Wind power plant clustering method based on transient state and steady state rotating speed data

Abstract

The invention relates to the technical field of wind power generation, and discloses a wind power plant clustering method based on transient and steady-state rotating speed data, which comprises the following steps: collecting dynamic data based on fault time sequence key nodes, and extracting five-dimensional feature vectors of transient and steady-state rotating speeds of a doubly-fed wind generator as clustering indexes; noise data is adaptively separated and classified through a density clustering algorithm, and abnormal value interference is eliminated; the homogeneous units are clustered and combined into an equivalent model in combination with a hierarchical modeling strategy, and a detailed model is reserved for heterogeneous units; dynamically fusing multi-dimensional parameters by adopting a capacity weighting method, a wind energy conservation principle and a power weight, calculating equivalent generator, transformer, wind speed and system parameters, and evaluating and optimizing model precision through errors; according to the method, the problems of insufficient dynamic response characterization, abnormal value sensitivity and parameter coupling deficiency in a traditional method are solved, and the simulation precision of a wind power plant equivalent model and the engineering applicability of interaction analysis of a power system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power generation technology, and in particular to a wind farm clustering method based on transient and steady state speed data. BACKGROUND

[0002] With the development of wind power generation technology, the scale of wind farms is expanding. If each wind turbine is modeled separately, the complexity and computation time of the power system simulation model will be greatly increased, and even the risk of "curse of dimensionality" may be faced. However, the operating characteristics and dynamic responses of each wind turbine in a large-scale wind farm are not completely consistent. Therefore, in order to accurately analyze and evaluate the interaction and influence between large-scale wind farms and power systems, it is of great significance to study and find suitable equivalent methods and models for wind farms.

[0003] Domestic and foreign scholars have conducted a lot of research on the equivalent modeling of wind farms. For example, one study proposes a method of using generator speed as a clustering index before the fault. Another method considers different input wind speeds and generator speeds and selects specific wind speeds and generator speeds to construct a feature vector as a clustering index. However, these methods only reflect the operating state of the doubly-fed induction generator (DFIG) before the fault, and do not indicate the influence of fault factors on the operating state. Based on the coherence method, one study selects the generator armature swing curve as a clustering index to improve the effectiveness of clustering division, but this method is difficult to obtain the swing curve of the wind turbine in actual engineering, and therefore is difficult to apply. Existing methods such as K-means clustering algorithm, support vector machine and two-step clustering have been widely used in wind farm clustering. However, through the analysis of the selected clustering data, it can be found that there are often some outliers in the measured data of the wind farm, and it is difficult to predict the distribution shape of high-dimensional data in space, which has a great influence on the accuracy of the clustering results. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a wind farm clustering method and system based on transient and steady state speed data, which combines the transient and steady state data in wind farm simulation, selects the transient and steady state data of the wind turbine speed vector to form a five-dimensional vector as a clustering index, to more accurately reflect the dynamic response characteristics of the wind farm. And using the DBSCAN algorithm based on density, it can effectively identify outliers and classify them into a separate group, thereby improving the accuracy of the clustering results. The method of the present application has general applicability and can accurately represent the external characteristics of wind farms of different scales, and has significant engineering application value.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the application provides a wind farm clustering method based on transient and steady state speed data, comprising:

[0008] Dynamic data acquisition based on key time nodes for first-level index extraction;

[0009] Adaptive separation and classification of key data through algorithm definition;

[0010] Selecting grouping standards for same-type merging modeling;

[0011] Based on the same-type merging modeling results, a model is obtained for parameter calculation.

[0012] As a preferred scheme of the wind farm clustering method based on transient and steady state speed data, wherein: dynamic data acquisition based on key time nodes for first-level index extraction, comprising:

[0013] Synchronous acquisition of measurement data at key nodes in different periods;

[0014] Integrating and extracting measurement data at different time nodes.

[0015] As a preferred scheme of the wind farm clustering method based on transient and steady state speed data, wherein: adaptive separation and classification of key data through algorithm definition, comprising:

[0016] Core point recognition according to a preset threshold;

[0017] Marking data points that do not meet the threshold conditions and independently classifying them.

[0018] As a preferred scheme of the wind farm clustering method based on transient and steady state speed data, wherein: the algorithm definition comprises:

[0019] Finding core points and expanding based on the overall data set;

[0020] Wherein, the expansion includes finding data points connected to the core points and traversing.

[0021] As a preferred scheme of the wind farm clustering method based on transient and steady state speed data, wherein: selecting grouping standards for same-type merging modeling, comprising:

[0022] Selecting grouping standards and collecting clustering indicators to form a data set;

[0023] Classifying the data set using an algorithm;

[0024] Outputting clustering results and outliers, and separately modeling according to the outliers.

[0025] As a preferred scheme of the wind farm clustering method based on transient and steady speed data provided by the application, wherein: the first index extraction comprises:

[0026] Collecting the binary data of the wind turbine speed vector in the wind farm;

[0027] Considering the operating characteristics and dynamic response of the unit, the five-dimensional vector of the two selected time points before the fault, at the beginning of the fault, when the fault is cleared, and after the fault is cleared is selected as the clustering index.

[0028] As a preferred scheme of the wind farm clustering method based on transient and steady speed data provided by the application, wherein: based on the same type of merging modeling results, the model is obtained, and the parameter calculation is performed, comprising:

[0029] Based on the clustering results, the parameters of the same type of units are integrated to generate the equivalent model basic parameters;

[0030] Through capacity and power weight distribution, combined with wind energy conservation, the equivalent wind speed and electrical parameters are calculated;

[0031] Dynamic fusion of multi-dimensional parameters, optimization of equivalent model output characteristics and precision.

[0032] Secondly, the application provides a wind farm clustering system based on transient and steady speed data, comprising:

[0033] The time sequence acquisition module is based on dynamic data acquisition of key time nodes to perform first index extraction;

[0034] The adaptive clustering module realizes adaptive separation and classification of key data through algorithm definition;

[0035] The cluster modeling module selects grouping standards to perform same type merging modeling;

[0036] The equivalent parameter module obtains the model based on the same type of merging modeling results, and performs parameter calculation.

[0037] Thirdly, the application provides an electronic device, comprising:

[0038] A memory and a processor;

[0039] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions, which realizes the steps of the wind farm clustering method based on transient and steady speed data when the computer executable instructions are executed by the processor.

[0040] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which, when executed by a processor, implement the steps of the wind farm clustering method based on transient and steady state speed data.

[0041] Compared with the prior art, the present application has the following beneficial effects: the present application extracts a five-dimensional feature vector of transient and steady state speed of a doubly-fed wind turbine as a clustering index by performing dynamic data acquisition at a key node in time sequence of a wind farm fault, accurately characterizes the dynamic characteristics of each unit, and solves the problem that a traditional modeling method cannot accurately reflect the characteristic differences of units under transient disturbance. In actual wind farm short-circuit faults, the speed change characteristics of different units differ significantly, and the clustering index of the present application can provide high-discrimination multi-dimensional features for subsequent analysis. At the same time, the density clustering algorithm is used to realize adaptive separation and classification of key data, effectively identify and eliminate outliers, solve the problem that traditional clustering algorithms are sensitive to outliers, and ensure the reliability of the clustering results. In addition, combined with the hierarchical modeling strategy and multi-physical quantity collaborative calculation, the balance between homogeneous unit merging modeling and heterogeneous unit independent modeling is realized, the output characteristics and accuracy of the equivalent model are optimized, the reliability and efficiency of power system simulation analysis are improved, the simulation calculation cost of the wind farm connected to the power grid is reduced, and more accurate technical support is provided for power grid dispatching and wind farm planning. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] Figure 1 The wind farm equivalent modeling flowchart of the wind farm clustering method based on transient and steady state speed data according to an embodiment of the present application.

[0044] Figure 2 The wind farm topology structure diagram of the wind farm clustering method based on transient and steady state speed data according to an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0046] Embodiment 1, refer to Figure 1 For an embodiment of the present application, a wind farm clustering method based on transient and steady state speed data is provided, comprising:

[0047] S1: dynamic data acquisition based on key time nodes, first level index extraction;

[0048] S2: self-adaptive separation and classification of key data through algorithm definition;

[0049] S3: selection of grouping standards, same type merging modeling;

[0050] S4: based on the same type merging modeling result, the model is obtained, and parameter calculation is performed.

[0051] It should be noted that the existing wind farm equivalent modeling method depends on single steady state data or ignores the fault dynamic response, which leads to the fact that the model cannot accurately reflect the characteristic differences of the unit under transient disturbance; at the same time, the traditional clustering algorithm is sensitive to abnormal values and difficult to process high-dimensional data distribution, which is easy to introduce modeling deviation. In addition, the coupling relationship between the dynamic characteristics of the unit and the physical constraints is not fully considered in the parameter calculation, which further affects the engineering applicability of the equivalent model.

[0052] Therefore, in order to solve the problems of insufficient dynamic response representation, abnormal value interference and missing parameter fusion mechanism, through the steps of S1-S4, first, in S1, five-dimensional speed vectors are extracted based on key nodes in time sequence under fault, which can fully capture the dynamic characteristics of the unit; S2 uses density clustering algorithm to adaptively separate noise data, eliminating the influence of abnormal values on clustering results; S3 implements the hierarchical strategy of same type unit merging modeling and heterogeneous unit independent modeling according to the clustering results, which improves the model accuracy; S4 realizes dynamic equivalent parameter fusion through capacity weighting, wind energy conservation and power weight distribution, etc. Finally, a high-fidelity wind farm equivalent model is constructed, which significantly improves the reliability and efficiency of power system simulation analysis.

[0053] Embodiment 2, refer to Figure 1 For an embodiment of the present application, based on the above embodiment, a wind farm clustering method based on transient and steady state speed data is provided.

[0054] In the present application, in step S1, dynamic data acquisition based on key time nodes, first level index extraction, through synchronous acquisition of speed data of double-fed wind generator at five time nodes before fault, initial stage of fault, fault clearing, 0.2 seconds and 0.4 seconds after fault clearing, multi-time point measurement values are integrated to construct five-dimensional vector representing dynamic response as clustering index, and preprocessing is performed on abnormal values and high-dimensional distribution characteristics in the data, forming a core feature set reflecting the dynamic characteristics and fault response differences of wind farm units.

[0055] In an alternative embodiment, the dynamic data acquisition based on the key time node in step S1, the first-level index extraction can also be performed by acquiring the voltage-speed joint parameters before the fault, the transient peak in the fault and the recovery period after the fault, combining the data change rate in different time windows, constructing a multi-dimensional dynamic feature matrix, and eliminating short-time disturbance noise through a sliding window filtering algorithm to form a fusion index set representing the dynamic response of the unit.

[0056] In another alternative embodiment, the dynamic data acquisition based on the key time node in step S1, the first-level index extraction can also be performed by extracting the mechanical torque and electromagnetic torque difference sequence before the fault, during the fault and in the steady state period after the fault, extracting multi-scale frequency domain features based on wavelet transform, generating low-dimensional feature vectors after dimensionality reduction based on principal component analysis (PCA), and inputting the low-dimensional feature vectors as the input index of the clustering analysis.

[0057] In the embodiments of the present application, the dynamic data acquisition based on the key time node in step S1, the first-level index extraction also includes:

[0058] The extraction of the clustering index is mainly aimed at the dynamic response characteristics of the doubly-fed wind generator. When a short-circuit fault occurs on the system side, the change of the terminal voltage will cause the change of the electromagnetic torque, while the mechanical torque is usually considered to remain unchanged. Under the action of the unbalanced torque, the speed of the generator will change, and therefore the operating characteristics of the wind farm can be represented by the speed of the asynchronous generator. Since the system usually experiences a sudden change when the fault is cleared, the dynamic response of the wind farm is the strongest, and the differences between different units are also the most obvious. Therefore, considering the operating characteristics and dynamic response of the units, a five-dimensional vector at five time points, i.e., before the fault, at the beginning of the fault, when the fault is cleared, 0.2 seconds and 0.4 seconds after the fault is cleared, is selected as the clustering index.

[0059] In the embodiments of the present application, the adaptive separation and classification of the key data in step S2 are realized by the algorithm definition. Through the spatial clustering algorithm of density, the neighborhood radius and the density threshold are defined to dynamically divide the data density distribution, identify the core points and expand their density reachable regions to form high-density clustering clusters. At the same time, the non-core points and noise points that do not meet the density condition are independently marked and classified to realize the automatic separation of abnormal data, eliminate the interference of abnormal values in high-dimensional data on the clustering results, and thus improve the accuracy and clustering reliability of the representation of the dynamic response characteristics of the wind farm.

[0060] In an alternative embodiment, the adaptive separation and classification of key data defined by the algorithm in step S2 can also be achieved by a dynamic threshold adjustment algorithm based on hierarchical clustering, which automatically divides the hierarchical structure according to the data distribution density, merges the low-density region data clusters layer by layer and eliminates isolated points, dynamically optimizes the clustering boundary by combining the sliding window mechanism, and realizes the multi-scale classification and outlier filtering of high-dimensional rotating speed data.

[0061] In another alternative embodiment, the adaptive separation and classification of key data defined by the algorithm in step S2 can also be achieved by a graph partitioning algorithm based on spectral clustering, which constructs a similarity matrix of wind farm rotating speed data, maps a low-dimensional space using eigenvalue decomposition of the Laplacian matrix, identifies data subgroups and separates outlier nodes through graph partitioning, and optimizes the classification robustness under nonlinear data distribution by combining kernel functions.

[0062] In the embodiments of the present application, the adaptive separation and classification of key data defined by the algorithm in step S2 further includes:

[0063] The DBSCAN algorithm based on wind farms is a classic density-based clustering algorithm, which defines a class as a high-density target region separated by a low-density region in the data space. First, the algorithm involves several core concepts: Eps represents the neighborhood radius of the spatial data points, MinPts represents the density threshold, dist(p,q) represents the distance between points p and q, and D represents the data set. The definition of "neighborhood" is the set of points contained in the circular region with Eps as the radius in the data space; "core point and non-core point" describes the number of points in the neighborhood of a point p under a given density threshold MinPts. If the density of point p is greater than MinPts, the point is called a core point, otherwise it is called a non-core point. The definition of "directly density reachable" is that if point p is in the neighborhood of point q and q is a core point, then point p is directly density reachable to point q. "Density reachable" means that if there is a string of objects p1, p2,..., p n , where pi can be directly density reachable to p(i+1), and if p1=p, p n =q, then q is density reachable to p. The definition of "density connected" is that if there is a point o from which point p and q are both density reachable, then point p and q are density connected. The definition of "class" is that the non-empty set C of the data set D is a class, which satisfies the following conditions: for any points p and q in the space, if p belongs to C and q is density reachable to p, then q also belongs to C; for any points p and q in the space, if p belongs to C and q belongs to C, then p and q are density connected. Finally, "noise points" are described as points in the data set D that do not belong to any class and are called noise points.

[0064] The basic idea of DBSCAN algorithm is to scan the entire data set D, find any core point, and then expand the core point. The expansion method is to find all the data points that are density connected to the core point, traverse all other core points in the core point neighborhood, find points that are density connected to these data points, and stop until there are no expandable data points. Then re-scan the clusters that do not form data clusters, find the core points that are not clustered, repeat the above steps to expand the core points, until there are no new core points in the data set. The points in the data set that are not included in any cluster are considered as noise points, i.e. outliers. The noise points are the abnormal points of the wind turbine, and their running state is greatly different from that of any wind turbine group, so they should be classified as a separate group.

[0065] In the embodiments of the present application, the grouping criteria are selected in step S3, and the same type of merging modeling further includes:

[0066] First, the grouping criteria are selected and the clustering indicators are collected, i.e. the transient and steady-state data of the wind turbine speed vector in the entire wind farm are collected, and the data set D is formed by the data of each wind turbine. Next, the data set of the wind farm is classified using the DBSCAN algorithm, which is a classic clustering method based on density, by finding core points and expanding core points to identify all density-connected data points. Then, the clustering results and outliers are output, the equivalent model of the same type of unit is constructed, and for the deviated unit, it is modeled separately according to its detailed model and connected to the power system using the same topology structure as the detailed model. Finally, the equivalent parameters are calculated, the same model and parameters are used for the wind power units connected to the same bus, and the capacity-weighted method is used to calculate the equivalent parameters of the generators and transformers, and the equivalent wind speed is calculated based on the principle that the total input wind energy of the wind turbines is equal, and then the equivalent parameters of the equivalent system are calculated.

[0067] The main process is as follows:

[0068] The grouping criteria are selected and the clustering indicators are collected: the transient and steady-state data of the wind turbine speed vector in the entire wind farm are collected, and the data set D is formed by the data of each wind turbine, and the data of each wind turbine is taken as a data point.

[0069] The DBSCAN algorithm is used to classify the wind farm data set D: DBSCAN is a density-based clustering algorithm that classifies data points in the data set by identifying core points and expanding core points.

[0070] The clustering results and outliers are output: according to the clustering results, the equivalent model of the same type of unit is constructed; for the deviated unit, it is modeled separately according to its detailed model; the equivalent system uses the same topology structure as the detailed model to connect to the power system.

[0071] In the embodiment of the present application, the model is obtained based on the same type merging modeling result in step S4, and the parameter calculation further comprises:

[0072] The equivalent parameters of the generator and transformer, the equivalent parameters of the shaft, the equivalent parameters of the wind speed, the equivalent parameters of the equivalent system, and the equivalent model evaluation index;

[0073] The equivalent parameters of the generator and transformer are:

[0074] For the wind power generator units connected to the same bus, the same model and parameters are used. The capacity weighting method is used to calculate the equivalent parameters of the generator and transformer. The calculation of the equivalent wind speed is based on the principle that the total input wind energy of the equivalent wind turbine is equal to that before the equivalent. The calculation of the equivalent parameters of the equivalent system is weighted by the power of each wind turbine, and the integral voltage of the equivalent wind turbine is equal to the weighted average of the integral voltage of the same group of wind turbines in the original wind farm. The specific calculation formula is as follows:

[0075]

[0076] wherein S Gi and Z Gi represent the rated capacity and generator impedance of the i th unit, respectively; S Ti and Z Ti represent the rated capacity and generator impedance of the i th unit, respectively; n represents the number of units.

[0077] The equivalent parameters of the shaft are represented as:

[0078]

[0079] wherein H ti , H gi , K i and D i represent the inertia time constant, rotor inertia time constant, shaft stiffness coefficient and shaft damping coefficient of the wind power generator unit i, respectively.

[0080] The equivalent parameters of the wind speed are:

[0081] The equivalent parameters of the wind speed are calculated based on the principle that the total input wind energy of the wind power generator is equal, and the formula is as follows:

[0082]

[0083] wherein A represents the swept area of the equivalent wind power generator; A i , c pi and v i represent the swept area, wind energy utilization coefficient and input wind speed of the unit i, respectively.

[0084] The equivalent parameters of the equivalent system are:

[0085] The equivalent parameters of the equivalent system are calculated by considering the power weight of each wind generator, specifically, multiplying the branch line impedance of each unit by the total power passing through it, and dividing by the total power passing through the equivalent impedance, as follows:

[0086]

[0087] Wherein, Z li represents the branch line impedance of unit i; P Zi represents the total power passing through the impedance Z li ; P Zs represents the total power passing through the equivalent impedance Z eq .

[0088] Wherein, the equivalent model evaluation index is:

[0089] The error evaluation index of the equivalent model of the wind farm determined according to the detailed simulation results of the wind farm is

[0090]

[0091] Wherein P i , represent the active power at the outlet of the detailed model and the equivalent model.

[0092] In summary, the present application proposes a high-precision wind farm equivalent modeling method by fusing five-dimensional dynamic speed data acquisition of double-fed wind generators at key nodes in the fault time sequence, adaptive outlier separation and classification based on density clustering (DBSCAN), hierarchical equivalent modeling, and multi-physical quantity collaborative parameter calculation. This method uses capacity weighting method to calculate equivalent generator and transformer parameters, deduces equivalent wind speed based on the principle of wind energy conservation, combines power weight distribution branch line impedance, and introduces inertia time constant, shaft stiffness coefficient and other dynamic parameters for weighted fusion, to realize accurate calculation of multi-dimensional equivalent parameters. At the same time, through the error evaluation index of the active power at the outlet of the equivalent model and the detailed model, the modeling process is dynamically optimized, effectively solving the problems of insufficient dynamic response representation, abnormal value interference and missing parameter coupling in traditional methods, and significantly improving the simulation accuracy and engineering applicability of the equivalent model of the wind farm.

[0093] Embodiment 3, refer to Figure 2 As an embodiment of the application, a double high characteristic power transient stability analysis method is provided. In order to verify the beneficial effects of the present application, economic benefit calculation and simulation experiments are carried out for scientific demonstration.

[0094] The wind farm consists of 36 doubly-fed wind turbines of the same model. The doubly-fed wind turbines use constant power factor control, and the interval between adjacent doubly-fed wind turbines is 400 meters. The generator terminal voltage is 690V. After being boosted to 35kV by the transformer, it is connected to the intermediate voltage bus by the overhead line, and then boosted to 220kV by the main transformer of the wind farm, and finally connected to the power system through a double circuit. On the DIgSILENT / PowerFactory platform, a detailed model and an equivalent model of the wind farm were established. The detailed model includes the single-unit model of the 36 units in the wind farm, the equivalent model between units, the generator terminal transformer and the main transformer model. The wind farm topology is as follows: Figure 2 As shown in the figure, the input wind speed is 12 m / s. Transient simulations typically do not consider wind speed variations, but only consider the distribution of wind farms at different locations and the distribution of wind turbines at different wind speeds. Therefore, it is necessary to consider the impact of the wake effect between turbines on the input wind speed. Point B is set at the wind farm outlet. A three-phase short-circuit fault occurs within 5 seconds and clears after 150 ms. Transient steady-state data is then collected for the velocity vectors of 36 wind turbines.

[0095] The DBSCAN clustering method is used to select the transient and steady-state data of the velocity vector as the clustering criterion to classify the wind turbines. The clustering results are shown in Table 1.

[0096] Table 1 Clustering results based on DBSCAN clustering (36 DFIGs)

[0097]

[0098] As shown in Table 2, the power error of the wind farm equivalent model using the DBSCAN clustering method is smaller than that of the single model.

[0099] Table 2 Error evaluation indicators of multi-machine equivalent model and single-machine model based on DBSCAN

[0100]

[0101] The results show that the DBSCAN clustering method can effectively solve the problem of outlier sets and noise points. The wind farm equivalent model based on these clustering results will be more accurate and can be used to more accurately describe the characteristics of the wind farm.

[0102] Embodiment 4, the above is a schematic scheme of a wind farm clustering method based on transient and steady state speed data. It should be noted that the technical scheme of the wind farm clustering system based on transient and steady state speed data is the same as the technical scheme of the wind farm clustering method based on transient and steady state speed data described above, and the technical details of the wind farm clustering system based on transient and steady state speed data in this embodiment are not described in detail. The description of the technical scheme of the wind farm clustering method based on transient and steady state speed data described above can be referred to.

[0103] The embodiment also provides a wind farm clustering system based on transient and steady state speed data, comprising:

[0104] The time sequence acquisition module performs first-level index extraction based on dynamic data acquisition of key time nodes.

[0105] The adaptive clustering module realizes adaptive separation and classification of key data through algorithm definition.

[0106] The cluster modeling module selects grouping standards and performs same-type merging modeling.

[0107] The equivalent parameter module obtains a model based on the same-type merging modeling result and performs parameter calculation.

[0108] The embodiment also provides an electronic device suitable for the wind farm clustering based on transient and steady state speed data, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the wind farm clustering method based on transient and steady state speed data proposed in the above embodiment.

[0109] The embodiment also provides a storage medium having a computer program stored thereon, and the program is executed by a processor to realize the wind farm clustering method based on transient and steady state speed data proposed in the above embodiment.

[0110] The storage medium proposed in the embodiment and the wind farm clustering method based on transient and steady state speed data proposed in the above embodiment belong to the same inventive concept, and the technical details not described in detail in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0111] Those skilled in the art can clearly understand the present application by the above description of the embodiments, and the present application can be realized by software and necessary general hardware, and of course, can also be realized by hardware. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk, or an optical disc, and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of various embodiments of the present application.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. A wind farm clustering method based on transient and steady-state speed data, characterized in that: include: Based on dynamic data collection at key time nodes, primary indicators are extracted; Adaptive separation and classification of key data is achieved through algorithm definition; Select grouping criteria and conduct similar merger modeling; Based on the similar merge modeling results, the model is obtained and the parameters are calculated.

2. The wind farm clustering method based on transient and steady-state speed data according to claim 1, characterized in that: Based on dynamic data collection at key time nodes, primary indicators are extracted, including: Synchronously collect measurement data at key points in different periods; Integrate and extract measurement data from different time nodes.

3. The wind farm clustering method based on transient and steady-state speed data according to claim 2, characterized in that: Adaptive separation and classification of key data is achieved through algorithm definition, including: Perform core point identification based on the preset threshold; Data points that do not meet the threshold conditions are marked and classified independently.

4. The wind farm clustering method based on transient and steady-state speed data according to claim 3, characterized in that: The algorithm definition includes: Based on the entire data set, find the core points and expand them; The expansion includes finding data points connected to the core points and traversing them.

5. The wind farm clustering method based on transient and steady-state speed data according to claim 4, characterized in that: Select grouping criteria and conduct similar merger modeling, including: Select grouping criteria and collect clustering indicators to form a data set; Use algorithms to classify datasets; Output clustering results and outliers, and perform separate modeling based on the outliers.

6. The wind farm clustering method based on transient and steady-state speed data according to claim 5, characterized in that: The first-level indicator extraction includes: Collecting binary data of wind turbine speed vectors in a wind farm; Taking into account the operating characteristics and dynamic response of the unit, the five-dimensional vectors at two selected time points, namely before the fault, at the initial stage of the fault, when the fault is cleared, and after the fault is cleared, are selected as clustering indicators.

7. The wind farm clustering method based on transient and steady-state speed data according to claim 6, characterized in that: Based on the similar merge modeling results, the model is obtained and the parameters are calculated, including: Based on the clustering results, the parameters of the same type of units are integrated to generate the basic parameters of the equivalent model; By allocating capacity and power weights and combining wind energy conservation, equivalent wind speed and electrical parameters are calculated; Dynamically fuse multi-dimensional parameters to optimize the output characteristics and accuracy of equivalent models.

8. A wind farm clustering system based on transient and steady-state speed data, applying the method according to any one of claims 1 to 7, characterized in that: include: The time series acquisition module collects dynamic data at key time nodes and extracts primary indicators; Adaptive clustering module, which realizes adaptive separation and classification of key data through algorithm definition; Cluster modeling module, select grouping criteria and conduct similar merging modeling; The equivalent parameter module obtains the model based on the similar merge modeling results and performs parameter calculation.

9. An electronic device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the wind farm clustering method based on transient and steady-state speed data according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the wind farm clustering method based on transient and steady-state speed data according to any one of claims 1 to 7.

Images