A wind farm stability judgment method and system based on stable domain quantification

Abstract

The application discloses a wind farm stability judgment method and system based on a stable domain quantification, relates to the technical field of wind farm stability judgment, and comprises the following steps: uniformly quantifying an amplitude margin and a phase margin; constructing an affine partition of the wind farm, and screening out candidate partitions for stability judgment; in the candidate partitions, an inner point method based on multiple starting points is used to solve the amplitude margin and the phase margin, the minimum quantity between the amplitude margin and the phase margin is taken as a stability margin, a stability margin sample set is constructed, boundary sample points in the stability margin sample set are selected, and thus, a stability boundary and a stability domain of the wind farm are determined, and a stability judgment result of the wind farm is obtained. The application solves the problems of poor calculation performance and high complexity of online wind farm stability domain calculation, and is used for online stability judgment and field control guidance of the wind farm.

Description

Technical Field

[0001] This invention relates to the field of wind farm stability assessment technology, and in particular to a method and system for assessing wind farm stability based on stability domain quantization. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The impedance method uses the generalized Nyquist stability criterion to determine the stability of the wind farm at a given operating point. Absolute stability is determined by graphically observing whether the Nyquist curve encircles the critical point (-1, j0). However, the Middlebrook analysis method has the problem of being too conservative in terms of the stability margin of the Nyquist curve. Although the gain margin and phase margin methods reduce the conservatism, the stability margin of the system can only be measured by judging the stability margin of the system by judging the gain margin and phase margin separately. Both sets of indicators must be used together to measure the stability margin of the system.

[0004] As the operating point changes, two sets of indicators may change in opposite directions, making it impossible to determine a uniform quantitative stability margin at a given operating point. Based on the stability at a given operating point, existing research investigates the stability domain under different parameters and operating points to guide wind farm operation and control. Series compensation is considered a significant cause of subsynchronous oscillations, and a method of calculating stability at different series compensation values ​​point-by-point is used to establish a stability domain for these values. While the point-by-point method is feasible for three-variable cases like series compensation, the computational complexity increases exponentially with the number of parameters when establishing a multi-parameter stability domain. The impact of PI parameters of the rotor converter and the phase-locked loop on the stability domain is also analyzed.

[0005] The above studies all rely on the assumption that the grid impedance is known and can be equivalent to an RLC (Representational Logic Cell) form, enabling offline point-by-point stability calculations. However, in actual operation, the grid impedance changes over time and cannot be simply simplified using RLC; it is impossible to traverse all the dynamics of the grid offline. Therefore, it is necessary to calculate and construct the stability region of the wind farm online based on online measured grid impedance.

[0006] To establish an online, multi-parameter wind farm stability domain, the method of calculating the stability boundary point-by-point can be transformed into a classification problem of whether the operating point is stable or unstable, thus significantly reducing the computational load. Support Vector Machines (SVMs) are easy to model and are a classification method well-suited for complex and high-dimensional data. SVMs are typically trained offline, with the classification results applied online to predict the system's state. However, in SVM training, for some heavily interleaved and high-dimensional sample sets, finding the optimal hyperplane requires considering every sample point, thus solving an optimization problem. This leads to a complex classification process, high computational burden, and makes it difficult to apply to online training. Summary of the Invention

[0007] To address the aforementioned issues, this invention proposes a method and system for judging the stability of wind farms based on stability domain quantization. This method solves the problems of poor stability domain computation performance and high complexity in online wind farms, and is used for online stability judgment and field control guidance of wind farms.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for determining the stability of a wind farm based on stability domain quantization, comprising:

[0010] Define the gain margin and phase margin for uniform quantization under the same unit;

[0011] Construct piecewise affine partitions of the wind farm, determine the position of the Nyquist curve corresponding to the partition based on the eigenvalues ​​of the complex state variables at the endpoints of the partition, and thereby screen out candidate partitions for stability assessment.

[0012] In the candidate partitions, the magnitude margin and phase margin are solved using an interior-point method based on multiple starting points. The minimum value between the magnitude margin and the phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The interior-point method based on multiple starting points selects several optimization initial points within the constraints of the complex state variables of the candidate partitions, and obtains the smallest global optimal solution by comparing different solutions.

[0013] A stability margin sample set is constructed, and boundary sample points in the stability margin sample set are selected to determine the stability boundary and stability domain of the wind farm, thereby obtaining the stability judgment result of the wind farm.

[0014] As an alternative implementation, the unified quantization is as follows: the left real axis of (-1, j0) is defined as the forbidden region, and the gain margin and phase margin are uniformly represented by the distance from the Nyquist curve to the forbidden region; wherein, the gain margin is in the right part of the forbidden region, the phase margin is in the upper part of the forbidden region, and the distance from the minimum real axis intersection point of the Nyquist curve to the forbidden region is the gain margin or the phase margin.

[0015] As an alternative implementation method, the candidate partition selection process includes: for magnitude margin, classifying the partitions according to the characteristic values ​​of complex state variables in the current and previous partitions, based on whether they have opposite signs, are the same positive, or are the same negative, specifically:

[0016] If the eigenvalues ​​of the complex state variables at both ends of the current partition are located above and below the real axis, then the filtering condition is:

[0017] Im(λ m,1,2 (s m Im(λ) m,1,2 (s m+1 ))≤0

[0018] If the eigenvalues ​​of the complex state variables at both ends of the current partition are all above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are all below the real axis; or, if the eigenvalues ​​of the complex state variables at both ends of the current partition are all below the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are all above the real axis, then the filtering condition is:

[0019] Im(λ m,1,2 (s m Im(λ) m-1,1,2 (s m ))≤0

[0020] Where, λ m,1,2 (s m ), λ m-1,1,2 (s m ) represents the Nyquist curves in the m-th and (m-1)-th partitions.

[0021] As an alternative implementation, if the eigenvalues ​​of the complex state variables at both ends of the current partition are both above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are both above the real axis, then it is considered that there is no intersection between the two partitions.

[0022] If the eigenvalues ​​of the complex state variables at both ends of the current partition are both above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are respectively above and below the real axis, then there is an intersection point in the previous partition, so the judgment is not repeated.

[0023] As an alternative implementation, the candidate partition selection process includes: for phase margin, the selection criteria are:

[0024]

[0025] Where, λ m,1,2 (s) represents the Nyquist curve in the m-th partition, and L represents the set filtering range.

[0026] As an alternative implementation method, the process of solving for the gain margin and phase margin using the interior-point method based on multiple starting points includes:

[0027] The optimization function established for the gain margin is as follows:

[0028] s G,m =arg min s {Re(λ m,1,2 (s))+w*Im(λ m,1,2 (s))}

[0029] The optimization function established for phase margin is as follows:

[0030]

[0031] Where t is the penalty coefficient, w is the penalty coefficient, and λ is the penalty coefficient. m,1,2 (s) is the Nyquist curve in the m-th partition.

[0032] As an alternative implementation method, the process of selecting boundary sample points from the stability margin sample set to determine the stability boundary and stability region of the wind farm specifically includes:

[0033] Determine the Euclidean distance between each pair of samples, calculate the k-nearest neighbors of the same class and the k-nearest neighbors of different classes for each sample point, and thus determine the average distance between the sample point and the two classes of samples.

[0034] Boundary sample points are extracted based on the average distance between the two types of samples and a set stability threshold, and non-boundary points are deleted.

[0035] For the remaining stability margin sample set, based on the k-nearest samples of the same type for each sample point, if the distance is less than a given threshold, then the sample set is reduced.

[0036] Based on boundary sample points, support vector machines are used to determine the stability boundary and stability region of the wind farm.

[0037] Secondly, the present invention provides a wind farm stability assessment system based on stability domain quantization, comprising:

[0038] The quantization module is configured to define the gain margin and phase margin of uniform quantization under the same unit;

[0039] The filtering module is configured to construct piecewise affine partitions of the wind farm, and determine the position of the Nyquist curve corresponding to the partition based on the characteristic values ​​of the complex state variables at the endpoints of the partition, thereby filtering out candidate partitions for stability assessment.

[0040] The solution module is configured to solve for the gain margin and phase margin in the candidate partitions using a multi-starting-point interior-point method. The minimum value between the gain margin and phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The multi-starting-point interior-point method involves selecting several optimization initial points within the constraints of the complex state variables of the candidate partitions and obtaining the smallest global optimal solution by comparing different solutions.

[0041] The judgment module is configured to construct a stability margin sample set, select boundary sample points in the stability margin sample set, thereby determining the stability boundary and stability domain of the wind farm and obtaining the stability judgment result of the wind farm.

[0042] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0043] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention addresses the problems of poor computational performance and high complexity in online wind farm stability domain calculations by proposing a wind farm stability assessment method and system based on stability domain quantization for online stability assessment and wind farm control guidance. First, a simplified wind farm stability criterion is proposed, unifying the quantification of stability margin. Second, based on affine model partitioning, regions with potentially minimum stability margins are selected, avoiding the need to solve for stability margins across the entire frequency band. Within the selected partitions, a multi-starting-point interior-point method is used to solve the simplified wind farm stability criterion online, offering significant advantages in global optimization and computational speed. This allows for the online calculation of stability margins at different operating points to form a sample set. Finally, a k-nearest neighbor support vector machine (SVM) is used to construct the wind farm's stability boundary and stability domain. The k-nearest neighbor method selects boundary sample points from the sample set for SVM training, significantly reducing the computational burden of online stability domain construction.

[0046] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0047] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0048] Figure 1 This is a flowchart of the wind farm stability judgment method based on stability domain quantization provided in Embodiment 1 of the present invention;

[0049] Figure 2 This is a generalized Nyquist plot provided in Embodiment 1 of the present invention;

[0050] Figure 3 This is a schematic diagram of the generalized Nyquist curve criterion provided in Embodiment 1 of the present invention;

[0051] Figure 4 This is a schematic diagram of a wind farm zoning set provided in Embodiment 1 of the present invention;

[0052] Figure 5 (a)- Figure 5 (d) is a schematic diagram of the Nyquist curve partition endpoints provided in Embodiment 1 of the present invention;

[0053] Figure 6 This is a schematic diagram of stable and unstable samples provided in Embodiment 1 of the present invention;

[0054] Figure 7 This is a schematic diagram of the single-unit wind speed and power stability domain provided in Embodiment 1 of the present invention;

[0055] Figure 8 This is a time-domain simulation diagram of the wind speed and power changes of a single unit provided in Embodiment 1 of the present invention;

[0056] Figure 9 This is a schematic diagram of the single-machine power variation stability region provided in Embodiment 1 of the present invention;

[0057] Figure 10 This is a schematic diagram of the time-domain simulation of the active power variation of a single machine provided in Embodiment 1 of the present invention;

[0058] Figure 11 This is a schematic diagram of the stability region of wind speed and power variation of a single wind turbine in a wind farm provided in Embodiment 1 of the present invention;

[0059] Figure 12 This is a schematic diagram of the stability region of wind speed and power variation of a single wind turbine in a wind farm provided in Embodiment 1 of the present invention;

[0060] Figure 13 This is a schematic diagram of the total power of the wind farm and the power stability region of wind turbine 1 provided in Embodiment 1 of the present invention;

[0061] Figure 14This is a time-domain simulation diagram of the total power of the wind farm and the power of wind turbine 1 provided in Embodiment 1 of the present invention. Detailed Implementation

[0062] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0063] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, 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.

[0064] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0065] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0066] Example 1

[0067] This embodiment provides a method for judging the stability of wind farms based on stability domain quantization, such as... Figure 1 As shown, it includes:

[0068] Define the gain margin and phase margin for uniform quantization under the same unit;

[0069] Construct piecewise affine partitions of the wind farm, determine the position of the Nyquist curve corresponding to the partition based on the eigenvalues ​​of the complex state variables at the endpoints of the partition, and thereby screen out candidate partitions for stability assessment.

[0070] In the candidate partitions, the magnitude margin and phase margin are solved using an interior-point method based on multiple starting points. The minimum value between the magnitude margin and phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The interior-point method based on multiple starting points involves selecting several optimization initial points within the constraints of the complex state variables of the candidate partitions, and obtaining the smallest global optimal solution by comparing different solutions.

[0071] A stability margin sample set is constructed, and boundary sample points in the stability margin sample set are selected to determine the stability boundary and stability domain of the wind farm, thereby obtaining the stability judgment result of the wind farm.

[0072] This embodiment first introduces the GMPM stability criterion for the generalized Nyquist curve, and then proposes a simplified stability criterion and a method for quantifying the stability margin based on this criterion.

[0073] For determining the stability of a wind farm at a specific operating point, the generalized Nyquist curve is calculated as follows:

[0074]

[0075] Among them, Z grid (s) is the wind farm impedance, Z farm (s) represents the grid impedance, E is the identity matrix, and s is the complex state variable.

[0076] Under different values ​​of s, there are different solutions λ. 1,2 Therefore, λ 1,2 The values ​​of can form a curve, also called the Nyquist curve. The traditional generalized Nyquist stability curve is as follows: Figure 2 As shown, the eigenvalue λ of the Nyquist curve across the entire frequency band is determined graphically. 1,2 Whether the area encloses (-1, j0) is used to determine the stability of the wind farm. Figure 2 The Nyquist curve in the left subgraph does not enclose the critical point, indicating that the system is stable. However, this method cannot quantify the stability margin; it can only determine whether the system is stable or unstable.

[0077] To quantify stability margin, the GMPM criterion defines both gain margin and phase margin, such as... Figure 3 As shown, the boundary of the stable region is defined, that is, as long as λ 1,2 If the eigenvalue locus at any frequency s does not enter the unstable region, it can be guaranteed that the eigenvalue locus does not cross the left real axis of -, thus not encircling the critical point (-1, j0).

[0078] Based on the definition of the stable region, the stability boundaries of this criterion are the magnitude margin. And phase margin PM, so the gain margin and phase margin can be calculated separately based on the Nyquist curve; however, the determination of gain margin and phase margin is independent and based on different unit definitions. The calculation of the two stability margins cannot determine a unified stability margin at the current operating point. In actual use, there is no definite basis for whether to use gain margin or phase margin, and the two are not comparable because their units are different.

[0079] To unify and quantify stability margin and phase margin, and to provide a unified margin index for a given operating point, this embodiment defines simplified stability margin and phase margin under the same unit.

[0080] First, the left real axis of (-1, j0) is defined as the forbidden region. If the Nyquist curve crosses the forbidden region, then both the gain margin and phase margin should be considered absolutely unstable. The range of the forbidden region is:

[0081]

[0082] Here, ε approaches 0 infinitely, representing the height of the forbidden region.

[0083] Secondly, to avoid using different units for phase margin and gain margin, which would prevent the stability margin from being unified, the simplified gain margin and phase margin proposed in this embodiment are uniformly represented by the distance from the Nyquist curve to the forbidden region.

[0084] The simplified gain margin is calculated from right to left on the right side of the forbidden region. Specifically, it is the distance from the Nyquist curve to the forbidden region on the real axis, which is the distance from the least real axis intersection point of the Nyquist curve to the forbidden region.

[0085] M G =min(Re(λ) 1,2 (s))+1)subject to:Im(λ 1,2 (s))=0 (3)

[0086] Among them, M G For the sake of simplification, a value greater than zero indicates stability, and a larger value indicates a greater stability margin.

[0087] The simplified phase margin is the margin of distance from the forbidden region from top to bottom in the upper part of the forbidden region. Specifically, it is the distance from the Nyquist curve to the forbidden region on the real axis, which is the distance from the least real axis intersection point of the Nyquist curve to the forbidden region:

[0088] M P =min(Im(λ) 1,2 (s)))subject to:Re(λ 1,2 (s))<-1 (4)

[0089] Among them, M P For the sake of simplification, a phase margin greater than zero indicates stability, and a larger value indicates a greater stability margin.

[0090] After determining M respectively G and M PSince both margins are defined in the same unit, their sizes can be compared, and the smaller margin can be used to represent the current stability margin M. GP :

[0091] M GP =min(M) G M P (5)

[0092] When M GP When M is less than 0, it indicates that the Nyquist curve has crossed the forbidden region, and the system is unstable; the smaller the value, the more unstable the system. GP When the value is greater than 0, it indicates that the Nyquist curve does not cross the forbidden region and the system is stable. The larger the value, the greater the stability margin.

[0093] In this embodiment, a simplified stability criterion is applied to the wind farm impedance based on a piecewise affine model. First, an affine partition of the wind farm in the s-domain is established, and a screening method for the piecewise affine partition is proposed to select candidate partitions that may be determined to be stable. Second, within the candidate partitions, a simplified gain margin and phase margin are solved using an interior-point method based on multiple starting points.

[0094] In this embodiment, an affine partition of the wind farm is constructed at a given operating point; given the operating points x of all wind turbines... farm ={x DFIG,1 ,...,x DFIG,k ,...,x DFIG,K}, where the subscript k represents the k-th wind turbine; for any wind turbine k, the partitioning condition of its piecewise affine model is:

[0095] χ k,n ={F k,n [x DFIG,k ,s] T +g k,n ≤0} (6)

[0096] Where the subscript n represents the nth partition;

[0097] The boundary conditions for a single wind turbine then become a first-order inequality in s. By combining the boundary conditions of all wind turbines and recombining the boundaries, we obtain the affine partition χ of the wind farm in the s dimension. F,1 ,...,χ F,m ,...,χ F,M ,like Figure 4 As shown, each wind farm zone χ F,m Different zone combinations are selected corresponding to the impedance of a single unit. The subscript m indicates the m-th zone of the wind farm. The wind farm has a total of M zones. χ F,m The set is:

[0098] χ F,m={s|s m ≤s≤s m+1} (7)

[0099] Among them, s m and s m+1 The state variables represent the start and end points within the m-th partition.

[0100] Finally, a piecewise affine model of s is obtained.

[0101] In this embodiment, since the set of partitions results in a large number of partitions, the partitions are screened before being applied to the stability criterion. Since the eigenvalues ​​of the complex state variables s at the endpoints of the partitions are known, the location of the partitions is determined based on the eigenvalues ​​of the two endpoints s of the partitions in order to quickly determine the location of the partitions. Candidate partitions are screened based on the location of the Nyquist curve corresponding to the partitions, avoiding the need to calculate eigenvalues ​​point by point across the entire frequency band for stability margin determination, reducing the computational load of solving the stability criterion, and improving the computational speed.

[0102] For the simplified gain margin, it is necessary to find regions where real axis intersections may exist. The simplest way to determine this is that the eigenvalues ​​of 's' at both ends of the partition are located above and below the real axis, respectively. Figure 5 As shown in (a), there must be intersection points within the partitions; however, the Nyquist curve obtained by piecewise affine is divided according to the partitions, and there may be cases where the curve actually crossing zero is broken between two adjacent partitions, such as... Figure 5 As shown in (b), it is not enough to judge the current region alone; it is necessary to combine the feature values ​​of the two ends s of the previous region for joint judgment.

[0103] Based on the feature values ​​of the current partition and the previous partition, and categorized according to different signs, same positive, and same negative within the partition, the following detailed analysis is performed:

[0104] (1) The eigenvalues ​​of the two ends s of the current region are located above and below the real axis, so no matter what the previous region is, there must be zero crossings in these two regions. This includes 3 cases. Figure 5 (a) is one such case. The determination condition is:

[0105] Im(λ m,1,2 (s m Im(λ) m,1,2 (s m+1 ))≤0 (8)

[0106] (2) The eigenvalues ​​of both ends of the current region s are above the real axis, while the eigenvalues ​​of both ends of the previous region s are above the real axis. At this time, it is considered that there is no intersection between the two regions.

[0107] (3) The eigenvalues ​​of both ends of the current region s are above the real axis, while the eigenvalues ​​of both ends of the previous region s are above and below the real axis respectively. At this time, there must be an intersection point in the previous region. However, the first case has already been satisfied when the previous region was judged, and it has been judged once. Therefore, the current judgment will not be repeated.

[0108] (4) The eigenvalues ​​of both ends 's' of the current region are above the real axis, while the eigenvalues ​​of both ends 's' of the previous region are below the real axis, such as... Figure 5 (b)- Figure 5 (c) It is determined that there may be an intersection point between the two areas; the determination condition is:

[0109] Im(λ m,1,2 (s m Im(λ) m-1,1,2 (s m ))≤0 (9)

[0110] (5) The case where the eigenvalues ​​of both ends of the current partition are below the real axis is exactly the same as (2)-(4), so I will not repeat it. The case where the eigenvalues ​​of both ends of the current partition are below the real axis is the same as the case where the eigenvalues ​​of both ends of the previous partition are above the real axis. We only need to use the same judgment formula.

[0111] For the simplified phase margin, determine whether the distance between the eigenvalues ​​at both ends of the partition 's' and the real axis is within the filtering range L, such as... Figure 5 As shown in (d), the initial value of the filtering range L is 2. If the Nyquist curve does not enter this range, the stability is very strong. If no partition can be found within the range, L needs to be expanded.

[0112]

[0113] Where, λ m,1,2 (s) represents the Nyquist curve in the m-th partition.

[0114] In summary, based on the endpoint symbols of the current partition and the previous partition, regions with potentially small amplitude margins are filtered according to equations (8)-(9), with specific filtering criteria shown in Table 1; based on the endpoint symbols of the current partition, regions with potentially small phase margins are filtered according to equation (10); therefore, among all M partitions of s, only the filtered c partitions χ need to be filtered. F,c The internal calculation stability criterion avoids solving point by point across the entire frequency band, thus reducing the amount of computation.

[0115] Table 1. Gradient Margin Zoning Screening Table

[0116]

[0117] In a partitioned frequency band χF,m Although the piecewise affine doubly-fed wind turbine impedance model is linear, solving the Nyquist curve, after calculating the series and parallel impedances of the wind farm and the eigenvalues ​​of the Nyquist curve, is a nonlinear optimization problem involving equality and inequality constraints. For the simplified gain margin calculation, the Nyquist curve may cross the real axis multiple times within the region, potentially leading to local optima. Similarly, for the phase margin calculation, local optima may also exist within the region. While a point-by-point traversal method could find the optimal solution by iterating through all frequencies within the region, the computational load is enormous, making online application difficult.

[0118] To accelerate the solution process and avoid getting trapped in local optima, this embodiment employs a multi-starting-point interior-point method to calculate the gain margin and phase margin. The multi-starting-point interior-point method is a globally optimal algorithm, offering a significant advantage in computational speed compared to the traversal method, which aids in determining the stability of online applications.

[0119] For the gain margin criterion in equation (3), the equality constraint Im(λ) is applied. m,1,2 Write (s))=0 into the cost function, and establish the optimization problem of equation (11):

[0120] s G,m =arg min s {Re(λ m,1,2 (s))+w*Im(λ m,1,2 (s))} (11)

[0121] Where w is the penalty coefficient, which takes the value of 100. By increasing the cost of the imaginary part of the eigenvalue, the imaginary part of the eigenvalue is forced to be zero or as close to 0 as possible.

[0122] For the phase margin criterion in equation (4), the inequality constraint Re(λ) is applied. m,1,2 Write (s)<-1 into the cost function, and establish the optimization problem of equation (12):

[0123]

[0124] Where t is the penalty coefficient, taking a value of 0.5, by making Re(λ m,1,2 When (s) approaches -1, the numerical value of the penalty term increases exponentially, forcing the optimization equation to exist in the interval of inequality constraints.

[0125] Based on the optimization problem of equations (11)-(12), the KKT equation system is solved using the gradient descent method, and the solution within the range is obtained based on the initial optimization point. To avoid getting trapped in local optima, this embodiment selects several initial optimization points within the complex state variable constraint range of the candidate partition using a multi-starting point approach, and compares different solutions to find the smallest global optimal solution s. G,m and s P,m Furthermore, considering the computational speed of online solutions, multiple starting points can lead to longer solution times and slower convergence. Therefore, this embodiment adopts a parallel computing method to further improve the online computational speed.

[0126] The solution to the optimization problem based on gain margin and phase margin G,m and s P,m M G,m and M P,m :

[0127] M G,m =Re(λ) 1,2 (s G,m ))+1 (13)

[0128] M P,m =Im(λ 1,2 (s P,m (14)

[0129] Determine the minimum of the gain margin and phase margin to obtain the current partition χ. F,m stability margin M within GP,m :

[0130] M GP,m =min(M) G,m M P,m (15)

[0131] Then use all candidate partitions χ F,c The minimum value of the stability margin is taken as the current operating point x. farm Stability margin M GP :

[0132] M GP (x farm ) = min m∈{1,2...c...C} M GP,m (16)

[0133] This embodiment is based on the stability margin of the operating point. First, a stability domain dataset is established. Second, to improve the online computing speed and accelerate the classification optimization process of the support vector machine, k-nearest neighbor is used to select boundary sample points in the stability domain dataset. Finally, based on the boundary sample points, the support vector machine is used to establish the stability domain and stability boundary of the system, so that the stability of the wind farm can be judged.

[0134] There are many operating points in a wind farm, and changes in these operating points will affect the stability margin M. GP The changes. In order to establish the stability domain of the wind farm, it is necessary to select some of the operating points as state variables and treat the other quantities as constants; the selection principles of state variables are: (1) select wind farm state variables that are easy to measure or directly obtain; (2) the stability domain formed by the state variables can guide the actual regulation and operation.

[0135] The state variable of a wind farm is defined as x farm The stability margin is M GP (x farm ), in x farm Within the range of values, based on the operating point X farm ={x farm,1 ,...,x farm,H}, thus obtaining the stability margin sample set:

[0136] M farm ={M GP (x farm,h )|x farm,h ∈X farm} (17)

[0137] Wherein, the subscript h represents the h-th stability margin sample, and the subscript H represents the total number of stability margin samples (H).

[0138] To determine the boundary between stable and unstable conditions in the stability margin sample set, that is, M GP =0 boundary, all M GP Samples with values ​​less than 0 are defined as unstable samples, and all M samples are considered unstable samples. GP Samples with a value greater than 0 are defined as stable samples. Furthermore, this embodiment can not only form stable and unstable boundaries, but also create multiple margin boundaries by redefining stable samples according to stability margin requirements. To obtain a more conservative stable boundary, all M... GP Samples with a value < 0.1 are defined as unstable samples, and all M... GP Samples with a value greater than 0.1 are defined as stable samples, thus shifting the stability boundary from... Figure 6 The L2 in the model is moved to L3. Therefore, the method in this embodiment can set different stability margins to form a multi-level stability boundary.

[0139] Support Vector Machine (SVM) is a commonly used data classification method. However, traditional SVMs are complex for some heavily interleaved sample sets because they need to consider all sample points when searching for the optimal hyperplane. This leads to a large amount of memory usage, slow training speed, and difficulty in meeting the computational speed requirements for building stable regions online.

[0140] Therefore, this embodiment employs a k-nearest neighbor-based support vector machine algorithm to filter out boundary sample points while simultaneously reducing the sample set based on the number of similar and dissimilar samples in the k-nearest neighbor set of each sample, thereby improving computational efficiency. The specific process is as follows:

[0141] (1) Extraction of k-nearest neighbor boundary sample points: First, define two sample x farm,h and x farm,t Euclidean distance between:

[0142] d 2 (x farm,h ,x farm,t )=(x farm,h -x farm,t (x) farm,h -x farm,t ) T (18)

[0143] Where the subscripts h and t represent the sample numbers in the sample set.

[0144] Secondly, for a known stable sample x farm,h Calculate its k stable nearest neighbors T of the same kind. k (x farm,h ) and its k unstable nearest neighbors Then calculate the average distance between this sample point and samples of the same class and samples of different classes:

[0145]

[0146]

[0147] Where, x farm,t This represents a sample from the neighboring class.

[0148] Set a stability threshold ξ, if D T (x farm,h )<ξ and Then determine x farm,h These are boundary sample points. For example... Figure 6 As shown, k is 3. The 3-nearest neighbors of the same class and the 3-nearest neighbors of different classes of sample point A are relatively small, which means that point A is close to both stable and unstable samples. Therefore, point A is a boundary sample point. The 3-nearest neighbors of the same class of sample point B are relatively small, while the 3-nearest neighbors of different classes are relatively large, which means that B is far from unstable samples. Therefore, point B is a non-boundary sample point.

[0149] (2) Simplification of k-nearest neighbor boundary sample points: After extracting the boundary sample points, the k-nearest neighbor method is used to further reduce the boundary sample set, removing the very close parts of the boundary samples and retaining samples with strong features. While ensuring accuracy, the data size can be reduced and the computational efficiency can be improved.

[0150] Based on the nearest neighbor T of each sample k (x farm,h If the distance between samples of the same type is less than a given threshold, Right now:

[0151]

[0152] That means sample x farm,h With x farm,t If the samples are of the same type and very close in distance, then x... farm,h The samples are removed from the sample set, and the remaining samples are used as training samples to train the support vector machine.

[0153] Because the stability margin samples are quite complex and cannot be completely separated by a linear decomposition surface, the input sample X is... farm Mapping to a high-dimensional feature vector space:

[0154] φ(X farm )=(φ1(X farm ),φ2(X farm ),…,φ l (X farm )) T (twenty one)

[0155] In this context, the subscript 'l' indicates the dimension of the high-dimensional space.

[0156] With the eigenvector φ(X) farm (replace input variable X) farm To find the optimal interface in high-dimensional space;

[0157] f(x)=ωφ(X farm )+b=0 (22)

[0158] To obtain the optimal hyperplane that correctly classifies all samples, the following optimization problem needs to be solved:

[0159]

[0160] Among them, y i To represent the stability of the sample, M GP If y > 0, then y i =1,M GP If y < 0, then y i =-1.

[0161] To compute the inner product of samples in a high-dimensional space, direct computation in the high-dimensional space is very complex. Therefore, a kernel function is defined to compute the function value in the original space:

[0162] k(X farm,i ,X farm,j )=φ(X farm,i ) T φ(X farm,j ) (twenty four)

[0163] The hyperplane dual problem using kernel functions is as follows:

[0164]

[0165]

[0166] Finally, the stable boundary in the original dimension is obtained:

[0167]

[0168] In convex quadratic optimization, optimization is performed not only on support vectors but also on non-support vectors. Therefore, using k-nearest neighbor boundary point selection and training the support vector machine only with boundary points can significantly improve computational efficiency. The steps for implementing a k-nearest neighbor support vector machine are as follows:

[0169] (1) Determine the sample state variables. Based on the piecewise affine model, determine the partition range of s according to equation (7).

[0170] (2) Based on equations (9)-(10), perform partitioning and screening to select regions that may be judged for stability.

[0171] (3) Within the partition, the magnitude margin and phase margin are solved using the multi-point interior method to determine the stability margin.

[0172] (4) Spread points within the range of state variables to form a stability margin sample set.

[0173] (5) Calculate the Euclidean distance between each pair of samples according to equation (18);

[0174] (6) Calculate the same-class k-neighbors and different-class k-neighbors for each sample point;

[0175] (7) Extract the boundary sample points according to equation (19) and delete the non-boundary points;

[0176] (8) Use the k-nearest neighbor method to reduce the sample set;

[0177] (9) The reduced sample set is trained using a support vector machine.

[0178] The effectiveness of the stability region construction method proposed in this embodiment is verified through numerical examples. First, in a single-unit scenario, stability regions for wind speed and active power, and active power and reactive power are considered separately, and time-domain simulations are performed for verification. Second, in a wind farm, a stability region is established based on the wind speed and active power of a single unit, and the impact of a single unit on the stability of the wind farm is analyzed. Furthermore, a stability region is established based on the total power of the wind farm and the power distribution within the farm, and the impact of total power and power distribution on the stability of the wind farm is analyzed. The accuracy of the stability region established in this embodiment is verified through time-domain simulations.

[0179] A. Single-machine stability domain case analysis

[0180] Adopting a doubly fed wind turbine system structure, the first step is to establish a system based on active power P. DFIG Wind speed V w The stable region has a power range of 0.3MW to 2.7MW, a wind speed range of 8m / s to 12m / s, and a grid impedance short-circuit ratio of 1.76.

[0181] like Figure 7 As shown, the change in stability margin is affected by P DFIG and V w The combined effects of these factors. Specifically, throughout the entire stability region, P DFIG The larger the value of V, the more unstable the system tends to be; for V w In P DFIG When it is less than 0.9MW, V w The change in has almost no effect on the stability margin, while in P DFIG When it is greater than 0.9MW, V w The larger the value of P, the more stable the system. This is because P... DFIG When it is large, change V w The rotation speed can be changed, thus affecting the stability of the unit. Figure 7 In the middle, State1's P DFIG It is 1.956MW, V w The velocity is 11 m / s, at which point the system is stable; V w The speed changes, decreasing to 10 m / s, and the operating point changes from State 1 to State 2, at which point the system is unstable; this can be addressed by changing the P output from the single machine. DFIG This can bring the system back to stability, and P DFIG After reducing the power from 1.95MW to 1.5MW, the system was able to return to stability.

[0182] against Figure 7 The time-domain model verification of the single-machine stability domain design is shown below. Figure 8 As shown, before 10 seconds, the system operated stably with a short-circuit ratio of 1.76, and the system P... DFIG Setting 1.95MW, Vw The speed is 11 m / s, in State 1; at 10 s, V w When the speed is reduced to 10 m / s, the system becomes unstable due to a small disturbance. DFIG and reactive power Q DFIG The oscillation amplitude gradually increases, and the phase a current i sa The system also oscillates and diverges, entering State 2, where it becomes unstable. To suppress the continued expansion of the unstable oscillations in the power grid, the active power setpoint is reduced to 1.5MW at 12s. The oscillation amplitude of the system gradually decreases, entering State 3, where it tends to stabilize. This verifies the results of the stability domain: the lower the active power, the more stable the system. The system can return to a stable state through active power adjustment. Thus, the time-domain simulation of the system corresponds to the stability domain of the single unit, proving the reliability of its stability domain under the piecewise affine model.

[0183] Establish a standalone P DFIG and Q DFIG The stability region was determined with a wind speed of 11 m / s and a short-circuit ratio of 1.7. For example... Figure 9 As shown, the change in stability performance is more significantly affected by the active power command, exhibiting P DFIG The larger the output, the more unstable the system. When outputting P... DFIG When it is higher, Q is emitted. DFIG The more stable the system, the smaller the impact. This is because during operation, changing Q... DFIG It will not change the motor speed, while P DFIG It directly affects the rotational speed and has a significant impact on system stability. Figure 9 In the middle, State1's P DFIG It is 2.4MW, Q DFIG When the value is 0MVar, the system is unstable; P DFIG The voltage decreased to 1.8 MW, and the operating point changed from State 1 to State 2, at which point the system was stable. Therefore, the voltage could be adjusted by... DFIG To change the stability of a single machine, but Q DFIG The impact is relatively small, and it is more difficult to achieve this by adjusting Q. DFIG This can alter the system's stability. Therefore, in the following analysis of the wind farm, adjusting Q will not be considered. DFIG The situation.

[0184] against Figure 9 The time-domain model verification of the single-machine stability domain design is shown below. Figure 10 As shown, before 10 seconds, the system operated stably with a short-circuit ratio of 1.76, and the system P... DFIG Set at 2.4MW, Q DFIGThe initial value is 0 MVar. At 10s, changes occur in the grid-side system; the grid short-circuit ratio changes from 1.76 to 1.7, the grid weakens, and instability occurs due to a small disturbance, entering State 1, where the oscillation amplitude gradually increases. To suppress the continued expansion of the unstable oscillations of the grid, P is adjusted at 11s. DFIG When the setpoint was reduced to 1.8MW, the corresponding motor speed decreased from 1391 r / min to 1184 r / min. The system oscillation amplitude gradually decreased, entering State 2 and the stable region. This verifies... Figure 9 As a result, the lower the active power, the more stable the system, and its power setpoint corresponds to the stability region. Therefore, for changes in grid impedance or due to V... w In the event of instability caused by changes, the doubly fed wind turbine is within its own power regulation range and can stabilize or return to stability through active power regulation.

[0185] B. Case Study of Wind Farm Stability Region

[0186] The stability region of a wind farm is analyzed according to the method proposed in this embodiment. Figure 11 The wind farm structure is considered in two scenarios: the power of a single wind turbine changes independently and the power of all wind turbines in the farm changes in tandem. Stability domains of the wind farm are established for each scenario to study the stability of the wind farm.

[0187] First, establish a single wind turbine P within the wind farm. DFIG and V w This study investigates the stability domain of a wind farm under varying conditions, examining the impact of different wind turbines on the stability of the wind farm's grid-connected system. To comprehensively analyze the influence of a single turbine on the wind farm, the wind speed range is set from 11 m / s to 16 m / s in the calculation example to avoid situations where the operating point cannot be reached. Additionally, the power range is set from 1.5 MW to 3 MW to simulate the basic operating point range of the wind farm during operation. For each turbine in the wind farm, when its output power is changed, the power output of the remaining turbines remains constant at 2.4 MW.

[0188] The individual power variation stability region of each wind turbine in the wind farm is as follows: Figure 12 As shown, the stability region of a single wind turbine in a wind farm differs significantly from that of a single-unit grid-connected system. Firstly, the stability region of a single wind turbine is much smaller, and the stability margin M... GP The variation range is within 0.2, indicating that the impact of a single wind turbine on the wind farm is much smaller than that of a single turbine. Secondly, the impact on stability varies depending on the turbine's location within the wind farm; the P values ​​for turbines 1, 2, and 3... DFIG and V w The impact on stability is basically the same as that on a single machine, and it is also P. DFIG The larger the value, the less stable it is. wThe larger the power, the more stable it is. For wind turbines 3, 4, and 5, the larger the power, the more stable they are. This provides guidance for the power allocation of wind farms. When considering stability, wind turbines 3, 4, and 5 can generate more power, while wind turbines 1, 2, and 3 can generate less power.

[0189] The stability region of a wind farm is analyzed when the power of all wind turbines changes in tandem. With the total output power of the wind farm remaining constant, a change in the power of a single turbine (e.g., T1) proportionally affects the power setpoints of the remaining turbines. Similarly, a change in the total output power setpoint affects the power setpoints of each individual turbine. During the overall power change, the following can be observed: Figure 13 During the transition from state 1 to state 2, the stability of the wind farm gradually deteriorates. When the total power increases, the power of each individual turbine increases according to the distribution. Since the wind speed is the same, this leads to an increase in turbine rotational speed. Based on the individual turbine example, the increase in turbine power leads to a decrease in stability, consistent with the pattern observed in the wind farm. Therefore, according to the power distribution within the farm, the power of turbine 1 is increased, while the power of other turbines decreases. This transition from state 2 to state 3 allows the system to return to a stable region.

[0190] Finally, the conclusions of the wind farm stability analysis were verified through time-domain simulation. According to the wind farm's stability domain, the system is stable at this point. During the simulation, the power was varied: the total power was changed at 10 seconds, and the power distribution of individual turbine 1 was changed at 14 seconds. The simulation results are as follows: Figure 14 As shown, when the power distribution changes, the system transitions from a steady state to an unstable state. However, by altering the power distribution within the field, the system regains stability. This is consistent with the results of the wind farm stability analysis, verifying the correctness of the conclusion.

[0191] Example 2

[0192] This embodiment provides a wind farm stability assessment system based on stability domain quantization, including:

[0193] The quantization module is configured to define the gain margin and phase margin of uniform quantization under the same unit;

[0194] The filtering module is configured to construct piecewise affine partitions of the wind farm, and determine the position of the Nyquist curve corresponding to the partition based on the characteristic values ​​of the complex state variables at the endpoints of the partition, thereby filtering out candidate partitions for stability assessment.

[0195] The solution module is configured to solve for the gain margin and phase margin in the candidate partitions using an interior-point method based on multiple starting points. The minimum value between the gain margin and phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The interior-point method based on multiple starting points involves selecting several optimization initial points within the constraints of the complex state variables of the candidate partitions and obtaining the smallest global optimal solution by comparing different solutions.

[0196] The judgment module is configured to construct a stability margin sample set, select boundary sample points in the stability margin sample set, thereby determining the stability boundary and stability domain of the wind farm and obtaining the stability judgment result of the wind farm.

[0197] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0198] In further embodiments, the following is also provided:

[0199] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0200] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0201] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0202] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0203] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0204] Those skilled in the art will recognize that the units, i.e., algorithm steps, of the various examples described in connection with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0205] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for judging the stability of wind farms based on stability region quantization, characterized in that, include: Define the gain margin and phase margin for uniform quantization under the same unit; Affine partitions of the wind farm are constructed, and the position of the Nyquist curve corresponding to the partition is determined based on the eigenvalues ​​of the complex state variables at the endpoints of the partition, thereby selecting candidate partitions for stability assessment. In the candidate partitions, the magnitude margin and phase margin are solved using an interior-point method based on multiple starting points. The minimum value between the magnitude margin and the phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The interior-point method based on multiple starting points selects several optimization initial points within the constraints of the complex state variables of the candidate partitions, and obtains the smallest global optimal solution by comparing different solutions. A stability margin sample set is constructed, and boundary sample points in the stability margin sample set are selected to determine the stability boundary and stability domain of the wind farm, thereby obtaining the stability judgment result of the wind farm.

2. The method for determining the stability of a wind farm based on stability region quantization as described in claim 1, characterized in that, The unified quantization is as follows: the left real axis of (-1, j0) is defined as the forbidden region, and the gain margin and phase margin are uniformly represented by the distance from the Nyquist curve to the forbidden region; where the gain margin is in the right part of the forbidden region, the phase margin is in the upper part of the forbidden region, and the distance from the minimum real axis intersection point of the Nyquist curve to the forbidden region is the gain margin or the phase margin.

3. The method for judging the stability of a wind farm based on stability region quantization as described in claim 1, characterized in that, The candidate partition selection process includes: for magnitude margin, based on the characteristic values ​​of complex state variables in the current and previous partitions, classification is performed according to whether the signs are opposite, both are positive, or both are negative within the partition, specifically: If the eigenvalues ​​of the complex state variables at both ends of the current partition are located above and below the real axis, then the filtering condition is: If the eigenvalues ​​of the complex state variables at both ends of the current partition are all above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are all below the real axis; or, if the eigenvalues ​​of the complex state variables at both ends of the current partition are all below the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are all above the real axis, then the filtering condition is: in, and Representing the The beginning and end of the complex state variables within each partition; and No. The complex state variables inherent in each partition begin and the final complex state variable The points corresponding to the Nyquist curve; For the first -1 partitions containing complex state variables The corresponding point on the Nyquist curve.

4. The method for determining the stability of a wind farm based on stability region quantization as described in claim 3, characterized in that, If the eigenvalues ​​of the complex state variables at both ends of the current partition are both above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are both above the real axis, then it is considered that there is no intersection between the two partitions. If the eigenvalues ​​of the complex state variables at both ends of the current partition are both above the real axis, and the eigenvalues ​​of the complex state variables at both ends of the previous partition are respectively above and below the real axis, then there is an intersection point in the previous partition, so the judgment is not repeated.

5. The method for determining the stability of a wind farm based on stability region quantization as described in claim 1, characterized in that, The candidate partition selection process includes the following: For phase margin, the selection criteria are as follows: in, For the first The Nyquist curve of each partition under the complex state variable S, where L is the set filtering range; and Representing the The beginning and end of the complex state variables within each partition; and No. The complex state variables inherent in each partition begin and the final complex state variable The corresponding point on the Nyquist curve.

6. The method for determining the stability of a wind farm based on stability region quantization as described in claim 1, characterized in that, The process of solving for the gain margin and phase margin using the interior-point method based on multiple starting points includes: The optimization function established for the gain margin is as follows: The optimization function established for phase margin is as follows: in, The penalty coefficient is... The penalty coefficient is... For the first Nyquist curves within each partition.

7. The method for determining the stability of a wind farm based on stability region quantization as described in claim 1, characterized in that, The process of selecting boundary sample points from the stability margin sample set to determine the stability boundary and stability region of the wind farm specifically includes: Determine the Euclidean distance between each pair of samples, calculate the k-nearest neighbors of the same class and the k-nearest neighbors of different classes for each sample point, and thus determine the average distance between the sample point and the two classes of samples. Boundary sample points are extracted based on the average distance between the two types of samples and a set stability threshold, and non-boundary points are deleted. For the remaining stability margin sample set, based on the k-nearest samples of the same type for each sample point, if the distance is less than a given threshold, then the sample set is reduced. Based on boundary sample points, support vector machines are used to determine the stability boundary and stability region of the wind farm.

8. A wind farm stability assessment system based on stability domain quantization, characterized in that, include: The quantization module is configured to define the gain margin and phase margin of uniform quantization under the same unit; The filtering module is configured to construct piecewise affine partitions of the wind farm, and determine the position of the Nyquist curve corresponding to the partition based on the characteristic values ​​of the complex state variables at the endpoints of the partition, thereby filtering out candidate partitions for stability assessment. The solution module is configured to solve for the gain margin and phase margin in the candidate partitions using a multi-starting-point interior-point method. The minimum value between the gain margin and phase margin is taken as the stability margin of the current candidate partition, and the minimum value of the stability margin among all candidate partitions is taken as the stability margin at the current operating point. The multi-starting-point interior-point method involves selecting several optimization initial points within the constraints of the complex state variables of the candidate partitions and obtaining the smallest global optimal solution by comparing different solutions. The judgment module is configured to construct a stability margin sample set, select boundary sample points in the stability margin sample set, thereby determining the stability boundary and stability domain of the wind farm and obtaining the stability judgment result of the wind farm.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-7.

Images