Direct-drive wind power plant low voltage ride through equivalent modeling method and system based on AP clustering
By using AP clustering, the turbines of a direct-drive wind farm are divided into multiple clusters, and an equivalent model is established. This solves the problem of excessively long simulation time for large-scale wind farms, achieves efficient equivalent modeling of wind farms, and improves simulation accuracy and operational efficiency.
Patent Information
- Application Number
- CN202410890941.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-04
- Publication Date
- 2026-01-13
AI Technical Summary
Detailed modeling of large-scale wind farms leads to excessively long simulation times and low operational efficiency, failing to meet engineering requirements. Furthermore, existing technologies have not been able to effectively perform equivalent modeling of wind farms to reflect their dynamic characteristics.
Using AP-based clustering, the turbines of a direct-drive wind farm are divided into multiple groups through initial and secondary clustering. An equivalent model is established, and the clustering is performed using the operation of the unloading circuit and the characteristics of the turbine voltage and active power. The effectiveness of the model is verified by combining it with an electromechanical transient simulation model.
It improves the accuracy and efficiency of the simulation model, accurately reflects the low-voltage ride-through characteristics of wind farms, and shortens the simulation time.
Smart Images
Figure CN121328255A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of wind farm low voltage ride through equivalent modeling, in particular to a direct-drive wind farm low voltage ride through equivalent modeling method and system based on AP clustering. BACKGROUND
[0002] At present, with the increasing scale of wind farms, due to the great difference in the operation state of each wind turbine, if a large wind farm is modeled in detail, the simulation time will be too long and the operation efficiency will be too low, which does not meet the actual engineering requirements. The "Wind Power Grid Operation Control Technology Regulation" clearly proposes that in the simulation calculation, the wind farm equivalent modeling reflecting the dynamic characteristics of the wind farm should be adopted for a single wind farm. The wind farm equivalent modeling can greatly reduce the complexity of the model and shorten the simulation time under the premise of more accurately expressing the dynamic characteristics of the wind farm. Therefore, establishing an equivalent model of the wind farm meeting the accuracy requirements is a problem to be solved at present. SUMMARY
[0003] In view of the above technical problems, the present application provides a direct-drive wind farm low voltage ride through equivalent modeling method based on AP clustering, comprising:
[0004] determining the characteristic quantity of the direct-drive wind farm unit; according to the characteristic quantity, the direct-drive wind farm unit is divided into a unit group whose unloading circuit has been turned on and a unit group whose unloading circuit has not been turned on through primary grouping;
[0005] for the unit group whose unloading circuit has been turned on, the AP clustering algorithm is used for secondary grouping based on the voltage and active power characteristics of the wind turbine, so as to divide the unit group whose unloading circuit has been turned on into multiple unit groups; and for the unit group whose unloading circuit has not been turned on, it is classified into the same unit group;
[0006] an equivalent model of the direct-drive wind farm low voltage ride through is established according to the equivalent parameters of each unit group.
[0007] Further, the grouping index of the primary grouping is the action of the unloading circuit; and the grouping index of the secondary grouping is the voltage at the end of the wind turbine during the low voltage ride through and the active power output of the single wind turbine before the fault.
[0008] Further, the characteristic quantity of the direct-drive wind farm unit comprises the action of the unloading circuit and the operation state of each wind turbine during the low voltage ride through.
[0009] Further, for the unit group whose unloading circuit has been turned on, the AP clustering algorithm is used for secondary grouping based on the voltage and active power characteristics of the wind turbine, so as to divide the unit group whose unloading circuit has been turned on into multiple unit groups, comprising:
[0010] For the unloading circuit has been turned on unit, collect the low voltage ride through data of each wind turbine, form a data set {X1, X2, …, XN} N};
[0011] Calculate the similarity matrix S between each wind turbine, and the diagonal element s(i,j) is expressed as s(i,j)=-||X i -X j || 2 , wherein the greater the diagonal element of s matrix, the greater the possibility of becoming a cluster center;
[0012] Define the attraction degree r(i,k) to represent the appropriate degree of data point k as the cluster center of data point i, and the appropriate degree of data point i selecting data point k as its cluster center is represented by the defined membership information a(i,k), and the cluster center is determined in an iterative manner, and the specific expression for iteration is:
[0013] r(i,k)=s(i,k)-max k′≠k (a(i,k′)+s(i,k′))
[0014]
[0015] In the formula, k' and i' are data point numbers;
[0016] In each round of iteration, all data points with r(k,k)+a(k,k)>0 are regarded as potential cluster centers, and the iteration stops when the cluster center remains unchanged for a specified number of consecutive iterations or reaches the maximum number of iterations;
[0017] The points in the data set that are not cluster centers will be assigned to the cluster center with the highest similarity, and the clustering of the data set is completed.
[0018] Further, after the step of establishing the low voltage ride through equivalent model of the direct drive wind farm, it further comprises:
[0019] Collect fault simulation waveform data under different working conditions through the electromechanical transient simulation model;
[0020] Compare the data of the simulation model and the equivalent model under different fault conditions to verify the effectiveness of the equivalent method.
[0021] The application also provides a direct drive wind farm low voltage ride through equivalent modeling system based on AP clustering, comprising:
[0022] The initial clustering module is used to determine the characteristic quantity of the direct drive wind farm unit, and the direct drive wind farm unit is divided into unloading circuit already turned on units and unloading circuit not yet turned on units through initial clustering according to the characteristic quantity;
[0023] a secondary clustering module, configured to, for the wind turbine units whose unloading circuits have been turned on, perform secondary clustering by an AP clustering algorithm based on the voltage and active power characteristics of the wind turbines, so as to divide the wind turbine units whose unloading circuits have been turned on into a plurality of groups; and for the wind turbine units whose unloading circuits have not been turned on, group them into one group;
[0024] an equivalent model establishing module, configured to establish an equivalent model of low voltage ride through of the direct-drive wind farm according to the equivalent parameters of each group.
[0025] Further, the clustering index of the primary clustering is the action condition of the unloading circuit; and the clustering index of the secondary clustering is the voltage at the end of the wind turbine during the low voltage ride through and the active power of the single wind turbine before the fault.
[0026] Further, the characteristic quantity of the wind turbine unit of the direct-drive wind farm includes the action condition of the unloading circuit and the operation state of each wind turbine during the low voltage ride through.
[0027] Further, the secondary clustering module includes:
[0028] a data set forming submodule, configured to, for the wind turbine units whose unloading circuits have been turned on, collect low voltage ride through data of each wind turbine unit to form a data set {X1, X2, …, Xn} of the wind turbine units; N};
[0029] a similarity matrix calculating submodule, configured to calculate a similarity matrix S between each wind turbine unit, wherein a non-diagonal element s(i,j) is expressed as s(i,j) = -||X i -X j || 2 , wherein the greater the diagonal element of the s matrix, the greater the possibility of becoming a clustering center;
[0030] a clustering center determining submodule, configured to define an attraction degree r(i,k) for the appropriate degree of data point k as a clustering center of data point i, and the appropriate degree of data point i selecting data point k as a clustering center is represented by defined belonging degree information a(i,k), and the clustering center is determined in an iterative manner, and a specific expression for iteration is:
[0031] r(i,k) = s(i,k) - max k′≠k (a(i,k') + s(i,k'))
[0032]
[0033] In the formula, k' and i' are data point numbers;
[0034] The iterative submodule is used in each iteration to treat all data points r(k,k)+a(k,k)>0 as potential cluster centers. The iteration stops when the cluster centers remain unchanged for a specified number of consecutive iterations or when the maximum number of iterations is reached.
[0035] The clustering submodule is used to assign points in the dataset that are not cluster centers to the cluster centers with the highest similarity to them, thus completing the clustering of the dataset.
[0036] Furthermore, it also includes:
[0037] The waveform data collection module is used to collect fault simulation waveform data under different operating conditions through the electromechanical transient simulation model.
[0038] The verification submodule is used to compare the data of the simulation model and the equivalent model under different fault conditions to verify the effectiveness of the equivalent method.
[0039] This invention provides a method and system for low voltage ride-through equivalent modeling of direct-drive wind farms based on AP clustering. By using the AP clustering algorithm, the generator units of a direct-drive wind farm are divided into multiple groups and equivalent modeling is performed, which improves the simulation accuracy and operating efficiency of the simulation model. Attached Figure Description
[0040] Figure 1 This is a flowchart illustrating an equivalent modeling method for low-voltage ride-through in direct-drive wind farms based on AP clustering, provided by the present invention.
[0041] Figure 2 This invention relates to a direct-drive fan structure;
[0042] Figure 3 This is a schematic diagram of the equivalent unit structure involved in the present invention;
[0043] Figure 4 This invention relates to a cluster partitioning flowchart;
[0044] Figure 5 This is a schematic diagram of the structure of an equivalent modeling system for low voltage ride-through of a direct-drive wind farm based on AP clustering provided by the present invention. Detailed Implementation
[0045] Numerous specific details are set forth in the following description to provide a full understanding of the invention. However, the invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0046] Figure 1This is a flowchart illustrating an equivalent modeling method for low-voltage ride-through in direct-drive wind farms based on AP clustering, provided by this invention. The following is a summary of the process. Figure 1 The method provided by this invention will be described in detail below. For example... Figure 1 As shown, it includes the following steps:
[0047] Step S101: Determine the characteristic quantities of the direct-drive wind farm units; based on the characteristic quantities, the direct-drive wind farm units are initially grouped into units whose unloading circuit has been turned on and units whose unloading circuit has not been turned on.
[0048] direct drive fan structure as follows Figure 2 As shown, firstly, the characteristic quantities of each unit in the direct-drive wind farm are determined, such as the operation status of the unloading circuit, to reflect the operating status of each turbine during low-voltage ride-through. Based on the characteristic quantities, such as the voltage difference on the DC bus capacitor or the operation status of the unloading circuit, initial grouping is performed. The grouping process is as follows: Figure 4 As shown, the wind turbine is divided into two parts: one where the unloading circuit is already connected and the other where it is not. The initial grouping index is the operation status of the unloading circuit.
[0049] Regarding the selection of clustering indicators, when a wind farm experiences a fault, the presence of full-power back-to-back power electronic converters isolates the PMSG from the grid side. At the same time, the wind turbine has a large inertia, and the speed of the wind turbine and permanent magnet synchronous generator changes little under different operating conditions. Therefore, using the unit's speed as a clustering indicator is difficult to accurately represent the unit's operating characteristics.
[0050] The resulting unbalanced power varies depending on the depth of the fault, leading to significant differences in the voltage across the DC bus capacitor of each wind turbine under different operating conditions. The operation of the unit's unloading circuit can fully reflect the energy storage status of the DC bus capacitor and the consumption of unbalanced power on both sides of the DC bus by the unloading circuit.
[0051] When the fault occurs at low wind speeds, the mechanical energy captured by the wind turbine is relatively small, resulting in a small unbalanced power across the DC bus capacitor. The voltage across the DC bus capacitor does not reach the voltage threshold of the unloading circuit, indicating that the GSC still has the ability to adjust active power. However, when the fault occurs at high wind speeds, the mechanical power captured by the wind turbine is larger, leading to a larger unbalanced power across the DC bus capacitor. After charging, the DC bus capacitor generates an overvoltage, reaching the unloading circuit's activation threshold. This indicates that the GSC has lost control of active power due to the presence of the current-limiting circuit. Therefore, in this paper, the initial clustering index is selected as the operation status of the unloading circuit, dividing the wind turbine into two parts: one where the unloading circuit is already conducting and the other where it is not.
[0052] When short-circuit fault occurs in the grid side, because of the isolation of the back-to-back converter, the terminal voltage of each unit remains unchanged. At the same time, due to the different electrical distances between each wind turbine and the point of common coupling (PCC) in the wind farm, the voltage drop of each unit during the fault period is different. The dynamic characteristics of the unit during the low voltage ride through period can be represented by the terminal voltage of the wind turbine. Similarly, the active power output of the wind turbine before the fault has a great influence on the duration of the active power recovery period. Therefore, for the units whose unloading circuits are not turned on, the secondary grouping index is selected as the terminal voltage of the wind turbine during the low voltage ride through period and the active power output of the single wind turbine before the fault.
[0053] In step S102, for the units whose unloading circuits have been turned on, the AP clustering algorithm is used for secondary grouping based on the voltage and active power characteristics of the wind turbine, and the units whose unloading circuits have been turned on are divided into multiple groups; for the units whose unloading circuits have not been turned on, they are grouped into the same group.
[0054] The secondary grouping index is the terminal voltage of the wind turbine during the low voltage ride through period and the active power output of the single wind turbine before the fault.
[0055] The group division process is as shown in Figure 4 For the units whose unloading circuits have not been turned on, they are grouped into the same group.
[0056] For the units whose unloading circuits have been turned on, the terminal voltage of the wind turbine and the active power output before the fault are used as the secondary grouping index. The AP (Affinity Propagation) algorithm is used for secondary grouping based on the voltage and active power characteristics of the wind turbine to determine the clustering attribution of each data point (i.e. each wind turbine). The specific steps are as follows:
[0057] For the units whose unloading circuits have been turned on, the low voltage ride through data of each wind turbine is collected to form a data set {X1, X2, …, Xn} ; N};
[0058] The similarity matrix S between each wind turbine is calculated, and the non-singular s(i,j) is expressed as s(i,j) = -||X i -X j || 2 , wherein the greater the diagonal element of the s matrix, the greater the possibility of becoming a clustering center;
[0059] The attraction degree r(i,k) is defined to represent the appropriate degree of data point k as the clustering center of data point i, and the appropriate degree of data point i selecting data point k as its clustering center is represented by the defined attribution information a(i,k). The clustering center is determined by iteration, and the specific expression for iteration is:
[0060] r(i,k) = s(i,k) - maxk′≠k (a(i,k') + s(i,k'))
[0061]
[0062] wherein k' and i' are data point numbers;
[0063] In each iteration, all data points with r(k,k) + a(k,k) > 0 are considered as potential cluster centers, and the iteration stops when the cluster centers remain unchanged for a specified number of consecutive iterations or the maximum number of iterations is reached;
[0064] The points in the data set that are not cluster centers will be assigned to the cluster center with the highest similarity, and the clustering of the data set is completed.
[0065] In step S103, an equivalent model of low voltage ride through of the direct-driven wind farm is established according to the equivalent parameters of each cluster.
[0066] The equivalent parameters of each cluster are calculated, and an equivalent unit of each cluster and an equivalent model of low voltage ride through of the direct-driven wind farm are established according to the equivalent parameters. The equivalent unit structure is shown in Figure 3 Specifically, an electromechanical transient model can be established in simulation software such as DIgSILENT / PowerFactory, and simulation waveform data is collected through fault simulation under different conditions. By comparing the responses of the detailed model and the equivalent model under different fault conditions, the effectiveness of the equivalent method is verified, and it is ensured that the model can accurately reflect the low voltage ride through characteristics.
[0067] Based on the same inventive concept, the application also provides an AP clustering-based equivalent modeling system 500 for low voltage ride through of a direct-driven wind farm, as shown in Figure 5 The system comprises:
[0068] A primary clustering module 510 is configured to determine characteristic quantities of units of the direct-driven wind farm, and divide the units into units with turned-on unloading circuits and units with not-yet-turned-on unloading circuits through primary clustering according to the characteristic quantities.
[0069] A secondary clustering module 520 is configured to divide the units with turned-on unloading circuits into multiple clusters through AP clustering algorithm based on voltage and active power characteristics of the wind turbines, and divide the units with not-yet-turned-on unloading circuits into a same cluster.
[0070] An equivalent model establishing module 530 is configured to establish an equivalent model of low voltage ride through of the direct-driven wind farm according to equivalent parameters of each cluster.
[0071] Further, the grouping index of the primary grouping is the action condition of the unloading circuit; the grouping index of the secondary grouping is the fan terminal voltage during low voltage ride through and the active power of the single fan before the fault.
[0072] Further, the characteristic quantity of the direct-drive wind farm unit includes: the action condition of the unloading circuit, and the operation state of each fan during low voltage ride through.
[0073] Further, the secondary grouping module includes:
[0074] The data set forming submodule is configured to collect low voltage ride through data of each wind power unit for the unit whose unloading circuit has been turned on, and form a data set {X1, X2, …, Xn} of the wind power units. N};
[0075] The similarity matrix calculating submodule is configured to calculate a similarity matrix S between the wind power units, and a non-diagonal element s(i,j) of the similarity matrix S is expressed as s(i,j)=-||X i -X j || 2 wherein the greater the diagonal element of the s matrix, the greater the possibility of becoming a clustering center;
[0076] The clustering center determining submodule is configured to define an attraction degree r(i,k) for the appropriate degree of data point k as a clustering center of data point i, and the appropriate degree of data point i selecting data point k as a clustering center is represented by defined belonging degree information a(i,k), and the clustering center is determined in an iterative manner, and a specific expression for iteration is:
[0077] r(i,k)=s(i,k)-max k′≠k (a(i,k′)+s(i,k′))
[0078]
[0079] wherein k' and i' are data point numbers;
[0080] The iteration submodule is configured to regard all data points with r(k,k)+a(k,k)>0 as potential clustering centers in each round of iteration, and the iteration is stopped when the clustering center remains unchanged in continuous specified number of iterations or reaches a maximum number of iterations.
[0081] The clustering submodule is configured to assign the points without clustering centers in the data set to the clustering center with the highest similarity, and complete clustering of the wind power units in the data set.
[0082] Further, the secondary grouping module includes:
[0083] A waveform data collection module is configured to collect fault simulation waveform data under different working conditions through the electromechanical transient simulation model.
[0084] A verification submodule is configured to compare the data of the simulation model and the equivalent model under different fault conditions to verify the effectiveness of the equivalent method.
[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement should be covered in the scope of the claims of the present application.
Claims
1. A method for equivalent modeling of low voltage ride through (LVRT) of direct drive wind farm based on AP clustering, characterized in that, The method comprises the following steps: determining characteristic quantities of the direct-drive wind turbine generators; grouping the direct-drive wind turbine generators into two groups, i.e., a group of wind turbine generators with the unloading circuit turned on and a group of wind turbine generators with the unloading circuit turned off, according to the characteristic quantities; grouping the wind turbine generators with the unloading circuit turned on into multiple groups according to the voltage and active power characteristics of the wind turbine generators by using an AP clustering algorithm; grouping the wind turbine generators with the unloading circuit turned off into one group; 2. The method of claim 1, wherein, establishing a low-voltage ride-through equivalent model of the direct-drive wind farm according to the equivalent parameters of each group.
3. The method of claim 1, wherein, The grouping index of the primary grouping is the action of the unloading circuit, and the grouping index of the secondary grouping is the voltage at the end of the wind turbine during the low-voltage ride-through and the active power of the single wind turbine before the fault.
4. The method of claim 1, wherein, The characteristic quantities of the direct-drive wind turbine generators include the action of the unloading circuit and the operating state of each wind turbine during the low-voltage ride-through. For the unloading circuit has been turned on unit, the low voltage ride through data of each wind turbine is collected to form a data set {X1, X2, …, X N} A similarity matrix S between the wind turbines is calculated, with elements s(i,j) expressed as s(i,j) = -||X i -X j ‖ 2 where the greater the diagonal elements of the s matrix, the greater the likelihood of becoming a cluster center. The grouping of the wind turbine generators with the unloading circuit turned on into multiple groups according to the voltage and active power characteristics of the wind turbine generators by using an AP clustering algorithm comprises the following steps: r(i,k) = s(i,k) - max k′≠k (a(i,k') + s(i,k')) defining an attraction degree r(i, k) to represent the appropriateness of data point k as the clustering center of data point i, and the appropriateness of data point i selecting data point k as the clustering center is represented by the defined membership information a(i, k), and the clustering center is determined in an iterative manner, and the specific expression for iteration is: wherein k' and i' are data point numbers; In each round of iteration, all data points with r(k, k) + a(k, k) > 0 are regarded as potential clustering centers, and the iteration stops when the clustering center remains unchanged for a continuous specified number of iterations or reaches the maximum number of iterations.
5. The method of claim 1, wherein, The points that are not clustering centers in the data set are assigned to the clustering center with the highest similarity, and the clustering groups of the data set are completed. After the step of establishing the low-voltage ride-through equivalent model of the direct-drive wind farm, the method further comprises the following steps: collecting fault simulation waveform data under different operating conditions through an electromechanical transient simulation model; 6. A direct-drive wind farm low voltage ride through equivalent modeling system based on AP clustering, characterized in that, comparing the data of the simulation model and the equivalent model under different fault conditions to verify the effectiveness of the equivalent method. The method comprises the following steps: a primary grouping module for determining characteristic quantities of the direct-drive wind turbine generators; grouping the direct-drive wind turbine generators into two groups, i.e., a group of wind turbine generators with the unloading circuit turned on and a group of wind turbine generators with the unloading circuit turned off, according to the characteristic quantities; a secondary grouping module for grouping the wind turbine generators with the unloading circuit turned on into multiple groups according to the voltage and active power characteristics of the wind turbine generators by using an AP clustering algorithm, and grouping the wind turbine generators with the unloading circuit turned off into one group; 7. The system of claim 6, wherein, an equivalent model establishing module for establishing a low-voltage ride-through equivalent model of the direct-drive wind farm according to the equivalent parameters of each group.
8. The system of claim 6, wherein, The grouping index of the primary grouping is the action of the unloading circuit, and the grouping index of the secondary grouping is the voltage at the end of the wind turbine during the low-voltage ride-through and the active power of the single wind turbine before the fault.
9. The system of claim 6, wherein, The characteristic quantities of the direct-drive wind turbine generators include the action of the unloading circuit and the operating state of each wind turbine during the low-voltage ride-through. The secondary grouping module comprises the following steps: The data set forming sub-module is configured to collect low-voltage ride-through data of each wind turbine for the wind turbine group in which the unloading circuit has been turned on, and form a data set {X1, X2, …, XN} of the wind turbine group. N} The similarity matrix calculation submodule is configured to calculate a similarity matrix S between the wind turbines, wherein a non-diagonal element s(i,j) is expressed as s(i,j)=-||X i -X j ‖ 2 The greater the diagonal element s of the matrix, the greater the possibility of becoming a clustering center. The clustering center determining sub-module is configured to define an attraction degree r(i, k) for representing the degree of appropriateness of the data point k as the clustering center of the data point i, and the degree of appropriateness of the data point i selecting the data point k as the clustering center is represented by the defined belonging degree information a(i, k), and the clustering center is determined in an iterative manner, and the specific expression of the iteration is as follows: r(i,k) = s(i,k) - max k′≠k (a(i,k') + s(i,k')) In the formula, k' and i' are data point numbers; The iteration sub-module is configured to regard all data points with r(k, k)+a(k, k)>0 as potential clustering centers in each iteration, and the iteration is stopped when the clustering center remains unchanged for a continuous specified number of iterations or reaches a maximum number of iterations; The clustering sub-module is configured to assign the points without clustering centers in the data set to the clustering center with the highest similarity, and complete the clustering of the data set.
10. The system of claim 6, wherein, Further comprising: The waveform data collecting module is configured to collect fault simulation waveform data under different working conditions through the electromechanical transient simulation model; The verification sub-module is configured to compare the data of the simulation model and the equivalent model under different fault conditions to verify the effectiveness of the equivalent method.