Wind power plant power dynamic optimization method and system based on spectral clustering partition

Abstract

The invention discloses a wind power plant power dynamic optimization method and system based on spectral clustering partition, and the method comprises the steps: building a long-time scale wind power plant power maximization steady-state optimization problem based on an engineering wake flow model according to the average wind speed and wind direction, and solving an optimal yaw angle; calculating a wake flow coupling relation of the wind power plant in the current wind direction according to the optimal yaw angle to construct a wake flow diagram; clustering the wake flow diagram based on a spectral clustering method, and dividing the wind power plant into a plurality of sub-regions; and constructing a short-time scale wind power plant power dynamic optimization problem with optimization targets including maximum wind power plant total power and minimum wind turbine generator thrust fluctuation for different subareas, and solving the short-time scale wind power plant power dynamic optimization problem to obtain the optimal control quantity of each generator. The wind power plant maximum power output optimization method aims at overcoming the defects of a traditional static optimization strategy in a dynamic environment, and the overall generating capacity of the wind power plant is dynamically improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind farm power dynamic optimization, and in particular to a wind farm power dynamic optimization method and system based on spectral clustering partitioning. Background Art

[0002] As an important part of clean energy, the technology of wind power generation has witnessed continuous rapid growth in terms of technological progress and installed capacity. After years of development, wind power technology is moving towards large-scale and intelligent directions, but this is accompanied by more complex operation and control challenges. Especially in large-scale wind farms, due to the wake effect among wind turbines, this complex aerodynamic coupling makes key parameters such as wind speed, power output, and mechanical loads all exhibit obvious dynamic change characteristics, which poses higher requirements for the optimal regulation of the overall power of the wind farm. Traditional wind farm optimization control methods are mostly based on static models, usually assuming that the wind speed remains constant or simply using deterministic models to mathematically express the power of the units, thereby transforming the power generation optimization problem into a high-dimensional non-linear non-convex problem for solution. However, this method often ignores the uncertainty of wind speed and the complex dynamic coupling problem among wind turbines, resulting in difficulty in achieving the expected improvement in power generation efficiency during actual operation and possibly exacerbating the fatigue load of the units. Summary of the Invention

[0003] The technical problem to be solved by the present invention: In view of the above problems of the prior art, a wind farm power dynamic optimization method and system based on spectral clustering partitioning are provided. The present invention aims to make up for the deficiencies of traditional static optimization strategies in a dynamic environment and dynamically improve the overall power generation of the wind farm.

[0004] To solve the above technical problems, the technical solutions adopted by the present invention are as follows: A wind farm power dynamic optimization method based on spectral clustering partitioning, comprising the following steps: Based on the average wind speed and wind direction, construct a long-time scale wind farm power maximization steady-state optimization problem based on the engineering wake model and solve for the optimal yaw angle; Calculate the wake coupling relationship of the wind farm under the current wind direction with the optimal yaw angle to construct a wake map; Cluster the wake map based on the spectral clustering method to divide the wind farm into multiple sub-regions; Construct optimization objectives for different sub-regions respectively, including short-time scale wind farm power dynamic optimization problems with the maximum total power of the wind farm and the minimum thrust fluctuation of the wind turbines, and solve the short-time scale wind farm power dynamic optimization problems to obtain the optimal control quantities of each unit.

[0005] Optionally, the construction of the engineering wake model includes: S101, for the wind turbine i For the wind turbine nThe wake effect generated will be due to the wind turbine i The yaw movement of which will cause the deflection of its wake center, defined as yaw deflection. Determine the wake deflection distance of the yaw deflection as: , wherein, represents the initial deflection angle of the wind turbine i , represents the impeller radius of the wind turbine, represents the deflection expansion coefficient, represents the wind turbine i for the wind turbine n the deflection calculation constant of the generated wake, and there is: , , , wherein, represents the geographical distance along the wind direction between the wind turbine i and the wind turbine n , represents the yaw angle of the wind turbine i , represents the thrust coefficient of the wind turbine i , represents the wind turbine i axial induction factor; S102. Define that the rotation of the impeller itself will also cause the deflection of the wake center, which is defined as rotational deflection. Calculate the rotational deflection i generated by the wind turbine n for the wind turbine as: , wherein, and are constants; S103. Calculate the coordinate i on the y-axis of the center of the wake effect generated by the wind turbine n for the wind turbine according to the following formula: , wherein, represents the i axis coordinate where the geographical location of the wind turbine is located; y S104. The wake effect will gradually expand during the propagation process. The wake expansion radius i generated by the wind turbine n for the wind turbine is: ​, Among them, represents taking the maximum value, represents the wake expansion coefficient, represents the wake region q of the scaling constant, q represents the wake region, q = 1, 2, 3 respectively represent the near wake region, far wake region and mixed region; S105, calculate the wind turbine according to the following formula i for the wind turbine n generated wake wind speed : , , , , , , , , Among them, represents the wind turbine i inflow wind speed of, represents the wind turbine i for the wind turbine n wind speed decay rate of, represents the wake decay coefficient, represents the wind turbine i for the wind turbine n caused wake region q of the wake overlap area, is the impeller swept area; represents the wake decay correction parameter, , and are all constants; represents the wind turbine i for the wind turbine n caused wake region q of the expansion radius; and respectively represent the wind turbine i for the wind turbine n caused wake region q of the wake overlap area for calculating the area of the two angles, represents the coordinate of the wind turbine n in the y-axis direction and the wind turbine i for the wind turbine n caused distance between the wake center, represents the wind turbinen The y-axis coordinate of

[0006] Optionally, constructing a long-term scale wind farm power maximization steady-state optimization problem based on the engineering wake model according to the average wind speed and direction, and solving for the optimal yaw angle includes: S201, calculating the power output of the wind turbine according to the following formula i : : , , where represents the axial induction factor of the wind turbine i , represents the yaw angle of the wind turbine i , represents the incoming wind speed of the wind turbine i , represents the air density, represents the impeller swept area, represents the power coefficient of the wind turbine i , represents the generator efficiency, p represents the power loss coefficient caused by yaw; S202, considering the existence of wake effects inside the wind farm, the incoming wind speed of each wind turbine is related to the free wind speed, the axial induction factor and the yaw angle of the upstream wind turbine, and constructing the total power output function of the wind farm : , where is the incoming wind speed of the wind farm, represents the set of all wind turbine axial induction factors, represents the set of all wind turbine yaw angles, represents the number of wind turbines, represents the power of wind turbine i; S203, constructing the long-term scale wind farm power maximization steady-state optimization problem shown in the following formula: , , where represents the upper and lower limits of the axial induction factor of the wind turbine , represents the upper and lower limits of the yaw angle of the wind turbine , represents the upper and lower limits of the active power output of the wind turbine ; S204. For the steady-state optimization problem of maximizing the wind farm power on a long time scale, heuristic algorithms such as the particle swarm algorithm are used to solve for the optimal yaw angle.

[0007] Optionally, the construction of the wake coupling relationship of the wind farm in the current wind direction with the optimal yaw angle to construct a wake map includes: constructing a matrix of the following size using the wind speed attenuation rate of each wind turbine in the wind farm calculated when solving for the optimal yaw angle: , where any element in the i-th row and j-th column of this matrix represents the wind speed attenuation rate of wind turbine i on wind turbine j ; respectively, taking the wind turbines as nodes in a weighted undirected graph and the wind speed attenuation rate between the wind turbines as the edges in the weighted undirected graph to construct a weighted undirected graph : , where, represents the edge of the weighted undirected graph , if there is a wake coupling between wind turbine i and wind turbine j , then , otherwise = 0; n is the number of wind turbines.

[0008] Optionally, the clustering of the wake map based on the spectral clustering method to divide the wind farm into multiple sub-regions includes: S301. Define the degree i of the node corresponding to wind turbine as the sum of all edges of the weighted undirected graph according to the following formula: , where, n represents the number of wind turbines, represents the edge of the weighted undirected graph ; Define the degree matrix i of the node corresponding to all wind turbines according to the following formula: : ; S302. Calculate the Laplacian matrix according to the following formula: , where, represents the weighted undirected graph; calculate the normalized Laplacian matrix according to the following formula : , S303, calculate the normalized Laplacian matrix of the eigenvalues e , and calculate e of the second-order difference : , where, represents the i +2nd eigenvalue and the i +1st eigenvalue difference, represents the i +1st eigenvalue and the i th eigenvalue difference, represents the i +2nd eigenvalue, represents the i th eigenvalue, represents the i +1st eigenvalue; S304, design the condition threshold, starting from i =3, determine the number of clusters when the eigenvalue meets the following conditions K = i : , where, represents the i -1st second-order difference, is an adjustable parameter, represents the j th second-order difference, represents the j -1st second-order difference; S305, according to the clustering idea, establish the minimum cut problem: , where, represents the cut, ~ respectively represent the sets of nodes belonging to subsets 1~ K , represents and the cut weight between, represents the set of points belonging to subset , represents 's complement; Select the "normalized cut" method to solve this minimum cut problem: , Among them, represents the "Normalized Cuts" method, represents the sum of the weights of the edges of, and through the spectral clustering method, the above minimum cut problem is transformed into the following form: , Among them, represents the matrix to be solved, and the in the superscript indicates the transpose, represents n the -dimensional identity matrix, represents the normalized Laplacian matrix; S306, by solving the first eigenvectors of the normalized Laplacian matrix K to obtain the matrix to be solved, and then based on the k-means clustering method to finally obtain the clustering results of the fan nodes in the wind farm.

[0009] Optionally, the optimization objectives constructed for different sub-regions respectively include the short-time scale wind farm power dynamic optimization problem of maximizing the total power of the wind farm and minimizing the thrust fluctuation of the wind turbines, including: S401, determine the thrust i received by the fan according to the following formula: , Among them, is the axial induction factor of the fan i , is the yaw angle of the fan i , is the inflow wind speed of the fan i , is the air density, is the impeller swept area, is the thrust coefficient of the fan i , is the thrust loss coefficient in the yaw state of the fan; S402, considering that the thrust received by the fan changes continuously with the wind speed during dynamic operation, while increasing the total power of the wind farm, suppressing the fluctuation of the thrust received by the fan, and constructing the following short-time scale wind farm power dynamic optimization problem: , Among them, represents the weight of the optimization objective, represents the number of fans, represents the optimal yaw angle of the fan i , represents the fan iPower output at the optimal yaw angle represents the wind turbine i thrust received at the optimal yaw angle; represents the wind turbine i initial value of the thrust received; S403. According to the sub-regions into which the wind farm has been divided in the spectral clustering result N c sub-regions, the optimization problem of each sub-region can be split and calculated separately, and the short-term scale wind farm power dynamic optimization problem is rewritten into the following sub-region form: , s.t. , wherein represents the axial induction factor of the wind turbine belonging to the k th subset, is the axial induction factor of the wind turbine in the k th subset i of the wind turbine, represents the number of wind turbines in the k th subset, represents the wind turbine i upper and lower limits of the axial induction factor, represents the wind turbine i upper and lower limits of the active power output.

[0010] Optionally, when solving the short-term scale wind farm power dynamic optimization problem to obtain the optimal control quantity of each unit, the optimal control quantity includes the axial induction factor of each wind turbine.

[0011] In addition, the present invention also provides a wind farm power dynamic optimization system based on spectral clustering partitioning, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning.

[0012] In addition, the present invention also provides a computer-readable storage medium, in which a computer program or instruction is stored, and the computer program or instruction is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning through a processor.

[0013] In addition, the present invention also provides a computer program product, including a computer program or instruction, and the computer program or instruction is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning through a processor.

[0014] Compared with the prior art, the present invention can mainly achieve the following beneficial effects: The wind farm power dynamic optimization method based on spectral clustering partitioning first constructs an optimization problem for maximizing the total power of the wind farm on a long time scale according to the average wind speed and direction, calculates the optimal yaw angle, and then considers the change of the inflow wind speed. By combining two optimization objectives of maximizing the total power of the wind farm and minimizing the thrust fluctuation of wind turbines, it constructs a power dynamic optimization problem for the wind farm on a short time scale, and introduces the spectral clustering method to divide the wind farm into multiple sub-regions according to the wake coupling relationship, divides the original optimization problem into multiple sub-problems with smaller dimensions for parallel solution, and greatly improves the solution speed to adapt to the time-varying wind speed environment. The wind farm power dynamic optimization method based on spectral clustering partitioning of the present invention uses spectral clustering technology to partition the wind turbines in the wind farm, divides the turbines with strong wake coupling into the same sub-region, realizes the coordinated control of the turbines within the sub-region, can greatly reduce the calculation time of the original optimization problem. This method can not only make up for the deficiencies of traditional static optimization strategies in dynamic environments, but also provide a new technical path for the intelligent control of large-scale wind farms. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the basic process of the method in the embodiment of the present invention.

[0016] Figure 2 It is a schematic diagram of the structure of the engineering wake model in the embodiment of the present invention.

[0017] Figure 3 It is a flowchart of the wind farm partitioning method based on spectral clustering in the embodiment of the present invention.

[0018] Figure 4 It is a clustering result diagram and a partitioning result diagram of the wind farm at 0 degrees wind direction in the embodiment of the present invention, where (a) is the clustering result diagram and (b) is the wind farm partitioning schematic diagram.

[0019] Figure 5 It is a clustering result diagram and a partitioning result diagram of the wind farm at 10 degrees wind direction in the embodiment of the present invention, where (a) is the clustering result diagram and (b) is the wind farm partitioning schematic diagram.

[0020] Figure 6 It is a clustering result diagram and a partitioning result diagram of the wind farm at 20 degrees wind direction in the embodiment of the present invention, where (a) is the clustering result diagram and (b) is the wind farm partitioning schematic diagram.

[0021] Figure 7 It is a simulation verification diagram of the total power improvement of the wind farm at 0 degrees wind direction in the embodiment of the present invention.

[0022] Figure 8 It is a simulation verification diagram of the total power improvement of the wind farm at 20 degrees wind direction in the embodiment of the present invention.

[0023] Figure 9 This is the simulation verification diagram for suppressing the total thrust fluctuation of the wind farm under the 0-degree wind direction in the embodiment of the present invention.

[0024] Figure 10 This is the simulation verification diagram for suppressing the total thrust fluctuation of the wind farm under the 20-degree wind direction in the embodiment of the present invention.

[0025] Figure 11 This is the simulation verification diagram for suppressing the thrust fluctuation of wind turbines #1 and #36 in the embodiment of the present invention, where (a) is the simulation verification diagram for suppressing the thrust fluctuation of wind turbine #1, and (b) is the simulation verification diagram for suppressing the thrust fluctuation of wind turbine #36.

[0026] Figure 12 This is the average time comparison diagram of the method in the embodiment of the present invention compared with the centralized control.

[0027] Figure 13 This is the simulation verification diagram for the calculation time saving effect of the method in the embodiment of the present invention compared with the centralized control. Detailed implementation manners

[0028] In order to enable those skilled in the art of the present technology to better understand the technical solution of the present invention, the technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0029] As Figure 1 shown, the wind farm power dynamic optimization method based on spectral clustering partitioning in this embodiment includes the following steps: S1. According to the average wind speed and wind direction, construct a long-time scale wind farm power maximization steady-state optimization problem based on the engineering wake model and solve for the optimal yaw angle; S2. Calculate the wake coupling relationship of the wind farm under the current wind direction with the optimal yaw angle to construct a wake diagram; S3. Cluster the wake diagram based on the spectral clustering method and divide the wind farm into multiple sub-regions; S4. Respectively construct short-time scale wind farm power dynamic optimization problems with optimization objectives including maximizing the total power of the wind farm and minimizing the thrust fluctuation of the wind turbines for different sub-regions, and solve the short-time scale wind farm power dynamic optimization problems to obtain the optimal control quantities of each unit.

[0030] As Figure 2 shown, for any wind turbine i in terms of, three wake regions will be formed on its leeward side: the near wake region, the far wake region, and the mixed region, which are respectively represented by q = 1, 2, 3; the wake effect will gradually expand during the propagation process, and the wake expansion radius generated by the wind turbine i on the wind turbine n produced 。 Representing the wind turbine i For the wind turbine n The wake area caused q The wake overlap area. Since the wind turbine i The yaw movement will cause the deflection of its wake center, which is defined as yaw deflection. Determine the wake deflection distance of the yaw deflection . In step S1 of this embodiment, the construction of the engineering wake model includes: S101, for the wind turbine i For the wind turbine n The wake effect generated. The yaw movement of the wind turbine will cause the deflection of its wake center, which is defined as yaw deflection. Determine the wake deflection distance of the yaw deflection i Is: As: , Wherein, Represents the wind turbine i The initial deflection angle, Represents the impeller radius of the wind turbine, Represents the deflection expansion coefficient, Represents the wind turbine i For the wind turbine n The deflection calculation constant of the generated wake, and there is: , , , Wherein, Represents the wind turbine i And the wind turbine n The geographical distance along the wind direction between, Represents the wind turbine i The yaw angle, Represents the wind turbine i The thrust coefficient (related to the axial induction factor), Represents the wind turbine i The axial induction factor; S102, define that the rotation of the impeller itself will also cause the deflection of the wake center, which is defined as rotational deflection. Calculate the rotational deflection i Generated by the wind turbine n For the wind turbine Is: , Wherein, And Are constants; S103, calculate the wind turbine according to the following formula i Generated by the wind turbine nThe coordinate of the center of the generated wake effect on the y-axis : , wherein, represents the i axial coordinate of the geographical location of the wind turbine y ; S104, the wake effect will gradually expand during the propagation process. The i wake expansion radius generated by the wind turbine n for the wind turbine : , wherein, represents taking the maximum value, represents the wake expansion coefficient, represents the q scale constant of the wake region q represents the wake region, q =1, 2, 3 respectively represent the near wake region, far wake region and mixed region; S105, calculate the i wake wind speed generated by the wind turbine n for the wind turbine according to the following formula: , , , , , , , , wherein, represents the i inflow wind speed of the wind turbine, represents the i wind speed decay rate of the wind turbine n for the wind turbine represents the wake decay coefficient, represents the i wake overlap area of the n wake region caused by the wind turbine q , is the impeller swept area; represents the wake decay correction parameter, , and are all constants; represents the wind turbinei For the wind turbine n induced wake region q expansion radius; and respectively represent two angles used to calculate the area in the wake overlap area of the wind turbine i for the wind turbine n induced wake region q ; represents the coordinate of wind turbine n in the y-axis direction and the distance between the wake center of the wind turbine i for the wind turbine n induced; represents the wind turbine n y-axis coordinate.

[0031] In step S1 of this embodiment, according to the average wind speed and wind direction, based on the engineering wake model, a long-time-scale wind farm power maximization steady-state optimization problem is constructed and the optimal yaw angle is solved, including: S201, calculate the power output of the wind turbine i according to the following formula : , , wherein, represents the axial induction factor of the wind turbine i ; represents the yaw angle of the wind turbine i ; represents the inflow wind speed of the wind turbine i ; represents the air density, represents the impeller swept area, represents the power coefficient of the wind turbine i ; represents the generator efficiency, p represents the power loss coefficient caused by yaw. This power loss coefficient is a constant parameter and can be selected according to actual needs. For example, in this embodiment, the value is 1.88; S202, considering the wake effect in the wind farm, the inflow wind speed of each wind turbine is related to the free wind speed, the axial induction factor and the yaw angle of the upstream wind turbine, and construct the total power output function of the wind farm : , wherein, is the inflow wind speed of the wind farm, represents the set of all wind turbine axial induction factors, represents the set of all wind turbine yaw angles, represents the number of wind turbines, represents the power of wind turbine i; S203. Construct the long - time - scale wind farm power maximization steady - state optimization problem shown by the following formula: , , where, represents the upper and lower limits of the axial induction factor of wind turbine ; represents the upper and lower limits of the yaw angle of wind turbine ; represents the upper and lower limits of the active power output of wind turbine ; S204. For the long - time - scale wind farm power maximization steady - state optimization problem, use heuristic algorithms such as the particle swarm algorithm to solve it to obtain the optimal yaw angle. Since the wake model is discontinuous under the condition of uncertain yaw angle, the above - mentioned optimization problem can be solved by using heuristic algorithms such as the particle swarm algorithm to obtain the optimal yaw angle.

[0032] In step S2 of this embodiment, constructing the wake coupling relationship of the wind farm under the current wind direction with the optimal yaw angle to build a wake map includes: constructing a matrix of size shown by the following formula with the wind speed decay rate of each wind turbine in the wind farm calculated when solving the optimal yaw angle: , Any element in the i - th row and j - th column of this matrix represents the wind speed decay rate of wind turbine i on wind turbine j ; Respectively take the wind turbines as nodes in the weighted undirected graph and the wind speed decay rate between the wind turbines as the edges in the weighted undirected graph to construct the weighted undirected graph : , where, represents the edge of the weighted undirected graph . If there is wake coupling between wind turbine i and wind turbine j , then , otherwise = 0; n is the number of wind turbines.

[0033] As shown in Figure 3 , in step S3 of this embodiment, clustering the wake map based on the spectral clustering method to divide the wind farm into multiple sub - regions includes: S301. Define the degree i of the node corresponding to wind turbine as the weighted undirected graph Sum of all edges: , where n represents the number of fans, represents a weighted undirected graph edge; According to all fans i degree of the corresponding node Define the degree matrix : ; S302. Calculate the Laplacian matrix according to the following formula : , where represents a weighted undirected graph; Calculate the normalized Laplacian matrix according to the following formula : , S303. Calculate the eigenvalues of the normalized Laplacian matrix , and calculate e second-order difference of e : : , where represents the difference between the i +2-th eigenvalue and the i +1-th eigenvalue, represents the difference between the i +1-th eigenvalue and the i -th eigenvalue, represents the i +2-th eigenvalue, represents the i -th eigenvalue, represents the i +1-th eigenvalue; S304. Design the condition threshold. Starting from i = 3, determine the number of clusters when the eigenvalue satisfies the following conditions K = i : , where represents the i -1-th second-order difference, is an adjustable parameter, represents the j -th second-order difference, represents the j-1 second-order difference; S305. According to the clustering idea, establish the minimum cut problem: , where, represents the cut, ~ respectively represent the sets of nodes belonging to subsets 1~ K , represents and the cut weight between, represents the set of points belonging to subset , represents the complement of; Select to solve this minimum cut problem by the "normalized cut" method: , where, represents the "normalized cut" method, represents the sum of the edge weights of, Through the spectral clustering method, the above minimum cut problem is transformed into the following form: , where, represents the matrix to be solved, and the in the superscript represents the transpose, represents n the dimensional identity matrix, S306. By solving the first eigenvectors of the normalized Laplacian matrix K to obtain the matrix to be solved, and then based on the k-means clustering method to finally obtain the clustering results of the fan nodes in the wind farm.

[0034] In step S4 of this embodiment, the optimization objectives are respectively constructed for different sub-regions, and the short-time scale wind farm power dynamic optimization problem including the maximum total power of the wind farm and the minimum thrust fluctuation of the wind turbines includes: S401. Determine the thrust i received by the fan according to the following formula: , where, is the axial induction factor of the fan i , is the yaw angle of the fan i , is the inflow wind speed of the fan i , is the air density, is the impeller swept area, is the fan i thrust coefficient, is the thrust loss coefficient in the yaw state of the fan. This thrust loss coefficient is a constant parameter and can be set according to actual needs. For example, in this embodiment, it is set to 2; S402. Considering that the thrust on the fan changes continuously with the wind speed during dynamic operation, while increasing the total power of the wind farm, suppressing the fluctuation of the thrust on the fan, the following short-term wind farm power dynamic optimization problem is constructed: , wherein, represents the weight of the optimization objective, represents the number of fans, represents the fan i optimal yaw angle, represents the fan i power output at the optimal yaw angle, represents the fan i thrust received at the optimal yaw angle; represents the fan i initial value of the thrust received; S403. According to the N c sub-regions into which the wind farm has been divided in the spectral clustering result, the optimization problem of each sub-region can be split and calculated separately, and the short-term wind farm power dynamic optimization problem is rewritten into the following sub-region form: , s.t. , wherein, represents the axial induction factor of the fan belonging to the k th sub-set, is the axial induction factor of the fan in the k th sub-set, i axial induction factor of the fan, represents the number of fans in the k th sub-set, represents the fan i axial induction factor upper and lower limits, represents the fan i active power output upper and lower limits. In this way, the original optimization problem is decomposed into multiple sub-problems with smaller dimensions, which can be solved in parallel, greatly reducing the solution time to be applicable to real-time dynamic control.

[0035] In step S4 of this embodiment, when obtaining the optimal control quantities of each unit by solving the short-term scale wind farm power dynamic optimization problem, the optimal control quantities include the axial induction factor of each fan, that is, the axial induction factors of each fan i axial induction factor .

[0036] Figure 4 , Figure 5 and Figure 6 respectively show the partitioning results of the wind farm by the method proposed in this embodiment (abbreviation: "proposed method") under different wind directions (0°, 10° and 20° wind directions). Figure 4 , Figure 5 and Figure 6 In the subgraph (a), the short solid lines of the same color represent the fan groups divided into the same area. In the corresponding subgraph (b), the fan groups divided into the same area are represented by black dashed boxes. It can be seen that changing the wind direction will cause different wake distributions in the wind farm, so the partitioning results will also change. Compared with the centralized optimization method that optimizes 36 fans in the whole field at the same time, the proposed method optimizes the fans in different independent sub-regions, thus greatly reducing the calculation time.

[0037] Figure 7 and Figure 8 respectively show the total active power of the wind farm under the proposed method in this embodiment and other existing control methods under 0° and 20° wind directions. Among other existing control methods: the "centralized method" is a method that optimizes 36 fans in the whole field at the same time, the "static control method" is a method that uses the optimal value obtained from the long-term scale wind farm power maximization steady-state optimization problem as the control quantity, and the "greedy control" is a control method in which all the wind turbines in the whole field do not yaw and the axial induction factor takes the maximum value. It can be seen that under different wind directions, the proposed method, the centralized method and the static control in this embodiment all have higher total active power than the greedy control. Since the optimization objective is only to maximize the wind farm power, the total active power of the wind farm under the static control method is slightly higher than that of the proposed method and the centralized method in this embodiment; the objectives of minimizing the thrust fluctuation of the fans are added in the proposed method and the centralized method in this embodiment, so the total active power is slightly lower. In addition, since the proposed method in this embodiment partitions the wind farm, the possible wake coupling between the fans in different sub-regions is ignored, so the total active power is slightly lower than that of the centralized method.

[0038] Figure 9 and Figure 10 respectively show the total thrust of the wind farm under different control methods under 0° and 20° wind directions; Figure 11 shows the thrust of fan #1 and fan #36 (the 36 fans in the whole field are respectively denoted as fan #1 to fan #36) under 0° wind direction. FromFigures 9 to 11 It can be seen that due to the addition of the goal of minimizing the wind turbine thrust fluctuation, the total thrust of the wind farm and the thrust of individual wind turbines under the control of the method proposed in this embodiment and the centralized method are smoother and have smaller fluctuations, which is beneficial to improving the service life of the wind turbines and even the wind farm and reducing the operation and maintenance costs. In addition, the centralized method shows a better thrust fluctuation suppression effect, which is also because the method proposed in this embodiment ignores the possible wake coupling between the wind turbines in different sub-regions.

[0039] Figure 12 and Figure 13 shows the time required for each optimization calculation of the method proposed in this embodiment and the centralized method under different wind directions, as well as the effect of the method proposed in this embodiment on saving calculation time compared with the centralized method. According to the previous analysis, although the control performance of the method proposed in this embodiment is not as good as that of the centralized method, the calculation time of the method proposed in this embodiment is less than 0.5 seconds under all wind directions, while the centralized method requires longer calculation time, and even reaches nearly 4 seconds under some wind directions. At a wind direction of 20°, the calculation time required by the method proposed in this embodiment is only 3.42% of that of the centralized method, and at the worst wind direction of 70°, it is only 18.37%, which undoubtedly greatly improves the performance of real-time control.

[0040] In summary, the wind farm power dynamic optimization method based on spectral clustering partitioning in this embodiment includes: based on the average wind speed and wind direction, constructing a long-term scale wind farm power maximization steady-state optimization problem based on the engineering wake model, and solving the optimal yaw angle; calculating the wake coupling relationship of the wind farm under the current wind direction with the optimal yaw angle, and constructing a wake map; clustering the wake map based on the spectral clustering method, and dividing the wind farm into multiple sub-regions; respectively constructing short-term scale wind farm power dynamic optimization problems for different sub-regions, and the optimization goals include maximizing the total power of the wind farm and minimizing the thrust fluctuation of the wind turbines, and obtaining the optimal control quantities of each unit. This embodiment can dynamically improve the overall power generation of the wind farm considering the influence of the wake effect, bring higher economic benefits to the operator, and has a calculation speed of seconds, ensuring the timeliness of control.

[0041] In addition, this embodiment also provides a wind farm power dynamic optimization system based on spectral clustering partitioning, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning.

[0042] In addition, this embodiment also provides a computer-readable storage medium, in which a computer program or instruction is stored, and the computer program or instruction is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning through a processor.

[0043] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning through a processor.

[0044] Those skilled in the art should understand that the technical solution provided by the present invention can be in the form of a method, a system, or a computer program product. Therefore, the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product on one or more computer-readable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code. The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks. These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0045] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A dynamic optimization method for wind farm power based on spectral clustering partitioning, characterized in that, Including the following steps: According to the average wind speed and direction, based on the engineering wake model, construct a long-time scale wind farm power maximization steady-state optimization problem and solve for the optimal yaw angle; Calculate the wake coupling relationship of the wind farm under the current wind direction with the optimal yaw angle to construct a wake map; Cluster the wake map based on the spectral clustering method and divide the wind farm into multiple sub-regions; Construct optimization objectives for different sub-regions respectively, including a short-time scale wind farm power dynamic optimization problem with the maximum total power of the wind farm and the minimum thrust fluctuation of wind turbines, and solve the short-time scale wind farm power dynamic optimization problem to obtain the optimal control quantity of each unit.

2. The method for dynamically optimizing the power of a wind farm based on spectral clustering partitioning according to claim 1, wherein The construction of the engineering wake model includes: S101, for the wind turbine i For the wind turbine n Regarding the wake effect generated by the wind turbine i The yaw action of the wind turbine will cause the deflection of its wake center, which is defined as yaw deflection, and determine the wake deflection distance of the yaw deflection as follows: ， Among them, represents the initial deflection angle of the fan i , represents the impeller radius of the fan represents the deflection expansion coefficient represents the fan i for the fan n the deflection calculation constant for generating wake, and there is: ， ， ， Among them, represents the wind turbine i and the wind turbine n the geographical distance along the wind direction between them, represents the yaw angle of the wind turbine i ; represents the thrust coefficient of the wind turbine i ; represents the axial induction factor of the wind turbine i ; S102. Define the deflection of the wake center caused by the rotation of the impeller itself as rotational deflection, and calculate the fan i for the fan n to generate rotational deflection as follows: ， Among them, and are constants; S103, calculate the coordinates of the center of the wake effect generated by the fan on the y-axis according to the following formula i For the fan n The coordinates of the center of the wake effect generated : ， Among them, represents the i axial coordinate of the y geographical location where the fan is located; S104, the wake effect will gradually expand during the propagation process, and the fan for the fan the wake expansion radius generated : ， Among them, represents taking the maximum value, represents the wake expansion coefficient, represents the wake region q of the scaling constant, q represents the wake region, q = 1, 2, 3 respectively represent the near wake region, far wake region, and mixed region; S105, calculate the wind turbine according to the following formula For the wind turbine The wake wind speed generated : ， ， ， ， ， ， ， ， Among them, represents the inflow wind speed of the fan i , represents the wind speed attenuation rate of the fan i to the fan n . represents the wake attenuation coefficient, represents the fan i to the fan n causing the wake overlap area of the wake region q ; is the impeller swept area; represents the wake attenuation correction parameter, , and are all constants; represents the expansion radius of the wake region i caused by the fan n to the fan q ; and respectively represent the two angles used to calculate the area in the wake overlap area of the wake region i caused by the fan n to the fan q ; represents the coordinate of the fan n in the y-axis direction and the distance between the fan i to the fan n causing the wake center; represents the y-axis coordinate of the fan n .

3. The method for dynamically optimizing the power of a wind farm based on spectral clustering partitioning according to claim 2, wherein The step of constructing a long-time scale wind farm power maximization steady-state optimization problem and solving for the optimal yaw angle according to the average wind speed and direction, based on the engineering wake model, includes: S201, calculate the power output of the fan according to the following formula i :​ ， ， Among them, represents the axial induction factor of the wind turbine i , represents the yaw angle of the wind turbine i , represents the incoming flow wind speed of the wind turbine i , represents the air density, represents the impeller swept area, represents the power coefficient of the wind turbine i , represents the generator efficiency, p represents the power loss coefficient caused by yaw; S202. Considering the wake effect within the wind farm, the incoming wind speed of each wind turbine is related to the free wind speed, the axial induction factor of the upstream wind turbine, and the yaw angle. A total power output function of the wind farm is constructed. : ， Among them, is the incoming wind speed of the wind farm, represents the set of axial induction factors of all wind turbines, represents the set of yaw angles of all wind turbines, represents the number of wind turbines, represents the power of wind turbine i; S203, construct a long-time scale wind farm power maximization steady-state optimization problem shown by the following formula: ， ， Among them, represents the upper and lower limits of the axial induction factor of the fan , represents the upper and lower limits of the yaw angle of the fan , represents the upper and lower limits of the active power output of the fan . S204, for the long-time scale wind farm power maximization steady-state optimization problem, use heuristic algorithms such as the particle swarm algorithm to solve for the optimal yaw angle.

4. The wind farm power dynamic optimization method based on spectral clustering partitioning according to claim 1, characterized in that The construction of the wake map for the wake coupling relationship of the wind farm in the current wind direction with the optimal yaw angle includes: constructing a matrix of the size shown in the following formula with the wind speed attenuation rate of each wind turbine in the wind farm calculated when solving the optimal yaw angle as : ， Any element in the \(i\)-th row and \(j\)-th column of the matrix represents a wind turbine i For the wind turbine j the wind speed attenuation rate; respectively, taking the wind turbines as nodes in an undirected weighted graph and the wind speed attenuation rates between the wind turbines as the edges in the undirected weighted graph to construct an undirected weighted graph : ， Among them, represents the edge of the weighted undirected graph If there is wake coupling between wind turbines i and wind turbine j , then , otherwise = 0; n is the number of wind turbines.

5. The method for dynamically optimizing the power of a wind farm based on spectral clustering partitioning according to claim 4, wherein The step of clustering the wake map based on the spectral clustering method and dividing the wind farm into multiple sub-regions includes: S301, define the fan according to the following formula i Degree of the corresponding node is a weighted undirected graph Sum of all edges of: ， Among them, n represents the number of fans, represents an edge of the weighted undirected graph ; According to all the wind turbines i The degree of the corresponding node Define the degree matrix : ； S302, calculate the Laplacian matrix according to the following formula :[[]]END]] ， Among them, represents a weighted undirected graph; calculate the normalized Laplacian matrix according to the following formula : ， S303, calculate the normalized Laplacian matrix of the eigenvalues e , and calculate e the second-order difference of : ， Among them, represents the difference between the i +(2) eigenvalues and the i +(1) eigenvalue, represents the difference between the i +(1) eigenvalue and the i th eigenvalue, represents the i +(2) eigenvalues, represents the i th eigenvalue, represents the i +(1) eigenvalue; S304, Design condition threshold, starting from i = 3, determine the number of clusters when the eigenvalue satisfies the following conditions K = i : ， Among them, represents the i -1th second-order difference, is an adjustable parameter, represents the j th second-order difference, represents the j -1th second-order difference; S305, according to the clustering idea, establish a minimum cut problem: ， Among them, represents a split, ~ respectively represent the sets of nodes belonging to subsets 1 to K , represents and the split weight between, represents the set of points belonging to subset , represents the complement of; choose to solve this minimum cut problem by the "normalized cut" method: ， Among them, represents the "normalized cut" method, represents the sum of the weights of the edges of, and through the spectral clustering method, the above minimum cut problem is transformed into the following form: ， Among them, represents the matrix to be solved, and the superscript represents the transpose, represents n the \(n\times n\) identity matrix, represents the normalized Laplacian matrix; S306, obtaining the matrix to be solved by solving the first K eigenvectors of the normalized Laplacian matrix , and finally obtaining the clustering results of the fan nodes in the wind farm based on the k-means clustering method. the first K eigenvectors to obtain the matrix to be solved and then finally obtaining the clustering results of the fan nodes in the wind farm based on the k-means clustering method.

6. The method for dynamically optimizing the power of a wind farm based on spectral clustering partitioning according to claim 1, wherein The step of constructing optimization objectives for different sub-regions respectively, including a short-time scale wind farm power dynamic optimization problem with the maximum total power of the wind farm and the minimum thrust fluctuation of wind turbines, includes: S401, determine the thrust received by the fan according to the following formula i :​ ， Among them, is the axial induction factor of the wind turbine i , is the yaw angle of the wind turbine i , is the incoming flow wind speed of the wind turbine i , is the air density, is the impeller swept area, is the thrust coefficient of the wind turbine i , is the thrust loss coefficient in the yaw state of the wind turbine; S402, considering that the thrust received by the wind turbine changes continuously with the wind speed during dynamic operation, while increasing the total power of the wind farm, suppress the fluctuation of the thrust received by the wind turbine, and construct the following short-time scale wind farm power dynamic optimization problem: ， Among them, represents the weight of the optimization objective, represents the number of wind turbines, represents the wind turbine i 's optimal yaw angle, represents the wind turbine i 's power output at the optimal yaw angle, represents the wind turbine i 's thrust at the optimal yaw angle; represents the wind turbine i 's initial thrust value; S403. According to the sub-areas into which the wind farm has been divided based on the spectral clustering results, the optimization problem of each sub-area can be split and calculated separately, and the short-term scale wind farm power dynamic optimization problem is rewritten in the following sub-area form: N c ​ ， s.t. , Among them, represents the axial induction factor of the fan belonging to the k th subset, is the axial induction factor of the fan in the k th subset, i and represents the number of fans in the k th subset, represents the upper and lower limits of the axial induction factor of the fan i , and represents the upper and lower limits of the active power output of the fan i .

7. The wind farm power dynamic optimization method based on spectral clustering partitioning according to claim 1, characterized in that When solving the short-time scale wind farm power dynamic optimization problem to obtain the optimal control quantity of each unit, the optimal control quantity includes the axial induction factor of each wind turbine.

8. A wind farm power dynamic optimization system based on spectral clustering partitioning, comprising a microprocessor and a memory connected to each other, characterized in that, The microprocessor is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program or instructions, characterized in that, The computer program or instruction is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning described in any one of claims 1 to 7 through a processor.

10. A computer program product, comprising a computer program or instructions, characterized in that, The computer program or instruction is programmed or configured to execute the wind farm power dynamic optimization method based on spectral clustering partitioning described in any one of claims 1 to 7 through a processor.

Images