A method and device for decentralized coordinated control of an offshore wind farm cluster

Abstract

The application discloses a kind of offshore wind farm group's decentralized coordination control method and device, its method includes: the tail flow wind turbine power and thrust load balance optimization and control model of considering tail flow influence of single wind turbine group are constructed;Original tail flow directed graph is constructed, and original tail flow directed graph is decomposed into completely uncoupled sparse sub-tail flow directed graph using graph weight pruning algorithm and graph depth-first search algorithm, sparse wind farm group-field-machine multi-layer decentralized control system is constructed, and decentralized wind farm power and thrust load balance optimization model is constructed, and the optimal value of power and thrust load balance control parameter is solved.The method and device of the application establish power and thrust balance optimization model, realize a kind of communication burden low, less and scalable wind turbine group decentralized coordination control method and device.

Description

[Technical Field]

[0001] This invention relates to artificial intelligence and machine learning algorithms, particularly to the field of optimized scheduling methods and intelligent control technologies for offshore wind farm clusters. Specifically, it relates to a distributed coordination control method and device for offshore wind farm clusters. [Background Technology]

[0002] Currently, offshore wind power is developing towards larger scale and greater scale. With the increasing size of wind farms, new challenges arise for wind farm efficiency optimization and control technologies. Large offshore wind farms are complex optimization and control systems characterized by high dimensionality, strong time-varying properties, and strong nonlinearity. Due to the "deep array" wake effect, downstream turbines located within the wake region of upstream turbines can experience power losses of up to 40%. The uncertainty of turbulent changes increases the fatigue load on the turbines. Simultaneously, as the number of wind turbines in a wind farm increases, the control variables also grow exponentially. This leads to problems such as high communication pressure and high computational costs for traditional centralized controllers, making it difficult to ensure a sufficiently fast response to wake changes between wind turbines when optimizing the scheduling of the wind farm, thus failing to guarantee optimal coordinated control results.

[0003] As wind farms expand in scale, the optimal scheduling of large-scale wind farms is gradually evolving towards coordinated control of clustered turbines. Research indicates that the wake relationship between upstream and downstream turbines has a significant impact on wind farm optimization and control. The wake cascading relationship between turbines along the windward direction exhibits unidirectional coupling characteristics, and these turbines possess cluster characteristics. This necessitates moving beyond simple independent control strategies for individual turbines and instead implementing coordinated optimization and control of the cluster. Simultaneously, with the emergence of smart wind farms, exploring the wake coupling effect between wind turbines in complex meteorological environments within large-scale wind farms using new artificial intelligence and machine learning algorithms, optimizing power acquisition and thrust fatigue load distribution, and achieving simultaneous coordinated optimization of wind energy capture and turbine lifespan are critical issues that urgently need to be addressed in wind farm optimal scheduling and control technology. [Summary of the Invention]

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a decentralized coordinated control method and device for large-scale offshore wind farm clusters. For large-scale offshore wind farm clusters operating normally, a directed wake graph is constructed by analyzing the wake coupling relationship and distribution characteristics of the wind turbine generators. Graph pruning is then used to sparsify the topology of the wind farm communication network formed by this wake graph. A graph clustering algorithm is then applied to disperse and cluster the wind turbine generators, establishing an integrated simulation model for wind farm cluster power acquisition, thrust fatigue load, and control characteristic analysis. This results in a decentralized coordinated control method and device for wind turbine generator clusters with low communication burden, low computational load, and scalability.

[0005] To achieve the above objectives, in a first aspect, the technical solution adopted by the present invention is: a decentralized coordinated control method for offshore wind farm clusters, comprising:

[0006] Construct a calculation model for the power and thrust load of a single wind turbine in an offshore wind farm, taking into account the wake effect;

[0007] The location, wind direction, and wind speed information of all wind turbines in the offshore wind farm are sequentially input into the wake turbine power and thrust load calculation model that considers the wake effect to obtain the wake coupling relationship of the offshore wind farm. This relationship is then modeled as an original wake directed graph. In this graph, a single wind turbine in the offshore wind farm constitutes a node, and the wake coupling relationship between two adjacent single wind turbines in the offshore wind farm constitutes an edge. The weight of this edge is defined as the equivalent wind speed attenuation caused by the wake.

[0008] The original wake directed graph is decomposed into a completely uncoupled sparse sub-wake directed graph using graph weight pruning algorithm and graph depth-first search algorithm. Based on the sparse sub-wake directed graph, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm group-farm-turbine multi-layer distributed control system.

[0009] Based on the wake turbine power and thrust load calculation model considering the wake effect, and combined with the sub-wind farm, the power and thrust calculation functions of multiple clustered sub-wind farm turbine clusters within the offshore wind farm are obtained, and the turbine control variable x is defined. i :=[γ i ] T A distributed wind farm power and thrust load balance optimization model is constructed to solve for the optimal values ​​of the power and thrust load balance control parameters. The distributed wind farm power and thrust load balance optimization model is expressed as follows:

[0010]

[0011] Where f(x) represents the sum of the power and thrust loss functions of all the sub-wind farms in the offshore wind farm; i represents the i-th group of the clustered sub-wind farms, M represents the total number of the clustered sub-wind farms in the offshore wind farm; j represents the j-th turbine in the clustered sub-wind farms, and K represents the total number of turbines in the clustered sub-wind farms; γ i，j -γ represents the yaw angle of the j-th turbine in the i-th cluster of sub-wind farms. min γ represents the minimum yaw angle value. max Indicates the maximum yaw angle; p i，j p represents the power of the j-th turbine in the i-th cluster of sub-wind farms. low p represents the minimum power value.rate f represents the rated power value. i，j f represents the thrust load of the j-th turbine in the i-th cluster of sub-wind farms. low f represents the minimum thrust load. rate ξ represents the rated thrust load, and ξ represents the balance parameter.

[0012] Based on the above method, in a second aspect, the technical solution adopted by the present invention is: a decentralized coordinated control device for an offshore wind farm cluster, comprising:

[0013] The single wind turbine model building module is used to build a calculation model of the power and thrust load of a single wind turbine in an offshore wind farm, taking into account the wake effect.

[0014] The wake analysis module is used to sequentially input the location, wind direction, and wind speed information of all wind turbines in the offshore wind farm into the wake turbine power and thrust load calculation model that considers the wake effect, to obtain the wake coupling relationship of the offshore wind farm, and model it as an original wake directed graph. In this graph, a single wind turbine in the offshore wind farm constitutes a node, and the wake coupling relationship between two adjacent single wind turbines in the offshore wind farm constitutes an edge. The weight value of this edge is defined as the equivalent wind speed attenuation caused by the wake.

[0015] The system construction module is used to decompose the original wake directed graph into a completely uncoupled sparse sub-wake directed graph using graph weight pruning algorithm and graph depth-first search algorithm. Based on the sparse sub-wake directed graph, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm group-farm-turbine multi-layer distributed control system.

[0016] The coordination and optimization module is used to obtain the power and thrust calculation functions of multiple clustered sub-wind farm wind turbine clusters within the offshore wind farm, based on the wake turbine power and thrust load calculation model that considers the wake effect, combined with the sub-wind farm, define wind turbine control variables, construct a distributed wind farm power and thrust load balance optimization model, and solve for the optimal values ​​of power and thrust load balance control parameters.

[0017] The advantages of this invention are:

[0018] 1. The method of this invention analyzes the wake coupling relationship and distribution characteristics of the wind farm in the offshore wind farm cluster by using the wake turbine power and thrust load model of a single wind turbine unit. It adopts the graph weight pruning algorithm and the graph depth-first search algorithm to sparsify and disperse the wake distribution, and constructs a sparse wind farm cluster-field-turbine multi-layer decentralized control system, thereby realizing decentralized management and coordinated control of the entire wind farm cluster.

[0019] 2. The method of this invention constructs a power and thrust balance model for the wind turbine wake based on the influence of the wind turbine's yaw angle on the wake. It uses graph clustering to form highly correlated and strongly coupled intelligent agents, and introduces a decentralized control approach to achieve global objectives through local control coordination. Furthermore, it employs graph pruning and graph depth-first search algorithms to construct a decentralized wind farm communication framework, facilitating parallel computation and deployment across multiple computing nodes. Based on the yaw wake effect, it constructs a power and thrust load balance optimization model for the clustered sub-wind farms, and uses sequential quadratic programming to solve for the optimal values ​​of power and thrust load balance optimization, achieving optimized control of wind farm power and thrust load balance. The established decentralized power and thrust optimization framework can reduce computational complexity, improving the total power generation of the wind farm while simultaneously reducing the wind turbine thrust load value. [Attached Image Description]

[0020] Appendix Figure 1 This is a flowchart illustrating one embodiment of a decentralized coordinated control method for offshore wind farm clusters according to the present invention.

[0021] Appendix Figure 2 This is a flowchart illustrating the calculation model of wake turbine power and thrust load for a single wind turbine unit, considering the wake effect, in the decentralized coordinated control method for offshore wind farm groups of the present invention.

[0022] Appendix Figure 3 This is a flowchart illustrating the construction of a sparsed sub-wake directed graph in a decentralized coordinated control method for offshore wind farm clusters according to the present invention.

[0023] Appendix Figure 4 It is attached Figure 3 A magnified view of a portion of Figure (b) in the figure;

[0024] Appendix Figure 5 It is attached Figure 3 The figure shows a magnified view of the pruned wake directed graph after k=0 in Figure (c).

[0025] Appendix Figure 6 It is attached Figure 3 The figure shows a magnified view of the pruned wake direction graph after pruning, where k = 0.035;

[0026] Appendix Figure 7 It is attached Figure 3 The figure shows a magnified view of the pruned wake directed graph after k=0.055 in Figure (c).

[0027] Appendix Figure 8 It is attached Figure 3 The figure shows a magnified view of the sparsified wake directed graph (d) where k=0;

[0028] Appendix Figure 9 It is attached Figure 3The figure shows a magnified view of the sparsified wake directed graph (d) with k = 0.035.

[0029] Appendix Figure 10 It is attached Figure 3 The figure shows a magnified view of the sparsified wake directed graph (d) with k = 0.055.

[0030] Appendix Figure 11 This is a schematic diagram of the construction of a sparse wind farm cluster-farm-machine multi-layer distributed control system in one embodiment of the present invention;

[0031] Appendix Figure 12 This is a schematic block diagram of the structure of a distributed coordination control device for an offshore wind farm group according to one embodiment of the present invention.

Detailed Implementation Methods

[0032] This invention provides a decentralized coordinated control method and apparatus for offshore wind farm clusters, primarily addressing problems in the optimal scheduling and coordinated control of large-scale offshore wind farm clusters. The technical concept of this invention is as follows: First, a calculation model for the wake turbine power and thrust load of a single wind turbine in the wind farm cluster, considering the wake effect, is constructed. Based on this model, the wake coupling relationship and distribution characteristics in the wind farm are analyzed, and an original wake directed graph is constructed. Further, a graph pruning sparsification algorithm and a graph depth-first search algorithm are used to process the original wake directed graph, constructing a sparse sub-wake directed graph. This makes the entire wind farm cluster constitute a sparsed wind farm cluster-farm-turbine multi-layer decentralized control system. Furthermore, this method addresses the highly correlated and strongly coupled components in the offshore wind farm. The wind turbine clusters are clustered into multiple sub-wind farms, constructing a sparse wind farm cluster-farm-turbine multi-layer decentralized control system. Combined with the wake turbine power and thrust load calculation model that considers the wake effect, a decentralized wind farm power and thrust load balance optimization model is constructed. Using this model, a decentralized optimization framework and a computationally efficient wake-directing wind farm control strategy are built. This strategy uses the yaw angle of the wind turbine to change the wake behavior of the wind turbine and minimize the wake interaction of the wind turbine, thereby improving the power generation of the wind farm while considering the fatigue load of the wind farm.

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the embodiments described in this specification are only a part of the implementation of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should be considered to fall within the protection scope of this invention.

[0034] As attached Figure 1As shown, this invention provides a decentralized coordinated control method for offshore wind farm clusters, applicable to large-scale wind farms with multiple wind turbine generators, particularly suitable for large offshore wind farm clusters. It employs a strategy to simultaneously optimize wind farm power and thrust load by adjusting the yaw angle of the wind turbines. (See attached diagram.) Figure 1 The specific implementation process of the method of the present invention is described below, and the technical solution of the method of the present invention is described as follows:

[0035] Step S101. Construct a wake turbine power and thrust load calculation model for a single wind turbine unit in an offshore wind farm, considering the wake effect. The wake turbine power and thrust load calculation model can be understood as constructing a calculation model of the wind speed distribution of a single wind turbine unit in the wake region. Current extensive wind tunnel and field experimental data show that the wind speed distribution in the wake region exhibits a Gaussian distribution, with wind shear effects in the incoming wind and complex variations in turbulence intensity in the wake region. Therefore, in practice, a Gaussian wake model is preferred to describe the wind speed changes in the wake region, improving the prediction accuracy of the wake wind speed distribution in the wind turbine hub height plane.

[0036] Step S102. Input the location, wind direction, and wind speed information of all wind turbines in the offshore wind farm into the wake turbine power and thrust load calculation model considering the wake effect in sequence to obtain the wake coupling relationship of the offshore wind farm, and model it as an original wake directed graph. In this graph, a single wind turbine in the offshore wind farm constitutes a node of the original wake directed graph, and the wake coupling relationship between two adjacent single wind turbines in the offshore wind farm constitutes an edge of the original wake directed graph. The weight value of this edge is defined as the equivalent wind speed attenuation caused by the wake effect.

[0037] The specific implementation process of step S102 above can be described in detail by formula: the original wake directed graph is expressed as follows: Where v represents a node and ε represents an edge. This represents the weight on the node connection, and the weight value is expressed as (w ij ) N×N It is defined as the weight value of the line connecting upstream wind turbine j and downstream wind turbine i in a wind farm with N wind turbines, and is expressed as a matrix:

[0038]

[0039] in, This represents the equivalent wind speed reduction at the downstream wind turbine location, where N represents the total number of wind turbines in a wind farm.

[0040] In this embodiment, the idea of ​​constructing the wake directed graph is as follows: a wind farm capable of autonomy is constructed through an intelligent graph representation learning model. In this wind farm, each individual wind turbine is regarded as an intelligent agent in the model, and the entire wind farm is regarded as an intelligent agent system in the model. Through this approach to constructing an intelligent wind farm control system, the wake coupling relationship in the wind farm can be constructed as a wake directed graph. The network topology structure is constructed through the directed graph to realize the digital management of the wind farm.

[0041] Step S103. Using the graph weight pruning algorithm and the graph depth-first search algorithm, the original wake directed graph is decomposed into completely uncoupled sparse sub-wake directed graphs. Based on the sparse sub-wake directed graphs, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm group-farm-turbine multi-layer distributed control system. In this control system, "machine" refers to a single wind turbine in a wind farm cluster, which forms the bottom layer of the control system in the form of a subset of wind turbines in the distributed control system of the wind farm cluster. "Field" refers to the clustered sub-wind farms obtained by dispersively clustering the wind farm cluster, which form the middle layer of the control system in the form of wind turbine clusters in the distributed control system of the wind farm cluster. The wind turbines in each clustered sub-wind farm are analyzed and controlled by distributed sub-controllers. "Cluster" refers to the wind farm cluster, that is, the entire large-scale offshore wind farm cluster, which is usually centrally processed and coordinated by a central controller.

[0042] By employing a graph weighted pruning algorithm, considering the maximum shared node in-degree value, the original wake directed graph is decomposed into completely decoupled sparse sub-wake directed graphs, reducing the complexity of the sub-graphs. A graph depth-first search algorithm is used to cluster highly correlated and strongly coupled wind turbine units. This algorithm searches the tree branches as deeply as possible; when all edges containing a node have been explored, the search backtracks to the starting node of the edge that discovered the node. This process continues until all nodes reachable from the source node have been found. The algorithm finds all nodes in the sub-directed graph and defines them as sub-wind farms, facilitating hierarchical control of several wind farm groups.

[0043] Step S104. Based on the wake turbine power and thrust load calculation model considering the wake effect, and combined with the sub-wind farm, obtain the power and thrust calculation functions for the turbine clusters of multiple clustered sub-wind farms within the offshore wind farm, and define the turbine control variable x. i :=[γ i ] T A distributed wind farm power and thrust load balance optimization model is constructed to solve for the optimal values ​​of the power and thrust load balance control parameters. The distributed wind farm power and thrust load balance optimization model is expressed as follows:

[0044]

[0045] Where f(x) represents the sum of the power and thrust loss functions of all the sub-wind farms in the offshore wind farm; i represents the i-th group of the clustered sub-wind farms, M represents the total number of the clustered sub-wind farms in the offshore wind farm; j represents the j-th turbine in the clustered sub-wind farms, and K represents the total number of turbines in the clustered sub-wind farms; γ i，j -γ represents the yaw angle of the j-th turbine in the i-th cluster of sub-wind farms. min γ represents the minimum yaw angle value. max Indicates the maximum yaw angle; p i，j p represents the power of the j-th turbine in the i-th cluster of sub-wind farms. low p represents the minimum power value. rate f represents the rated power value. i，j f represents the thrust load of the j-th turbine in the i-th cluster of sub-wind farms. low f represents the minimum thrust load. rate ξ represents the rated thrust load, and ξ represents the balance parameter.

[0046] The above technical solution analyzes the distribution characteristics and impact effects of wind farm wakes based on the influence of wind turbine yaw angle on wind turbine wakes. It adopts a decentralized control approach, from individual wind turbines to local control and coordination of the entire large-scale wind farm group, to achieve the global objective of the whole. Digital algorithms and analysis methods are used to construct a power and thrust balance optimization model of the wind farm group. By calculating and solving the optimal values ​​of power and thrust load balance control parameters, the global control and optimization coordination of the large-scale offshore wind farm is achieved.

[0047] In a preferred embodiment, combined with the appendix Figure 2 In step S101 above, a Gaussian wake calculation model oriented towards control is preferably used to construct a wake turbine power and thrust load calculation model for the single wind turbine unit considering the wake effect. The specific implementation steps include:

[0048] Step S111. Substitute the wake expansion coefficient, axial induction factor, and rotor radius of the single wind turbine unit into the Gaussian wake calculation model to calculate the wind speed u(d,r,α) and wind speed attenuation δu(d,r,α) flowing into a certain point of the downstream single wind turbine i under the influence of the wake of the upstream wind turbine j; In the wind farm, when the free wind U flows past a certain wind turbine, at a downstream wake distance of d and a radial distance r from the wind turbine, the wind speed u(d,r,α) can be expressed as:

[0049] u(d,r,α)=(1-δu(d,r,α))U, (6)

[0050] Where δu(d, r, α) represents the quantified wind speed attenuation within the wake region, and α represents the axial induction factor of a single wind turbine. The wind speed attenuation δu(d, r, α) is obtained using a Gaussian wake model, which is expressed as:

[0051]

[0052] Where R0 represents the radius of the fan rotor, and κ represents the wake expansion coefficient.

[0053] Step S112. According to the momentum theorem, integrate the wind speed and wind speed attenuation on the entire rotor of the downstream single fan under the influence of the upstream fan wake to calculate the equivalent wind speed and equivalent wind speed attenuation on the rotor of the downstream single fan, thus obtaining the wind speed flowing into the entire rotor of the single fan. In step S111 above, the wind speed and wind speed attenuation at a certain point in space are obtained through a single Gaussian wake model, that is, the wind speed u(d,r,α) and wind speed attenuation δu(d,r,α) at ​​a certain point of the downstream single fan i. In order to calculate the wind speed flowing into the entire rotor of the fan, that is, to find the equivalent wind speed flowing into the rotor of the downstream single fan i under the influence of the upstream fan j wake, it is necessary to further integrate the wind flow flowing into the rotor of the downstream fan i according to the momentum theorem to obtain the equivalent wind speed flowing into the rotor of the downstream fan i. and equivalent wind speed attenuation

[0054] Preferably, the wind speed u(d,r,α) and wind speed attenuation δu(d,r,α) flowing into the entire rotor of the downstream single wind turbine i, which are affected by the wake of the upstream wind turbine j, are integrated to obtain the equivalent wind speed flowing into the rotor of the downstream wind turbine i. and equivalent wind speed attenuation The specific steps include:

[0055] Obtain the swept area of ​​the rotor of the downstream wind turbine i, as well as the blade length and rotation center point of the downstream wind turbine i that form the swept area;

[0056] Calculate the wake distance and downstream radial distance from the blade rotation center point of the upstream wind turbine j to the downstream wind turbine i, respectively;

[0057] Based on the wake distance of the blade rotation center point of the downstream wind turbine i, the downstream radial distance, and the axial sensing factor of the wind turbine, the wind speed attenuation of the upstream wind turbine j on the downstream wind turbine i is obtained, and the wind speed attenuation is integrated to obtain the equivalent wind speed and the equivalent wind speed attenuation.

[0058] Specifically, the equivalent wind speed attenuation is expressed as... Its calculation formula is based on the law of conservation of momentum, and calculates the equivalent wind speed attenuation at downstream wind turbine i from upstream wind turbine j. The formula is expressed as follows:

[0059]

[0060] Among them, R i A represents the rotor radius of the downstream wind turbine unit i. i Let d represent the swept area of ​​the rotor of downstream wind turbine i, q represent the radius of the swept area of ​​downstream wind turbine i (or the blade length of downstream wind turbine i), and d represent the sweeping area of ​​downstream wind turbine i. qj R represents the distance from the upstream wind turbine j to the downstream wake. qj α represents the radial distance from upstream wind turbine j to downstream wind turbine j. j The axial induction factor of the upstream wind turbine j is represented by the integral in the above formula (8) to average the wind speed attenuation of the area swept by the upstream wind turbine j on the rotor of the downstream wind turbine i.

[0061] The wind speed is equivalently represented as Calculate the equivalent wind speed of upstream wind turbine j at downstream wind turbine i. The formula is expressed as follows:

[0062]

[0063] Where U represents the free wind flowing into the wind farm. The equivalent wind speed attenuation is given. Thus, the wind speed flowing into the entire rotor of the downstream wind turbine i can be obtained.

[0064] Step S113. Using the law of conservation of kinetic energy, calculate the convergent wind speed and the decrease in convergent wind speed of all upstream wind turbines affecting the wake of a downstream wind turbine. In a wind farm, downstream wind turbine i will be affected by the wakes of multiple upstream wind turbines j. To calculate the impact of the wakes generated by all upstream wind turbines j on the inflow wind speed of a downstream wind turbine i, the law of conservation of kinetic energy is used to calculate the convergent wind speed flowing into that downstream wind turbine i. and aggregate wind speed attenuation

[0065] Specifically, the aggregate wind speed attenuation of downstream wind turbine i is expressed as... Its calculation formula is based on the law of conservation of kinetic energy, which allows for the superposition of the effects of wakes generated by multiple upstream wind turbine units j on downstream wind turbine unit i, and the aggregated wind speed attenuation. The calculation formula is expressed as follows:

[0066]

[0067] Among them, [j|W i，j=1} is the set of upstream wind turbines j that affect downstream wind turbine i. This represents the equivalent wind speed attenuation at downstream wind turbine i from upstream wind turbine j.

[0068] Based on this, the aggregate wind speed of the downstream wind turbine i is calculated. The calculation formula is as follows:

[0069]

[0070] Thus, we can analyze and obtain the impact of the wake generated by all upstream wind turbines j on the inflow wind velocity of a certain downstream wind turbine i.

[0071] Step S114. Based on the above calculation results, construct a calculation model for the power and thrust load of the single wind turbine unit considering the wake effect. Substitute the aggregate wind speed into the model to obtain the power and thrust of the single wind turbine as follows:

[0072]

[0073]

[0074] Where ρ represents the air density in the environment where the offshore wind farm is located, A represents the rotor area of ​​the downstream wind turbine i, and C P C represents the power coefficient of the wind turbine. t Represents the wind turbine thrust coefficient, cos(γ) i ) 1.5 γ represents the wake offset correction factor when the wind turbine blades are misaligned during yaw. i This represents the yaw angle of wind turbine i. The above calculation model also includes the convergent wind speed of downstream wind turbine i. It can also be understood as the average wind speed flowing into the downstream wind turbine i of the wind farm.

[0075] In the above embodiments, combined with photographic attachment Figure 3 (a) and appendix Figure 3 (b) (or appendix) Figure 4 It can be seen that the process of constructing the original wake directed graph is based on the equivalent wind speed attenuation in the wake region of the offshore wind farm. The wake coupling relationship within the wind farm is modeled as a directed graph of the original wake. The wake of the upstream wind turbine causes a loss (reduction) in the wind speed of the downstream wind turbine. Therefore, the wake coupling strength between wind turbines can be represented by the equivalent wind speed attenuation. The larger the equivalent wind speed attenuation value, the greater the loss in free wind speed U caused by the wake.

[0076] In a preferred embodiment, the step S103 above, which uses a graph weight pruning algorithm and a graph depth-first search algorithm to decompose the original wake directed graph into completely uncoupled sparse sub-wake directed graphs, is described in the appendix. Figure 3 (b) Appendix Figure 3 (c) and appendix Figure 3 (d) is explained in detail below:

[0077] Step S131. Based on the original wake directed graph, a graph weight pruning algorithm is used to obtain the pruned wake directed graph, i.e., the wake sub-directed graph.

[0078] Step S132. Use a graph depth-first search algorithm to obtain the directed graph of the wake sub-branch for each pruned branch. The set of nodes is used to obtain the directed graph of each of the said wake sub-elements. The shared nodes between node V and each of the directed graphs of the wake sub-elements;

[0079] Step S133. Calculate the total in-degree value C of the shared node in each of the directed graphs of the wake sub-graphs. in-deg The calculation formula is as follows:

[0080]

[0081] Where v represents a shared node, W l (v) represents the weight value of the in-degree, and L represents the number of in-degrees flowing into the node.

[0082] Step S134. Sort the total in-degree values ​​according to their numerical values, and retain the shared nodes in the wake sub-directed graph with the largest total in-degree value, thus decomposing the original wake directed graph into completely decoupled sparse sub-wake directed graphs.

[0083] Preferably, in step S131 above, a graph weight pruning algorithm is used to obtain a pruned directed wake graph, i.e., a wake sub-directed graph, based on the original directed wake graph. The specific implementation process includes:

[0084] Based on the original wake directed graph Determine the original wake weight matrix Based on the sparsed sub-wake directed graph Determine the sparse wake matrix Specifically, it is expressed as follows:

[0085]

[0086] Calculate the matrix sparsity coefficients of the original wake weight matrix respectively. and the matrix sparsity coefficient of the sparse wake matrix The formula for calculating the sparsity coefficient of a graph matrix is:

[0087]

[0088] Where sparseness(x) represents the sparsity coefficient of the tail weight matrix x, N represents the dimension of the tail weight matrix x, and i represents the number of matrix elements. i This represents a matrix element. The sparseness coefficient of the matrix has a range of values ​​(0≤sparseness(x)≤1). The sparser the matrix, the larger the value of the sparseness coefficient.

[0089] Based on the matrix sparsity coefficient of the original wake weight matrix and the matrix sparsity coefficient of the sparse wake matrix Calculate the graph sparsity relative coefficient ε of the sparsified sub-wake directed graph. d The calculation formula is:

[0090]

[0091] To ensure that not too much graph information is lost, ε is typically defined. d <0.1.

[0092] Based on the relative coefficient of sparsity ε in the graph d The optimal hyperparameter k is determined by the variation pattern of the hyperparameter, and the weight threshold ε on the node connection line in the directed graph of the pruned tail sub-flow is calculated based on the optimal hyperparameter. k The weight threshold ε k The calculation method is expressed as: ε k =k*ε, where k is a hyperparameter, and N is the dimension of the matrix.

[0093] Substitute the weight threshold into the original wake directed graph. Constructing a pruned wake sub-directed graph

[0094] The wake weight matrix of the pruned wake subdirected graph Through the weight threshold ε k Defined as follows:

[0095]

[0096] in, w represents the threshold activation function. ij This represents the edge weight value, and the weight threshold ε k Choose and adjust the hyperparameter k.

[0097] Based on the above implementation methods, the pruned wake directed graphs under the hyperparameter thresholds of K=0, 0.035, and 0.055 in this embodiment are shown in the attached figure. Figure 3 (c) and appendix Figure 3As shown in (d), T in the attached figure represents the node (wind turbine) number. Referring to the attached figure, it can be seen that in this embodiment, our goal is to find the set of nodes in the wake sub-directed graph using a graph depth-first search algorithm based on the wake directed graph. For example, in... Figure 3 In (c), the network topology analysis of the pruned wake directed graph (k = 0.035) shows that the node sets with dominant nodes (wind turbines) T1 and T2 are N1 = {T1 | T8, T15, T22, T29, T21, T28} and N2 = {T2 | T9, T16, T23, T30, T22, T29}, respectively. In both node sets N1 and N2, each subset contains node T22, which is defined as a shared node, and it has an in-degree value C in each subset. in-deg (T22); In subset N1, its C in-deg (T22) = 0.0093133, in subset N2, its C in-deg (T22) = 0.003249. Generally, the in-degree value C of a node... in-deg The larger (T22) is, the more important the node is.

[0098] The hyperparameter k is determined by the variation law of the relative coefficient of graph sparsity. The optimal graph weight threshold is obtained and substituted into the original directed wake graph to construct the wake sub-directed graph, thus obtaining the sparsified sub-wake directed graph. Table 1 below details how this embodiment changes the relative coefficient ε of graph sparsity by changing the hyperparameter k. d The quantitative changes in hyperparameter k were used to determine the most important topology in the wind farm wake graph. Table 1 shows the relative coefficient ε of graph sparsity as the hyperparameter k changes. d The sparsity varies in a stepwise manner. When k is in the ranges of (0, 0.035), (0.035, 0.055), and (0.055, 0.1), the relative coefficient ε of the graph sparsity increases. d The coefficients of the sparse wake weight matrix remain constant at 0.892504725, 0.892505148, and 0.915151713. This implies a sparse directed graph. Within these ranges, the network topology is stable. When determining the final sparsified sub-wake graph, we consider whether the network topology is both sparse and can express the most important topological relationships of the wake graph. Observing the sparsified sub-wake directed graphs obtained from different hyperparameter k values, we can find that when k = 0.035, the wake graph becomes sparser than when k = 0, but it is not as sparse as the sparse graph obtained when k = 0.055. Based on experience, we choose k = 0.035 as the most suitable hyperparameter k value.

[0099] Table 1: Graph sparsity index ε under varying hyperparameter K in this embodiment d Quantitative changes

[0100]

[0101] Based on the selection results in the table above, we obtain Figure 3 The results shown in (d) illustrate the network topology of the sparse wake directed graph. (See attached...) Figure 3 (d) Selecting the optimal hyperparameter k = 0.035, we finally obtain 10 sub-wind farm turbine clusters: the main turbines are T6, T25, T19, T13, T7, T5, T4, T3, T2 and T1, and the sub-wind farm turbine clusters are respectively... N3={T19∣T26}, N4={T13∣T20,T27}, N5={T7∣T14,T21,T28}, N6={T5∣T12}, N7={T4∣T11 ,T18},N8={T3∣T10,T17,T24},N9={T2∣T9,T16,T23,T30},N10={T1∣T8,T15,T22,T29}.

[0102] In a preferred embodiment, in step S103, a graph weight pruning algorithm and a graph depth-first search algorithm are used to decompose the original wake directed graph into completely uncoupled sparse sub-wake directed graphs. Based on these sparse sub-wake directed graphs, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms, constructing a sparse wind farm cluster-farm-turbine multi-layer distributed control system. According to the wake turbine power and thrust load calculation model considering the wake influence, combined with the sub-wind farms, the power and thrust calculation functions of the wind turbine clusters of multiple clustered sub-wind farms within the offshore wind farm are obtained, and the wind turbine control variable x is defined. i :=[γ i ] T To construct a distributed wind farm power and thrust load balance optimization model and solve for the optimal values ​​of the power and thrust load balance control parameters, a distributed sequential quadratic programming algorithm is used to calculate, update, and recalculate the yaw angle in the distributed wind farm power and thrust load balance optimization model until the overall power variation in the offshore wind farm reaches its minimum, thus obtaining the optimal power value.

[0103] The following is in conjunction with the appendix Figure 3-10 The steps S103 and S104 of this invention will be described in detail below. In this embodiment, based on K=0.035, the 10 sub-wind farm turbine clusters are determined, and the distributed power optimization calculation of the wind farm is performed to obtain the attached... Figure 11 The attached diagram shows the parameter mapping relationship of the sparse wind farm cluster-farm-turbine multi-layer distributed control system in this embodiment. Figure 11 The specific meanings it embodies are as follows:

[0104] F(x1) = f(x6); F(x2) = f(x6) 25 ); F(x3)=f(x 19 x 26 )

[0105] F(x4)=f(x 13 x 20 x 27 ); F(x5) = f(x7, x 14 x 21 x 28 ); F(x6) = f(x5, x 12 )

[0106] F(x7)=f(x4,x 11 x 18 ); F(x8) = f(x3, x 10 x 17 x 24 ); F(x9) = f(x2, x9, x 16 x 23 x 30 )

[0107] F(x 10 )=f(x1,x8,x 15 x 22 x 29 )

[0108] The above F(x) i ) represents the power function of the sub-wind farm, x i Let f(x) represent the control parameters of the i-th sub-wind farm, where i represents the number of sub-wind farms. i ) represents the sum of the power and thrust loss functions of all wind farm groups.

[0109] The above f(x) i The distributed wind farm power and thrust load balance optimization model is defined as follows:

[0110]

[0111]

[0112] Where i represents the i-th cluster of sub-wind farms, M represents the total number of clusters of sub-wind farms in the offshore wind farm; j represents the j-th turbine in the cluster of sub-wind farms, and K represents the total number of turbines in the cluster of sub-wind farms; γ i，j -γ represents the yaw angle of the j-th turbine in the i-th cluster of sub-wind farms. min γ represents the minimum yaw angle value. maxIndicates the maximum yaw angle; p i，j p represents the power of the j-th turbine in the i-th cluster of sub-wind farms. low p represents the minimum power value. rate f represents the rated power value. i，j f represents the thrust load of the j-th turbine in the i-th cluster of sub-wind farms. low f represents the minimum thrust load. rate ξ represents the rated thrust load, and ξ represents the balance parameter. The balance parameter 0 ≤ ξ ≤ 1 can adjust the ratio between the power and thrust loss functions. When ξ = 0 defines the single loss function of the thrust load, the power and any intermediate value of ξ = 1 define the multi-objective loss function.

[0113] The parameter mapping relationship diagram of the distributed power and thrust balance function of the electric field is obtained to obtain the shared wind turbine set between each group of the clustered sub-wind farms. According to Algorithm 1 "Solving the nonlinear wind farm power function f(x) based on the distributed sequential quadratic programming algorithm", the distributed wind farm power and thrust balance optimization model (1) is solved for the most control variables using the sequential quadratic programming method. The algorithm process is explained as follows:

[0114]

[0115] Following the calculation sequence of the above algorithm and steps, the yaw angle in the distributed wind farm power and thrust balance optimization model (1) is continuously calculated and updated. When the range of yaw angle variation is set to [-30°, 30°], the above calculation process continues until the overall power change rate of the wind farm is less than 10%. -3 .

[0116] To verify the computational optimization effect of the method of this invention, the computation time, power, and thrust optimization results of centralized and distributed algorithms considering yaw angle optimization were compared. The greedy algorithm uses parameter settings that maximize the power of a single wind turbine, with the yaw angle of all 30 wind turbines in the wind farm being 0 and the wind speed being 8 m / s. The centralized algorithm directly optimizes the parameters of the 30 wind turbines together; the distributed algorithm optimizes in parallel for the 10 sub-wind farms obtained in step S103.

[0117] Table 2 below presents the centralized and decentralized optimization strategies for power generation, thrust load, and computation time rate under different balance parameters ξ in the distributed wind farm power and thrust balance optimization model of this invention. Under a wind direction with an angle of 30° relative to the x-axis, the total power generation and thrust load of all wind turbines in the offshore wind farm cluster were optimized and measured. The results show that when the balance parameter ξ = 1, it is a power function optimization; when the balance parameter ξ = {0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1}, it is a combined power and thrust function optimization; and when the balance parameter ξ = 0, it is a thrust load function optimization. When ξ = {0, 0.1, 0.2}, the thrust load decreases significantly, accompanied by a substantial reduction in power output, lower than the power obtained by the greedy method, which is not the expected result of increased power and increased thrust load. Therefore, the balance parameter ξ = {0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1} is a range of values ​​that can be considered to balance increasing power generation and reducing thrust load, thereby using the power increment to reduce thrust load when the wind farm forgoes some power increment.

[0118] Table 2: Comparison of power, thrust, and computation time under different balance parameters ξ (relative to greedy gain)

[0119]

[0120] Table 2 shows that, within the equilibrium parameter range ξ = {0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3}, regarding power and thrust load changes, compared to the greedy method, the distributed wind farm cluster coordinated optimization increases power output and reduces thrust load. Compared to centralized control, the distributed wind farm cluster's power output and thrust load are relatively reduced, but the maximum average rate of change in power is less than 0.3%, and the average rate of change in thrust load is less than 0.6%. The computation time for distributed optimization is only 40% of that for centralized optimization. From the perspective of real-time control and reduced computation time, the results of distributed power and thrust load optimization are acceptable. These results indicate that the distributed wind farm cluster control method can reduce computation time without sacrificing too much power and can reduce thrust load.

[0121] Based on the same concept as the above embodiments of the distributed coordinated control method for offshore wind farm clusters of the present invention, the distributed coordinated control device for offshore wind farm clusters provided by the embodiments of the present invention will be described below. The distributed coordinated control device for offshore wind farm clusters described below can be referred to in correspondence with the distributed coordinated control method for offshore wind farm clusters described above. Please refer to the appendix. Figure 12This invention provides a decentralized coordinated control system for offshore wind farm clusters. The device includes: a single wind turbine model construction module 101, a wake analysis module 102, a system construction module 103, and a coordination optimization module 104.

[0122] The single-turbine model construction module 101 is used to construct a wake turbine power and thrust load calculation model for a single wind turbine unit in an offshore wind farm, considering the wake effect. The wake analysis module 102 is used to input the location, wind direction, and wind speed information of all wind turbine units in the offshore wind farm into the wake turbine power and thrust load calculation model considering the wake effect, obtain the wake coupling relationship of the offshore wind farm, and model it as an original wake directed graph. The single wind turbine units in the offshore wind farm constitute the nodes of the original wake directed graph, and the wake coupling relationship between two adjacent single wind turbine units in the offshore wind farm constitutes the edge of the original wake directed graph. The weight value of this edge is defined as the equivalent wind speed attenuation caused by the wake effect. The system construction module... Block 103 is used to decompose the original wake directed graph into completely uncoupled sparse sub-wake directed graphs using graph weight pruning algorithm and graph depth-first search algorithm. Based on the sparse sub-wake directed graphs, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm cluster-farm-turbine multi-layer decentralized control system. The coordination optimization module 104 is used to obtain the power and thrust calculation functions of the wind turbine clusters of multiple clustered sub-wind farms in the offshore wind farm according to the wake wind turbine power and thrust load calculation model considering the wake influence, combined with the sub-wind farms, define wind turbine control variables, construct a decentralized wind farm power and thrust load balance optimization model, and solve for the optimal values ​​of power and thrust load balance control parameters.

[0123] Further, the single-fan model construction module 101 includes: a first calculation unit 111, used to substitute the wake expansion coefficient, axial induction factor, and rotor radius of the single fan unit into the Gaussian wake calculation model to calculate the wind speed and wind speed attenuation at a certain point of the downstream single fan affected by the wake of the upstream fan; and a second calculation unit 112, used to integrate the wind speed and wind speed attenuation on the entire rotor of the downstream single fan affected by the wake of the upstream fan according to the momentum theorem to calculate the equivalent wind speed and equivalent wind speed on the rotor of the downstream single fan. Wind speed attenuation; the third calculation unit 113 is used to calculate the aggregate wind speed and aggregate wind speed attenuation of all upstream wind turbines flowing into the downstream wind turbine, which affect the wake of a certain downstream wind turbine, using the law of conservation of kinetic energy; the fourth calculation unit 114 is used to substitute the aggregate speed attenuation factor into the wake steady-state wind speed function to calculate the inflow equivalent wind speed of the downstream wind turbine group; the model building unit 115 is used to build a wake turbine power and thrust load calculation model of the single wind turbine group considering the wake influence based on the calculation results of the first to the fourth units.

[0124] Furthermore, the system construction module 103 includes: a first graph calculation unit 131, used to calculate a wake sub-directed graph based on the original wake directed graph using a graph weight pruning algorithm; and a second graph calculation unit 132, used to calculate a sparsified sub-wake directed graph based on the wake sub-directed graph using a graph depth-first search algorithm.

[0125] In this embodiment, the distributed coordination control device for offshore wind farm clusters is used to implement the aforementioned distributed coordination control method for offshore wind farm clusters. Therefore, the specific implementation of the distributed coordination control device for offshore wind farm clusters can be referred to the embodiment section of the distributed coordination control method for offshore wind farm clusters mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0126] The distributed coordination control method for offshore wind farm clusters of the present invention can be implemented by an electronic device equipped with a computer system. At the hardware level, the electronic device includes a processor and optionally also includes an internal bus, a network interface, and a memory.

[0127] The apparatus, module, or unit described in the above embodiments can be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0128] For ease of description, the above apparatus is described by dividing it into various units or modules according to their functions. Of course, in implementing this invention, the functions of each unit or module can be implemented in one or more software and / or hardware.

[0129] This invention is described with reference to flowchart illustrations and / or block diagrams of methods and apparatus according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0131] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0132] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the method of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A decentralized coordinated control method for an offshore wind farm cluster, characterized in that, include: Construct a calculation model for the power and thrust load of a single wind turbine in an offshore wind farm, taking into account the wake effect; The location, wind direction, and wind speed information of all wind turbines in the offshore wind farm are sequentially input into the wake turbine power and thrust load calculation model that considers the wake effect to obtain the wake coupling relationship of the offshore wind farm. This relationship is then modeled as an original wake directed graph. In this graph, a single wind turbine in the offshore wind farm constitutes a node, and the wake coupling relationship between two adjacent single wind turbines in the offshore wind farm constitutes an edge. The weight of this edge is defined as the equivalent wind speed attenuation due to the wake effect. The original wake directed graph is decomposed into a completely uncoupled sparse sub-wake directed graph using graph weight pruning algorithm and graph depth-first search algorithm. Based on the sparse sub-wake directed graph, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm group-farm-turbine multi-layer distributed control system. Based on the wake turbine power and thrust load calculation model considering the wake effect, and combined with the sub-wind farm, the power and thrust calculation functions of multiple clustered sub-wind farm turbine clusters within the offshore wind farm are obtained, and turbine control variables are defined. A distributed wind farm power and thrust load balance optimization model is constructed to solve for the optimal values ​​of the power and thrust load balance control parameters. The distributed wind farm power and thrust load balance optimization model is expressed as follows: （1） ； Where f(x) represents the sum of the power and thrust loss functions of all the sub-wind farms in the offshore wind farm; i Represented as the first i The term "clustered sub-wind farms" refers to the group of sub-wind farms, where M represents the total number of clustered sub-wind farms in the offshore wind farm. j Represented as the first in the clustered sub-wind farm j γ represents the total number of turbines in the clustered sub-wind farms; i,j Indicates the first i The first group of clustered sub-wind farms j The yaw angle of each turbine, -γ min γ represents the minimum yaw angle value. max Indicates the maximum yaw angle; p i,j Indicates the first i The first group of clustered sub-wind farms j The power of each turbine, p low p represents the minimum power value. rate Indicates the rated power value. f i,j Indicates the first i The first group of clustered sub-wind farms j The thrust load of each turbine f low Indicates the minimum thrust load. f rate ξ represents the rated thrust load, and ξ represents the balance parameter.

2. The decentralized coordinated control method for offshore wind farm clusters according to claim 1, characterized in that, A control-oriented Gaussian wake calculation model is used to construct a calculation model for the wake turbine power and thrust load of the single wind turbine unit, considering the wake effect. The model includes the following steps: Substitute the wake expansion coefficient, axial induction factor and rotor radius of the single wind turbine unit into the Gaussian wake calculation model to calculate the wind speed and wind speed attenuation at a certain point of the downstream single wind turbine affected by the wake of the upstream wind turbine. According to the momentum theorem, the wind speed and wind speed attenuation on the entire wheel of the downstream single fan affected by the wake of the upstream fan are integrated to calculate the equivalent wind speed and equivalent wind speed attenuation on the wheel of the downstream single fan, and thus obtain the wind speed flowing into the entire wheel of the single fan. Using the law of conservation of kinetic energy, calculate the aggregate wind speed and the amount of aggregate wind speed attenuation of all upstream wind turbines flowing into the downstream wind turbine that affect the wake of a certain downstream wind turbine. Based on the above calculation results, a calculation model for the power and thrust load of a single wind turbine unit considering the wake effect is constructed. Substituting the aggregated wind speed into this model, the power and thrust of a single wind turbine are obtained as follows: ， （2） ， （3） Where, p i Indicates wind turbine unit i power, f i Indicates wind turbine unit i Thrust load, ū i Indicates the flow of air into the fan unit i The convergence wind speed, ρ This indicates the air density in the offshore wind farm. A Indicates wind turbine unit i The swept area of ​​the rotor, C p C represents the power coefficient of the wind turbine. t Represents the wind turbine thrust coefficient, cos(γ) i ) 1.5 γ represents the wake offset correction factor when the wind turbine blades are misaligned during yaw. i Indicates wind turbine unit i Yaw angle.

3. The decentralized coordinated control method for offshore wind farm clusters according to claim 2, characterized in that, The step of integrating the wind speed and wind speed attenuation on the entire rotor of the downstream single wind turbine affected by the wake of the upstream wind turbine includes: Obtain the swept area of ​​the rotor of the downstream wind turbine unit, as well as the blade length and rotation center point of the downstream wind turbine unit that forms the swept area; Calculate the wake distance and downstream radial distance from the upstream wind turbine unit to the blade rotation center point of the downstream wind turbine unit, respectively. Based on the wake distance of the blade rotation center point of the downstream wind turbine, the downstream radial distance, and the axial sensing factor of the wind turbine, the wind speed attenuation of the upstream wind turbine to the downstream wind turbine is obtained, and the wind speed attenuation is integrated to obtain the equivalent wind speed and the equivalent wind speed attenuation.

4. The decentralized coordinated control method for offshore wind farm clusters according to claim 1, characterized in that, The steps of decomposing the original wake directed graph into completely decoupled sparse sub-wake directed graphs using the graph weight pruning algorithm and the graph depth-first search algorithm include: Based on the original wake directed graph, a graph weight pruning algorithm is used to obtain the wake sub-directed graph; The node set of each pruned wake subdirected graph is obtained by using a graph depth-first search algorithm, and the nodes of each wake subdirected graph and the shared nodes between each wake subdirected graph are obtained. Calculate the total in-degree value of the shared node in each of the directed graphs of the wake sub-graphs; The total in-degree values ​​are sorted according to their numerical values, and the shared nodes are retained in the wake sub-directed graph with the largest total in-degree value. The original wake directed graph is then decomposed into a completely uncoupled sparsed sub-wake directed graph.

5. The decentralized coordinated control method for offshore wind farm clusters according to claim 4, characterized in that, The steps of obtaining the wake sub-directed graph based on the original wake directed graph using the graph weight pruning algorithm include: The original wake weight matrix is ​​determined based on the original wake directed graph, and the sparse wake matrix is ​​determined based on the sparsed sub-wake directed graph. Calculate the matrix sparsity coefficient of the original wake weight matrix and the matrix sparsity coefficient of the sparse wake matrix respectively; The relative coefficient of graph sparsity of the sparsed sub-wake directed graph is calculated based on the matrix sparsity coefficient of the original wake weight matrix and the matrix sparsity coefficient of the sparse wake matrix. The hyperparameters are determined based on the variation law of the relative coefficient of sparsity in the graph, and the weight thresholds on the node connections in the directed graph of the wake are calculated. Substitute the weight threshold into the original wake directed graph to construct a wake subdirected graph.

6. The decentralized coordinated control method for offshore wind farm clusters according to claim 4, characterized in that, The steps for calculating the in-degree value of the shared node in each pruned tail sub-directed graph include: Obtain the number of in-degrees and the in-degree weight value of the shared nodes; Substituting the in-degree count and in-degree weight into the formula for calculating the node's in-degree value, the following is an example: （4） Where v represents a shared node, W1(v) represents the weight value of the in-degree of the shared node, and L represents the number of in-degrees flowing into the shared node.

7. The decentralized coordinated control method for offshore wind farm clusters according to claim 1, characterized in that, The optimal values ​​of the power and thrust load balance control parameters are obtained by using a distributed sequential quadratic programming algorithm to calculate, update, and recalculate the yaw angle in the distributed wind farm power and thrust load balance optimization model until the overall power variation in the offshore wind farm reaches its minimum.

8. A distributed coordination control device for an offshore wind farm cluster, characterized in that, include: The single wind turbine model building module is used to build a calculation model of the power and thrust load of a single wind turbine in an offshore wind farm, taking into account the wake effect. The wake analysis module is used to sequentially input the location, wind direction, and wind speed information of all wind turbines in the offshore wind farm into the wake turbine power and thrust load calculation model that considers the wake effect, to obtain the wake coupling relationship of the offshore wind farm, and to model it as an original wake directed graph. In this graph, a single wind turbine in the offshore wind farm constitutes a node, and the wake coupling relationship between two adjacent single wind turbines in the offshore wind farm constitutes an edge. The weight value of this edge is defined as the equivalent wind speed attenuation caused by the wake. The system construction module is used to decompose the original wake directed graph into a completely uncoupled sparse sub-wake directed graph using graph weight pruning algorithm and graph depth-first search algorithm. Based on the sparse sub-wake directed graph, highly correlated and strongly coupled wind turbine groups are clustered into multiple sub-wind farms to construct a sparse wind farm group-farm-turbine multi-layer distributed control system. The coordination and optimization module is used to obtain the power and thrust calculation functions of multiple clustered sub-wind farm wind turbine clusters within the offshore wind farm, based on the wake turbine power and thrust load calculation model that considers the wake effect, combined with the sub-wind farm, define wind turbine control variables, construct a distributed wind farm power and thrust load balance optimization model, and solve for the optimal values ​​of power and thrust load balance control parameters.

9. The distributed coordination control device for offshore wind farm clusters according to claim 8, characterized in that, The single wind turbine model construction module includes: The first calculation unit is used to substitute the wake expansion coefficient, axial induction factor and rotor radius of the single wind turbine into the Gaussian wake calculation model to calculate the wind speed and wind speed attenuation at a certain point of the downstream single wind turbine affected by the wake of the upstream wind turbine. The second calculation unit is used to integrate the wind speed and wind speed attenuation on the entire wheel of the downstream single fan affected by the wake of the upstream fan, according to the momentum theorem, and to calculate the equivalent wind speed and equivalent wind speed attenuation on the wheel of the downstream single fan. The third calculation unit is used to calculate the aggregate wind speed and aggregate wind speed attenuation of all upstream wind turbines flowing into the downstream wind turbine that affect the wake of a certain downstream wind turbine, using the law of conservation of kinetic energy. The fourth calculation unit is used to substitute the aggregation velocity decay factor into the wake steady-state wind speed function to calculate the inflow equivalent wind speed of the downstream wind turbine. The model building unit is used to build a calculation model of the wake turbine power and thrust load of the single wind turbine unit considering the wake effect based on the calculation results of the first to the fourth calculation units.

10. The distributed coordination control device for offshore wind farm clusters according to claim 8, characterized in that, The system construction module includes: The first graph calculation unit is used to calculate the wake sub-directed graph based on the original wake directed graph using the graph weight pruning algorithm; The second graph calculation unit is used to calculate the sparsified sub-wake directed graph based on the wake sub-directed graph using a graph depth-first search algorithm.

Images