Wind power plant optimization control method and device based on multi-working-condition model dimensionality reduction

Abstract

The invention relates to a wind power plant optimization control method and device based on multi-working-condition model dimensionality reduction, and the method comprises the steps: analyzing the wake flow influence relation between fans in a wind power plant, constructing a wake flow influence adjacency matrix, judging the wake flow coupling relation between the fans, and carrying out the partitioning of the wind power plant based on the wake flow influence relation; a wind power plant is divided into a plurality of sub-wind power plants by adopting directed graph connected component analysis. And according to the parameters of each fan in the wind power plant, constructing a wind power plant wake flow optimization control mathematical model. And through linearization processing, a linear state space model of the fan model is obtained. A gap measurement theory is adopted to carry out dimension reduction processing on fan models, the fan models with similar dynamic characteristics are combined, and the number of control models is reduced. Through the construction of a multi-model control matrix, the optimization control of the wind power plant under different wind speed working conditions is realized. According to the method, the complexity of the control model is effectively reduced, the scheduling efficiency and energy output of the wind power plant are improved, and the method has relatively high practical application value and popularization prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind farm control, and particularly relates to a wind farm optimization control method and device based on multi-condition model dimensionality reduction. Background Art

[0002] In recent years, the global wind power installed capacity has increased rapidly, the scale of wind farms has been continuously expanding, and the layout of wind turbines has gradually become denser. On the one hand, the construction of such large-scale and centralized wind farms has improved the wind energy utilization efficiency and power generation scale benefits. On the other hand, it has also brought new technical problems. Especially in the problem of wake interaction between wind turbines, it has gradually become a key factor affecting the overall performance and stability of wind farms.

[0003] The wake effect refers to the interference phenomenon of the wake generated by the upstream wind turbine during power generation on the downstream wind turbine. This effect not only directly reduces the wind energy utilization efficiency and power generation performance of the downstream wind turbine, but also increases the load of the unit and exacerbates structural fatigue, thus threatening the overall operation safety and stability of the wind farm. Therefore, how to effectively reduce the wake effect and improve the collaborative power generation efficiency between wind turbines has become an important research direction in the wind power field.

[0004] To reduce the negative impact brought by the wake effect of wind farms, the prior art usually adopts an overall wind farm optimization control strategy. This strategy coordinates the operating states of each fan and controls parameters such as the yaw angle, pitch angle, and rotational speed of the wind turbine, aiming to reduce wake losses. However, with the continuous expansion of the wind farm scale and the increase in the number of wind turbines, the complexity and computational amount of the overall optimization control model increase exponentially, and real-time control faces unprecedented challenges. The complexity of the wind farm control model is mainly reflected in the following aspects: the number of model variables is huge, the wake interaction relationship between wind turbines is complex, and the management and optimization difficulty of the control model increase under different wind speed conditions. The traditional single optimization control model often leads to slow control response due to its inability to effectively meet the real-time dynamic control requirements of large-scale wind farms, affecting the control efficiency and accuracy in practical applications.

[0005] In the prior art, Chinese Patent CN114841090A discloses a wind farm grouping optimization control method, system, device and medium, aiming to solve the problems of excessive calculation cost and communication cost between wind turbines during the optimization control process of wind farms, as well as the poor applicability of existing optimization control methods. This method characterizes the wake relationship degree between wind turbines in a wind farm by establishing a wake undirected graph, and divides the wind farm into multiple sub-wind farms based on this graph. By optimizing the control of wind turbines in each sub-wind farm, the calculation cost and communication cost can be significantly reduced, and the influence of spatial variation of boundary conditions on optimization control can be reduced. However, although the optimization control strategy is improved by dividing the spatial structure of the wind farm, which reduces part of the calculation burden, this method still has the problem of high complexity of the control model. Especially in the application of large-scale wind farms, the sharp increase in the amount of calculation and the dimension of the model restricts the execution efficiency and real-time performance of the control system.

[0006] In summary, although the prior art has made certain progress in wind farm grouping and optimization control, it still faces problems such as high complexity of the control model, large amount of calculation, and poor real-time response ability. In order to further improve the efficiency and feasibility of wind farm optimization control, it is urgent to innovate in the dimension reduction and simplification of the control model to cope with the increasingly large scale of wind farms and the increasingly complex control requirements. Summary of the Invention

[0007] The purpose of the present invention is to overcome the above-mentioned defects existing in the prior art and provide a wind farm optimization control method and device based on multi-condition model dimension reduction.

[0008] The purpose of the present invention can be achieved through the following technical solutions:

[0009] On the one hand, the present invention provides a wind farm optimization control method based on multi-condition model dimension reduction, including the following steps:

[0010] Step S1: According to the overlapping area of wake influence between wind turbines in the wind farm, judge the wake influence relationship between each pair of wind turbines in the wind farm, and establish a wake influence adjacency matrix;

[0011] Step S2: Based on the wake influence adjacency matrix, partition the wind farm. By using directed graph connected component analysis, divide the wind turbines in the wind farm into multiple sub-wind farms according to the wake influence relationship, where the wake coupling relationship between wind turbines within each sub-wind farm is strong, while the wake influence between sub-wind farms can be ignored;

[0012] Step S3: According to the parameters of each wind turbine in the wind farm, construct a wake mathematical model and calculate the equivalent inflow wind speed of each wind turbine;

[0013] Step S4: According to the equivalent inflow wind speed of each wind turbine, construct the wind turbine models of each wind turbine in each sub-wind farm, linearize the wind turbine models in each sub-wind farm, and use the Taylor expansion method to obtain the linear state space models of the wind turbine models in each sub-wind farm under different wind speed conditions;

[0014] Step S5: Use the gap metric theory to reduce the dimension of the linear state space models of the wind turbine models in each sub-wind farm under different wind speed conditions, merge the wind turbine models with similar dynamic characteristics, and reduce the number of control models;

[0015] Step S6: Divide and merge the dimension-reduced wind turbine control models according to sub-wind farms to form a multi-model control matrix, and realize the optimal control of the wind farm adapted to different wind speed conditions.

[0016] Furthermore, the elements in the wake influence adjacency matrix represent whether there is a significant wake influence relationship between wind turbines.

[0017] Furthermore, the wake influence adjacency matrix is:

[0018]

[0019] where N represents the number of wind turbines in the wind farm, and α ij represents the wake influence relationship between wind turbine i and wind turbine j, specifically:

[0020]

[0021] where S i is the rotor swept area of wind turbine i, and S j,i is the overlapping area between the wake of upstream wind turbine j and the rotor swept area of downstream wind turbine i.

[0022] Furthermore, for the partitioning of the wind farm based on the wake influence adjacency matrix, directed graph connected component analysis is adopted, and the wind turbines in the wind farm are divided into multiple sub-wind farms according to the wake influence relationship, specifically including:

[0023] Regard each wind turbine in the wind farm as a node in a graph. According to the element α ji in the wake influence adjacency matrix L, assign weights to the edges in the graph. If α jj = 1, then establish an edge between wind turbine i and wind turbine j, indicating that there is a significant wake influence relationship between the two;

[0024] Adopt the directed graph connected component analysis algorithm to detect all the connected components in the graph. Specifically, use the depth-first search DFS or breadth-first search BFS algorithm to traverse the adjacency matrix and identify the wind turbine groups with strong wake coupling relationships;

[0025] All connected components are regarded as a sub-wind farm, where the wake coupling relationship between the wind turbines within each sub-wind farm is strong, while the wake influence between sub-wind farms can be ignored.

[0026] Further, based on the parameters of each wind turbine in the wind farm, a wake mathematical model is constructed to calculate the equivalent inflow wind speed of each wind turbine, specifically including:

[0027] Collect the parameters of each wind turbine in the wind farm, including the inflow wind speed, thrust coefficient, tip speed ratio, pitch angle, rotor radius, and the distance between wind turbines;

[0028] According to the parameters of the wind turbine, establish a wake mathematical model to calculate the wake influence wind speed of each wind turbine affected by the wake;

[0029] Comprehensively consider the wake influence of all upstream wind turbines on the downstream wind turbines, and use the wake wind speed superposition model to calculate the equivalent inflow wind speed of the wind turbines;

[0030] By calculating the equivalent inflow wind speed of each wind turbine, obtain the actual wind speed input of this wind turbine.

[0031] Further, the formula for calculating the wake influence wind speed of each wind turbine affected by the wake is:

[0032]

[0033] R j,i =R j +kd j,i

[0034] where ν i,j represents the wake influence wind speed of wind turbine i under the wake influence of wind turbine j, v j is the inflow wind speed of wind turbine j, C t,j is the thrust coefficient of wind turbine j, λ j is the tip speed ratio of wind turbine j, β j is the pitch angle of wind turbine j, R i is the rotor radius of wind turbine i, R j,i is the wake expansion radius of wind turbine j at the position of wind turbine i, S i is the rotor swept area of wind turbine i, S j,i is the overlapping area of the wake of upstream wind turbine j and the rotor swept area of downstream wind turbine i, k is the wake expansion coefficient of wind turbine i, d j,i is the distance between wind turbine j and wind turbine i.

[0035] Further, the formula for calculating the equivalent inflow wind speed of the wind turbine is:

[0036]

[0037] Among them, Γ is the set of all upstream wind turbines that have a wake effect on the downstream wind turbine i, which is obtained according to the wake influence adjacency matrix L, and V ∞ is the background wind speed of the wind farm, and δν i,j is the wind speed loss.

[0038] Furthermore, the root step S4 specifically includes:

[0039] Near the equivalent inflow wind speed V i , perform a first-order Taylor expansion on the non-linear model of the wind turbine to obtain a linearized model;

[0040] According to the linearization result, construct a linear state space model, and the linear state space model includes:

[0041]

[0042] Among them, the specific expressions of the state space matrices A, B, C, E, and W are:

[0043]

[0044] d = V i

[0045] x = (ω r ω e ψ T e ) T

[0046]

[0047] y = (P e M t T s C t ) T

[0048] Among them, A is the system matrix, B is the input matrix, C is the output matrix, E is the input disturbance matrix, W is the output disturbance matrix, x = (ω r ω e ψ T e ) T is the state vector of the wind turbine, ω r is the rotational speed of the wind rotor, ω e is the rotational speed of the generator, ψ is the torsional angle, T e is the electromagnetic torque, β is the pitch angle, ref is the abbreviation of reference, representing the meaning of reference value, u is the control input, y = (P e M t T s C t ) Tis the output vector, P e is the electrical power output of the wind turbine, M t is the tower load, T s is the tower height, C t is the thrust coefficient is the main shaft damping coefficient, N gear is the gearbox speed ratio, J r is the rotor inertia, J g is the generator inertia, K sp is the main shaft elastic coefficient, τ e is the generator inertia time constant, τ β is the pitch angle inertia time constant.

[0049] Furthermore, it specifically includes:

[0050] After obtaining the linear state space model of each wind turbine in different equivalent inflow wind speed conditions in each sub-wind farm, based on the system state matrix A, input matrix B, and output matrix C, the state space expression is converted into the transfer function form through Laplace transform to obtain the transfer function model of the wind turbine control model:

[0051] K(s) = C(sI - A) -1 B

[0052] Among them, K(s) is the transfer function model of the wind turbine, s is the Laplace complex frequency domain variable, and I is the identity matrix;

[0053] Applying the gap metric theory to the transfer function model K(s) under multiple wind speed conditions in each sub-wind farm specifically includes:

[0054] The transfer function model is expressed in the following form:

[0055] K(s) = N(s)M(s) -1

[0056] Among them, M(s) and N(s) are the input transfer matrix and output transfer matrix of the system respectively, satisfying The symbol ~ represents conjugate transpose;

[0057] Calculate the symmetric gap metric between any two wind turbines i and j in the sub-wind farm:

[0058]

[0059] Among them, δ(K i , K j ) is the symmetric gap metric between the two models of wind turbine i and wind turbine j, represents the wind turbine model K i relative to Kj Directional gap metric, K i , K j respectively represent the transfer function models of wind turbine i and wind turbine j, M i , N i , M j , N j are respectively the system input transfer matrix and output transfer matrix of the models of wind turbine i and wind turbine j. inf represents finding the minimum value. Q ∈ H ∞ represents in H ∞ space, a selected stable transformation matrix is used to approximate the mapping relationship between two systems, ||.|| ∞ represents the H ∞ norm;

[0060] Based on the symmetric gap metric δ(K i , K j ) among all wind turbines in the sub-wind farm, a gap metric matrix of the sub-wind farm is constructed, and a threshold-based hierarchical clustering method is used to divide the wind turbines into multiple model subsets with similar dynamic characteristics. In each subset, the wind turbine model with the smallest sum of gap metric distances is selected as the representative model of the subset for unified control modeling of this type of wind turbine.

[0061] On the other hand, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements a wind farm optimization control method based on multi-condition model reduction as described in any one of the above.

[0062] Compared with the prior art, the present invention has the following advantages:

[0063] (1) Through the adjacency matrix and connected component analysis, the present invention realizes the automatic partitioning of the wind farm. This technical means can intelligently divide the wind farm into multiple sub-wind farms according to the wake influence relationship between wind turbines, and ensure the minimization of wake interference between wind turbines in each sub-wind farm. Compared with the prior art, this method not only reduces the manual intervention in the wind farm partitioning process, but also improves the accuracy and efficiency of the partitioning process.

[0064] (2) The present invention innovatively introduces a system dynamic response similarity evaluation mechanism, and through the gap metric theory, the linear models of wind turbines under multiple wind speed conditions are uniformly characterized, greatly reducing the dimension of the control model. This technical means enables the control model of wind turbines to greatly reduce the amount of calculation and model complexity while ensuring accuracy when facing different wind speed conditions. Compared with the traditional O(N 2 ) model, the complexity of the control model of the present invention has been reduced to O(N / 6) 2 , and the calculation time is shortened by more than 80%.

[0065] (3) Through two-stage dimensionality reduction and simplification processing, the present invention significantly improves the computational efficiency of the optimization control of wind farms. Especially in the real-time control scenarios of large-scale and intensive wind farms, it can achieve more efficient and accurate dynamic responses. In traditional methods, as the scale of wind farms expands, the computational amount and complexity of the control model increase sharply, resulting in low real-time control efficiency. However, the innovative dimensionality reduction technology of the present invention effectively addresses this challenge, ensuring the real-time performance and high efficiency of the control system.

[0066] (4) The optimization control method of the present invention has wide applicability, especially suitable for the intelligent scheduling and control of large-scale and intensive wind farms. Through model dimensionality reduction and simplification, it can greatly improve the scheduling and control efficiency of wind farms while ensuring the system accuracy. Description of the Drawings

[0067] Figure 1 Schematic diagram of the layout of the wind farm according to the embodiment of the present invention;

[0068] Figure 2 Inflow wind speed diagram according to the embodiment of the present invention;

[0069] Figure 3 Flowchart of the method according to the embodiment of the present invention;

[0070] Figure 4 Diagram of the unified model result within a partial wind speed range according to the embodiment of the present invention;

[0071] Figure 5 Control model diagram under the low wind speed simulation period according to the embodiment of the present invention;

[0072] Figure 6 Control model diagram under the high wind speed simulation period according to the embodiment of the present invention. Detailed Embodiments

[0073] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0074] Embodiment 1:

[0075] This embodiment proposes a two-stage control model dimensionality reduction method applicable to the optimization control of large-scale wind farms. By scientifically and effectively partitioning the wind farm and unifying the multi-model representation of wind turbines, the complexity of the control model is significantly reduced, and the efficiency of real-time solution and dynamic control is improved. It can provide a reference for the fast control method of large-scale wind farms.

[0076] This embodiment aims to construct an efficient wind farm optimization control method. By reasonably dividing the wind farm area and reducing the dimensions of multiple models of wind turbines, the real-time performance of control calculations is improved, and the computational complexity of optimization solutions is reduced, thereby achieving fast and accurate control of the wind farm. The specific schematic diagram is as Figure 3 shown.

[0077] The wind farm optimization control method based on two-stage dimensionality reduction proposed in this embodiment can effectively identify and divide the wake influence areas of wind turbines, reduce the scale of wind farm optimization calculations, and at the same time use the gap metric theory to uniformly characterize the multi-model characteristics of wind turbines, reducing the computational burden caused by wind speed changes. This can not only significantly improve the real-time performance and computational efficiency of wind farm optimization control, but also reduce the wake loss of wind turbines and improve the wind energy utilization rate, providing strong support for the intelligent scheduling and efficient operation of wind farms.

[0078] It aims to improve the real-time performance and computational efficiency of wind farm control and reduce the computational complexity of wind farm optimization solutions. The specific steps include constructing a wake mathematical model suitable for rapid analysis of wind farms, as well as region division of wind farms and dimensionality reduction of multiple models of wind turbines based on the model. The implementation process of the present invention is described in detail below.

[0079] This embodiment provides a wind farm optimization control method based on multi-condition model dimensionality reduction, including the following steps:

[0080] Step S1: According to the overlapping area of wake influence between wind turbines in the wind farm, judge the wake influence relationship between each pair of wind turbines in the wind farm, and establish a wake influence adjacency matrix;

[0081] Step S2: Based on the wake influence adjacency matrix, partition the wind farm. Using directed graph connected component analysis, divide the wind turbines in the wind farm into multiple sub-wind farms according to the wake influence relationship. Among them, the wake coupling relationship between the wind turbines within each sub-wind farm is relatively strong, while the wake influence between sub-wind farms can be ignored;

[0082] Step S3: According to the parameters of each wind turbine in the wind farm, construct a wake mathematical model and calculate the equivalent inflow wind speed of each wind turbine;

[0083] Step S4: According to the equivalent inflow wind speed of each wind turbine, construct a wind turbine model for each wind turbine within each sub-wind farm, linearize the wind turbine models within each sub-wind farm, and use the Taylor expansion method to obtain the linear state space model of the wind turbine models within each sub-wind farm under different wind speed conditions;

[0084] Step S5: Use the gap metric theory to perform dimensionality reduction on the linear state - space models of the fan models in each sub - wind farm under different wind speed conditions, merge the fan models with similar dynamic characteristics, and reduce the number of control models;

[0085] Step S6: Divide and merge the dimensionality - reduced wind turbine control models according to sub - wind farms to form a multi - model control matrix, and achieve the optimal control of the wind farm adapted to different wind speed conditions.

[0086] Furthermore, the elements in the wake influence adjacency matrix represent whether there is a significant wake influence relationship between the wind turbines.

[0087] Furthermore, the wake influence adjacency matrix is:

[0088]

[0089] where N represents the number of wind turbines in the wind farm, and α ij represents the wake influence relationship between wind turbine i and wind turbine j, specifically:

[0090]

[0091] where S i is the rotor swept area of wind turbine i, and S j,i is the overlapping area between the wake of the upstream wind turbine j and the rotor swept area of the downstream wind turbine i.

[0092] In a wind farm, the operating state of wind turbines is affected by the wake effect. In particular, the upstream wind turbines have a significant impact on the wind speed and energy acquisition of the downstream wind turbines. To effectively optimize the control of the wind farm, it is first necessary to clarify the relationship of the wake effect between the wind turbines. Therefore, the main purpose of step S1 is to judge the wake influence between each wind turbine by calculating the overlapping area of the wake influence between the wind turbines and establish a wake influence adjacency matrix. This process is the basis for the optimal control of the wind farm.

[0093] This adjacency matrix L is an N×N binary matrix, where the rows and columns respectively correspond to the numbers of each wind turbine in the wind farm, that is, each wind turbine represents a node. The elements of matrix L are encoded with 0 and 1, indicating whether there is a significant wake coupling relationship between the wind turbines. The diagonal elements are all 0, indicating that the wind turbine has no wake influence on itself. If wind turbine i has a significant wake influence on wind turbine j, it is recorded as 1 at the i - th row and j - th column of the matrix. For example, if wind turbine 1 has a wake influence on wind turbine 2, it is marked as 1 at the 2nd column and 1st row. Through this adjacency matrix, the wake influence map inside the wind farm can be constructed. Under a wind direction, the front - rear positions of the wind turbines are fixed, and the first - row wind turbines are used as independent items for dividing subsets, that is, there are at least as many subsets as there are first - row wind turbines.

[0094] By performing directed graph connected component analysis on the adjacency matrix L (the first row of fan columns are all zero terms), the fans with a connected relationship (including one term) are grouped into one category, so that subsets with strong coupling between fans can be identified, thereby realizing the automatic zoning of the wind farm and dividing the wind farm into i subsets. These subsets can be regarded as several independent "sub-wind farms". The wake coupling within each sub-wind farm is strong, while the wake influence between sub-wind farms can be approximately ignored. Based on this zoning result, a more targeted and clearer-structured distributed control model design can be realized.

[0095] Further, for the zoning of the wind farm based on the wake influence adjacency matrix, directed graph connected component analysis is adopted, and the fans in the wind farm are divided into multiple sub-wind farms according to the wake influence relationship, specifically including:

[0096] Regarding each fan in the wind farm as a node in a graph, and according to the element α in the wake influence adjacency matrix L ij , assign weights to the edges in the graph. If α ij = 1, then establish an edge between fan i and fan j, indicating that there is a significant wake influence relationship between the two;

[0097] Adopt the directed graph connected component analysis algorithm to detect all connected components in the graph. Specifically, use the depth-first search DFS or breadth-first search BFS algorithm to traverse the adjacency matrix and identify the fan groups with strong wake coupling relationships;

[0098] Regard all connected components as a sub-wind farm, where the wake coupling relationship between the fans within each sub-wind farm is strong, while the wake influence between sub-wind farms can be ignored.

[0099] In a wind farm, the wake effect between fans will have an important impact on the operation of downstream fans. Therefore, accurately dividing the area of the wind farm and identifying the wake coupling relationship between fans is the key to realizing effective control. The purpose of step S2 is to reasonably divide the area of the wind farm based on the wake influence adjacency matrix through the directed graph connected component analysis method. This step not only provides a more efficient calculation framework for the optimal control of the wind farm, but also improves the real-time performance and accuracy of the control strategy.

[0100] Further, for constructing a wake mathematical model based on the parameters of each wind turbine in the wind farm and calculating the equivalent inflow wind speed of each fan, specifically including:

[0101] Collect the parameters of each fan in the wind farm, including the inflow wind speed, thrust coefficient, tip speed ratio, pitch angle, rotor radius, and the distance between fans;

[0102] According to the parameters of the fan, establish a wake mathematical model and calculate the wake influence wind speed of each fan affected by the wake;

[0103] Taking into account the wake effects of all upstream wind turbines on downstream wind turbines, a wake wind speed superposition model is used to calculate the equivalent inflow wind speed of the wind turbines.

[0104] By calculating the equivalent inflow wind speed of each wind turbine, the actual wind speed input of the wind turbine is obtained.

[0105] Furthermore, the formula for calculating the wake influence wind speed of each wind turbine affected by the wake is:

[0106]

[0107] R j,i =R j +kd j,i

[0108] where ν i,j represents the wake influence wind speed of wind turbine i under the wake influence of wind turbine j, v j is the inflow wind speed of wind turbine j, C t,j is the thrust coefficient of wind turbine j, λ j is the tip speed ratio of wind turbine j, β j is the pitch angle of wind turbine j, R i is the rotor radius of wind turbine i, R j,i is the wake expansion radius of wind turbine j at the position of wind turbine i, S i is the rotor swept area of wind turbine i, S j,i is the overlapping area between the wake of upstream wind turbine j and the rotor swept area of downstream wind turbine i, k is the wake expansion coefficient of wind turbine i, d j,i is the distance between wind turbine j and wind turbine i.

[0109] Furthermore, the formula for calculating the equivalent inflow wind speed of the wind turbine is:

[0110]

[0111] where Γ is the set of all upstream wind turbines that have a wake influence on downstream wind turbine i, obtained according to the wake influence adjacency matrix L, V ∞ is the background wind speed of the wind farm, and δν i,j is the wind speed loss.

[0112] Furthermore, the root step S4 specifically includes:

[0113] Near the equivalent inflow wind speed V i a first-order Taylor expansion is performed on the non-linear model of the wind turbine to obtain a linearized model;

[0114] According to the linearization result, a linear state space model is constructed, and the linear state space model includes:

[0115]

[0116] Among them, the specific expressions of the state space matrices A, B, C, E, and W are as follows:

[0117]

[0118] d = V i

[0119] x = (ω r ω e ψT e β) T

[0120]

[0121] y = (P e M t T s C t ) T

[0122] Among them, A is the system matrix, B is the input matrix, C is the output matrix, E is the input disturbance matrix, W is the output disturbance matrix, x = (ω r ω e ψT e β) T is the state vector of the wind turbine, ω r is the wind turbine rotor speed, ω e is the generator speed, ψ is the twist angle, T e is the electromagnetic torque, β is the pitch angle, ref is the abbreviation of reference, representing the meaning of reference value, u is the control input, y = (P e M t T s C t ) T is the output vector, P e is the electrical power output of the wind turbine, M t is the tower load, T s is the tower height, C t is the thrust coefficient, is the main shaft damping coefficient, N gear is the gearbox speed ratio, J r is the wind turbine rotor inertia, J g is the generator inertia, K sp is the main shaft elastic coefficient, τ e is the generator inertia time constant, τ β is the pitch angle inertia time constant.

[0123] The purposes of steps S3 and S4 are to accurately calculate the equivalent inflow wind speed of the wind turbines and construct the corresponding state - space model through the wake mathematical model and the linearized wind turbine model, so as to achieve precise control and optimization of the wind farm. In step S3, first, according to the relevant parameters of the wind turbine (such as thrust coefficient, tip - speed ratio, pitch angle, etc.), a wake mathematical model is established to calculate the wake - affected wind speed of each wind turbine. This process involves superimposing the wake effects of multiple upstream wind turbines, and then obtaining the equivalent inflow wind speed of each wind turbine. The wake effect is the key to optimizing the performance of the wind farm because the wake interaction between wind turbines will affect the wind speed of downstream wind turbines, thus affecting the overall power generation efficiency.

[0124] In step S4, the wind turbine model of each sub - wind farm is further linearized by the Taylor expansion method to obtain a linear state - space model. The technical effect of this is to transform the complex non - linear wind turbine dynamic model into a simplified linear form, which is convenient for the design and analysis of the control system. By constructing the linear state - space model, the control and optimization of the wind turbines can be carried out more efficiently under actual wind speed conditions, improving the overall response speed and power output stability of the wind farm. Specifically, after linearization, the operation model of the wind turbine becomes easy to handle and calculate, enabling more precise adjustment of the control strategy for each wind turbine, thus maximizing the power generation efficiency of the wind farm when the wind speed changes, reducing the computational complexity of the system, and enhancing the real - time control effect.

[0125] Furthermore, it specifically includes:

[0126] After obtaining the linear state - space models of each wind turbine in different equivalent inflow wind speed conditions in each sub - wind farm, based on the system state matrix A, input matrix B, and output matrix C, the state - space expression is converted into the transfer - function form through Laplace transform to obtain the transfer - function model of the wind turbine control model:

[0127] K(s)=C(sI - A) -1 B

[0128] where K(s) is the transfer - function model of the wind turbine, s is the Laplace complex - frequency domain variable, and I is the identity matrix;

[0129] Applying the gap metric theory to the transfer - function models K(s) under multiple wind speed conditions in each sub - wind farm specifically includes:

[0130] The transfer - function model is expressed in the following form:

[0131] K(s)=N(s)M(s) -1

[0132] where M(s) and N(s) are the input transfer matrix and output transfer matrix of the system respectively, satisfying The symbol ~ represents conjugate transpose;

[0133] Calculate the symmetric gap metric between any two wind turbines i and j in the sub-wind farm:

[0134]

[0135] where δ(K i , K j ) is the symmetric gap metric between the two models of wind turbine i and wind turbine j, represents the directional gap metric of wind turbine model K i relative to K j , K i , K j represent the transfer function models of wind turbine i and wind turbine j respectively, M i , N i , M j , N j are the system input transfer matrix and output transfer matrix of the models of wind turbine i and wind turbine j respectively, inf represents finding the minimum value, Q ∈ H ∞ represents a stable transformation matrix selected in the H ∞ space, which is used to approximate the mapping relationship between the two systems, ||.|| ∞ represents the H ∞ norm;

[0136] Based on the symmetric gap metric δ(K i , K j ) between all wind turbines in the sub-wind farm, construct the gap metric matrix of the sub-wind farm, and use the threshold-based hierarchical clustering method to divide the wind turbines into multiple model subsets with similar dynamic characteristics. In each subset, select the wind turbine model with the smallest sum of gap metric distances as the representative model of this subset, which is used for the unified control modeling of this type of wind turbine.

[0137] The purpose of steps S5 and S6 is to simplify the control models of the wind turbines in the wind farm through dimensionality reduction and clustering processing, so that the wind farm can perform efficient optimal control under different wind speed conditions. The dimensionality reduction processing in step S5 is based on the gap metric theory, mainly by comparing the dynamic characteristics between the control models of the wind turbines and merging those wind turbine models with similar characteristics. The main reason for doing this is that in a wind farm, the control model of each wind turbine may be very complex, and the control requirements and dynamic response characteristics between different wind turbines may be very similar. Therefore, through dimensionality reduction, the control models of multiple wind turbines can be merged into fewer representative models. This not only reduces the number of control models in the system, but also reduces the computational amount and the complexity of the control system, thereby improving the control efficiency.

[0138] After dimensionality reduction, in step S6, multiple wind turbine models are integrated according to the division of sub-wind farms in the wind farm to form a multi-model control matrix. This matrix can adapt to different wind turbine behaviors under different wind speed conditions, thereby achieving global optimal control of the wind farm. By constructing a unified control model for each sub-wind farm, the wind turbines in each subsystem within the wind farm can be more effectively scheduled and managed, avoiding the problem of excessive computational burden caused by too many control models. In addition, when using the gap metric theory for clustering, the wind turbine models are classified into different subsets of dynamic characteristics based on similarity, enabling the wind turbine control to be adjusted according to the representative models of different subsets, further improving the performance and stability of the overall control system of the wind farm.

[0139] By adopting this method of dimensionality reduction and merging, the wind farm can flexibly respond to different working condition changes, while ensuring the real-time response ability and optimization performance of the system, thus improving the economic benefits and energy production efficiency in large-scale wind farms.

[0140] Through the dimensionality reduction process in this stage, the unified modeling of linear models under multiple wind speed conditions within the sub-wind farm is successfully achieved. Specifically, for the model differences caused by wind speed changes, the gap metric theory is used to classify models with similar dynamic characteristics into one category, enabling some rear-row wind turbines to share a unified control model, while the front-row wind turbines or individuals with significant dynamic differences still retain independent modeling. By this method, the number of models is significantly reduced, and the complexity and real-time computational burden of the overall control system are decreased.

[0141] On this basis, each sub-wind farm finally corresponds to a finite number of representative control models, forming a multi-model control strategy with clear structure and switchability. This simplified control model not only retains the main characteristics of the system's dynamic response but also can effectively adapt to the working condition changes brought about by wind speed fluctuations, providing a unified and efficient model basis for subsequent distributed optimal control. As shown in the following formula, some of the original wind turbine models in A WF1 are unified, and the simplified model is used for the control of the sub-wind farm. All the divided sub-wind farms are controlled through this process to achieve the purpose of controlling large-scale wind farms.

[0142] Embodiment 2:

[0143] This embodiment selects the wind farm shown in Figure 1 for simulation analysis. This wind farm includes 16 5MW wind turbines, with a total installed capacity of 80MW. The wind direction is east (blowing from west to east), and the wind speed magnitude is as shown in Figure 2 . The simulation step is set to 1s, and the total simulation duration is 300s.

[0144] (1) Wake zone analysis based on the adjacency matrix

[0145] During the wake zone division process of a wind farm, first, the upstream and downstream relationships of the wind turbines are determined based on the wind direction information, and a wake influence network between the wind turbines is established. Subsequently, the wake influence relationship between the wind turbines is constructed based on the adjacency matrix method. By calculating the wake influence degree between the wind turbines, a 16×16 wake adjacency matrix is formed, and the connected component algorithm is used for regional division. In this case, considering the wake overlap range and influence degree between the wind turbines, the wind farm is finally divided into six independent regions as Figure 5 , Figure 6 shown.

[0146] (2) Analysis of the unified characterization model of the gap metric theory

[0147] It can be seen from Figure 4 that the method for simplifying the control model based on the gap metric theory proposed in this embodiment can unify the characterization of multi-wind speed linear models in a large range from 12 m / s to 25 m / s into four models by selecting a gap degree of 0.1, namely 12 m / s, 15 m / s, 19 m / s, and 23 m / s.

[0148] (3) Analysis of computational efficiency

[0149] For the above two-stage dimensionality reduction method, the fast control performance of a typical wind farm is analyzed. In the first stage, the wake influence between the wind turbines is analyzed through the adjacency matrix, and the wind farm is divided into six independent sub-wind farms, enabling the calculation to be carried out only in local areas, thereby reducing the dimension of the model and the computational scale. In the second stage, the gap metric theory is used to uniformly characterize the linearized models of the wind turbines under different wind speed conditions, reducing the number of models to be processed during the optimization process and avoiding computational redundancy caused by excessive model switching.

[0150] Table 1: Solution results

[0151]

[0152] According to Table 1, it can be seen from this embodiment that the computational complexity of this method is reduced from O(N 2 ) to O(N / 6) 2 compared with the traditional method, and the computational time is shortened by about 83.1%, significantly improving the real-time performance and computational efficiency of the optimal control. The present invention provides an efficient and reliable technical solution for the fast optimal control model of large-scale wind farms.

[0153] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0154] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An optimized control method for a wind farm based on multi-condition model dimensionality reduction, characterized in that It includes the following steps: Step S1: According to the wake influence overlapping area between wind turbines in a wind farm, judge the wake influence relationship between each pair of wind turbines in the wind farm, and establish a wake influence adjacency matrix; Step S2: Based on the wake influence adjacency matrix, partition the wind farm. By using directed graph connected component analysis, divide the wind turbines in the wind farm into multiple sub-wind farms according to the wake influence relationship. Among them, the wake coupling relationship between the wind turbines within each sub-wind farm is strong, while the wake influence between sub-wind farms can be ignored; Step S3: According to the parameters of each wind turbine in the wind farm, construct a wake mathematical model and calculate the equivalent inflow wind speed of each wind turbine; Step S4: According to the equivalent inflow wind speed of each wind turbine, construct a wind turbine model for each wind turbine within each sub-wind farm, linearize the wind turbine models within each sub-wind farm, and use the Taylor expansion method to obtain the linear state space model of the wind turbine models within each sub-wind farm under different wind speed conditions; Step S5: Use the gap metric theory to perform dimensionality reduction processing on the linear state space models of the wind turbine models within each sub-wind farm under different wind speed conditions, merge the wind turbine models with similar dynamic characteristics, and reduce the number of control models; Step S6: Combine the dimensionality-reduced wind turbine control models according to sub-wind farms to form a multi-model control matrix, and realize the optimal control of the wind farm adapted to different wind speed conditions.

2. The optimized control method for a wind farm based on multi-condition model dimensionality reduction according to claim 1, characterized in that The elements in the wake influence adjacency matrix represent whether there is a significant wake influence relationship between wind turbines.

3. The optimized control method for a wind farm based on multi-condition model dimensionality reduction according to claim 2, wherein The wake influence adjacency matrix is: Among them, N represents the number of wind turbines in the wind farm, and α ij represents the wake influence relationship between wind turbine i and wind turbine j, specifically: Among them, S i is the rotor swept area of wind turbine i, and S j,i is the overlapping area between the wake of upstream wind turbine j and the rotor swept area of downstream wind turbine i.

4. The optimized control method for a wind farm based on multi - operating condition model dimension reduction according to claim 1, characterized in that, For the partition of the wind farm based on the wake influence adjacency matrix, using directed graph connected component analysis to divide the wind turbines in the wind farm into multiple sub-wind farms according to the wake influence relationship, specifically including: Each wind turbine in a wind farm is regarded as a node in a graph, and according to the element α in the wake influence adjacency matrix L ij , weights are assigned to the edges in the graph. If α ij = 1, an edge is established between wind turbine i and wind turbine j, indicating a significant wake influence relationship between the two; Adopt the directed graph connected component analysis algorithm to detect all connected components in the graph. Specifically, use the depth-first search DFS or breadth-first search BFS algorithm to traverse the adjacency matrix and identify the wind turbine groups with strong wake coupling relationships; Regard all connected components as a sub-wind farm. Among them, the wake coupling relationship between the wind turbines within each sub-wind farm is strong, while the wake influence between sub-wind farms can be ignored.

5. A wind farm optimization control method based on multi-condition model dimension reduction according to claim 1, characterized in that For constructing the wake mathematical model according to the parameters of each wind turbine in the wind farm and calculating the equivalent inflow wind speed of each wind turbine, specifically including: Collect the parameters of each wind turbine in the wind farm, including the inflow wind speed, thrust coefficient, tip speed ratio, pitch angle, rotor radius, and the distance between wind turbines; [[ID=I4]]According to the parameters of the wind turbine, establish a wake mathematical model and calculate the wake influence wind speed of each wind turbine affected by the wake; Comprehensively consider the wake influence of all upstream wind turbines on downstream wind turbines, and use the wake wind speed superposition model to calculate the equivalent inflow wind speed of the wind turbine; By calculating the equivalent inflow wind speed of each wind turbine, obtain the actual wind speed input of this wind turbine.

6. The optimized control method for a wind farm based on multi - operating condition model dimension reduction according to claim 5, wherein The formula for calculating the wake influence wind speed of each wind turbine affected by the wake is: R j,i = R j + kd j,i Among them, ν i,j represents the wake influence wind speed of wind turbine i under the wake influence of wind turbine j, ν j is the inflow wind speed of wind turbine j, C t,j is the thrust coefficient of wind turbine j, λ j is the tip speed ratio of wind turbine j, β j is the pitch angle of wind turbine j, R i is the rotor radius of wind turbine i, R j,i is the wake expansion radius of wind turbine j at the position of wind turbine i, S i is the rotor swept area of wind turbine i, S j,i is the overlapping area between the wake of the upstream wind turbine j and the rotor swept area of the downstream wind turbine i, k is the wake expansion coefficient of wind turbine i, d j,i is the distance between wind turbine j and wind turbine i.

7. A wind farm optimization control method based on multi - operating condition model dimensionality reduction according to claim 5, characterized in that, The formula for calculating the equivalent inflow wind speed of the wind turbine is: Among them, Γ is the set of all upstream wind turbines that have a wake effect on the downstream wind turbine i, obtained according to the wake effect adjacency matrix L, V ∞ is the background wind speed of the wind farm, and δν i,j is the wind speed loss.

8. A wind farm optimization control method based on multi-condition model dimension reduction according to claim 1, characterized in that Regarding step S4, specifically including: Near the equivalent incoming flow velocity V i Perform a first-order Taylor expansion on the non-linear model of the wind turbine to obtain a linearized model; According to the linearization result, construct a linear state space model. The linear state space model includes: Among them, the specific expressions of the state space matrices A, B, C, E, and W are as follows: d = V i x = (ω r ω e ψT e β) T y = (P e M t T s C t ) T where, A is the system matrix, B is the input matrix, C is the output matrix, E is the input disturbance matrix, W is the output disturbance matrix, x = (ω r ω e ψT e β) T is the state vector of the wind turbine, ω r is the wind turbine rotor speed, ω e is the generator speed, ψ is the twist angle, T e is the electromagnetic torque, β is the pitch angle, ref is the abbreviation of reference representing the meaning of reference value, u is the control input, y = (P e M t T s C t ) T is the output vector, P e is the electrical power output of the wind turbine, M t is the tower load, T s is the tower height, C t is the thrust coefficient, is the main shaft damping coefficient, N gear is the gearbox speed ratio, H r is the wind turbine rotor inertia, J g is the generator inertia, K sp is the main shaft elastic coefficient, τ e is the generator inertia time constant, τ β is the pitch angle inertia time constant.

9. The optimization control method for a wind farm based on multi-condition model dimensionality reduction according to claim 1, wherein, The specific steps of step S5 include: After obtaining the linear state space models of each wind turbine in different equivalent inflow wind speed conditions in each sub-wind farm, based on the system state matrix A, input matrix B, and output matrix C, the state space expression is converted into the transfer function form through Laplace transform to obtain the transfer function model of the wind turbine control model: K(s) = C(sI - A) -1 B Among them, K(s) is the transfer function model of the wind turbine, s is the Laplace complex frequency domain variable, and I is the identity matrix; Applying the gap metric theory to the transfer function models K(s) under multiple wind speed conditions in each sub-wind farm specifically includes: Express the transfer function model in the following form: K(s) = N(s)M(s) -1 where \(M(s)\) and \(N(s)\) are the input transfer matrix and output transfer matrix of the system respectively, and satisfy The symbol ~ represents conjugate transpose; Calculate the symmetric gap metric between any two wind turbines i and j in the sub-wind farm: where, δ(K i ,K j ) is the symmetric gap metric between two models of fan i and fan j, denotes the directional gap metric of fan model K i relative to K j , K i ,K j respectively denote the transfer function models of fan i and fan j, M i , N i , M j , N j are respectively the system input transfer matrix and output transfer matrix of the models of fan i and fan j, inf represents finding the minimum value, Q ∈ H ∞ denotes the stable transformation matrix selected in the H ∞ space to approximate the mapping relationship between two systems, ||.|| ∞ denotes the H ∞ norm; Based on the symmetric gap metric δ(K i , K j ) among all the wind turbines in the sub-wind farm, a gap metric matrix of the sub-wind farm is constructed, and a hierarchical clustering method based on a threshold is used to divide the wind turbines into multiple model subsets with similar dynamic characteristics. In each subset, the wind turbine model with the smallest sum of gap metric distances is selected as the representative model of the subset for the unified control modeling of this type of wind turbine.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1 to 9.

Images