Wind power plant rapid frequency response method based on distributed control strategy

Abstract

The invention discloses a wind power plant rapid frequency response method based on a distributed control strategy. Relates to the technical field of wind power plant frequency optimization, in particular to the technical field of cloth control wind power plant frequency response optimization. The method effectively solves the problems that the wind power plant is low in frequency response speed, poor in robustness and high in control complexity, and improves the frequency response optimization efficiency of the wind power plant. The method comprises the following steps: establishing a wind power plant partition model, and dividing a plurality of wind turbines into independent control units; determining an optimal classification threshold value, and confirming optimal fuzzy classification; performing problem modeling; decomposing problems; and performing iterative optimization. The method is remarkably different from the prior art in the aspect of improving the frequency response speed and flexibility of the wind power plant, and the problems of response delay, high control complexity, insufficient system stability and the like in an existing method are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind farm frequency optimization, and in particular to the technical field of wind farm frequency response optimization under distributed control. Background Art

[0002] As the global demand for renewable energy continues to grow, wind power, as an important component of renewable energy, continues to expand in scope and scale. However, the randomness and intermittent characteristics of wind power pose a huge challenge to the frequency stability of the power system. Traditional power systems mainly rely on traditional power generation methods such as thermal power and hydropower to regulate frequency, but these methods often perform poorly when faced with large-scale wind power access, and the speed and stability of frequency response are difficult to meet the requirements.

[0003] When wind farms are connected to the grid, the volatility of grid frequency increases. Traditional centralized frequency regulation methods usually rely on a single central controller for global control. This approach faces the following major challenges:

[0004] Slow response: Due to communication delays and complex control structures, centralized control does not respond quickly enough in large-scale wind farms to quickly respond to frequency changes.

[0005] Poor system robustness: The central controller is the core of the system. If it fails, the frequency response of the entire system will be seriously affected.

[0006] High control complexity: As the scale of wind farms increases, the implementation of centralized control strategies becomes more complex and unstable. Summary of the invention

[0007] In response to the above problems, the present invention discloses a wind farm fast frequency response method based on a distributed control strategy, which relates to the technical field of wind farm frequency optimization, effectively solves the problems of slow frequency response speed, poor robustness and high control complexity of wind farms, and improves the efficiency of wind farm frequency response optimization.

[0008] The method comprises the following steps:

[0009] S1. Establish a wind farm zoning model to divide multiple wind turbines into independent control units;

[0010] S2, determine the best classification threshold and confirm the best fuzzy classification;

[0011] S3, problem modeling;

[0012] S4, problem decomposition;

[0013] S5. Iterative optimization.

[0014] Furthermore, the establishment of the wind farm zoning model is specifically: using a fuzzy clustering method to partition the wind farm and divide multiple wind turbines into independent control units; the fuzzy clustering analysis method is specifically: constructing a feature matrix based on the electrical distance between nodes, constructing a fuzzy similarity matrix based on the feature matrix, and then performing fuzzy clustering based on the fuzzy similarity matrix.

[0015] Furthermore, the electrical distance between the nodes is calculated by their physical positions and electrical parameters; the characteristic matrix is ​​specifically: X=(x ij ) n×m , the x ij represents the electrical distance between node i and node j, x represents the electrical distance, n represents the number of features, and m represents the total number of nodes.

[0016] Furthermore, the fuzzy similarity matrix is ​​specifically: The c represents the optimal number of clusters, the k represents an index variable, and the k=[1,2,3,…,m], and the x ik represents the value of node i at the kth eigenvalue, the x jk Represents the value of node j at the kth eigenvalue; the fuzzy clustering is specifically: the fuzzy equivalent matrix is ​​obtained by the successive square method of the fuzzy similarity matrix.

[0017] Furthermore, the optimal classification threshold is determine; described represents the average value of the i-th type node on the k-th eigenvalue, the c i represents the i-th category, Shown represents the number of nodes in the i-th category, and j represents the j-th node in the i-th category; when and When they are equal, the current fuzzy classification is the best fuzzy classification. Represents the average value of the kth feature in all nodes.

[0018] Furthermore, the problem modeling is specifically as follows: the global frequency response problem is modeled as a global optimization problem, and the global optimization problem is specifically as follows: The f i () represents the local objective function of the i-th wind turbine, and the w i represents the frequency response parameter of the i-th fan, the subject to represents the linear constraint condition, the A represents an m×n dimensional matrix, the w represents an n×1 column vector, and the b represents the coupling constraint of the system.

[0019] Furthermore, the problem decomposition is specifically as follows: decomposing the global optimization problem into sub-problems by means of the alternating direction multiplier method; the sub-problems are specifically as follows: The lambda T represents the transpose of the Lagrange multiplier, λ represents the Lagrange multiplier, ρ represents the coefficient of the penalty term, and ||()|| 2 Represents the square of the norm.

[0020] Furthermore, the iterative optimization is specifically as follows: each wind turbine independently optimizes the sub-problem and updates w i , according to w i Update the global variable z, and then update the Lagrange multiplier λ according to the global variable z; the update w i Specifically: Said Indicates that w is the k+1th iteration i The updated value of represents f i (w i ) minimizes w i The value of z k represents the global variable of the kth iteration; i Update the global variable z as follows: The z k+1 represents the updated value of z at the k+1th iteration; the Lagrange multiplier updated according to the global variable z is specifically: k+1 =λ k +ρ(w k+1 -z k+1 ), the λ k+1 represents the Lagrange multiplier of the k+1th iteration, the λ k represents the Lagrange multiplier of the kth iteration; when When the optimization iteration ends, the ||Aw k -b|| 2 represents the original residual, the w k represents the variable vector of the kth iteration, Indicates the preset accuracy threshold.

[0021] The beneficial effects of the present invention are:

[0022] The wind farm rapid frequency response method based on the distributed control strategy proposed in the present invention is not only innovative in terms of control strategy, but also focuses on optimizing the design for the wind farm to rapidly respond to grid frequency fluctuations. The distributed control strategy allows the wind turbines in the wind farm to respond autonomously according to local conditions, while ensuring the consistency and efficiency of the frequency response of the entire farm through information exchange and collaborative control between modules. This method not only improves the response speed, but also retains control over the stability of the system, avoiding the single point failure and communication delay problems that may exist in existing centralized control. The present invention is significantly different from the prior art in terms of improving the frequency response speed and flexibility of wind farms, and solves the problems of response hysteresis, high control complexity, and insufficient system stability in the existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of wind farm zoning according to an embodiment of the present invention;

[0024] Figure 2 Iterative optimization flow chart for an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of dynamic clustering according to an embodiment of the present invention;

[0026] Figure 4 Schematic diagram of coordination optimization of independent control units in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] This embodiment provides a wind farm fast frequency response method based on a distributed control strategy, the method comprising the following steps:

[0029] S1. Establish a wind farm zoning model to divide multiple wind turbines into independent control units.

[0030] The relevant operations of step S1 are introduced with a specific example:

[0031] In large-scale wind farms, distributed control strategies can be used to coordinate the frequency response of each wind turbine to achieve overall optimization of the system. Figure 1As shown in the figure, first, the wind farm is divided into multiple independent control units through clustering algorithm, and each unit can optimize the frequency response separately. Considering that the hard division method may not be suitable for objects with fuzzy or changing features, the fuzzy clustering analysis method can divide the control area more reasonably, thereby improving the response efficiency and accuracy of the system.

[0032] The fuzzy clustering analysis method specifically comprises: constructing a feature matrix according to the electrical distance between nodes, constructing a fuzzy similarity matrix according to the feature matrix, and then performing fuzzy clustering according to the fuzzy similarity matrix.

[0033] Each wind farm has n features, and each wind farm can be represented by a vector containing m nodes. This process is called calibration. The calibration is to quantify the features of the wind farm to be classified into specific data. The electrical distance between the nodes is calculated by their physical position and electrical parameters. The feature matrix is ​​specifically: X = (x ij ) n×m , the x ij represents the electrical distance between node i and node h, and x represents the electrical distance.

[0034] The i and j represent node indexes, which are characterized by the electrical distance between each node and other nodes. The characteristic matrix of the wind farm to be classified is composed of the electrical distances between nodes. To optimize its classification process, it is necessary to fully consider these electrical distance characteristics to ensure accurate identification and classification when analyzing the relationship between nodes, thereby improving the overall performance and efficiency of the system. In order to measure the similarity between wind farms to be classified, the absolute value subtraction method is used to construct a fuzzy similarity matrix to analyze the similarity between nodes.

[0035] The fuzzy similarity matrix is ​​specifically: The c represents the optimal number of clusters, so that r ij The value of is scattered between [0, 1], k represents the index variable, k = [1, 2, 3, ..., m], x ik represents the value of node i at the kth eigenvalue, the x jk Represents the value of node j at the kth eigenvalue; the fuzzy clustering is specifically: the fuzzy similarity matrix is ​​calculated through the successive square method to obtain its transitive closure matrix, and the transitive closure matrix is ​​a fuzzy equivalence matrix.

[0036] Fuzzy cluster analysis is a classification method based on fuzzy mathematics. It classifies objective things by analyzing their different characteristics, closeness, similarity and other relationships. It can effectively deal with classification problems with fuzziness and uncertainty.

[0037] S2. Determine the optimal classification threshold and confirm the optimal fuzzy classification.

[0038] The relevant operations of step S2 are introduced with a specific example:

[0039] The optimal classification threshold is determine; described represents the average value of the i-th type node on the k-th eigenvalue, the c i represents the i-th category, Shown represents the number of nodes in the i-th category, and j represents the j-th node in the i-th category; Figure 3 As shown, the fuzzy equivalence matrix is ​​truncated according to an appropriate threshold value, and the appropriate threshold value is adjusted by dynamically adjusting and observing the fuzzy clustering results, optimizing the compactness within the fuzzy clusters and the separation between the fuzzy clusters, and calculating the classification error and combining the application scenario with the experimental results. ij When the value of decreases from 1 to 0, the fuzzy classification gradually merges from fine to coarse, forming a dynamic clustering graph, which is the dynamic clustering result. and When they are equal, the current fuzzy classification is the best fuzzy classification. represents the average value of the kth feature in all nodes.

[0040] S3. Problem modeling.

[0041] The relevant operations of step S3 are introduced with a specific example:

[0042] Wind farms involve a variety of complex constraints, including: wind turbine power output limits, grid frequency security constraints, and wind turbine response speed and capacity limits. Figure 2 As shown in the figure, the distributed optimization algorithm ADMM decomposes the constraints and maps the global constraints to each local sub-problem, so that the wind turbine can be optimized independently while satisfying its own constraints, and can flexibly deal with various complex constraints in the system. The global frequency response problem is modeled as a global optimization problem with the goal of adjusting the frequency response of each wind turbine to minimize the frequency deviation of the entire power system while satisfying the constraints of the wind farm and wind turbines. The global optimization problem is specifically: The f i () represents the local objective function of the i-th wind turbine, and the w i represents the frequency response parameter of the i-th fan, the subject to represents the linear constraint condition, the A represents an m×n dimensional matrix, the w represents an n×1 column vector, and the b represents the coupling constraint of the system.

[0043] This distributed architecture greatly improves the scalability of computing and is suitable for large wind farms and systems consisting of multiple wind farms. When the number of wind farms increases, the algorithm remains efficient and does not significantly reduce performance due to the growth of the system scale. The communication network is used to support distributed optimization, and the distributed optimization algorithm coordinates and optimizes through data exchange between the dispatch center and each wind turbine.

[0044] S4. Problem decomposition.

[0045] The relevant operations of step S4 are introduced with a specific example:

[0046] The distributed optimization algorithm can decompose the global optimization problem into sub-problems of multiple wind turbines, and each wind turbine can process the sub-problems in parallel.

[0047] The problem decomposition is specifically as follows: decomposing the global optimization problem into sub-problems by means of the alternating direction multiplier method; the sub-problems are specifically as follows: The lambda T represents the transpose of the Lagrange multiplier, λ represents the Lagrange multiplier, ρ represents the coefficient of the penalty term, and ∥()∥ 2 Represents the square of the norm.

[0048] S5. Iterative optimization.

[0049] The relevant operations of step S5 are introduced with a specific example:

[0050] The iterative optimization is as follows: each wind turbine independently optimizes the sub-problem and updates w i , according to w i Update the global variable z, and then update the Lagrange multiplier λ according to the global variable z; the update w i Specifically: Said Indicates that w is the k+1th iteration i The updated value of represents f i (w i ) minimizes w i The value of z k represents the global variable of the kth iteration; i Update the global variable z as follows: The z k+1 represents the updated value of z at the k+1th iteration; the Lagrange multiplier updated according to the global variable z is specifically: k+1 =λ k +ρ(w k+1 -z k+1 ), the λ k+1 represents the Lagrange multiplier of the k+1th iteration, the λ krepresents the Lagrange multiplier of the kth iteration; when When the optimization iteration ends, the ||Aw k -b|| 2 represents the original residual, the w k represents the variable vector of the kth iteration, Indicates the preset accuracy threshold.

[0051] like Figure 4 As shown in the figure, each wind turbine can independently adjust its output power based on the local grid frequency and real-time operating status. This design enables the wind turbine to respond quickly to frequency fluctuations, and because each wind turbine is a relatively independent control unit, it no longer relies on a single central controller, which greatly reduces communication delays and centralized control bottlenecks. This autonomous response design not only improves the response speed, but also enhances the system's fault tolerance. Even if some wind turbines fail, other wind turbines can continue to work and maintain the stability of the overall frequency response.

[0052] Although each wind turbine can adjust itself, in order to ensure that the frequency response of the entire wind farm is coordinated, each wind turbine needs to exchange information and coordinate control, and a communication network between wind turbines is required to share key information such as real-time grid frequency and local operating status. This information sharing enables each wind turbine to not only optimize itself, but also coordinate optimization based on the global grid status to ensure the consistency of the overall frequency response.

Claims

1. A wind farm fast frequency response method based on distributed control strategy, characterized in that: The method comprises the following steps: S1. Establish a wind farm zoning model to divide multiple wind turbines into independent control units; S2, determine the best classification threshold and confirm the best fuzzy classification; S3, problem modeling; S4, problem decomposition; S5. Iterative optimization.

2. A wind farm fast frequency response method based on a distributed control strategy according to claim 1, characterized in that: The specific steps of establishing a wind farm zoning model are as follows: using a fuzzy clustering method to partition the wind farm and dividing multiple wind turbines into independent control units; the specific steps of the fuzzy clustering analysis method are as follows: constructing a feature matrix based on the electrical distance between nodes, constructing a fuzzy similarity matrix based on the feature matrix, and then performing fuzzy clustering based on the fuzzy similarity matrix.

3. A wind farm fast frequency response method based on distributed control strategy according to claim 2, characterized in that: The electrical distance between the nodes is calculated by their physical position and electrical parameters; the characteristic matrix is ​​specifically: X = (x ij ) n×m , the x ij represents the electrical distance between node i and node j, x represents the electrical distance, n represents the number of features, and m represents the total number of nodes.

4. A wind farm fast frequency response method based on a distributed control strategy according to claim 3, characterized in that: The fuzzy similarity matrix is ​​specifically: The c represents the optimal number of clusters, the k represents an index variable, and the k=[1,2,3,…,m], and the x ik represents the value of node i at the kth eigenvalue, the x jk Represents the value of node j at the kth eigenvalue; the fuzzy clustering is specifically: the fuzzy equivalent matrix is ​​obtained by the successive square method of the fuzzy similarity matrix.

5. A wind farm fast frequency response method based on a distributed control strategy according to claim 4, characterized in that: The optimal classification threshold is determine; described represents the average value of the i-th type node on the k-th eigenvalue, the c i represents the i-th category, Shown represents the number of nodes in the i-th category, and j represents the j-th node in the i-th category; when and When they are equal, the current fuzzy classification is the best fuzzy classification. Represents the average value of the kth feature in all nodes.

6. The method for rapid frequency response of a wind farm based on a distributed control strategy according to claim 1, characterized in that: The problem modeling is specifically: modeling the global frequency response problem as a global optimization problem, and the global optimization problem is specifically: The f i () represents the local objective function of the i-th wind turbine, and the w i represents the frequency response parameter of the i-th fan, the subjectto represents the linear constraint, the A represents an m×n dimensional matrix, the w represents an n×1 column vector, and the b represents the coupling constraint of the system.

7. A wind farm fast frequency response method based on distributed control strategy according to claim 6, characterized in that: The problem decomposition is specifically as follows: decomposing the global optimization problem into sub-problems by means of the alternating direction multiplier method; the sub-problems are specifically as follows: The lambda T represents the transpose of the Lagrange multiplier, λ represents the Lagrange multiplier, ρ represents the coefficient of the penalty term, and ||()|| 2 Represents the square of the norm.

8. A wind farm fast frequency response method based on a distributed control strategy according to claim 7, characterized in that: The iterative optimization is as follows: each wind turbine independently optimizes the sub-problem and updates w i , according to w i Update the global variable z, and then update the Lagrange multiplier λ according to the global variable z; the update w i Specifically: Said Indicates that w is the k+1th iteration i The updated value of represents f i (w i ) minimizes w i The value of z k represents the global variable of the kth iteration; i Update the global variable z as follows: The z k+1 represents the updated value of z at the k+1th iteration; the Lagrange multiplier updated according to the global variable z is specifically: k+1 =λ k +ρ(w k+1 -z k+1 ), the λ k+1 represents the Lagrange multiplier of the k+1th iteration, the λ k represents the Lagrange multiplier of the kth iteration; when When the optimization iteration ends, the ||Aw k -b|| 2 represents the original residual, the w k represents the variable vector of the kth iteration, Indicates the preset accuracy threshold.

Images