A method and device for controlling power generation in a wind farm

By employing a distributed control method with a multi-agent system in a multi-wind turbine power plant, a wake influence matrix and directed graph between wind turbines are constructed, and the yaw angle is optimized to maximize the power of the wind farm. This solves the problem of insufficient adaptability and robustness of the single-agent system and improves the power generation efficiency and robustness of the wind farm.

CN117489528BActive Publication Date: 2026-11-10GUANGDONG POWER GRID CO LTD +1
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202311529404.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-15
Publication Date
2026-11-10
Estimated Expiration
2043-11-15

AI Technical Summary

Technical Problem

Existing wind farm control systems, due to their use of a single agent system, have poor adaptability and robustness in the event of environmental changes or failures, making it difficult to flexibly adjust control strategies.

Method used

Each wind turbine controller in a multi-turbine power plant is treated as an independent intelligent agent. By constructing the wake influence matrix and directed graph between wind turbines, the turbine depth and yaw decision sequence are determined. A distributed control method of a multi-agent system is adopted to optimize the yaw angle to maximize the power of the wind farm.

Benefits of technology

It improves the adaptability and robustness of wind farms, enhances their ability to cope with environmental changes and faults, and improves the power generation efficiency and power output of wind farms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117489528B_ABST
    Figure CN117489528B_ABST
Patent Text Reader

Abstract

This invention discloses a wind farm power generation control method and apparatus, applied to a multi-wind turbine power farm. In the multi-wind turbine power farm, the controller of each wind turbine is an independent intelligent agent, and communication connections are deployed between the intelligent agents. The method includes: constructing a wake influence matrix between each wind turbine in the target wind farm; determining the depth corresponding to each wind turbine based on the wake influence matrix; determining the yaw decision sequence corresponding to each wind turbine based on the depth; optimizing the first yaw angle corresponding to each wind turbine according to the yaw decision sequence, with the goal of maximizing the power of the target wind farm, to obtain the optimized second yaw angle corresponding to each wind turbine; and controlling each wind turbine in the target wind farm to generate electricity based on the second yaw angle, thereby improving the adaptability and robustness of wind farm power generation control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a wind farm power generation control method and device. Background Technology

[0002] In recent years, wind power generation has experienced rapid development as a clean energy source. Currently, wind farm control employs a single-agent system, where a single agent controls the yaw of the entire wind farm or a sub-farm. Even with multiple sub-farms, there is no interaction between the agents within each sub-farm; essentially, it's multiple single-agent systems. Because there's only one agent, the behavior and decision-making of a single-agent system are relatively simple and intuitive, easy to understand and master. However, its drawback lies in its lack of flexibility. When the environment changes, such as turbine failures or special scheduling needs, the agent's control strategy requires significant adjustments, limiting the adaptability and robustness of wind farm power generation control. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention discloses a wind farm power generation control method and apparatus, which improves the adaptability and robustness of wind farm power generation control.

[0004] To achieve the above objectives, in a first aspect, the present invention discloses a wind farm power generation control method applied to a multi-wind turbine power farm, wherein the controller of each wind turbine in the multi-wind turbine power farm is an independent intelligent agent, and communication connections are deployed between the intelligent agents. The method includes:

[0005] Construct the wake influence matrix between each wind turbine in the target wind farm;

[0006] The depth corresponding to each fan is determined based on the wake influence matrix between the fans.

[0007] The yaw decision sequence for each wind turbine is determined based on the depth.

[0008] With the goal of maximizing the power of the target wind farm as the optimal objective, the first yaw angle corresponding to each wind turbine is optimized sequentially according to the yaw decision order to obtain the second yaw angle corresponding to each wind turbine after optimization.

[0009] Based on the second yaw angle corresponding to each wind turbine, control each wind turbine in the target wind farm to generate electricity.

[0010] This invention discloses a wind farm power generation control method applicable to multi-wind turbine power farms. In this method, the controllers of each wind turbine are considered independent intelligent agents, and communication connections are deployed between these agents to form a multi-agent distributed system. Combining the distributed and decentralized characteristics of this multi-agent system, it can better adapt to environmental changes and fault occurrences, improving the robustness of wind farm control. Furthermore, to improve decision-making efficiency and avoid interference from redundant information, the method obtains the wake influence relationship matrix between each wind turbine in the target wind farm and determines the depth of each turbine based on this matrix. This determines whether information interaction is needed between the intelligent agents corresponding to each turbine. For two turbines without wake influence relationships, their intelligent agents do not need to interact. To improve decision-making efficiency and avoid interference from redundant information, the yaw decision-making order of each wind turbine is determined based on its depth. Wind turbines without influence are filtered out by calculating their depth, and the influence relationships between them are determined. The decision-making order is then determined based on these relationships. This reduces coupling between multiple agents, optimizes the decision-making order, and improves efficiency. With maximizing the power of the target wind farm as the optimization objective, the yaw angle of each wind turbine is optimized sequentially according to the yaw decision-making order to obtain the optimized yaw angle. The wind turbines are then controlled to generate electricity based on the optimized yaw angle. This solves the problem of poor control adaptability and robustness in existing single-agent control systems for wind farms, improving the adaptability and robustness of wind farm power generation control.

[0011] As a preferred example, the wake influence matrix between each wind turbine in the target wind farm includes:

[0012] Calculate several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wake of other wind turbines, according to the preset wind speed calculation formula;

[0013] Based on the aforementioned wind speed variation values, construct the wake influence matrix between each wind turbine in the target wind farm.

[0014] This invention characterizes the degree of influence between wind turbines by calculating the wind speed change value, so as to adapt to environmental changes and failures by calculating the degree of influence, thereby determining the decision sequence for each wind turbine in the later stage and improving the robustness of wind power control in wind farms.

[0015] As a preferred example, determining the depth corresponding to each wind turbine based on the wake influence matrix between wind turbines includes:

[0016] Based on the wake influence matrix between the wind turbines, a directed graph corresponding to each wind turbine is generated, and the target nodes in the directed graph are extracted using a preset extraction strategy; wherein, each node in the directed graph corresponds one-to-one with each wind turbine;

[0017] A depth-optimized search is performed starting from the target node, and the depths of all nodes in the directed graph except the target node are updated through the depth-optimized search.

[0018] This invention generates a directed graph corresponding to the matrix, wherein each node in the directed graph corresponds to a wind turbine. By performing a depth optimization search on the directed graph, the depth corresponding to each wind turbine can be obtained, that is, the degree of influence of each wind turbine and its corresponding adjacent wind turbines, so as to determine the yaw decision order based on the degree of influence, thereby improving the efficiency of decision-making.

[0019] As a preferred example, the extraction of target nodes in the directed graph using a preset extraction strategy includes:

[0020] By initializing all nodes in the directed graph, the depth of each node is controlled to be the initial depth value;

[0021] Traverse all nodes in the directed graph, extract all nodes with an in-degree of 0, and obtain the target node in the directed graph.

[0022] This invention improves the calculation efficiency and accuracy of the depth corresponding to each node and each wind turbine by extracting all nodes with an in-degree of 0 to find the upstream wind turbines, and then traversing downwards according to the upstream wind turbines.

[0023] As a preferred example, the process of performing a depth-optimized search starting from the target node, and updating the depths of all nodes in the directed graph except the target node through the depth-optimized search, includes:

[0024] Obtain each directed edge in the directed graph that points from the target node to any other node, and update the depth of the other nodes in the directed edge to the larger of a first depth and a second depth; wherein the first depth is the depth of the other nodes, and the second depth is the depth of the target node plus 1.

[0025] This invention determines the degree of influence of the target node (i.e., the target wind turbine) on the other nodes in the directed edges and influence chain corresponding to the target node (i.e., the target wind turbine). This allows for timely adjustment of the wind turbine power generation decision sequence based on the degree of influence, adapting to changes in the current environment or other factors, and improving the robustness and adaptability of wind farm power generation control.

[0026] As a preferred example, the optimization of the first yaw angle corresponding to each wind turbine according to the yaw decision order includes:

[0027] The wind farm information of all wind turbines in the target wind farm is obtained, the wind farm information is input into a preset decision model, the wind farm yaw strategy corresponding to the target wind farm is obtained, and then the first yaw angle corresponding to each wind turbine is determined according to the wind farm yaw strategy.

[0028] With the goal of maximizing the power of the target wind farm as the optimal objective, the first yaw angle corresponding to each wind turbine is optimized by a preset greedy algorithm. The current optimization result of each wind turbine is used as feedback to iterate the yaw angle optimization process until the current optimization result reaches a preset threshold and the iteration ends.

[0029] This invention takes into account that when the yaw angle of the downstream wind turbine changes, the optimal yaw angle of the upstream wind turbine will also shift. When determining the decision-making order based on the depth of the wind turbine, the shift can be minimized as much as possible to improve the optimization effect. At the same time, an iterative mechanism is designed in the optimization process, that is, the yaw angle is adjusted multiple times according to the above process to take into account the influence of the yaw angle of the downstream wind turbine on the optimal yaw angle of the upstream wind turbine, thereby improving the final optimization effect.

[0030] Secondly, the present invention discloses a wind farm power generation control device, which is applied to a multi-wind turbine power plant. The controller of each wind turbine in the multi-wind turbine power plant is an independent intelligent agent, and communication connections are deployed between the intelligent agents. The device includes a matrix construction module, a deep search module, a yaw decision module, a yaw optimization module, and an optimization control module.

[0031] The matrix construction module is used to construct the wake influence matrix between each wind turbine in the target wind farm.

[0032] The depth search module is used to determine the depth corresponding to each wind turbine based on the wake influence matrix between wind turbines.

[0033] The yaw decision module is used to determine the yaw decision sequence for each wind turbine based on the depth.

[0034] The yaw optimization module is used to optimize the first yaw angle corresponding to each wind turbine according to the yaw decision order, with the goal of maximizing the power of the target wind farm, to obtain the second yaw angle corresponding to each wind turbine after optimization.

[0035] The optimization control module is used to control each wind turbine in the target wind farm to generate electricity according to the second yaw angle corresponding to each wind turbine.

[0036] This invention discloses a wind farm power generation control device applicable to multi-wind turbine power farms. In this device, the controllers of each wind turbine are considered independent intelligent agents, and communication connections are deployed between these agents to form a multi-agent distributed system. Combining the distributed and decentralized characteristics of this multi-agent system, it can better adapt to environmental changes and fault occurrences, improving the robustness of wind farm control. Furthermore, to improve decision-making efficiency and avoid interference from redundant information, the device obtains the wake influence relationship matrix between each wind turbine in the target wind farm and determines the depth of each turbine based on this matrix. This determines whether information interaction is needed between the intelligent agents corresponding to each turbine. For two turbines without wake influence relationships, their intelligent agents do not need to interact. To improve decision-making efficiency and avoid interference from redundant information, the yaw decision-making order of each wind turbine is determined based on its depth. Wind turbines without influence are filtered out by calculating their depth, and the influence relationships between them are determined. The decision-making order is then determined based on these relationships. This reduces coupling between multiple agents, optimizes the decision-making order, and improves efficiency. With maximizing the power of the target wind farm as the optimization objective, the yaw angle of each wind turbine is optimized sequentially according to the yaw decision-making order to obtain the optimized yaw angle. The wind turbines are then controlled to generate electricity based on the optimized yaw angle. This solves the problem of poor control adaptability and robustness in existing single-agent control systems for wind farms, improving the adaptability and robustness of wind farm power generation control.

[0037] As a preferred example, the matrix construction module includes a wind speed calculation unit and a construction unit;

[0038] The wind speed calculation unit is used to calculate, according to a preset wind speed calculation formula, several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wake of other wind turbines.

[0039] The construction unit is used to construct the wake influence matrix between each wind turbine in the target wind farm based on the several wind speed change values.

[0040] This invention characterizes the degree of influence between wind turbines by calculating the wind speed change value, so as to adapt to environmental changes and failures by calculating the degree of influence, thereby determining the decision sequence for each wind turbine in the later stage and improving the robustness of wind power control in wind farms.

[0041] Thirdly, the present invention discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus;

[0042] The processor, communication interface, and memory communicate with each other through a communication bus.

[0043] Memory, used to store computer programs;

[0044] The processor, when executing a program stored in memory, implements a wind farm power generation control method as described in the first aspect.

[0045] Fourthly, the present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a wind farm power generation control method as described in the first aspect. Attached Figure Description

[0046] Figure 1 : A schematic flowchart of a wind farm power generation control method provided in an embodiment of the present invention;

[0047] Figure 2 : A schematic diagram of the structure of a wind farm power generation control device provided in an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the structural composition of a multi-wind turbine power plant provided in an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of a wind turbine yaw angle optimization process provided in an embodiment of the present invention.

[0050] Figure 5 : A depth diagram provided for an embodiment of the present invention;

[0051] Figure 6 : A schematic diagram of yaw optimization results provided in an embodiment of the present invention;

[0052] Figure 7 : A schematic diagram of an electronic device structure provided in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example

[0055] This embodiment discloses a wind farm power generation control method applied to a multi-wind turbine power farm. In this multi-wind turbine power farm, the controller for each wind turbine is an independent intelligent agent, and communication connections are deployed between the intelligent agents. For the specific implementation process of the control method, please refer to [reference needed]. Figure 1 It mainly includes steps 101 to 105, the steps being:

[0056] Step 101: Construct the wake influence matrix between each wind turbine in the target wind farm.

[0057] In this embodiment, the step mainly includes: calculating several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wake of other wind turbines according to a preset wind speed calculation formula; and constructing the wake influence matrix between each wind turbine in the target wind farm according to the several wind speed change values.

[0058] Specifically, in one embodiment provided in this example, the target wind farm can be an entire wind farm or a sub-wind farm after partitioning. An entire wind farm refers to a pristine, naturally formed wind farm that has not been partitioned. As the scale of the wind farm increases, the number of wind turbines in the target wind farm increases, leading to higher computational costs for simulation and more complex wake effects. Considering this, we can first analyze the wake effects among the wind turbines in the target wind farm, and then partition the multiple wind turbines in the target wind farm according to the analyzed wake effects. This breaks down the overall control optimization problem of the target wind farm into several sub-control optimization problems, thereby reducing the problem size and improving control efficiency.

[0059] Furthermore, the wind farm power generation control method can be applied to multi-wind turbine power plants. Firstly, the controller of each wind turbine in a multi-wind turbine power plant can be considered an independent intelligent agent. Communication connections are deployed between these intelligent agents to form a multi-agent system. The structural composition of the multi-wind turbine power plant can be referenced... Figure 3 ,like Figure 3 As shown, in a multi-agent system for wind farm optimization problems, the information interaction between agents is affected by the influence of wind turbine wake. For example... Figure 1 As shown in the diagram, for wind turbines 1 and 2, which have no wake influence relationship, their agents do not need to interact, thereby improving decision-making efficiency and avoiding unnecessary information from interfering with the decision-making process. Then, when generating the wake influence relationship matrix between each wind turbine in the target wind farm, several wind speed change values ​​of each wind turbine in the target wind farm affected by the wake of other wind turbines are first calculated, where the wind speed change value represents the degree of wake influence between two wind turbines; based on the wind speed change value, the wake influence relationship matrix between each wind turbine in the target wind farm is determined.

[0060] When calculating the wind speed change, the wind speeds of the wind turbines in the wind farm are coupled due to the wake effect. If the magnitude Δu of the wind speed decrease of turbine j due to the wake effect of turbine i can be obtained... ij Therefore, the magnitude of this value can reflect the degree of wake influence between the two wind turbines. Based on the change in wind speed, the wake influence matrix between the wind turbines can be calculated. This wake influence matrix can be represented as follows:

[0061]

[0062] Where, n r This refers to the number of wind turbines.

[0063] It should be noted that, since the wind speed attenuation effect of the wake can only be transferred unilaterally from the upstream wind turbine to the downstream wind turbine, when Δu ij When Δu > 0, ji =0 is always true.

[0064] Meanwhile, the calculation of the wind speed change value of each wind turbine in the target wind farm affected by the wake of other wind turbines can be achieved by creating a black observation point on the surface of each wind turbine, and the state inherited by the observation point is the state before the wind turbine yaws.

[0065] When the wind propagates from an upstream observation point to a downstream wind turbine, it affects the wind speed at the downstream wind turbine's observation point. The effective wind speed at an observation point is calculated as follows:

[0066]

[0067] Among them, u free,OP z represents the free wind speed at the observation point. OP The height of the observation point is indicated by z0, which represents the upper limit of the wind farm's height. OP / z0<1,(z OP / z0) α This indicates that the effective wind speed at the observation point decreases as the height of the observation point decreases. own n represents the wake attenuation factor of the observation point itself. T Indicates the number of wind turbines, r i It represents the wind speed attenuation factor of the observation point that is closest to the i-th wind turbine among all observation points belonging to the i-th wind turbine (i.e., the observation point where the effective wind speed is being calculated).

[0068] After obtaining the effective wind speed at the corresponding observation point of the fan, the effective wind speed at the fan is calculated by weighted summation of the effective wind speeds at the observation points on the fan surface. The specific calculation formula is as follows:

[0069]

[0070] Where, n OP w represents the number of observation points on the surface of the fan. i Indicates the weight.

[0071] Once the effective wind speed of the fan is obtained, the power of the fan can be calculated, as follows:

[0072]

[0073] Among them, C P η is the power coefficient, η is the efficiency, and p is a coefficient representing the influence of yaw angle on the power of the wind turbine.

[0074] According to equations (2) and (3) above, we can obtain:

[0075]

[0076] According to the above formula (5), the effective wind speed of the wind turbine is determined by all its upstream wind turbines.

[0077] Based on the magnitude Δu of the wind speed decrease of fan j due to the wake effect of fan i, ij Definition of Δu ij and u eff,T,i The following equation should be satisfied:

[0078]

[0079] Among them, u free Let $\frac{ ... ij .

[0080] Therefore, Δu can be considered. ij The estimated value is given by the following formula:

[0081]

[0082] Based on equation (7) above, the wind speed change Δu of each wind turbine in the target wind farm due to the wake of other wind turbines can be obtained. ij The estimated value.

[0083] In this embodiment, the wind speed change value of each wind turbine in the target wind farm affected by the wake of other wind turbines is calculated, and the wake influence relationship matrix between each wind turbine in the target wind farm is determined based on the wind speed change value.

[0084] In this embodiment, this step constructs a multi-agent system corresponding to the target wind farm. Each wind turbine is controlled by a separate intelligent agent. Each intelligent agent makes a decision on its own yaw angle based on information such as ambient wind speed, the location of other wind turbines, the yaw angle of other wind turbines, and the power of other wind turbines. With such a multi-agent system, when wind turbine failures or special scheduling needs occur, only some intelligent agents and the interaction information between intelligent agents need to be modified to complete the decision-making in the new environment, thereby improving the adaptability and robustness of wind farm power generation control.

[0085] Step 102: Determine the depth corresponding to each fan according to the wake influence matrix between the fans.

[0086] In this embodiment, the step mainly includes: generating a directed graph corresponding to each wind turbine based on the wake influence matrix between wind turbines, and extracting the target node in the directed graph through a preset extraction strategy; wherein, each node in the directed graph corresponds one-to-one with each wind turbine; performing a depth optimization search with the target node as the initial point, and updating the depth of the other nodes in the directed graph except for the target node through the depth optimization search.

[0087] Furthermore, the step of extracting the target node in the directed graph using a preset extraction strategy includes: initializing all nodes in the directed graph and controlling the depth of each node to be the initial depth value; traversing all nodes in the directed graph and extracting all nodes with an in-degree of 0 to obtain the target node in the directed graph.

[0088] After obtaining the target node, obtain each directed edge in the directed graph that points from the target node to any other node, and update the depth of the other nodes in the directed edge to the larger of the first depth and the second depth; wherein, the first depth is the depth of the other nodes, and the second depth is the depth of the target node plus 1.

[0089] Specifically, in one embodiment provided in this example, a directed graph D representing the influence relationship of the wind turbine wake is obtained. R = (V, E), where V is the set of nodes and E is the set of directed edges. Based on obtaining the wake influence relationship matrix between wind turbines, the transpose matrix R of the wake influence relationship matrix between wind turbines is... T As the adjacency matrix, the directed graph D of the wind farm is obtained. R Then initialize the D R The depth of all nodes in the array is 0, that is:

[0090]

[0091] Find D RFor all nodes with an in-degree of 0, update their depth to 1, i.e.:

[0092]

[0093] Here, in(v) represents the in-degree of node v.

[0094] Starting with each target node, a depth-optimized search is performed to update the depths of other nodes. The update rule is that for a directed edge from node u to node v, the depth of node v is updated to the larger of the depth of node v and the depth of node u plus 1. The expression is:

[0095]

[0096] In this embodiment, this step filters out wind turbines that do not have an impact by calculating the wind turbine depth, reduces the coupling between multiple agents, optimizes the decision-making order, and improves decision-making efficiency.

[0097] Step 103: Determine the yaw decision sequence for each wind turbine based on the depth.

[0098] In this embodiment, the main step is to make decisions on the yaw angle of each wind turbine in order of increasing depth. Since the wakes of wind turbines at the same depth do not affect each other, their optimization order can be arbitrarily determined.

[0099] Specifically, in this embodiment, a larger depth difference between two target wind turbines indicates a greater wake influence between them, while a depth of 0 indicates that the wakes of the two turbines do not affect each other. Thus, by sorting the depth differences from largest to smallest, the earlier the difference appears, the higher the priority for yaw decision-making. For example, if the depth difference between wind turbine A and wind turbine B is 1, the depth difference between wind turbine A and wind turbine C is 0.5, and the depth difference between wind turbine A and wind turbine D is 0, it means that the wake influence between wind turbine A and B is the greatest, followed by wind turbine A and C, and the wakes between wind turbine A and D do not affect each other. Therefore, when determining the yaw decision-making order, the yaw decision between wind turbine A and B has the highest priority, followed by the yaw decision between wind turbine A and C, and the yaw decision between wind turbine A and D does not need to be calculated.

[0100] In this embodiment, this step determines the yaw strategy order of the wind turbines based on the depth relationship between the two target wind turbines. That is, the greater the depth difference between the two target wind turbines, the greater the wake influence between them; conversely, when the depth difference is 0, the wakes of the two wind turbines do not affect each other. Thus, by sorting the turbines from largest to smallest depth difference, the earlier the difference is in the order of yaw decisions, thereby improving the robustness and adaptability of power generation control.

[0101] Step 104: Taking the maximization of the power of the target wind farm as the optimal objective, optimize the first yaw angle corresponding to each wind turbine in sequence according to the yaw decision order to obtain the second yaw angle corresponding to each wind turbine after optimization.

[0102] In this embodiment, the steps mainly include: acquiring wind farm information of all wind turbines in the target wind farm, inputting the wind farm information into a preset decision model to obtain the wind farm yaw strategy corresponding to the target wind farm, and then determining the first yaw angle corresponding to each wind turbine according to the wind farm yaw strategy; taking the maximization of the power of the target wind farm as the optimal objective, optimizing the first yaw angle corresponding to each wind turbine respectively through a preset greedy algorithm, and using the current optimization result of each wind turbine as feedback to iterate the yaw angle optimization process until the current optimization result reaches a preset threshold and the iteration ends.

[0103] Specifically, in one embodiment provided in this example, when optimizing the yaw angle of a wind turbine, a yaw strategy for the wind farm can be generated through reinforcement learning and other means to weaken the impact of the wake effect and improve the overall power generation capacity of the wind farm. Preferably, the wake optimization problem of the wind farm can be modeled, and a yaw strategy can be generated through the TD3 algorithm. The yaw angle can be optimized according to the yaw strategy. The process of generating the yaw strategy can be as follows: for each target wind farm, obtain the wind farm information of all wind turbines in the target wind farm. The wind farm information includes historical environmental wind speed, historical wind turbine yaw angle and current environmental wind speed. The wind farm information is then input into a pre-trained decision model to obtain the first wind turbine yaw angle corresponding to each target wind farm.

[0104] Then, after confirming the first yaw angle corresponding to each wind turbine, the yaw angles of other wind turbines are kept unchanged, and the yaw angles of wind turbines that have not yet been decided are initialized to 0°. With maximizing the power of the wind farm as the overall optimization objective, the optimization objective of each wind turbine agent is designed to maximize the sum of the power of that wind turbine and the wind turbines affected by its wake, that is:

[0105]

[0106] Reference Figure 4 If the overall optimization objective differs, requiring different optimization objectives for each wind turbine agent, the corresponding control algorithm is essentially a greedy control algorithm. This involves decomposing the yaw optimization problem of the entire wind farm into several individual wind turbine yaw optimization sub-problems. In each sub-problem, only turbines with wake effects are considered, significantly reducing the difficulty of solving the sub-problems. Finally, the solutions to the sub-problems are merged into the solution for the entire wind farm. Once the overall yaw optimization objective for the entire wind farm is determined, how to design the yaw optimization objective for each individual wind turbine will affect the final optimization effect of the algorithm.

[0107] During the optimization process, the current optimization result of each wind turbine is used as feedback to iterate the yaw angle optimization process until the current optimization result reaches a preset threshold and the iteration ends.

[0108] In this embodiment, this step takes into account that when the yaw angle of the downstream wind turbine changes, the optimal yaw angle of the upstream wind turbine will also shift. Determining the decision-making order based on the depth of the wind turbine is to minimize this shift, ensuring that the solution to the subproblem is as close as possible to the solution to the original problem. Based on this algorithm, an iterative mechanism can be further designed, that is, adjusting the yaw angle multiple times according to the above process to consider the influence of the downstream wind turbine's yaw angle on the optimal yaw angle of the upstream wind turbine, thereby improving the final optimization effect.

[0109] Step 105: Based on the second yaw angle corresponding to each wind turbine, control each wind turbine in the target wind farm to generate electricity.

[0110] In this embodiment, the main steps are as follows: after obtaining the yaw angles of each wind turbine after optimization, control each wind turbine in the target wind farm to generate electricity based on the yaw angle of each wind turbine.

[0111] Referring to the wind farm power generation control method provided in this embodiment, taking a 28-wind-turbine wind farm as an example, the wind turbine parameters corresponding to each wind turbine in the 28-wind-turbine wind farm are first calculated. The wind turbine parameters include depth and yaw angle. Based on the FLORIDyn model, the wake influence relationship between wind turbines is analyzed and the wind turbine power is calculated. Keeping the ambient wind speed and wind direction constant, the optimization objective is to maximize the overall power of the wind farm under the ambient wind conditions. When each wind turbine agent makes a decision, the optimal yaw angle is determined by the ambient wind conditions, the yaw angles of all wind turbines with depths less than that wind turbine (or the yaw angles of the wind turbines that have a wake influence on that wind turbine), and the power of all downstream wind turbines of that wind turbine. No iteration is performed.

[0112] First, calculate the depth of each node, and then calculate D. R Ignoring edges with weights less than 0.05, the depths of each wind turbine node are calculated as shown in Table 1. Figure 5 As shown, Table 1 is as follows:

[0113]

[0114]

[0115] Table 1

[0116] Based on the depth, the yaw angle that maximizes power generation is calculated sequentially using a search algorithm. u The yaw optimization results are shown in Table 2 and 3. Figure 6 As shown, Table 2 is as follows:

[0117] Fan serial number Yaw angle / ° Fan serial number Yaw angle / ° 1 3.47 15 0 2 6.43 16 5.62 3 -1.67 17 -1.67 4 0 18 -14.99 5 -8.55 19 0 6 -11.68 20 31.3 7 -8.55 21 0 8 21.74 22 0 9 -1.98 23 0 10 4.27 24 0 11 -0.54 25 0 12 0 26 0 13 -7.60 27 0 14 0 28 0

[0118] Table 2

[0119] Of these, the power output of wind turbine 28 using a direct-wind strategy was 46219.822 kW. After yaw control using the wind farm power generation control method provided in this embodiment, the power output was 47044.185 kW, representing an increase of 1.78%. This was achieved under 100 different environmental wind conditions (wind speed range [4, 12] m / s). -1 The wind direction range is [-180, 180]°. The above method was used to optimize the yaw of the 28 wind turbine wind farm, and its average power was increased by 1.27%, indicating that the above method can improve the output power of the wind farm.

[0120] On the other hand, the present invention also discloses a wind farm power generation control device, applied to a multi-wind turbine power plant. In this multi-wind turbine power plant, the controller for each wind turbine is an independent intelligent agent, and communication connections are deployed between the intelligent agents. The specific structural composition of the device can be found in [reference needed]. Figure 2 It mainly includes a matrix construction module 201, a depth search module 202, a yaw decision module 203, a yaw optimization module 204, and an optimization control module 205.

[0121] The matrix construction module 201 is used to construct the wake influence matrix between each wind turbine in the target wind farm.

[0122] The depth search module 202 is used to determine the depth corresponding to each wind turbine according to the wake influence matrix between wind turbines.

[0123] The yaw decision module 203 is used to determine the yaw decision sequence corresponding to each wind turbine based on the depth.

[0124] The yaw optimization module 204 is used to optimize the first yaw angle corresponding to each wind turbine according to the yaw decision order, with the goal of maximizing the power of the target wind farm, to obtain the second yaw angle corresponding to each wind turbine after optimization.

[0125] The optimization control module 205 is used to control each wind turbine in the target wind farm to generate electricity according to the second yaw angle corresponding to each wind turbine.

[0126] In this embodiment, the matrix construction module 201 includes a wind speed calculation unit and a construction unit.

[0127] The wind speed calculation unit is used to calculate, according to a preset wind speed calculation formula, several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wake of other wind turbines.

[0128] The construction unit is used to construct the wake influence matrix between each wind turbine in the target wind farm based on the several wind speed change values.

[0129] In this embodiment, the depth search module 202 further includes a map layout unit and a search unit.

[0130] The graph layout unit is used to generate a directed graph corresponding to each wind turbine based on the wake influence matrix between wind turbines, and to extract the target nodes in the directed graph through a preset extraction strategy; wherein, each node in the directed graph corresponds one-to-one with each wind turbine.

[0131] The search unit is used to perform a depth-optimized search with the target node as the initial point, and to update the depth of the other nodes in the directed graph except for the target node through the depth-optimized search.

[0132] In this embodiment, the graph layout unit further includes: initializing all nodes in the directed graph, controlling the depth of each node to be the initial depth value; traversing all nodes in the directed graph, extracting all nodes with an in-degree of 0, and obtaining the target node in the directed graph.

[0133] The search unit further includes: obtaining each directed edge in the directed graph from the target node to any other node, and updating the depth of the other nodes in the directed edge to the larger of a first depth and a second depth; wherein the first depth is the depth of the other nodes, and the second depth is the depth of the target node plus 1.

[0134] In this embodiment, the yaw optimization module 204 includes a decision-making unit and an optimization unit.

[0135] The decision unit is used to obtain wind farm information of all wind turbines in the target wind farm, input the wind farm information into a preset decision model, obtain the wind farm yaw strategy corresponding to the target wind farm, and then determine the first yaw angle corresponding to each wind turbine according to the wind farm yaw strategy.

[0136] The optimization unit is used to optimize the first yaw angle corresponding to each wind turbine by using a preset greedy algorithm with the goal of maximizing the power of the target wind farm. The current optimization result of each wind turbine is used as feedback to iterate the yaw angle optimization process until the current optimization result reaches a preset threshold and the iteration ends.

[0137] In addition to the methods and apparatus described above, this embodiment also provides an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communication bus 740. The processor 710, communications interface 720, and memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions from the memory 730 to execute a wind farm power generation control method provided in this embodiment. This method is applied to multi-turbine wind farms, where the controller of each wind turbine is considered an independent intelligent agent. Communication connections are deployed between these intelligent agents to form a multi-agent system.

[0138] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0139] On the other hand, this embodiment also provides a non-transitory computer-readable storage medium and a computer program product, on which a computer program is stored, and when the computer program is executed by a processor, it implements the wind farm power generation control method described in this embodiment.

[0140] This embodiment provides a wind farm power generation control method, device, electronic equipment, and storage medium. First, addressing the shortcomings of single-agent systems, a multi-agent system-based multi-wind turbine wind farm control framework is proposed. Second, based on this framework, a heuristic control algorithm is proposed. This algorithm, based on a greedy algorithm, determines the decision-making order among independent wind turbine agents according to the wake influence relationship between wind turbines to improve control optimization. Finally, based on the control optimization effect of a 28-wind-turbine wind farm under different wind conditions, a numerical example demonstrates that the control algorithm can effectively improve the output power of the wind farm.

[0141] In the above specific implementation, the wind farm power generation control method provided in this embodiment is applied to a multi-turbine wind farm. The controller of each wind turbine in the multi-turbine wind farm is regarded as an independent intelligent agent, and communication connections are deployed between the intelligent agents to form a multi-agent system. The method obtains the wake influence relationship matrix between wind turbines in the target wind farm, determines the depth of each wind turbine according to the wake influence relationship matrix, determines the yaw decision order of each wind turbine based on the depth of each wind turbine, optimizes the yaw angle of each wind turbine in sequence according to the yaw decision order, so as to obtain the optimization result of each wind turbine, and superimposes the optimization results of each wind turbine with the power maximization of the target wind farm as the optimization objective to obtain the optimization result of the wind farm.

[0142] Thus, combining the distributed and decentralized characteristics of multi-agent systems, they can better adapt to environmental changes and failures, exhibiting stronger robustness. In multi-agent systems for wind farm optimization problems, information interaction between agents is influenced by the wake effect of wind turbines. For two turbines without wake effects, their agents can avoid information interaction, thereby improving decision-making efficiency and preventing redundant information from interfering with the decision-making process. This solves the problem of poor control adaptability and robustness in existing single-agent control systems for wind farms, improving the adaptability and robustness of wind farm power generation control. Furthermore, by calculating the turbine depth, it filters out turbines without influence, reducing coupling between agents, optimizing the decision-making sequence, and improving decision-making efficiency.

[0143] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A wind farm power generation control method, characterized in that, Applied to multi-wind turbine power plants, where the controller of each wind turbine in the multi-wind turbine power plant is an independent intelligent agent, and communication connections are deployed between the intelligent agents, the method includes: Construct the wake influence matrix between each wind turbine in the target wind farm; wherein, according to the preset wind speed calculation formula, calculate several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wake of other wind turbines; and construct the wake influence matrix between each wind turbine in the target wind farm based on the several wind speed change values. The depth corresponding to each fan is determined based on the wake influence matrix between the fans. The yaw decision sequence for each wind turbine is determined based on the depth. With maximizing the power of the target wind farm as the optimal objective, the first yaw angle corresponding to each wind turbine is optimized sequentially according to the yaw decision order to obtain the second yaw angle corresponding to each wind turbine after optimization. Specifically, wind farm information of all wind turbines in the target wind farm is obtained, and the wind farm information is input into a preset decision model to obtain the wind farm yaw strategy corresponding to the target wind farm. Then, the first yaw angle corresponding to each wind turbine is determined according to the wind farm yaw strategy. With maximizing the power of the target wind farm as the optimal objective, the first yaw angle corresponding to each wind turbine is optimized separately using a preset greedy algorithm. The current optimization result of each wind turbine is used as feedback to iterate the yaw angle optimization process until the current optimization result reaches a preset threshold, at which point the iteration ends. Based on the second yaw angle corresponding to each wind turbine, control each wind turbine in the target wind farm to generate electricity.

2. The wind farm power generation control method as described in claim 1, characterized in that, The step of determining the depth corresponding to each wind turbine based on the wake influence matrix between wind turbines includes: Based on the wake influence matrix between the wind turbines, a directed graph corresponding to each wind turbine is generated, and the target nodes in the directed graph are extracted using a preset extraction strategy; wherein, each node in the directed graph corresponds one-to-one with each wind turbine; A depth-optimized search is performed starting from the target node, and the depths of all nodes in the directed graph except the target node are updated through the depth-optimized search.

3. The wind farm power generation control method as described in claim 2, characterized in that, The step of extracting target nodes from the directed graph using a preset extraction strategy includes: By initializing all nodes in the directed graph, the depth of each node is controlled to be the initial depth value; Traverse all nodes in the directed graph, extract all nodes with an in-degree of 0, and obtain the target node in the directed graph.

4. The wind farm power generation control method as described in claim 2, characterized in that, The step of performing a depth-optimized search starting from the target node, and updating the depths of all nodes in the directed graph except the target node through the depth-optimized search, includes: Obtain each directed edge in the directed graph that points from the target node to any other node, and update the depth of the other nodes in the directed edge to the larger of a first depth and a second depth; wherein the first depth is the depth of the other nodes, and the second depth is the depth of the target node plus 1.

5. A wind farm power generation control device, characterized in that, Applied to multi-wind turbine power plants, where the controller of each wind turbine in the multi-wind turbine power plant is an independent intelligent agent, and communication connections are deployed between intelligent agents, the device includes a matrix construction module, a deep search module, a yaw decision module, a yaw optimization module, and an optimization control module; The matrix construction module is used to construct the wake influence matrix between each wind turbine in the target wind farm. The matrix construction module includes a wind speed calculation unit and a construction unit. The wind speed calculation unit calculates several wind speed change values ​​corresponding to each wind turbine in the target wind farm after being affected by the wakes of other wind turbines, according to a preset wind speed calculation formula. The construction unit constructs the wake influence matrix between each wind turbine in the target wind farm based on the several wind speed change values. The depth search module is used to determine the depth corresponding to each wind turbine based on the wake influence matrix between wind turbines. The yaw decision module is used to determine the yaw decision sequence for each wind turbine based on the depth. The yaw optimization module is used to optimize the first yaw angle corresponding to each wind turbine according to the yaw decision order, with the goal of maximizing the power of the target wind farm, to obtain the second yaw angle corresponding to each wind turbine after optimization. Specifically, it acquires wind farm information of all wind turbines in the target wind farm, inputs the wind farm information into a preset decision model to obtain the wind farm yaw strategy corresponding to the target wind farm, and then determines the first yaw angle corresponding to each wind turbine according to the wind farm yaw strategy. With the goal of maximizing the power of the target wind farm, it optimizes the first yaw angle corresponding to each wind turbine using a preset greedy algorithm, and iterates the yaw angle optimization process using the current optimization result of each wind turbine as feedback until the current optimization result reaches a preset threshold, at which point the iteration ends. The optimization control module is used to control each wind turbine in the target wind farm to generate electricity according to the second yaw angle corresponding to each wind turbine.

6. An electronic device, characterized in that, Includes processor, communication interface, memory, and communication bus; The processor, communication interface, and memory communicate with each other through a communication bus. Memory, used to store computer programs; A processor, when executing a program stored in a memory, implements a wind farm power generation control method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a wind farm power generation control method according to any one of claims 1-4.

Citation Information

Patent Citations

  • Wind power plant grouping optimization control method, system, device and medium

    CN114841090A

  • Distributed coordination control method and device for offshore wind power plant group

    CN115663879A