Network following type wind power plant frequency response equivalence method based on VSG network construction type control strategy

Abstract

The invention discloses a network following type wind power plant frequency response equivalence method based on a VSG network construction type control strategy. The method comprises the following steps: acquiring active output, virtual inertia, primary frequency modulation coefficient and port active drop quantity of a fan based on a VSG control model, and constructing grouping index data for representing frequency dynamic characteristics; a fast DB-SCAN clustering algorithm with a particle swarm optimization mechanism introduced is adopted to conduct self-adaptive clustering analysis on the clustering indexes, and an optimal clustering result is obtained; and performing multi-machine multiplication aggregation modeling according to a clustering result to form a wind power plant frequency response equivalent model. According to the method, quantitative identification and rapid equivalent modeling of the frequency characteristics of the grid-forming wind power plant are realized, and the precision and efficiency of frequency dynamic analysis and stability evaluation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy power generation and power system control technology, in particular to a frequency response equivalent method of grid-following wind farm based on VSG grid-forming control strategy. BACKGROUND

[0002] Wind farms play an increasingly important role in high-proportion new energy power systems, and their grid-connected operation characteristics have a direct impact on the frequency stability of the power grid. Traditional wind turbine generators generally use grid-following control methods, which detect the grid voltage phase based on a phase-locked loop and achieve grid-connected operation through current command control. Although this control method is simple in structure, it lacks the ability to actively support the grid voltage and frequency, and when the system experiences frequency disturbances, the wind farm can only respond passively to grid changes, making it difficult to provide effective inertia support and primary frequency regulation in a short time. With the continuous increase in new energy penetration, the inertia of the power grid has decreased significantly, and the risk of frequency fluctuations has increased, leading to the emergence of virtual synchronous generator (VSG) control technology. This control strategy simulates the electromagnetic and mechanical motion equations of synchronous generators, introduces virtual inertia and damping coefficients, enabling wind turbines to have similar frequency response characteristics to synchronous generators and actively participate in frequency regulation, thereby improving the stability and dynamic performance of the power system.

[0003] In existing research, the equivalent analysis of wind farm frequency response characteristics mainly focuses on grid-following control scenarios. Common methods include single-machine equivalence and multi-machine aggregation equivalence, with the latter usually using frequency response characteristic parameters as the basis for grouping and using clustering algorithms to divide wind turbine groups with different dynamic response characteristics. However, these methods are mostly based on grid-following wind farm characteristics, and the selected grouping indicators are mainly limited to measured parameters such as steady-state power or frequency-active trajectory, without considering the decisive role of virtual inertia and primary frequency regulation in the frequency response of grid-forming control. At the same time, existing clustering algorithms mostly use K-means, fuzzy C-means, or Gaussian mixture model methods, which are highly sensitive to initial clustering centers and parameter settings, and have poor clustering stability in high-dimensional nonlinearly distributed data sets. Although the DB-SCAN algorithm can be applied to data of any shape, it has high complexity and long computation time in neighborhood search, making it difficult to meet the real-time requirements of large-scale wind farm modeling and analysis.

[0004] Based on the above technical status, existing methods still have significant deficiencies in modeling the frequency response of grid-forming wind farms, specifically in the following two aspects:

[0005] Firstly, there is a lack of a grouping indicator system for VSG grid-forming control characteristics, which cannot accurately represent the dynamic differences between wind turbines in the inertia response and primary frequency regulation stages.

[0006] Secondly, clustering algorithms have strong parameter dependencies and high computational complexity in high-dimensional index spaces, making it difficult to balance accuracy and efficiency.

[0007] Therefore, there is an urgent need to propose a method that can establish a cluster index system based on the VSG network control strategy and combine it with an efficient clustering algorithm to realize the equivalent modeling of wind farm frequency response, so as to improve the accuracy and operability of frequency dynamic analysis and stability assessment of network-type wind farms. Summary of the Invention

[0008] To address the problems of existing technologies, this invention provides an equivalent method for the frequency response of grid-connected wind farms based on a VSG grid-connected control strategy. The technical solution is as follows:

[0009] On the one hand, a method for valuing the frequency response of a grid-connected wind farm based on a VSG grid-connected control strategy is provided, including the following steps:

[0010] (1) Construction of cluster indicator data:

[0011] Based on the VSG control model, the active power output, virtual inertia, primary frequency regulation coefficient and active power drop at the port of each grid-type wind turbine in the wind farm are obtained. The above four parameters are used as clustering indices to form clustering index data for characterizing the frequency response characteristics of the grid-type wind farm.

[0012] (2) Cluster analysis:

[0013] The clustering index data is input into the fast DB-SCAN clustering algorithm based on particle swarm optimization mechanism. The optimal clustering result is obtained by iteratively optimizing the density threshold and neighborhood radius parameter, thereby realizing the division of wind turbine groups with similar frequency response characteristics.

[0014] (3) Multi-machine equivalent modeling:

[0015] Based on the optimal clustering results, single-unit multiplication aggregation modeling is performed on various wind turbine groups to form a multi-unit equivalent model, which characterizes the dynamic characteristics of grid-type wind farms in the inertial response and primary frequency regulation stages.

[0016] Furthermore, the virtual inertia and primary frequency regulation coefficient in the cluster index data are determined by the rotor motion relationship and speed control loop of the VSG control model.

[0017] Among them, virtual inertia is used to describe the inertial response characteristics of the wind turbine in the initial stage of frequency disturbance, and the primary frequency regulation coefficient is used to characterize the active power regulation capability of the wind turbine to the system frequency deviation during the speed regulation stage.

[0018] This is to ensure that the cluster index data can accurately reflect the dynamic characteristics of wind farm units at different response stages.

[0019] Furthermore, the clustering analysis is implemented using the fast DB-SCAN algorithm. When determining the density reachability of clustering index data, neighborhood search and clustering determination are achieved based on spatial partitioning of data features.

[0020] This aims to reduce the computational complexity of clustering large-scale wind turbine data and improve clustering efficiency and stability.

[0021] Furthermore, a particle swarm optimization mechanism is introduced into the clustering analysis to adaptively optimize the neighborhood radius and minimum number of points parameters in the fast DB-SCAN algorithm. During the iteration process, the particle position and velocity are dynamically updated according to the inertia weight, individual learning factor and group learning factor. The optimal solution of the clustering parameters is obtained by combining global and local search, so as to improve the stability and global convergence performance of the wind farm clustering results.

[0022] Furthermore, the clustering results in the clustering analysis are evaluated using a comprehensive index, which combines intra-cluster compactness and inter-cluster separation characteristics to determine the rationality and stability of the clustering results, so as to ensure that the clustering results of the clustering index data have high discriminativeness and consistency.

[0023] Furthermore, the spatial partitioning process in the clustering analysis includes:

[0024] The data intervals are divided according to the feature dimensions of the cluster index data to form a multi-dimensional spatial index. The division is terminated when the number of nodes reaches a preset threshold, so as to improve the ability to distinguish the spatial distribution characteristics of different wind turbine groups and improve the efficiency of clustering calculation.

[0025] Furthermore, the multi-machine equivalent modeling includes: based on the optimal clustering results, considering wind turbines within the same group as representative turbines with similar frequency response characteristics; performing a capacity-weighted average of the virtual inertia, primary frequency regulation coefficient, and active power output of the wind turbines within the group, and multiplying the representative turbine parameters according to the number of wind turbines in the group to ensure that the frequency characteristics of the obtained multi-machine equivalent model remain dynamically consistent with the original group in the inertial response and primary frequency regulation response stages.

[0026] Furthermore, the method is applicable to frequency dynamic response analysis of wind farms under operating conditions such as power deficit, grid connection disturbances, and voltage drops. Through the calculation and evaluation of the multi-machine equivalent model, the frequency support characteristics of the inertial response stage and the primary frequency regulation stage can be analyzed, which can be used to guide the frequency stability assessment and control strategy design of grid-type wind farms.

[0027] On the other hand, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, is used to perform the method described thereon.

[0028] On the other hand, an electronic device is provided, including at least one processor and at least one memory, wherein the memory stores a computer program that, when executed, causes the processor to perform the method.

[0029] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows:

[0030] This invention provides an equivalent method for the frequency response of grid-connected wind farms based on a VSG-based grid control strategy. By constructing a parameter system consistent with the frequency response characteristics of synchronous generators, it achieves accurate equivalence of the dynamic frequency characteristics of grid-connected wind farms. This method introduces virtual inertia, damping coefficients, and primary frequency regulation coefficients from VSG control to establish a multi-dimensional clustering index set, including active power output, virtual inertia, primary frequency regulation coefficients, and active power sag at the ports. This comprehensively reflects the dynamic differences between the inertial response and speed regulation phases of the units. Based on this, an improved fast DB-SCAN clustering algorithm is employed, introducing a particle swarm optimization mechanism to achieve adaptive optimization of the neighborhood radius and density threshold. Combined with a KD decision tree, efficient neighborhood search is achieved, thus significantly reducing algorithm complexity while ensuring clustering accuracy. Finally, a multi-unit equivalent model is established through a single-unit multiplication aggregation method, maintaining the consistency of the overall frequency response characteristics of the wind farm while significantly improving modeling and computational efficiency.

[0031] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in the following two aspects:

[0032] First, this invention proposes for the first time a cluster index system applicable to the frequency response of wind farms for the control characteristics of VSG grid-type wind farms, realizing a quantitative mapping from control parameters to frequency dynamic characteristics, and solving the problem that existing methods are difficult to accurately characterize the frequency characteristics of grid-type wind farms.

[0033] Secondly, the fast DB-SCAN clustering algorithm used in this invention has adaptive optimization capabilities and high-dimensional computational efficiency, overcoming the problems of strong parameter dependence and slow convergence speed of traditional clustering algorithms, making this method have good engineering applicability in the frequency dynamic analysis of large-scale wind farms.

[0034] Therefore, this invention has significant technical advantages and promotional value in the field of frequency characteristic modeling and system stability assessment of grid-type wind farms. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a diagram of the VSG-type grid-connected wind turbine grid connection system and control structure according to Embodiment 1 of the present invention;

[0037] Figure 2 This is a flowchart illustrating an equivalent method for the frequency response of a grid-connected wind farm based on a VSG grid-connected control strategy, according to Embodiment 1 of the present invention.

[0038] Figure 3 This is a flowchart of the VSG active-frequency control in Embodiment 1 of the present invention;

[0039] Figure 4 This is a diagram showing the frequency response characteristics of the VSG-type control system in Embodiment 1 of the present invention.

[0040] Figure 5 This is a flowchart illustrating the improved DBSCAN algorithm of Embodiment 1 of the present invention;

[0041] Figure 6 This is a schematic diagram of the equivalent circuit of Embodiment 1 of the present invention;

[0042] Figure 7 This is a typical wind farm topology diagram of Embodiment 1 of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0044] Example 1

[0045] This invention provides an equivalent method for the frequency response of grid-connected wind farms based on a VSG (Virtual Synchronous Generator) grid-connected control strategy. This method aims to address the problem of significant differences in frequency response between different types of turbines in existing wind farms, making it difficult to accurately characterize their overall dynamic characteristics. Based on the physical characteristics of VSG grid-connected control, this method simplifies and accurately describes the dynamic characteristics of wind farms through frequency response index extraction, cluster analysis, and equivalent modeling.

[0046] like Figure 1As shown, the grid-connected system of the VSG-type grid-connected wind turbine includes a converter control unit, a filter inductor, a grid connection point capacitor, and an impedance branch connected to the power grid. The control system consists of an active power-frequency link, a virtual rotor link, a speed control loop, and a voltage-current dual closed loop. The power outer loop simulates the inertial response process of a synchronous generator by introducing a virtual inertia J and a damping coefficient D, enabling the VSG-type grid-connected wind turbine to quickly adjust its output power during system frequency disturbances, achieving inertial support and primary frequency regulation control. This control structure ensures that the frequency response characteristics of the VSG-type grid-connected wind turbine are consistent with those of a traditional synchronous machine, providing a theoretical basis for the modeling method of this invention.

[0047] Based on the above control mechanism, the overall flow of the frequency response equivalent method proposed in this invention is as follows: Figure 2 As shown in the figure. This method mainly includes three stages: construction of cluster index data, cluster analysis, and multi-machine equivalent modeling. Specifically, firstly, feature data such as active power output, virtual inertia parameters, primary frequency regulation coefficient, and active power drop at the port of the grid-type wind turbines are extracted; then, the feature samples are input into the clustering algorithm model, and the adaptive clustering of the wind turbine frequency characteristics is achieved using the fast DB-SCAN clustering algorithm based on particle swarm optimization (hereinafter referred to as "PSK-DBSCAN algorithm"); finally, parameter aggregation and multiplication are performed based on the clustering results to establish a multi-machine equivalent model, realizing the system-level equivalence of the overall frequency response characteristics of the wind farm.

[0048] This method realizes a multi-layer mapping from individual wind turbine characteristics to the overall dynamic behavior of the wind farm. While ensuring the accuracy of frequency response, it significantly reduces the modeling complexity and provides a reliable analytical tool for the control optimization and frequency support evaluation of grid-type wind farms.

[0049] Based on the system structure and control characteristics of the aforementioned VSG grid-type wind turbines, this invention further proposes an equivalent modeling process for the frequency response of grid-type wind farms at the implementation level. This process is based on the VSG control equations and unfolds sequentially through three stages: parameter extraction, cluster analysis, and model equivalence.

[0050] The cluster index data construction stage is used to extract and quantify the inertial response and primary frequency regulation characteristics of wind turbines; the cluster clustering analysis stage uses the improved DB-SCAN algorithm to achieve automatic clustering of wind turbine characteristics; and the multi-machine equivalent modeling stage performs parameter aggregation and frequency response modeling on units of the same category to obtain an equivalent model that can represent the overall dynamic characteristics of the wind farm.

[0051] To clearly demonstrate the implementation process of this invention, the specific implementation methods of each step are described in detail.

[0052] Step 1: Constructing Cluster Indicator Data

[0053] This step aims to establish a cluster index data system that reflects the frequency dynamic characteristics of grid-connected wind farms, providing fundamental data support for subsequent cluster analysis and multi-machine equivalent modeling. The selection of the cluster index data is based on the VSG (Virtual Synchronous Generator) grid control principle, starting from the inertial response and primary frequency regulation characteristics of the units, extracting key control parameters, and establishing a quantifiable set of frequency response characteristics.

[0054] For grid-type wind turbines employing the VSG control strategy, their power outer loop achieves active power-frequency regulation through the rotor motion equations of a simulated synchronous generator and a primary frequency regulation mechanism. This control mechanism can maintain frequency stability by adjusting the output active power through virtual inertia and the primary frequency regulation coefficient when the system frequency is disturbed. To clarify the physical meaning and selection basis of each group index, the frequency response mechanism and index selection principle of grid-type wind turbines are further explained below.

[0055] The selection principle for the clustering index of grid-type wind farms in step one is as follows:

[0056] For VSG-controlled grid-type wind turbines:

[0057] The active frequency loop in the power outer loop achieves active frequency control by simulating the rotor motion equation and primary frequency regulation characteristics of a synchronous generator. Its control flowchart is shown below. Figure 3 As shown, its governing equations are:

[0058]

[0059] Where δ is the generator power angle, J is the rotor moment of inertia, and P ref Here, P is the active power reference value, D is the damping coefficient, ω is the system angular velocity, ω0 is the synchronous angular velocity, and k is the synchronous angular velocity. p This is the active power droop factor.

[0060] From equation (1), it can be seen that the VSG-type network control, by simulating the rotor motion equation of a synchronous generator and introducing virtual inertia J and damping coefficient D, allows the VSG to respond quickly and adjust its output active power to maintain system frequency stability when the system frequency changes. In terms of frequency response characteristics, it is consistent with the response characteristics of a synchronous generator.

[0061] The frequency response characteristics of a synchronous generator can be described as follows:

[0062] The inertial response characteristics of synchronous generator sets are as follows: when the system frequency changes, the synchronous generator set responds to the frequency change by releasing or absorbing kinetic energy through the rotor's rotational inertia. Its power-frequency dynamic relationship can be described by the rotor motion equations:

[0063]

[0064] Among them, T j P is the inertial time constant of the generator. T P represents the change in the power of the prime mover. e S is the change in electromagnetic power. Neq Where is the generator's rated capacity, and E is the generator's kinetic energy.

[0065] like Figure 4 As shown, taking the power deficit scenario as an example, the inertial response characteristics of the synchronous generator unit are mainly manifested in the inertial stage. At this time, the governor has not yet acted, and the frequency change is determined only by the rotor inertia and the load frequency characteristics. The frequency response curve decreases approximately linearly, and the rate of decrease is mainly affected by the generator kinetic energy. Its key parameter is the initial frequency change rate, which can be expressed as:

[0066]

[0067] The speed governor's response characteristics are as follows: the generator speed governor automatically adjusts the prime mover power according to the frequency deviation, i.e., primary frequency regulation, and its dynamic process can be approximated as a first-order inertial element.

[0068] ΔP L =K L Δf

[0069] Where: K G T is the droop coefficient of the speed governor. G This is the governor's response time constant.

[0070] Taking a power deficit scenario as an example, during the governor adjustment phase, the governor detects the frequency deviation and begins to increase the prime mover power, such as the steam turbine's steam intake and the turbine's guide vane opening. The frequency decrease rate gradually decreases, and the response curve exhibits exponential decay characteristics. Finally, under the action of primary frequency regulation and load adjustment, the frequency reaches a steady-state value. The key parameter is the final frequency change, which can be expressed as:

[0071]

[0072] The above analysis shows that the frequency response characteristics of VSG-based grid control are consistent with those of synchronous machines. Key characteristic 1, the inertial response stage, is mainly determined by its active power output and virtual moment of inertia, which can be used as a grouping index. Key characteristic 2, the governor operation stage, is mainly determined by the primary frequency regulation coefficient of the simulated governor, which can also be used as a grouping index. Finally, to establish the frequency equivalence for VSG-based grid wind farms, the grouping indices selected are: active power output of each wind turbine, virtual inertia of VSG control, primary frequency regulation coefficient of VSG control, and active power sag at the port.

[0073] After determining the core clustering indicators, in order to further construct a multidimensional dataset of the frequency response characteristics of wind farm units, this invention selects the active power output, virtual inertia, primary frequency regulation coefficient and active power drop at the port of the wind turbine as characteristic variables, which respectively characterize the steady-state power level, inertial support capability, speed regulation response performance and disturbance amplitude of the unit.

[0074] In the specific implementation, the virtual inertia and primary frequency regulation coefficient are determined by the rotor motion relationship and speed control loop of the VSG control model, and combined with wind turbine operation monitoring data to form a feature matrix. To eliminate the influence of capacity differences between different turbine models on the clustering results, each feature data is standardized to ensure that its distribution range is within a uniform range, thus guaranteeing the comparability of input samples and the stability of clustering results. The resulting cluster index dataset will serve as the input sample for subsequent fast DB-SCAN clustering analysis, used to automatically identify wind turbine groups with different frequency response characteristics.

[0075] Step 2: Cluster Analysis

[0076] After constructing the clustering index data, this step uses cluster analysis to group wind turbines with similar frequency response characteristics within the wind farm. This process employs the fast DB-SCAN clustering algorithm based on particle swarm optimization (PSO), using density reachability as the core criterion to automatically partition the wind turbine feature space, achieving adaptive identification of different types of wind turbines. Compared to traditional empirical clustering methods, this algorithm exhibits better global convergence and clustering stability when handling high-dimensional, nonlinear feature data, significantly improving the accuracy of frequency response isometric modeling.

[0077] In its implementation, the fast DB-SCAN algorithm determines the cluster boundary using two key parameters: neighborhood radius and minimum number of points. To avoid parameter uncertainty caused by manual setting, this invention introduces a particle swarm optimization algorithm to adaptively optimize these parameters during the clustering process. The particle swarm uses the clustering performance evaluation index as its fitness function, achieving global optimization through iterative updates of particle velocity and position, ensuring that the clustering parameters automatically converge to the optimal solution under different wind farm distribution characteristics. This method effectively balances intra-cluster compactness and inter-cluster separation, ensuring that the clustering results achieve the optimal balance between structural stability and resolution.

[0078] Specifically, this invention introduces a particle swarm optimization mechanism to achieve adaptive optimization of clustering parameters based on the traditional DBSCAN algorithm, forming the PSK-DBSCAN algorithm, the principle of which is explained below:

[0079] In this algorithm, the DBSCAN clustering parameters ε (neighborhood radius) and Minpints (minimum number of points) are adaptively obtained by introducing a particle swarm optimization (PSO) algorithm, avoiding the difficulty of manual setting and thus enhancing the algorithm's applicability in complex wind farms. The corresponding parameters in the PSO algorithm are defined as follows: the number of particles is N, the particle position data is X, and the velocity data is V.

[0080] The iterative formula for the particle is:

[0081]

[0082] Among them, v i k x is the velocity of the i-th particle in the k-th iteration. i k p is the position of the i-th particle in the k-th iteration. i,best Let p be the local optimal position of particle i after k iterations. g,best Let w be the global optimal position of the population after k iterations, w be the inertia weight that controls the tendency of particles to maintain their original velocity, and c be the global optimal position of the population after k iterations. 1, c2 is the learning factor, used to control the intensity of the particle's learning towards the individual optimum and the global optimum, r 1, r2 is a random number in the interval [0,1], used to enhance the randomness of the algorithm.

[0083] The particle position update formula is:

[0084]

[0085] The algorithm randomly generates the positions and velocities of N particles, and sets parameters such as the maximum number of iterations and inertia weights. It then calculates the fitness value of each particle to determine the optimal p for that individual. i,best and the global optimal p g,best After updating the particle state according to the velocity and position formulas, the fitness update is re-evaluated until the maximum number of iterations is reached or convergence is achieved. Finally, the optimal clustering parameters of the DBSCAN algorithm are obtained, ensuring the clustering effect and replacing manual setting.

[0086] The fitness function in this paper is designed based on the silhouette coefficient and DBI. The former directly characterizes the compactness of a single sample with its cluster and the separation from other clusters by calculating the matching degree between individuals and clusters and then taking a global average. The latter does not traverse the samples but calculates the average similarity between clusters from a global perspective to reduce the interference of noise points on the clustering effect and enhance robustness to outliers. Due to its low computational complexity, it is suitable for rapid evaluation of large-scale or high-dimensional data. In clustering calculation, the formula for calculating the silhouette coefficient is:

[0087]

[0088] Where a(i) is the average distance between the sample point and other data points in the cluster, and b(i) is the minimum average distance between the sample point and all data points in other clusters.

[0089] The silhouette coefficient ranges from [-1, 1], where 0 indicates the particle is in a boundary sample, -1 indicates an incorrect cluster, and 1 indicates a correct cluster. The overall silhouette coefficient is obtained by averaging the silhouette coefficients of all particles.

[0090]

[0091] The formula for calculating DBI is:

[0092]

[0093] Where σ is the average distance from cluster i to the cluster center, and c i For the center of cluster i, d(σ) i +σ i R is the inter-cluster distance. ij This represents the discreteness of a single cluster.

[0094] The final fitness function is:

[0095]

[0096] The fitness function combines the clustering performance of individuals and the whole, and has a certain robustness to outliers. The weighting factor α can be adjusted to adjust the proportion of individual and whole performance. In this paper, due to the large amount of data and the relatively uniform distribution of data points, the weighting factor α is set to 0.4.

[0097] The traditional DBSCAN algorithm achieves density reachability based on global linear neighborhood search. For each sample point, it needs to calculate all points in its ε-neighborhood, resulting in a time complexity of O(n^2). 2 This leads to low computational efficiency and a surge in computational load when processing large-scale data. In such cases, introducing a KD-tree can assist in fast neighborhood search, saving time and reducing computational complexity to O(nlogn). As a binary tree data search structure, each node of a KD-tree represents a hyperrectangular region in k-dimensional space. By alternately partitioning data across different dimensions—that is, through hyperplane segmentation—the search range is quickly narrowed. On one hand, range queries find all points clustered less than ε with the query point; on the other hand, if a node's region intersects with the query sphere, the subtree is skipped, prematurely terminating branches with a distance exceeding ε, reducing unnecessary distance calculations. The query data point starts from the root node and is compared with the range of values ​​in each corresponding dimension to determine whether a neighborhood search is successful or to continue iteratively searching the subtree. Segmentation stops when the number of points within a node is less than a threshold or the maximum depth is reached.

[0098] The calculation steps of the DB-SCAN algorithm based on KD-tree are as follows:

[0099] ① The KD tree of the dataset: Construct a KD tree by cyclically partitioning the PDW dataset according to its dimensions, such that the value of the left subtree in a given dimension is less than the value of the parent node, and the value of the right subtree is greater than the value of the parent node.

[0100] ② Perform density reachability search. For unvisited point p: use the range query of the KD tree to find all points in the ε neighborhood of p. If the number of neighborhood points is greater than or equal to Minpints, then p is the core point and a new cluster is formed.

[0101] ③ Recursively expand the cluster until no new core point can be found.

[0102] The final process for improving the DBSCAN algorithm is as follows: Figure 5 As shown.

[0103] To further improve the efficiency of clustering computation, this invention employs a KD decision tree structure to partition the multidimensional data space in the algorithm implementation. The KD tree, through recursive partitioning of the sample space and the establishment of a multi-layered index structure, significantly accelerates the neighborhood search process. Combined with a particle swarm optimization mechanism, this algorithm maintains high clustering accuracy while significantly reducing computational complexity, thus meeting the real-time clustering requirements of large-scale wind farm data.

[0104] The optimized clustering analysis process described above effectively distinguishes the frequency response characteristics of different types of wind turbines. Turbine groups with higher inertia parameters exhibit stronger transient support capabilities, while groups with larger frequency regulation coefficients demonstrate better regulation performance during the steady-state recovery phase. The final clustering results will serve as the input basis for subsequent multi-turbine equivalent modeling, realizing the mapping and correlation from turbine characteristics to system dynamic response, and providing a reliable basis for group division for equivalent modeling of the frequency response characteristics of grid-type wind farms.

[0105] Step 3: Multi-machine equivalent modeling

[0106] After completing the wind turbine group division, it is necessary to further establish a multi-turbine equivalent model to reflect the overall frequency response characteristics of the wind farm. Based on the optimal clustering results obtained in step two, this invention performs single-turbine multiplication processing on different wind turbine groups to construct a multi-turbine frequency response equivalent model, thereby improving computational efficiency while ensuring model accuracy.

[0107] like Figure 6The diagram shown is an equivalent circuit diagram of the present invention. This diagram reflects the electrical equivalence of the clustered wind turbine groups. Each group consists of VSG network-type wind turbines with similar frequency response characteristics. Through a multiplication modeling approach, the dynamic response characteristics of multiple wind turbines within a group under frequency disturbances are equivalently represented by a single representative unit. This equivalent model consists of the group's equivalent electromotive force, group's equivalent impedance, and grid connection point voltage. The group's equivalent electromotive force is obtained by the capacity-weighted average of the output voltages of each wind turbine, while the group's equivalent impedance reflects the comprehensive electrical characteristics of the wind turbines connected in parallel within the group.

[0108] In the equivalent modeling process, key control parameters of wind turbines within the group are equivalentized based on the cluster analysis results. For dynamic parameters such as virtual inertia, primary frequency regulation coefficient, and active power output, a capacity-weighted average method is used to determine representative parameters for the group. For active power drop at the port, the parameters are set based on the average characteristic value of the group samples to ensure dynamic consistency between the inertial response stage and the primary frequency regulation stage. Through this parameter equivalence processing, the resulting multi-turbine equivalent model can accurately map the response characteristics of the original wind farm under frequency disturbances, realizing the transformation from a complex multi-turbine system to a representative single-turbine model of frequency characteristics.

[0109] This model significantly reduces the complexity of system modeling and simulation calculations, while maintaining the overall inertial support and primary frequency regulation response characteristics of the wind farm, and has good engineering applicability and dynamic accuracy.

[0110] To facilitate understanding of the application scenarios of this invention in actual wind farms by those skilled in the art, the description is provided in conjunction with the structure of a typical wind farm.

[0111] like Figure 7 The diagram shows a typical wind farm topology. A typical wind farm topology consists of multiple VSG (Variable Retrieval System) grid-connected wind turbines connected to a 35kV collector bus via a step-up transformer and then connected to the upper-level power grid via a substation. The overall topology employs a dual-branch parallel connection method, with the wind turbine groups within the left and right branches distributed according to geographical location or power characteristics. The frequency response equivalence method based on the VSG grid-connected control strategy proposed in this invention can be directly applied to this type of topology. By grouping, clustering, and parameter equivalence processing of the wind turbines within the branches, a rapid equivalent expression of the overall frequency dynamic characteristics of the wind farm can be achieved.

[0112] In summary, this invention provides an equivalent method for the frequency response of grid-connected wind farms based on a VSG grid-connected control strategy. This method, based on the VSG control principle, systematically models the frequency response characteristics of wind farm turbines, forming an equivalent analysis process that balances physical consistency and computational efficiency.

[0113] This method first constructs cluster index data reflecting inertia and primary frequency regulation characteristics to quantify and compare the frequency response characteristics of wind turbines. Second, it introduces a fast DB-SCAN clustering algorithm based on particle swarm optimization to complete the adaptive clustering identification of units, overcoming the problems of strong parameter dependence and ambiguous clustering boundaries in traditional clustering methods. Finally, based on the clustering results, it conducts multi-unit equivalent modeling, and retains the inertial support and primary frequency regulation response characteristics of the groups through capacity weighting and multiplication, thereby achieving a high-precision equivalent expression of the frequency dynamic process of the wind farm.

[0114] This method effectively reduces the computational scale of frequency dynamic analysis in complex wind farms, avoids data redundancy and computational load caused by full-scale modeling, and maintains the consistency of system-level frequency response characteristics. Verification shows that the equivalent model established in this invention maintains high response accuracy in the inertial response stage, primary frequency regulation stage, and steady-state recovery stage, meeting the engineering application requirements for power system frequency stability analysis and control strategy optimization.

[0115] Therefore, this invention has significant engineering application value in the frequency response evaluation and control research of grid-type wind farms, and can be widely applied in fields such as new energy power system modeling, control parameter tuning, and grid frequency regulation strategy design.

[0116] Example 2

[0117] This embodiment provides an electronic device for implementing an equivalent method for the frequency response of a grid-connected wind farm based on a VSG network control strategy. The electronic device includes at least one processor and at least one memory. The memory stores a computer program that, when executed by the processor, performs feature extraction, cluster analysis, and equivalent modeling of the wind farm's frequency response.

[0118] The operation of the electronic device includes the following functional modules:

[0119] 1. Cluster Index Construction Module: This module is used to collect operating parameters of grid-type wind turbines within a wind farm, including active power output, virtual inertia, primary frequency regulation coefficient, and active power sag at the port. The module then normalizes and standardizes these parameters to form a multidimensional cluster index dataset for frequency response characteristic analysis.

[0120] 2. Clustering Analysis Module: Based on the aforementioned clustering index dataset, this module uses an improved fast DB-SCAN clustering algorithm to cluster the wind turbine samples. The clustering algorithm introduces a particle swarm optimization mechanism on top of the traditional DB-SCAN algorithm, adaptively optimizing the cluster radius and minimum sample number parameters to achieve automatic identification and classification of wind turbine groups.

[0121] 3. Equivalent Modeling Module: Based on the cluster analysis results, this module performs capacity weighting and multiplication on the key control parameters of wind turbines within each group to establish a multi-turbine equivalent model. This model, in the form of group representative turbines, characterizes the dynamic characteristics of the wind farm during the inertial response and primary frequency regulation phases, reflecting the overall frequency support capability of the wind farm.

[0122] 4. Results Output Module: Used to output frequency response characteristics and analysis results generated based on the equivalent model, in order to support frequency support performance evaluation, control strategy design and system operation optimization.

[0123] In this embodiment, each functional module can be implemented through independent software units or integrated programmatically within the same processor. The electronic device can be an industrial control host, an edge computing terminal, or a server device with data processing capabilities, capable of real-time processing and analysis of wind farm operation data.

[0124] Example 3

[0125] This embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a method for equivalence of the frequency response of a grid-connected wind farm based on a VSG grid-connected control strategy, including the following steps:

[0126] 1. Obtain the operating parameters of wind turbines of various grid types in the wind farm, and construct a cluster index dataset including active power output, virtual inertia, primary frequency regulation coefficient and active power drop at the port;

[0127] 2. Based on the fast DB-SCAN algorithm with particle swarm optimization mechanism, cluster analysis is performed on the cluster index dataset to complete the adaptive identification of wind turbine groups;

[0128] 3. Based on the clustering results, the parameters of the wind turbines within the group are weighted and multiplied by capacity to establish a multi-machine equivalent model and generate frequency response characteristic results.

[0129] The computer-readable storage medium can be a read-only memory, random access memory, flash memory, solid-state drive, disk, optical disk, or other medium capable of storing program code, all of which can implement the above method and process.

[0130] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A frequency response equivalent method for a follow-net type wind farm based on a VSG network type control strategy, characterized in that, The method comprises the following steps: (1) Grouping index data construction: Based on the VSG control model, the active power output, virtual inertia, primary frequency modulation coefficient and port active power drop of each grid-connected wind turbine in the wind farm are obtained, and the four parameters are taken as grouping indexes to form grouping index data for representing the frequency response characteristics of the grid-connected wind farm; (2) Grouping clustering analysis: The grouping index data is input into a fast DB-SCAN clustering algorithm based on a particle swarm optimization mechanism, the density threshold and neighborhood radius parameters are iteratively optimized, the optimal clustering result is obtained, and the wind turbine groups with similar frequency response characteristics are divided; (3) Multi-machine equivalent modeling: According to the optimal clustering result, single-machine multiple-aggregation modeling is performed on each type of wind turbine group to form a multi-machine equivalent model for representing the dynamic characteristics of the grid-connected wind farm in the inertia response and primary frequency modulation stage.

2. The method of claim 1, wherein, The virtual inertia and primary frequency modulation coefficient in the grouping index data are determined by the rotor motion relationship and speed control link of the VSG control model, wherein the virtual inertia is used to describe the inertia response characteristics of the wind turbine at the initial stage of frequency disturbance, and the primary frequency modulation coefficient is used to represent the active power regulation capability of the wind turbine to the system frequency deviation in the speed regulation stage, so as to ensure that the grouping index data can accurately reflect the dynamic characteristics of the wind farm units in different response stages.

3. The method of claim 2, wherein, The grouping clustering analysis is realized by the fast DB-SCAN algorithm, and when the density accessibility of the grouping index data is judged, the neighborhood search and clustering determination are realized based on the space division of the data characteristics, so as to reduce the calculation complexity of large-scale wind turbine data clustering and improve the clustering efficiency and stability.

4. The method of claim 3, wherein, In the grouping clustering analysis, a particle swarm optimization mechanism is introduced to adaptively optimize the neighborhood radius and minimum point number parameters in the fast DB-SCAN algorithm, and the particle position and speed are dynamically updated according to the inertia weight, individual learning factor and group learning factor in the iteration process. The optimal solution of the clustering parameters is obtained by combining global and local search to improve the stability and global convergence performance of the wind farm grouping result.

5. The method of claim 4, wherein, The clustering result in the grouping clustering analysis is evaluated by a comprehensive index, which combines the compactness within the cluster and the separation degree between the clusters to determine the rationality and stability of the clustering result, so as to ensure that the clustering division result of the grouping index data has high distinguishability and consistency.

6. The method of claim 5, wherein, The space division process in the grouping clustering analysis comprises: The data intervals are divided according to the feature dimensions of the grouping index data, a multi-dimensional space index is formed, and the division is terminated when the number of nodes reaches a preset threshold, so as to improve the distinguishability of the spatial distribution characteristics of different wind turbine groups and improve the clustering calculation efficiency.

7. The method of claim 6, wherein, The multi-machine equivalent modeling comprises: regarding the wind turbines in the same group as representative machines with similar frequency response characteristics according to the optimal clustering result; performing capacity-weighted average on the virtual inertia, primary frequency modulation coefficient and active power of the wind turbines in the group, and multiplying the representative machine parameters according to the number of wind turbines in the group, so as to ensure that the frequency characteristics of the obtained multi-machine equivalent model in the inertia response and primary frequency response stage are dynamically consistent with the original group.

8. The method of claim 7, wherein, The method is suitable for frequency dynamic response analysis of wind farms under power shortage, grid disturbance and voltage drop operation conditions. Through calculation and evaluation of the multi-machine equivalent model, frequency support characteristics in the inertia response stage and the primary frequency modulation stage can be analyzed, which is used for guiding frequency stability evaluation and control strategy design of grid-connected wind farms. 9.A computer readable storage medium having stored thereon a computer program, which, when executed by a processor, is configured to perform the method of any one of claims 1 to 8. 10.An electronic device comprising at least one processor and at least one memory having stored therein a computer program, which, when executed, is configured to cause the processor to perform the method of any one of claims 1 to 8.

Images