, this method solves the problem that existing methods cannot establish explicit stability boundaries containing virtual inertia, damping coefficients, and network parameters under complex grid topologies, and fills the technical gap of fast closed-loop solution of the stability boundary of virtual synchronous control of wind farms under weak grid conditions.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] 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.

[0019] Figure 1 A flowchart illustrating the small-signal steady-state discrimination method for a virtual synchronous machine-controlled doubly-fed wind farm; Figure 2 A structural diagram of a grid-connected system for a single doubly fed wind turbine controlled by a virtual synchronous machine; Figure 3 Here is the system structure diagram for the example; Figure 4 This is a structural diagram of a wind farm grid-connected system consisting of M doubly-fed wind turbines controlled by virtual synchronous generators; Figure label: I sd +jI sq Complex representation of stator current in dq coordinates; I rd +jI rq Representation of rotor or converter-side current in dq coordinates; I gd +jI gq Representation of grid-connected current in dq coordinates; P r Rotor-side active power flow; P g Grid-side active power flow; I d Direct-axis component of stator current; I q C. Stator current quadrature-axis component; V. Converter DC-side capacitance; dc DC bus voltage; U s ∠θ s 1. Voltage phasor of the external power grid at the grid connection point; P ref Active power reference; P; Active power measurement; Qref Reactive power reference; Q: Actual reactive power output; K q 1. Reactive power regulation coefficient; 1 / (2Hs); Schematic diagram of the dynamic link of the equivalent inertia of the virtual synchronous machine; H, virtual inertia; D, damping coefficient; 1 / s; Schematic diagram of the integral link from frequency to angle; θ r Reference phase angle; V r Rotor-side voltage phasor; X L 1. Grid-connected line reactance; ω1, instantaneous electrical angular frequency at grid connection point 1; ω r Rotor electrical angular velocity; ω slip Slip frequency; ω s Synchronous electrical angular frequency. Detailed Implementation

[0020] 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.

[0021] 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.

[0022] Example 1; like Figure 1 , Figure 2 and Figure 3 As shown, this application provides a method for small-signal steady-state discrimination in a doubly-fed wind farm controlled by a virtual synchronous machine. The method includes: S10: Establish a linearized state-space model for a doubly fed wind turbine controlled by a single virtual synchronous machine, including virtual inertia, damping coefficient, and grid parameters; S20: Based on the linear state-space model, write the characteristic polynomial and apply the Routh criterion to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane; S30: For a doubly fed wind farm including M units, construct a full-order state-space matrix including dynamic coupling between units and collector network impedance; S40: Decompose the full-order state space matrix into M independent subsystems according to the block diagonalization operation, and perform eigenvalue decomposition on the collector network impedance matrix to obtain the maximum eigenvalue as the equivalent network strength characterization. S50: Equivalent M independent subsystems to M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid; S60: Perform the Routh criterion derivation on the equivalent single-machine subsystem, and combine the maximum eigenvalue with the weak grid approximation condition to obtain the explicit stability criterion inequality of the doubly fed wind farm grid-connected system; S70: The real-time collected power grid strength characteristic value and wind turbine control parameters are input into an explicit stability inequality for online discrimination; when the inequality is true, a system stability signal is output, triggering normal wind turbine operation; when the inequality is false, an instability warning signal is output, triggering adaptive control of virtual inertia or damping coefficient.

[0023] Specifically, the first step is to linearize the doubly fed wind turbine controlled by a single virtual synchronous machine under a certain steady-state condition using small perturbation processing. The preferred steps are as follows: First, the operating point values ​​of each key variable are obtained using steady-state power flow or steady-state operating equations. Then, the equivalent inertia and damping terms introduced by the virtual synchronous machine control are incorporated into the active dynamics description of the virtual synchronous machine. State variables related to electromagnetic dynamics, such as current, voltage, and speed, are retained. Simultaneously, grid parameters on the grid side, including collector network impedance and the short-circuit capacity of the parallel equivalent power sources, are included as exogenous parameters in the model. Finally, small perturbations are applied to all the above state variables at the steady-state point, and linearization is performed to obtain the state-space description of the single machine. This linearized model... This includes virtual inertia, damping coefficient, converter setting parameters, and grid parameters. Based on this linearized state-space model, the characteristic equation of the closed-loop system is explicitly written according to traditional control theory procedures. The Routh stability criterion is used to determine the characteristic equation, and a single-machine stability inequality that guarantees the characteristic root lies in the left half of the complex plane is derived. This inequality expresses the stability margin using the combination of virtual inertia, damping, and grid-related equivalent impedance parameters. For a doubly-fed wind farm consisting of M units, it is preferable to establish a full-order state-space matrix including all dynamic coupling terms of the units and the impedance of the collector network according to the actual collector line structure. This matrix is ​​in symmetric coordinates or rectangular coordinates. The system can be constructed in any coordinate system, and the coupling between the units is manifested through the finite impedance of the collector lines and the public grid port. For ease of analysis and engineering implementation, the full-order matrix is ​​preferably transformed using the block diagonalization approach: First, the impedance or admittance matrix of the collector network is decomposed into eigenvalues ​​to obtain a set of orthogonal or linearly independent eigenvectors and their corresponding eigenvalues. Then, the full-order state matrix is ​​transformed in a coordinate system extended by these eigenvectors, which decomposes the coupling relationship into several independent subsystems. This operation equivalences the original coupling problem of M units to M independent subsystems divided by mode, where the largest eigenvalue of the network is taken as the characteristic of the system. The equivalent strength of the power grid, i.e., the maximum eigenvalue, is used as the descriptor of the equivalent network strength to measure the influence of the power collection network and the external power grid on the dynamics of the wind farm. Based on this, it is preferable to simplify the M independent subsystems into M-1 isolated subsystems that do not interact with the external power grid under the transformed coordinates, and an equivalent single-unit subsystem representing the interaction between the entire plant and the external power grid. The parameters of the equivalent single unit are obtained by the weighted combination of the original units under the mode transformation. Its equivalent inertia, equivalent damping, and equivalent converter control parameters can all be calculated from the original parameters and the components of the network eigenvector decomposition, thus preserving the influence of the differences between the units and the network distribution characteristics.Subsequently, the derivation process of single-machine linearization and Routh criterion is repeated for the equivalent single-machine subsystem. Under the weak grid approximation condition, the weak grid approximation is preferably understood as the maximum eigenvalue of the equivalent network strength being within a certain small range. This allows the coupling term of the network to the unit to be approximated using a first-order approximation. Combining the maximum eigenvalue as a characterization of network strength, an explicit stability criterion inequality for the wind farm grid-connected system is obtained. This inequality clearly gives the relationship between virtual inertia, damping coefficient and equivalent network strength, which is convenient for engineering to meet stability requirements by adjusting the virtual synchronous machine parameters or improving network access conditions. To facilitate engineering applications and verify the effectiveness of this method, an example is provided: In a sample wind farm consisting of 20 turbine units, the admittance matrix is ​​first constructed based on the measured collector impedance, and eigenvalue decomposition is performed. The maximum eigenvalue obtained is used to determine the network strength. In this example, when the virtual inertia of a single turbine unit is between approximately 1 and 3 seconds, the damping coefficient is between approximately 0.5 and 2, and the active power regulation loop bandwidth of the converter is set in the medium range, the explicit criterion obtained can be used to determine that the system can still maintain small-signal stability under weak grid conditions. If the criterion is not met, it can be remedied by increasing the virtual inertia, improving the damping, or improving the collector network, i.e., reducing the maximum eigenvalue.

[0024] The technical solution of this invention constructs a technical chain of single-machine Routh criterion, wind farm block diagonal decoupling, grid strength characteristic value quantification, and weak grid equivalent normalization. It solves the problem that existing methods cannot establish explicit stability boundaries containing virtual inertia, damping coefficients, and network parameters under complex grid topologies, and fills the technical gap of fast closed-loop solution of the stability boundary of virtual synchronous control of wind farms under weak grid conditions.

[0025] Furthermore, the parameters of the linearized state-space model include: The virtual inertia parameters and damping coefficient parameters defined in the virtual synchronous machine control loop; Resistance and inductance characteristics of the stator winding of a doubly-fed induction generator; Equivalent grid reactance parameters from the wind turbine connection point to the grid common coupling point; The reactive power control coefficient parameter used to adjust voltage deviation in the reactive power control loop of the virtual synchronous machine.

[0026] As a preferred embodiment, firstly, steady-state measurements are performed on each doubly-fed induction generator (DFIG) at its stable operating point after grid connection to obtain reference values ​​for linearization. Measurements include stator current, voltage, speed, power, and bus voltage amplitude and phase angle. For the virtual inertia and damping coefficient parameters defined in the virtual synchronous machine control, these are preferably initially selected by the system engineer during the controller design phase based on the wind farm size, grid short-circuit capacity, and required frequency response characteristics. Further identification and optimization are then performed on-site through small-amplitude disturbance tests or simulation sensitivity analysis. Specifically, a preferred approach is to apply a controllable active power pulse or frequency step at the rated operating point, record the unit's active power output and speed response, and then... The parameter identification method inversely derives the actual equivalent inertia and damping performance, thereby obtaining the values ​​of virtual inertia and damping coefficients for the linearization model. For the resistance and inductance characteristics of the stator windings of the doubly-fed induction generator (DFIG), it is preferable to use the manufacturer-provided design parameters as initial values, combined with on-site temperature correction and online impedance identification to obtain more accurate parameters. A preferred method for online identification includes applying a small voltage or current disturbance near rated operation and acquiring the stator port voltage and current waveforms, obtaining the equivalent resistance and inductance through frequency or time domain identification. Considering parameter variations caused by magnetic saturation or slip, it is preferable to establish separate parameter libraries for different operating conditions to select the closest parameter set during model linearization. Regarding the connection point of the wind turbine to the power grid... The equivalent grid reactance parameters at the common connection point are preferably determined in a two-step process during engineering implementation: first, by summarizing the known collector line parameters and transformer parameters according to the topology to obtain the initial static equivalent reactance value; second, by calibrating the equivalent reactance in conjunction with on-site short-circuit current testing or small disturbance voltage injection experiments. If the wind farm connection point is far from the public grid or there are multiple parallel lines, frequency domain admittance measurement is preferred, and frequency correlation is explained to obtain the equivalent reactance value in the small-signal analysis frequency band. For the reactive power control loop of the virtual synchronous machine, the reactive power regulation coefficient parameter used to regulate voltage deviation is preferably set by the controller tuning personnel before grid connection according to the voltage regulation target and grid connection voltage level, and then calibrated through steady-state and dynamic... Dynamic testing, such as applying a step reactive power reference change or external voltage disturbance, is used to observe the bus voltage response speed and overshoot. Closed-loop performance indicators are used to adjust this coefficient to balance voltage hold-up and system small-signal stability. For ease of engineering implementation, it is preferable to clearly define the order of parameter acquisition and uncertainty handling strategies in the model building process. Specifically, an initial linearized model is first established using manufacturer and line design parameters, then key parameters are corrected using field identification data, and reasonable uncertainty ranges are set for several parameters in the model for robustness analysis. Regarding parameter value recommendations, it is preferable to provide engineering experience ranges for reference. For example, in weak grid access scenarios, it is preferable to set the single-machine virtual inertia to an equivalent range of 1 to 5 seconds and the damping coefficient to 0.The range of 3 to 3 is designed to balance frequency support and local modal damping. Stator equivalent resistance and inductance are based on the manufacturer's nominal values ​​and are allowed adjustment within ±20% to match field measurements. The calibrated value of the equivalent grid reactance should reflect short-circuit capacity and line length; reactance is typically larger in remote locations where wind farms have weaker connections.

[0027] Furthermore, the single-machine stability criterion inequalities include: The damping coefficient parameter must be greater than the sum of the linear reference term based on the virtual inertia parameter and the product adjustment term based on the reactive power adjustment coefficient parameter and the equivalent grid reactance parameter; The proportional relationship of the linear reference term is determined by the resistance characteristic parameters of the stator winding, while the weighting relationship of the product adjustment term is determined by the inductance characteristic parameters of the stator winding.

[0028] As a preferred embodiment of the above, firstly, the doubly fed wind turbine controlled by a single virtual synchronous machine is linearized and modeled under steady-state conditions using conventional methods, and the coefficient expressions of each state variable and the closed-loop characteristic equation are clarified. Preferably, the criterion term of the characteristic equation is divided into two parts: one part is the linear reference term, which changes significantly with the virtual inertia, and its proportional coefficient is determined by the resistance characteristics of the stator winding; the other part is the product adjustment term, which consists of the product of the reactive power adjustment coefficient and the equivalent grid reactance, and its relative weight is determined by the inductance characteristics of the stator winding. Specifically, in engineering implementation, the single-machine stability criterion inequality should be obtained and applied according to the following steps: First The first step involves acquiring and confirming basic parameters. Reference values ​​for the stator equivalent resistance and inductance are obtained from manufacturer data and on-site measurements, and these are then identified and temperature-corrected online. Simultaneously, the equivalent reactance from the grid connection point to the point of common coupling is obtained through system design or short-circuit testing, and the reactive power regulation coefficient, virtual inertia, and damping coefficient settings in the virtual synchronous machine controller are read. The second step, based on a linearization model, describes the linear reference term in the criterion as a quantity amplified by the virtual inertia and proportionally adjusted by the stator resistance. This explains that if the stator resistance increases, the proportion of this reference term increases, thus requiring a larger damping coefficient to maintain the leftward shift of the characteristic root. Simultaneously, the product adjustment term is described as reactive power regulation... The adjustment effect, constituted by the product of the damping coefficient and the equivalent grid reactance, has a greater weight in the overall criterion as the stator inductance increases. That is, when the stator inductance is large, the coupling between the network and the reactive power loop has a more significant impact on small-signal stability. The third step provides a criterion verification and adjustment process. Substituting the above parameters into the criterion inequality checks whether the current damping coefficient meets the requirement that damping must be greater than the sum of the linear reference term and the product adjustment term. If not, adjustment measures are taken in priority order: preferably, the damping coefficient is moderately increased to improve modal damping; if limited, the virtual inertia is appropriately increased, but it must be considered that this will increase the linear reference term, potentially conversely increasing the damping effect. Requirements; at the same time, the product regulation term can be reduced by lowering the reactive power regulation coefficient or improving the equivalent reactance of the grid connection point, such as by modifying the collector line or using a parallel compensation device; the fourth step is to provide suggestions for engineering values ​​and test verification. In order to ensure the feasibility and safety margin of the project, it is preferable to set the virtual inertia of the single machine within the range of engineering experience under weak grid conditions and leave a margin of about 20% to 30% in combination with the damping coefficient. Before implementation, the consistency between the model calculation value and the actual closed-loop response is confirmed by small-amplitude active and reactive power step tests or frequency domain injection experiments. If necessary, the proportional coefficients of the two items in the fine-tuning criterion are identified online to reflect the actual stator impedance characteristics and network coupling.As an example, consider a wind turbine under a certain operating condition where the stator equivalent resistance is identified as relatively small and the stator inductance as medium. The equivalent grid reactance is in a weak grid range, and the reactive power regulation coefficient is set to a medium value. The sum of the calculated linear baseline term and the product regulation term represents a specific threshold. In engineering practice, the damping coefficient can be set above this threshold with a 20% margin to ensure stability even under fluctuating operating conditions or uncertain parameters. In this example, if increasing the reactive power regulation coefficient or grid reactance leads to a significant increase in the product regulation term, the additional damping requirement should be mitigated by reducing the reactive power regulation coefficient or improving network access conditions. If the network cannot be modified, the criteria must be met by increasing damping or redistributing the virtual inertia among the turbines.

[0029] Furthermore, eigenvalue decomposition is performed on the impedance matrix of the collector network to obtain the maximum eigenvalue, including: The collector network impedance matrix obtained by the block diagonalization operation is subjected to a symmetry transformation to obtain a symmetric matrix; Spectral analysis is used to calculate the symmetric matrix to obtain all eigenvalues ​​and generate an eigenvalue set structure. The eigencomponent with the largest value is selected from the eigenvalue set structure as the characterization of power grid strength.

[0030] As a preferred embodiment of the above, firstly, an original collector network impedance matrix is ​​constructed according to the engineering topology and component parameters. This matrix uses the unit grid connection point or the port after node reduction as rows and columns, and is filled with complex impedance or admittance elements. Preferably, after normalization and unification of the phase reference, if necessary, the network is first subjected to node reduction processing to eliminate the neutral point and internal intermediate nodes. That is, the equivalent processing commonly used in engineering is used to obtain a port-level impedance matrix that only reflects the coupling relationship between units. Secondly, in order to eliminate the asymmetry caused by modeling directionality or numerical noise and to ensure that the spectral decomposition result is a real eigenvalue and physically interpretable, it is preferred to perform symmetry processing on the port impedance matrix. Specifically, the original matrix and its conjugate transpose are weighted and averaged to obtain a Hermitian (i.e., conjugate symmetry). The matrix step is physically equivalent to preserving the symmetric portion of energy exchange in the network and filtering out small asymmetric errors. After symmetry, considering the frequency sensitivity of small-signal analysis, it is preferable to sample the impedance matrix at frequency points within the frequency band of interest for the small signal and repeat the following spectral analysis steps at each frequency point to obtain a set of frequency-dependent eigenvalues ​​or to obtain representative values ​​using a weighted average. Subsequently, a robust numerical spectral analysis method is used to calculate all eigenvalues ​​and corresponding eigenvectors of the symmetric matrix. It is preferable to use verified linear algebra numerical routines and employ double-precision arithmetic and strict convergence tolerances to ensure numerical accuracy. At the same time, condition number checks are performed on the matrix before solving, and ill-conditioned problems are mitigated by adding a minimal regularization to the diagonal terms when necessary. Spectral analysis After analysis, a set of real numbers, or eigenvalues ​​that should be a sequence of real numbers after symmetry, will be obtained, forming an eigenvalue set structure. Preferably, this set is sorted in descending order of numerical value, and the eigenvalue with the largest value and its corresponding eigenvector are selected. This largest eigenvalue is used in this method as a quantitative representation of grid strength; a larger value indicates a stronger coupling effect of the network on the generator units within the investigated frequency band, and a more significant equivalent rigidity or influence. Simultaneously, the corresponding eigenvector is used as modal participation information, which is used to weight and synthesize the original generator unit parameters to preserve spatial distribution characteristics when subsequently converting multiple generators to a single generator. For example, consider a wind farm consisting of ten generator units connected by several lines. After node reduction of the original line parameters, a tenth-order eigenvalue is obtained. The port impedance matrix is ​​averaged by its conjugate transpose to obtain a symmetric matrix. Then, all eigenvalues ​​calculated in the small signal frequency band are a set of column-by-column values. The largest eigenvalue is several and significantly larger than the others. The corresponding eigenvectors show that units 3 and 7 have the highest participation. Therefore, the largest eigenvalue is used as the equivalent network strength quantity and the normalized components of the corresponding mode vector are used to weight and synthesize the equivalent single unit's inertia and damping for use in subsequent single unit stability criteria. If, in this example, the largest eigenvalue is found to increase significantly with frequency, the largest value in the frequency response should be taken as a conservative criterion. In engineering, the largest eigenvalue should be reduced by improving the line, parallel compensation, or changing the access point, thereby enhancing the grid stability margin.

[0031] Furthermore, such as Figure 4 As shown, the M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem, including: The reactive coupling path of the interaction between units corresponding to the M-1 isolated subsystems is replaced with a static zero-impedance coupling model to achieve electrical decoupling. Based on the electrical decoupling state, the virtual inertia parameters and damping coefficient parameters of the original wind turbine system are kept unchanged for the equivalent single-machine subsystem; The maximum eigenvalue is coupled with the equivalent grid reactance parameter to generate the normalized equivalent reactance characterization parameter at the external grid port of the equivalent single-machine subsystem. The grid interface model of the equivalent single-machine subsystem is completed by replacing the grid reactance parameters with normalized equivalent reactance characterization parameters.

[0032] As a preferred embodiment of the above, based on the M independent subsystems obtained by block diagonalization, it is preferable to first identify M-1 subsystems that need to be treated as isolated subsystems and intentionally simplify the reactive coupling paths between them in the electrical modeling. Specifically, the line reactances reflecting the interaction between these subsystems are replaced in the model with a static zero-impedance coupling model, thereby achieving decoupling at the electrical level. In engineering implementation, to avoid numerical singularities, zero impedance can be replaced with a very small numerical quantity as an approximate zero value, and this is explained in the simulation and numerical solution. This decoupling operation is physically equivalent to treating the M-1 subsystems as being equipotentially connected to a common node through an ideal rigid connection, thereby making these isolated subsystems less susceptible to interference in small-signal dynamic analysis. Then, mutually coupled oscillating components are generated through network reactance. Simultaneously, to ensure the representativeness of the dynamics before and after equivalence, it is preferable to maintain the virtual inertia and damping coefficients of the virtual synchronous machine (representative unit or parameter set selected in the original wind turbine system) unchanged for the subsystem retained as the equivalent single-unit subsystem after decoupling. That is, the equivalent single unit directly adopts the pre-determined representative values ​​in the original system for virtual inertia and damping settings. These can be parameters from a single unit, or representative parameters obtained by weighting several unit parameters according to the initial modal participation, and these parameters are not adjusted after replacement. This ensures that the equivalent single unit retains the true response characteristics of the original system in terms of frequency support and damping. Subsequently, the maximum eigenvalue of the network obtained from the previous spectral analysis is compared with the equivalent power grid from the original grid connection point to the point of common coupling. Reactance parameters are used for network strength coupling calculations to generate normalized equivalent reactance characterization parameters for the external grid ports of the equivalent single-machine system. A preferred engineering implementation is to couple the maximum eigenvalue as an amplification or scaling factor with the equivalent reactance according to a reference value, thereby obtaining a normalized reactance value reflecting the strength of the network's coupling effect on the equivalent single-machine system in the frequency band of interest. In practice, a unified per-unit reference and a clearly defined base value can be established first, followed by obtaining the normalized value through numerical product or equivalent scaling. This normalized equivalent reactance characterization parameter then replaces various complex port impedances in the original network model, directly representing the influence of the external network on the equivalent single-machine system in the grid interface modeling of the equivalent single-machine subsystem. This simplifies the multi-machine coupling problem to a single-machine coupling with a single equivalent reactance. To address network issues and facilitate subsequent stability determination based on the Routh criterion or modal analysis, this embodiment also preferably provides examples and verification steps to facilitate engineering implementation and verify the rationality of the equivalence method: For example, in a wind farm consisting of 10 units, 10 independent subsystems are first obtained by diagonalizing the blocks according to the aforementioned method. Nine of these are selected as isolated subsystems, and the line reactance between them is replaced with an approximate zero value in the model to achieve electrical decoupling. The remaining subsystem is used as an equivalent single unit, and its virtual inertia is kept at 2 seconds and its damping coefficient at 1.2. If the equivalent reactance at the grid connection point is measured to be 0.6 per unit and the maximum eigenvalue obtained from spectral analysis is 2.5, then the normalized equivalent reactance value obtained by coupling the two is approximately 1.5 per unit, and this 1.5. Per-unit values ​​are used to replace the grid reactance at the equivalent single-unit external port for subsequent stability analysis. In numerical implementation, it is recommended to compare and verify the values ​​before and after the replacement, i.e., compare the eigenvalue distribution and main mode frequencies of the original system and the equivalent single-unit system under small disturbances. If the difference exceeds the allowable range, the selection of representative parameters needs to be adjusted or the zero-impedance approximation value needs to be increased to reflect residual coupling. Furthermore, it is recommended to perform sensitivity analysis after equivalence to assess the impact of measurement errors and model approximations on the normalized equivalent reactance. If necessary, the maximum eigenvalue or equivalent reactance can be reduced by improving the collector network, such as through parallel compensation or line modification, ultimately maintaining the steady-state small-signal margin required by the grid-connected system.

[0033] Furthermore, the derivation of the Routh-Hurwitz criterion for the equivalent single-machine subsystem includes: The normalized equivalent reactance characterization parameters completely replace the original grid reactance parameter terms included in the single-machine stability criterion inequality; Based on the above, the single-machine stability criterion inequality is simplified by parameter merging based on the linear approximation constraint rule of the collector network impedance parameters under weak grid operation conditions. By importing the maximum eigenvalue and the reference impedance component under weak grid conditions into the simplified single-unit stability criterion inequality, an explicit stability criterion inequality for the doubly fed wind farm grid-connected system is output.

[0034] As a preferred embodiment of the above, firstly, the normalized equivalent reactance characterization parameters obtained from the aforementioned equivalence are used to completely replace the grid-connected grid reactance terms originally involved in the single-machine stability criterion inequality. Preferably, the normalized equivalent reactance is converted into a dimension consistent with the single-machine linearized model according to the per-unit system and a unified reference voltage and reference power, and directly substituted into the criterion expression to ensure parameter consistency and comparability. Subsequently, based on this, the criterion is parameterized and simplified by combining the linear approximation constraint rules under the weak grid operating state. Specifically, the preferred approach is to: firstly, clarify the judgment conditions for the weak grid approximation and apply them to the model, that is, the maximum eigenvalue obtained by spectral analysis is located in the preset weak grid. The threshold range is used to treat the influence of network coupling on single-machine dynamics as a low-order disturbance term that can be approximated by a linear term, while ignoring or treating second-order and higher-order nonlinear or product terms related to network coupling as conservative margins; secondly, after substituting the normalized equivalent reactance, several network-related terms in the criterion are merged according to whether they contain a linear factor with a maximum eigenvalue, and the network terms multiplied by the virtual inertia, damping, and reactive power adjustment coefficients are extracted to form directly comparable equivalent network coupling coefficients, so as to reduce the originally complex multi-term interactions to a few quantities with clear physical meaning; thirdly, the maximum eigenvalue obtained from spectral analysis is compared with the values ​​under weak grid conditions. The reference impedance components are coupled and calculated using the same per-unit reference. Preferably, the amplification / scaling effect of the maximum eigenvalue on the equivalent reactance is described by a product or weighted product. This coupling result replaces the original grid reactance in the criterion, resulting in a simplified criterion that includes only internal unit parameters and a normalized network strength scalar. After parameter merging, explicit outputs are preferably provided for engineering operability. For example, the damping term should be greater than the sum of the virtual inertia reference term and the adjustment term determined by the reactive power regulation coefficient and the normalized equivalent reactance. The physical source and dimensional interpretation of each term should be noted so that engineers can directly verify and adjust them item by item. For ease of engineering... Verification and parameter tuning should preferably specify the order of parameter replacement, linear approximation and merging, the method for selecting the per-unit benchmark, and the conservative handling strategy for higher-order terms in the implementation process. It is also recommended to verify the predictive ability of the simplified criterion through small-disturbance modal analysis and time-domain transient simulation before practical application. If the difference between the simplified criterion prediction and the full-order model exceeds the allowable range, the weak network approximation threshold should be tightened or some neglected higher-order coupling terms should be compensated back to the criterion in the form of additional margin. As an example, taking the aforementioned ten-unit example: when the normalized equivalent reactance obtained after equivalence is about 1.5 per unit and the spectral analysis gives a maximum eigenvalue of 2.5, in engineering, the first step is to adjust this 1.5 per unit.The per-unit standard replaces the grid reactance term in the single-machine criterion. Following the weak grid approximation, only coupling terms linearly related to the maximum eigenvalue are retained, and these terms are merged with the reactive power regulation coefficient into a single network coupling coefficient. Substituting this into the baseline terms for virtual inertia and damping yields a directly verifiable stability threshold. If the test fails, the priority adjustment order is: increase the damping coefficient; adjust the reactive power regulation coefficient to reduce the product regulation term; and, under constrained conditions, increase the virtual inertia or reduce the normalized equivalent reactance or maximum eigenvalue through parallel compensation or line modification.

[0035] Furthermore, the linear approximation constraint rules under weak grid operating conditions include: Under the condition that the external grid connection impedance value is continuously greater than the total impedance of the wind farm collection network, the intrinsic reactance component of the transmission network included in the grid parameters is obtained as the basic reference quantity. Based on the topology of the wind farm's collector network, a network topology constant factor is generated, and a linear proportional constraint relationship is established between the network topology constant factor and the basic reference quantity. The linear proportional constraint relationship is used as the calculation criterion for the linear approximation constraint rule to generate the reference impedance component.

[0036] As a preferred embodiment of the above, the relative magnitudes of the external grid connection impedance and the total impedance of the wind farm collection network are first determined through measurement and identification methods before grid connection and during operation. Specifically, a preferred approach is to use methods such as short-circuit current testing, power flow calculation, and online voltage injection or response identification to obtain the equivalent impedance amplitude of the external grid at the grid connection point and the summed impedance amplitude of the collection network from each generator unit to the point of common coupling, respectively. The values ​​of the two are then continuously compared on the monitoring data to determine whether the weak network criterion of the external grid connection impedance being continuously greater than the total impedance of the collection network is met. When the weak network condition is met, it is preferable to use the transmission network model or existing... The intrinsic reactance component of the transmission network is extracted from the port data measured in the field and used as a basic reference quantity. This intrinsic reactance component can be obtained through spectral analysis of the transmission network admittance or impedance matrix. That is, modal analysis in the frequency domain is used to identify the impedance component with the greatest impact on the grid connection point and take its representative value. When selecting this reference quantity, the frequency range and per-unit standard are specified to ensure consistency with subsequent small-signal analysis. Secondly, based on the topology connection structure of the wind farm collection network, a network topology constant factor is generated to characterize the amplification or attenuation effect of the network structure on the equivalent impedance. Preferred methods for generating the topology constant factor include those based on node connectivity, number and length of lines, and branch connections. Factors such as connection status, transformer tapping or turns ratio, and bus segmentation degree are weighted and normalized. Specifically, the network can be abstracted into a topology at the port level. The relative reactance and length ratio of each branch, the degree of each node, and the network redundancy (whether it is a ring network or a multi-path parallel connection) are statistically analyzed. Then, these statistics are combined into a scaling factor according to a predetermined weighting rule. The larger the factor value, the more likely the topology makes the network to generate weak coupling or local cohesion modes, thus requiring the amplification of the base reference value to reflect the weak network effect. To ensure the linear operability of the rule, it is preferable to establish a linear ratio between the topology constant factor and the base reference value. The relationship serves as a linear approximation constraint, which means that the reference quantity is amplified or reduced in a manner proportional to the topology factor to obtain the reference impedance component. In this process, the initial selection and calibration method of the coefficients are given: the initial coefficients can be obtained by simulation statistics of typical topology cases. For example, by comparing the simulation results of multiple sets of network structures, the empirical factor intervals corresponding to ring networks, weak runoff networks, and strong parallel networks can be obtained. Then, the linear proportional coefficients are optimized to match the equivalent impedance value calculated by the full-order model well through field small disturbance tests or historical event playback. When optimizing, it is recommended to adopt a conservative strategy—if uncertain, take a factor slightly larger than the identification value to leave sufficient stability margin.To facilitate engineering applications and illustrative examples, a specific example is provided: Assume that during grid connection of a wind farm, a short-circuit test shows that the external grid connection impedance is significantly greater than the collection network impedance and meets the weak network criterion. Spectral analysis yields an intrinsic reactance representative value of 0.5 per unit for the transmission network. After topology statistics, the initial topology constant factor of the collection network is calculated to be 1.2, indicating the presence of longer branches and fewer parallel paths within the network. Therefore, the reference impedance component generated by coupling the base reference quantity with the topology constant factor according to linear proportional constraints is approximately the base reference quantity multiplied by this factor. In engineering practice, this reference impedance component is used as the reference impedance in subsequent weak network approximations. Simultaneously, periodic or event-triggered recalculation rules should be specified in implementation. When line modifications, grid connection structure changes, or external grid conditions change significantly, the recalculation should be performed, and the reference impedance component updated with a new topology constant factor and intrinsic reactance. Furthermore, it is recommended to conduct sensitivity analysis on the obtained reference impedance component to assess the impact of measurement errors and topology factor uncertainties on the final stability determination, and to introduce an adjustable conservative margin in the criterion to cover this uncertainty.

[0037] Furthermore, explicit stability criterion inequalities include: The maximum eigenvalue in the equivalent single-machine subsystem is correlated with the virtual inertia parameter and damping coefficient parameter to generate a power grid strength constraint relationship. Based on the weak grid approximation conditions, the grid reactance parameter in the single-machine stability criterion inequality is transformed into a network-dependent stability boundary variable. The power grid strength constraint relationship is integrated into the network-dependent stability boundary variable expression to form a three-dimensional stability criterion structure that includes virtual inertia boundary control terms, damping coefficient adjustment terms, and network strength dominant terms.

[0038] As a preferred embodiment of the above, after completing the aforementioned spectral analysis and equivalence steps, the normalized maximum eigenvalue and equivalent external grid reactance parameters are obtained, and they are standardized according to a unified per-unit reference and frequency range. Subsequently, based on the normalized maximum eigenvalue and unit parameters, a stable mapping relationship is generated using modal scanning and numerical testing methods. Specifically, in engineering, the discretization range of virtual inertia and damping coefficient is pre-defined, preferably with virtual inertia ranging from 1 to 5 seconds and damping coefficient ranging from 0.3 to 3, to cover common engineering scenarios. Under each pair of inertia and damping values, the normalized maximum eigenvalue is substituted into the equivalent single-machine model according to several representative values, and the Routh criterion or eigenvalue analysis is used to determine whether the system meets the small-signal stability requirement, thereby obtaining the three-dimensional... The stability or instability marker is determined at each point in the parameter space. A continuous boundary surface is obtained by applying smooth regression or surface fitting to the resulting discrete stability boundary. This boundary surface is the explicit representation of the network-dependent stability boundary variable. The network strength term is characterized by the normalized maximum eigenvalue or by a scalar obtained by coupling the maximum eigenvalue with the reference impedance component. The model explicitly states that this term monotonically tightens the acceptable lower limit of inertia and damping as the maximum eigenvalue increases, forming the so-called grid strength constraint relationship. Based on the weak grid approximation condition, a linear approximation order reduction is preferred for online engineering applications: the complex network term is linearized into a network-dependent stability boundary variable that can be directly multiplied and compared with inertia or damping. This variable is represented by the maximum eigenvalue. The linearly coupled boundary variables with the reference impedance components and topological constant factors according to a unified per-unit rule are obtained, and these linearized boundary variables are integrated into the criterion expression. The resulting explicit stability criterion consists of three identifiable terms: a virtual inertia boundary control term reflecting the amplification or weakening effect of inertia on the reference term; a damping coefficient adjustment term directly reflecting the constraint on modal damping; and a network strength dominant term, i.e., network constraint dominated by the maximum eigenvalue. These three constitute a three-dimensional stability discrimination structure that is easy for engineers to visually verify. The preferred engineering process for generating and applying this structure is as follows: first, obtain the boundary surface through numerical parameter scanning and save it as a lookup table or parameterized expression; then, during grid connection access or operational monitoring, use the current virtual inertia, damping, and the maximum eigenvalue obtained through real-time calculation or periodic identification as the input. The system checks whether the current operating point is within the boundary surface and has a preset margin. If the condition is not met, measures are taken according to priority: first, increase the damping coefficient; second, adjust the reactive power loop or allocate more virtual inertia to sensitive units; and third, improve grid-side conditions, such as parallel compensation or line modification to reduce the maximum eigenvalue. For ease of engineering understanding, an example is given: In a sample wind farm consisting of ten units, the normalized maximum eigenvalue obtained from spectral analysis is 2.5 and the equivalent normalized external grid reactance is 1.5 per unit. Based on the boundary surface obtained from the previous numerical parameter scan, the minimum safe damping corresponding to a virtual inertia of 2 seconds is approximately 1.2. In engineering, an additional 20% margin should be added, so the actual damping setting is recommended to be greater than approximately 1.44. If the current damping is only 1...A result of 0 indicates that the criterion is not met and triggers adjustment suggestions. During implementation, it is recommended to use sufficient parameter resolution and robust numerical fitting methods, such as regression band regularization or spline fitting, for boundary surface generation to avoid overfitting and to provide conservative treatment for approximate regions. Furthermore, before applying the criterion, its predictive ability should be verified through small-disturbance modal analysis, time-domain transient simulation, and necessary on-site small-disturbance tests. This three-dimensional criterion should be integrated into grid connection access checks, online monitoring, and event response strategies to achieve dynamic updates. For example, when grid-side conditions or topology change, the maximum eigenvalue should be recalculated and the stable boundary refreshed, ensuring that the criterion is both explicitly operable and conservative and adaptive during engineering operation.

[0039] Example 2; Based on the same inventive concept as the small-signal steady-state discrimination method for virtual synchronous machine-controlled doubly-fed wind farms in the foregoing embodiments, this invention also provides a small-signal steady-state discrimination system for virtual synchronous machine-controlled doubly-fed wind farms, the system comprising: The model building module establishes a linearized state-space model for a doubly fed wind turbine controlled by a single virtual synchronous machine, including virtual inertia, damping coefficient, and grid parameters. The Routh derivation module is based on the linear state-space model to write the characteristic polynomial and applies the Routh criterion to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane. The full-order matrix module constructs a full-order state-space model for a doubly fed wind farm with M units, including dynamic coupling between units and the impedance of the collector network. The diagonal operation module decomposes the full-order state space matrix into M independent subsystems according to the block diagonalization operation, and performs eigenvalue decomposition on the collector network impedance matrix to obtain the maximum eigenvalue as the equivalent network strength characterization. The equivalent system module transforms M independent subsystems into M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid. The criterion construction module performs Routh criterion derivation on the equivalent single-machine subsystem, and obtains the explicit stability criterion inequality of the doubly fed wind farm grid-connected system by combining the maximum eigenvalue and the weak grid approximation condition. The stability discrimination module inputs the real-time collected power grid strength characteristic values ​​and wind turbine control parameters into an explicit stability inequality for online discrimination. When the inequality is true, it outputs a system stability signal, triggering normal wind turbine operation; when the inequality is false, it outputs an instability warning signal, triggering adaptive adjustment of virtual inertia or damping coefficient.

[0040] The adjustment system described above in this invention can effectively realize the small-signal steady-state discrimination method for virtual synchronous machine control of doubly fed wind farms. The technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0041] Example 3; A computer-readable storage medium is characterized in that it stores a computer program, the computer program including program instructions, which, when executed by a processor, can realize a small-signal steady-state discrimination method for a doubly-fed wind farm controlled by a virtual synchronous machine.

[0042] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0043] 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 small-signal steady-state discrimination in a doubly-fed wind farm controlled by a virtual synchronous machine, characterized in that, The method includes: A linearized state-space model including virtual inertia, damping coefficient, and grid parameters is established for a doubly fed wind turbine controlled by a single virtual synchronous machine. Based on the linearized state-space model, the characteristic polynomial is written, and the Routh criterion is applied to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane. For a doubly fed wind farm with M generating units, construct a full-order state-space matrix that includes dynamic coupling between generating units and the impedance of the collector network; The full-order state space matrix is ​​decomposed into M independent subsystems according to the block diagonalization operation, and the impedance matrix of the collector network is decomposed into eigenvalues ​​to obtain the maximum eigenvalue as the equivalent network strength characterization. The M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid; The Routh criterion is derived by performing the equivalent single-machine subsystem on the equivalent single-machine subsystem. Combined with the maximum eigenvalue and the weak grid approximation condition, the explicit stability criterion inequality of the doubly fed wind farm grid-connected system is obtained. The real-time collected power grid strength characteristic value and wind turbine control parameters are input into the explicit stability criterion inequality for online discrimination; when the inequality is true, a system stability signal is output, triggering normal wind turbine operation; when the inequality is false, an instability warning signal is output, triggering adaptive adjustment of the virtual inertia or the damping coefficient.

2. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 1, characterized in that, The parameters of the linearized state-space model include: The virtual inertia parameters and damping coefficient parameters defined in the virtual synchronous machine control loop; The resistance and inductance characteristics of the stator winding of the doubly fed wind turbine; Equivalent grid reactance parameters from the wind turbine connection point to the grid common coupling point; The reactive power control loop of the virtual synchronous machine is used to adjust the reactive power regulation coefficient parameter for regulating voltage deviation.

3. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 2, characterized in that, The single-machine stability criterion inequality includes: The damping coefficient parameter must be greater than the sum of the linear reference term based on the virtual inertia parameter and the product adjustment term based on the reactive power adjustment coefficient parameter and the equivalent grid reactance parameter; The proportional relationship of the linear reference term is determined by the resistance characteristic parameter of the stator winding, and the weighting relationship of the product adjustment term is determined by the inductance characteristic parameter of the stator winding.

4. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 1, characterized in that, Eigenvalue decomposition of the collector network impedance matrix and obtaining the maximum eigenvalue include: The collector network impedance matrix obtained by the block diagonalization operation is subjected to a symmetry transformation to obtain a symmetric matrix; The symmetric matrix is ​​subjected to spectral analysis to obtain all eigenvalues ​​and generate an eigenvalue set structure; The feature component with the largest value is selected from the feature value set structure as the power grid strength characterization.

5. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 1, characterized in that, The M independent subsystems are equivalent to M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem, including: The reactive coupling path of the inter-unit interaction corresponding to the M-1 isolated subsystems is replaced with a static zero-impedance coupling model to achieve electrical decoupling; Based on the electrical decoupling state, the virtual inertia parameters and damping coefficient parameters of the original wind turbine system are kept unchanged for the equivalent single-machine subsystem. The maximum eigenvalue is coupled with the equivalent grid reactance parameter to perform network strength calculation, thereby generating the normalized equivalent reactance characterization parameter at the external grid port of the equivalent single-machine subsystem. The normalized equivalent reactance characterization parameters are used to replace the grid reactance parameters to complete the grid interface modeling of the equivalent single-machine subsystem.

6. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 5, characterized in that, The Routh criterion is derived for the equivalent single-machine subsystem, including: The normalized equivalent reactance characterization parameters completely replace the original grid reactance parameter terms included in the single-machine stability criterion inequality; Based on the above, the single-machine stability criterion inequality is simplified by parameter merging based on the linear approximation constraint rule of the collector network impedance parameters under weak grid operation conditions. The maximum eigenvalue and the reference impedance component under weak grid conditions are imported into the simplified single-machine stability criterion inequality, and the explicit stability criterion inequality of the doubly-fed wind farm grid-connected system is output.

7. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 6, characterized in that, Linear approximation constraint rules under weak grid operating conditions include: Under the condition that the external grid connection impedance value is continuously greater than the total impedance of the wind farm collection network, the intrinsic reactance component of the transmission network included in the grid parameters is obtained as a basic reference quantity. Based on the topology of the wind farm's power collection network, a network topology constant factor is generated, and a linear proportional constraint relationship is established between the network topology constant factor and the basic reference quantity. The linear proportional constraint relationship is used as the calculation criterion for the linear approximation constraint rule to generate the reference impedance component.

8. The method for small-signal steady-state discrimination of a doubly-fed wind farm controlled by a virtual synchronous machine according to claim 1, characterized in that, The explicit stability criterion inequalities include: The maximum eigenvalue in the equivalent single-machine subsystem is correlated with the virtual inertia parameter and damping coefficient parameter to generate a power grid strength constraint relationship by performing a stability correlation mapping operation. Based on the aforementioned weak grid approximation conditions, the grid reactance parameters in the single-machine stability criterion inequality are transformed into network-dependent stability boundary variables. The power grid strength constraint relationship is integrated into the network-dependent stability boundary variable expression to form a three-dimensional stability criterion structure that includes a virtual inertia boundary control term, a damping coefficient adjustment term, and a network strength dominant term.

9. A small-signal steady-state discrimination system for a doubly-fed wind farm controlled by a virtual synchronous machine, characterized in that, The system includes: The model building module establishes a linearized state-space model for a doubly fed wind turbine controlled by a single virtual synchronous machine, including virtual inertia, damping coefficient, and grid parameters. The Routh derivation module is based on the linear state-space model to write the characteristic polynomial and applies the Routh criterion to derive the single-machine stability criterion inequality that makes the characteristic roots lie in the left half of the complex plane. The full-order matrix module constructs a full-order state-space model for a doubly fed wind farm with M units, including dynamic coupling between units and the impedance of the collector network. The diagonal operation module decomposes the full-order state space matrix into M independent subsystems according to the block diagonalization operation, and performs eigenvalue decomposition on the collector network impedance matrix to obtain the maximum eigenvalue as the equivalent network strength characterization. The equivalent system module transforms M independent subsystems into M-1 isolated subsystems without network interaction and one equivalent single-machine subsystem that interacts with the power grid. The criterion construction module performs Routh criterion derivation on the equivalent single-machine subsystem, and obtains the explicit stability criterion inequality of the doubly fed wind farm grid-connected system by combining the maximum eigenvalue and the weak grid approximation condition. The stability discrimination module inputs the real-time collected power grid strength characteristic values ​​and wind turbine control parameters into an explicit stability inequality for online discrimination. When the inequality is true, it outputs a system stability signal, triggering normal wind turbine operation; when the inequality is false, it outputs an instability warning signal, triggering adaptive adjustment of virtual inertia or damping coefficient.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the small-signal steady-state discrimination method for virtual synchronous machine-controlled doubly-fed wind farms as described in any one of claims 1-8.

Images