CN117967499A - Wind power plant grouping wake optimization method and system - Google Patents
Wind power plant grouping wake optimization method and system Download PDFInfo
- Publication number
- CN117967499A CN117967499A CN202410391204.6A CN202410391204A CN117967499A CN 117967499 A CN117967499 A CN 117967499A CN 202410391204 A CN202410391204 A CN 202410391204A CN 117967499 A CN117967499 A CN 117967499A
- Authority
- CN
- China
- Prior art keywords
- fan
- wake
- constructing
- network
- yaw angle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 85
- 238000005457 optimization Methods 0.000 title claims abstract description 48
- 230000008569 process Effects 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 30
- 230000008878 coupling Effects 0.000 claims abstract description 26
- 238000010168 coupling process Methods 0.000 claims abstract description 26
- 238000005859 coupling reaction Methods 0.000 claims abstract description 26
- 238000011217 control strategy Methods 0.000 claims abstract description 24
- 238000010248 power generation Methods 0.000 claims abstract description 19
- 230000000694 effects Effects 0.000 claims abstract description 17
- 230000007246 mechanism Effects 0.000 claims abstract description 12
- 230000009471 action Effects 0.000 claims description 47
- 238000011144 upstream manufacturing Methods 0.000 claims description 45
- 230000006870 function Effects 0.000 claims description 37
- 239000003795 chemical substances by application Substances 0.000 claims description 33
- 238000004364 calculation method Methods 0.000 claims description 16
- 230000007613 environmental effect Effects 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 12
- 238000013461 design Methods 0.000 claims description 11
- 230000003993 interaction Effects 0.000 claims description 11
- 238000013139 quantization Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 5
- 230000007704 transition Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000011002 quantification Methods 0.000 claims 1
- 238000005452 bending Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 230000006698 induction Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000006735 deficit Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/72—Wind turbines with rotation axis in wind direction
Landscapes
- Wind Motors (AREA)
Abstract
The invention relates to the technical field of wind power generation, and particularly provides a wind farm grouping wake flow optimization method and system, comprising the following steps: dividing the whole wind farm into a plurality of subsystems according to the pneumatic coupling level between fans; constructing wake optimization as a Markov decision process, and constructing a depth deterministic strategy gradient network for each subsystem respectively; performing offline pre-training on the depth deterministic strategy gradient network to obtain an optimal fan yaw angle control strategy of each subsystem; and on-line adjustment of fan yaw angles of the corresponding subsystems is performed based on an optimal fan yaw angle control strategy. The method does not depend on a mechanism model of the wind power plant, and can avoid the problems of poor optimization effect and the like caused by low mechanism model precision, modeling errors and the like.
Description
Technical Field
The invention belongs to the technical field of wind power generation, and particularly relates to a wind farm grouping wake flow optimization method and system.
Background
The wind power generation field (wind power field for short) is characterized in that tens or even hundreds of wind turbines are intensively arranged at a field with abundant wind resources, are arrayed according to the topography and the main wind direction, and supply power to a power grid in a cluster mode. Most wind farms use a "greedy strategy" when generating electricity. This means that each fan operates individually to maximize its own power generation, ignoring potential impact on other fans. However, due to the wake effect in the wind power station, the front exhaust fan positioned in the incoming wind direction can form a wake area with reduced wind speed and increased turbulence at the downstream of the front exhaust fan when capturing wind energy, so that the power generation power of the downstream fan positioned in the wake area is reduced (more than 30 percent), and meanwhile, the fatigue load is obviously increased, and the service life of the unit is shortened. Thus, a greedy strategy is often not an optimal strategy.
Currently, mainstream wind farm wake optimization control methods can be divided into two categories: axial induction control and yaw control. Axial induction control refers to changing the axial induction set point of the fans to affect wake development, typically by reducing the energy capture of the upstream fans to mitigate wake effects in an effort to increase the energy capture of the fans in the downstream wake region. The method has obvious effect when adopting a simplified model to simulate. However, recent full-size field tests and wind tunnel tests show that the method has very limited effects on improving the power generation capacity of the wind power plant and reducing the fatigue load of the fan. The current advanced yaw control implementation scheme is mostly converted into a solution optimization problem. The first step is to characterize phenomena such as wake loss, wake deflection and the like of a single fan based on engineering analysis wake models (such as a two-dimensional Jensen model, a Mosaic model, a Frandsen model and the like). And secondly, characterizing wake coupling among fans by using a wake superposition model, establishing a whole wind power plant simulation model considering the influence of wake effect, and quantifying the relation between the states of all fans (for example, yaw angles of all fans) and the power generation amount of the whole wind power plant. And thirdly, designing different optimization algorithms to determine yaw angle input of each fan based on the model constructed in the second step by using simulation data or wind farm actual measurement data, namely, adjusting yaw angles of the fans in the wind farm to redirect wake flows and prevent downstream fans from being in an upstream fan wake flow area.
The disadvantages of the existing scheme are: 1. the existing optimization-based method needs to construct a wind power plant simulation model considering wake effects to calculate the power output of the whole wind power plant under different yawing, the model has limited fidelity, modeling errors cannot be avoided, and the optimization effect is unstable when the model is applied under actual environmental conditions. 2. The existing scheme mostly regards a wind farm as a whole to construct an optimization problem, but as the number of fans in the wind farm increases, control variables suddenly increase, and the method faces the dilemma of dimension disaster, so that calculation tasks are difficult to complete in one control period. 3. Most of the prior schemes focus on the maximization of the power output of the whole wind power plant, but neglect to reduce the fatigue load of the fan, and the lack of load reduction and synergy is unavoidable in the actual wind power plant operation.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a wind farm grouping wake optimizing method and system, which aim to solve the technical problems.
In a first aspect, the invention provides a wind farm clustered wake optimization method, comprising:
Dividing the whole wind farm into a plurality of subsystems according to the pneumatic coupling level between fans;
constructing wake optimization as a Markov decision process, and constructing a depth deterministic strategy gradient network for each subsystem respectively;
performing offline pre-training on the depth deterministic strategy gradient network to obtain an optimal fan yaw angle control strategy of each subsystem;
And on-line adjustment of fan yaw angles of the corresponding subsystems is performed based on an optimal fan yaw angle control strategy.
In an alternative embodiment, dividing the entire wind farm into a plurality of subsystems according to the level of aerodynamic coupling between fans, comprises:
Constructing a fan wake flow influence model;
Constructing a fan fatigue load quantization model;
constructing an influence factor of an upstream fan on a downstream fan based on a fan wake influence model;
converting the influence factors into influence weights of the upstream fans on the downstream fans;
Traversing all fan combinations to obtain influence weights among fans, and generating an undirected graph representing the pneumatic coupling degree among fans based on the influence weights.
In an alternative embodiment, constructing a fan wake impact model includes:
Constructing a downstream wind speed value calculation model: ；
，/>，/>；
Wherein, The wind speed value at the downstream distance x and the radial distance r of the fan is represented, v is the wind speed of the inflow wind,/>Representing the width of the wake plane,/>Represents the diameter of a wind wheel of a fan,/>Represents the fan thrust coefficient,/>Represents the yaw angle of the fan,/>Is constant, the value of which depends on the roughness of the ground;
Distance of deviation The calculation formula of (2) is as follows: /(I)；
Wherein,，/>For the initial wake angle,/>Is an uncertain model parameter that depends on the atmosphere.
In an alternative embodiment, constructing a fan fatigue load quantization model includes:
；
Wherein, Fatigue for wind turbine power generation,/>Is fatigue caused by fluctuation of the ambient wind speed,/>And k w is a weight factor that is a weight factor,For/>Generating power of fan at moment,/>For rated power generation of fan,/>For the design life of the fan, r is the action coefficient of repairing and maintaining the fan, and is/(For standard turbulence used in fan design,/>For/>The fan is subjected to environmental turbulence at any time.
In an alternative embodiment, constructing an influence factor of an upstream fan to a downstream fan based on a fan wake influence model includes:
Setting downstream fan The coordinates at the hub are/>If it is located at the upstream fan/>Within the wake region of (2) to be satisfied/>；
Construct upstream fanFor downstream blower/>Influence factor/>The calculation formula is as follows:
；
Wherein, And/>For the upstream blower/>Upper and lower limits of yaw angle set point,/>Is the weight factor of different yaw angles,/>Indicating yaw angle as/>Downstream blower/>Wind speed value at the hub.
In an alternative embodiment, converting the impact factor into an impact weight of the upstream fan on the downstream fan includes:
Upstream fan For point i, downstream blower/>For point j, upstream blower/>For downstream blower/>Influence factor/>Is the weight between point i and point j;
If the influence factor is ，/>If the pneumatic coupling cut-off value is a preset pneumatic coupling cut-off value, the point i and the point j are considered to be not communicated any more, and the weight is set to be 0;
traversing all wind turbine generators in sequence, obtaining fan numbers and influence weights of the fan numbers in downstream wake areas, and constructing an undirected graph representing pneumatic coupling degree among the fans;
Traversing the undirected graph by breadth-first search, and solving the connected components to realize wind farm group decoupling.
In an alternative embodiment, constructing the wake optimization as a markov decision process and constructing a depth deterministic strategy gradient network for each subsystem separately includes:
Constructing wake optimization control problems as a data-driven Markov decision process, transitioning states of a system through tuples Description of the preferred embodiments wherein/>The system state variables at the time t comprise yaw angles of all fans, power generated by all fans and fatigue degree/>, of all fansEtc./>The control action variables taken for time t, including the yaw angle of each fan,Signifying taking the control action/>After that, the updated state of the system,/>Rewards for evaluating the state transition quality at this time;
Aiming at each subsystem formed by decoupling in the first part of wind power plant grouping decoupling, respectively constructing a depth deterministic strategy gradient network to make a decision of a Markov process, wherein the depth deterministic strategy gradient network consists of an evaluation network and a strategy network, and each network comprises a current network and a target network;
agent in depth deterministic policy gradient network by sensing current system state Policy-based implementation of actionsThereby changing the environmental state and obtaining rewards/>The goal of the agent is to maximize the expected rewards, which represent all rewards that can be brought after performing an action, including instant rewards and future rewards at the current time;
Wherein rewards are awarded Comprises:
；
wherein the reward function First part/>Representing the difference value of the wind power plant after power and fatigue normalization, wherein the larger the rewards obtained by the intelligent agent is, the more obvious the load reduction and synergy effects are represented; reward function/>Second part/>Representing a penalty to dangerous actions of the agent; c p and C f are normalized weighting factors,/>For the power generated by the blower WT i at time t,/>For fatigue load of the blower WT j at time t,/>Penalty coefficient for yaw out-of-limit,/>And/>The upper limit and the lower limit of the yaw angle of the fan WT i are respectively set;
Constructing a state-action cost function ，/>Representing in stateIf according to policy/>Is required to perform the action/>Expected rewards that can be brought,/>Is a rewards discount factor, which means that the farther from the current moment, the lower the weight of rewards, E () represents mathematical expectations;
enabling an agent to learn an optimal strategy in interaction with an environment Maximizing the expected rewards; the optimal strategy is implemented by maximizing a state-action cost function, expressed as:。
In an alternative embodiment, training the depth deterministic strategy gradient network to obtain an optimal fan yaw control strategy for each subsystem includes:
initializing various parameters in a depth deterministic strategy gradient network;
In the training process, interacting with a simulated wind farm environment FAST.Farm, and storing tuples generated by interaction into an experience playback pool;
based on the tuple in the experience playback pool, performing distributed priority playback, performing iterative updating on the evaluation network parameters of the depth deterministic strategy gradient network by adopting a minimized loss function method, and performing iterative updating on the strategy network parameters of the depth deterministic strategy gradient network by adopting a gradient direction moving method of enabling the action strategy to the state-action cost function until the set iterative times are reached.
In an alternative embodiment, adjusting the fan yaw angle of the respective subsystem based on the optimal fan yaw angle control strategy includes:
The actual yaw angle of the fan is different from the optimal yaw angle of the fan, and a yaw value of the fan is obtained;
Setting a yaw action threshold value: ；
Representing the actual yaw angle of the current wind turbine i, when the optimum wind turbine yaw angle/> And/>When the difference is 0.0174 rad, the yaw mechanism can adjust the yaw angle to be/>。
In a second aspect, the present invention provides a wind farm clustered wake optimization system comprising:
the grouping decoupling module is used for dividing the whole wind power plant into a plurality of subsystems according to the pneumatic coupling level between the fans;
The network construction module is used for constructing wake optimization into a Markov decision process and constructing a depth deterministic strategy gradient network for each subsystem respectively;
the network training module is used for training the depth deterministic strategy gradient network and acquiring the optimal fan yaw angle control strategy of each subsystem;
And the yaw adjustment module is used for adjusting the fan yaw angles of the corresponding subsystems based on the optimal fan yaw angle control strategy.
In an alternative embodiment, dividing the entire wind farm into a plurality of subsystems according to the level of aerodynamic coupling between fans, comprises:
Constructing a fan wake flow influence model;
Constructing a fan fatigue load quantization model;
constructing an influence factor of an upstream fan on a downstream fan based on a fan wake influence model;
converting the influence factors into influence weights of the upstream fans on the downstream fans;
Traversing all fan combinations to obtain influence weights among fans, and generating an undirected graph representing the pneumatic coupling degree among fans based on the influence weights.
In an alternative embodiment, constructing a fan wake impact model includes:
Constructing a downstream wind speed value calculation model: ；
，/>，/>；
Wherein, The wind speed value at the downstream distance x and the radial distance r of the fan is represented, v is the wind speed of the inflow wind,/>Representing the width of the wake plane,/>Represents the diameter of a wind wheel of a fan,/>Represents the fan thrust coefficient,/>Represents the yaw angle of the fan,/>Is constant, the value of which depends on the roughness of the ground;
Distance of deviation The calculation formula of (2) is as follows: /(I)；
Wherein,，/>For the initial wake angle,/>Is an uncertain model parameter that depends on the atmosphere.
In an alternative embodiment, constructing a fan fatigue load quantization model includes:
；
Wherein, Fatigue for wind turbine power generation,/>Is fatigue caused by fluctuation of the ambient wind speed,/>And k w is a weight factor that is a weight factor,For/>Generating power of fan at moment,/>For rated power generation of fan,/>For the design life of the fan, r is the action coefficient of repairing and maintaining the fan, and is/(For standard turbulence used in fan design,/>For/>The fan is subjected to environmental turbulence at any time.
In an alternative embodiment, constructing an influence factor of an upstream fan to a downstream fan based on a fan wake influence model includes:
Setting downstream fan The coordinates at the hub are/>If it is located at the upstream fan/>Within the wake region of (2) to be satisfied/>；
Construct upstream fanFor downstream blower/>Influence factor/>The calculation formula is as follows:
；
Wherein, And/>For the upstream blower/>Upper and lower limits of yaw angle set point,/>Is the weight factor of different yaw angles,/>Indicating yaw angle as/>Downstream blower/>Wind speed value at the hub.
In an alternative embodiment, converting the impact factor into an impact weight of the upstream fan on the downstream fan includes:
Upstream fan For point i, downstream blower/>For point j, upstream blower/>For downstream blower/>Influence factor/>Is the weight between point i and point j;
If the influence factor is ，/>If the pneumatic coupling cut-off value is preset, the point i and the point j are considered to be no longer communicated, and the weight is set to 0.
In an alternative embodiment, constructing the wake optimization as a markov decision process and constructing a depth deterministic strategy gradient network for each subsystem separately includes:
Constructing wake optimization control problems as a data-driven Markov decision process, transitioning states of a system through tuples Description of the preferred embodiments wherein/>The system state variables at the time t comprise yaw angles of all fans, power generated by all fans and fatigue degree/>, of all fansEtc./>The control action variables taken for time t, including the yaw angle of each fan,Signifying taking the control action/>After that, the updated state of the system,/>Rewards for evaluating the state transition quality at this time;
Aiming at each subsystem formed by decoupling in the first part of wind power plant grouping decoupling, respectively constructing a depth deterministic strategy gradient network to make a decision of a Markov process, wherein the depth deterministic strategy gradient network consists of an evaluation network and a strategy network, and each network comprises a current network and a target network;
agent in depth deterministic policy gradient network by sensing current system state Policy-based implementation of actionsThereby changing the environmental state and obtaining rewards/>The goal of the agent is to maximize the expected rewards, which represent all rewards that can be brought after performing an action, including instant rewards and future rewards at the current time;
Wherein rewards are awarded Comprises:
；
wherein the reward function First part/>Representing the difference value of the wind power plant after power and fatigue normalization, wherein the larger the rewards obtained by the intelligent agent is, the more obvious the load reduction and synergy effects are represented; reward function/>Second part/>Representing a penalty to dangerous actions of the agent; c p and C f are normalized weighting factors,/>For the power generated by the blower WT i at time t,/>For fatigue load of the blower WT j at time t,/>Penalty coefficient for yaw out-of-limit,/>And/>The upper limit and the lower limit of the yaw angle of the fan WT i are respectively set;
Constructing a state-action cost function ，/>Representing in stateIf according to policy/>Is required to perform the action/>Expected rewards that can be brought,/>Is a rewards discount factor, which means that the farther from the current moment, the lower the weight of rewards, E () represents mathematical expectations;
enabling an agent to learn an optimal strategy in interaction with an environment Maximizing the expected rewards; the optimal strategy is implemented by maximizing a state-action cost function, expressed as:。
In an alternative embodiment, training the depth deterministic strategy gradient network to obtain an optimal fan yaw control strategy for each subsystem includes:
initializing various parameters in a depth deterministic strategy gradient network;
In the training process, interacting with a simulated wind farm environment FAST.Farm, and storing tuples generated by interaction into an experience playback pool;
based on the tuple in the experience playback pool, performing distributed priority playback, performing iterative updating on the evaluation network parameters of the depth deterministic strategy gradient network by adopting a minimized loss function method, and performing iterative updating on the strategy network parameters of the depth deterministic strategy gradient network by adopting a gradient direction moving method of enabling the action strategy to the state-action cost function until the set iterative times are reached.
In an alternative embodiment, adjusting the fan yaw angle of the respective subsystem based on the optimal fan yaw angle control strategy includes:
The actual yaw angle of the fan is different from the optimal yaw angle of the fan, and a yaw value of the fan is obtained;
Setting a yaw action threshold value: ；
Representing the actual yaw angle of the current wind turbine i, when the optimum wind turbine yaw angle/> And/>When the difference is 0.0174 rad, the yaw mechanism can adjust the yaw angle to be/>。
The wind power plant grouping wake optimization method and system have the beneficial effects that the wind power plant grouping decoupling order reduction is realized by correcting the pneumatic coupling degree between fans represented by the Gaussian wake model; the wake flow optimization control problem is built based on a Markov decision process, a DDPG algorithm frame is introduced, and the pre-training of DDPG intelligent agent is carried out based on FAST.Farm, so that the cooperative optimization of wind power plant power and fatigue is realized; and accessing the DDPG network after the pre-training to a wind power plant SCADA system to finally realize the optimal control of the wake flow of the wind power plant. Compared with the prior art, the wake flow optimization control is realized through interaction with the actual wind power plant, and the method does not depend on a mechanism model of the wind power plant, so that the problems of poor optimization effect and the like caused by low accuracy of the mechanism model and errors in modeling can be avoided. The method has the advantages that the optimization problem is not required to be constructed and solved, the solving speed of the wake flow optimization control problem is remarkably improved, and the control instruction can be directly given by using the strategy network of the intelligent agent. On the premise of ensuring the optimization effect, the method realizes 'divide-and-conquer' of the large-scale wind power plant, and compared with the prior art, the method can greatly reduce the calculation complexity and avoid the problems of 'dimension disaster', and the like. In the optimization control, not only the promotion of the power generated by the wind farm is considered, but also the power generation output fatigue of the fan and the fluctuation fatigue of the ambient wind speed are considered, and compared with the prior art, the load reduction and the synergy can be considered, so that the method is more in line with the actual application target.
In addition, the invention has reliable design principle, simple structure and very wide application prospect.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the description of the embodiments or the prior art will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of a method of one embodiment of the invention.
FIG. 2 is a schematic diagram of a fan wake impact model of a method of an embodiment of the invention.
FIG. 3 is a schematic diagram of a depth deterministic strategy gradient network of a method of one embodiment of the invention.
FIG. 4 is a schematic diagram of the environment interaction process of a depth deterministic strategy gradient network of the method of one embodiment of the present invention.
FIG. 5 is a schematic diagram of a training process of a depth deterministic strategy gradient network of a method of one embodiment of the invention.
Detailed Description
In order to make the technical solution of the present invention better understood by those skilled in the art, the technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
The wind farm grouping wake optimizing method provided by the embodiment of the invention is executed by the computer equipment, and correspondingly, the wind farm grouping wake optimizing system is operated in the computer equipment.
FIG. 1 is a schematic flow chart of a method of one embodiment of the invention. Wherein, the execution subject of fig. 1 may be a wind farm clustered wake optimization system. The order of the steps in the flow chart may be changed and some may be omitted according to different needs.
As shown in fig. 1, the method includes:
step 110, dividing the whole wind farm into a plurality of subsystems according to the pneumatic coupling level between fans;
step 120, constructing wake optimization as a Markov decision process, and constructing a depth deterministic strategy gradient network for each subsystem respectively;
130, performing offline pre-training on a depth deterministic strategy gradient network to obtain an optimal fan yaw angle control strategy of each subsystem;
And 140, carrying out online adjustment on the fan yaw angles of the corresponding subsystems based on the optimal fan yaw angle control strategy.
In order to facilitate understanding of the invention, the wind farm clustered wake optimization method provided by the invention is further described below by combining the process of optimizing the wind farm clustered wake in the embodiment.
Specifically, the wind farm grouping wake optimizing method comprises the following steps:
S1, wind farm grouping decoupling
Along with the continuous increase of the number of wind turbines in the wind power plant, the wind power plant is regarded as a whole to construct an optimization problem, a huge solving space is faced, and the solving time is multiplied. The aim of wind farm grouping decoupling is to divide the whole wind farm into small subsystems according to the pneumatic coupling level between fans, thereby reducing the computational complexity of deep reinforcement learning implementation. The grouping decoupling strategy mainly comprises the following steps:
S101, constructing wake flow influence models of all fans.
As shown in fig. 2, there is a wake region downstream of the fan where the wind speed is reduced compared to the inflow wind speed. Therefore, the wake wind speed deficit can be used to reflect the effect of the upstream fan, where a large wind speed deficit represents a large effect of the wake of the upstream fan. The fan wake influence model is shown in formulas (1) - (6):
-----(1)；
-----(2)；
-----(3)；
-----(4)；
Wherein, The wind speed value at the downstream distance x and the radial distance r of the fan is represented, v is the wind speed of the inflow wind,/>Representing the width of the wake plane,/>Represents the diameter of a wind wheel of a fan,/>Represents the fan thrust coefficient,/>Represents the yaw angle of the fan,/>Is constant, the value of which depends on the roughness of the ground, here taken to be 0.055.
In addition, because the fan yaw is opposite to the wind, the center line of the wake plane is deviated from the x-axis direction by a distanceCan be calculated from formulas (5) - (6):
-----(5)；
-----(6)；
Wherein, For the initial wake angle,/>For uncertain model parameters that depend on the atmosphere, 0.15 is taken here.
S102, constructing a fatigue load quantization model of each fan.
The traditional method for calculating the fatigue load of the fan is a rain flow counting method, and the method is high in calculation flow complexity and low in calculation speed and is generally used for post-processing analysis. For the optimized control of the fan fatigue, a simpler fan fatigue load quantization model is needed. Considering that the shaft torque, the tower bending moment and the blade bending moment of the fan are key reasons for the fatigue load of the fan, the shaft torque mainly depends on the power generation output of the fan, and the tower bending moment and the blade bending moment mainly depend on the fluctuation of the ambient wind speed. Therefore, in the fan fatigue load quantization model constructed by the invention, the fatigue degree of the fanFatigue generated by fanAnd environmental wind speed fluctuation fatigue/>The two parts are composed, and the specific formulas are shown as formulas (7) - (9):
-----(7)；
-----(8)；
-----(9)；
Wherein, Fatigue for wind turbine power generation,/>For the fatigue of fluctuation of the ambient wind speed, kp and kw are weight factors, here taken as 1,/>For/>Generating power of fan at moment,/>For rated power generation of fan,/>For the design life of the fan, 1 is taken as 25 years, r is the action coefficient of repairing and maintaining the fan, and 0.5/>, is taken asFor standard turbulence used in fan design, taken here as 1.824,/>For/>The fan is subjected to environmental turbulence at any time.
S103, grouping strategy based on breadth-first search
Based on the fan wake influence model constructed in S101, the range of the downstream wake area of the fan and the wind speed value of each point in the range can be obtained. Assuming downstream fansThe coordinates at the hub are/>If it is located at the upstream fan/>In the wake region of (2), the formula (10) is satisfied: /(I)-----(10)。/>
On the basis, an upstream fan is constructedFor downstream blower/>Influence factor/>As shown in formula (11):
-----(11)；
Wherein, And/>For the upstream blower/>The set range of yaw angles, here +30° and-30 °,Is the weight factor of different yaw angles, and is taken as 1/>, hereFor yaw angle/>Downstream blower/>Wind speed value at the hub.
On the basis, with an upstream fanFor point i, downstream blower/>For point j, upstream blower/>For downstream blower/>Influence factor/>Is the weight of the edge between point i and point j. If influencing factors/>，/>Taking 0.2 here, it is assumed that the point i and the point j are no longer connected, and the weight is set to 0.
According to the method, all wind turbines are traversed in turn, the numbers of fans in downstream wake areas and influence factors thereof are obtained, so that an undirected graph representing the pneumatic coupling degree among the fans can be obtained, and the undirected graph is stored in an adjacent matrix mode. And traversing the undirected graph by adopting breadth-first search, solving the number of connected components in the undirected graph, and realizing wind farm group decoupling.
S2, agent pre-training
S201, constructing wake optimization control problem based on Markov decision process
In order to achieve model-free optimal control of a wind farm, first, the wake optimization control problem is constructed as a data-driven markov decision process. Transferring state of a system through tuplesDescription of the preferred embodiments wherein/>The system state variables at the time t comprise yaw angles of all fans, power generated by all fans and fatigue degree/>, of all fansEtc./>The control action variables adopted for the moment t comprise the yaw angle of each fan,/>Signifying taking the control action/>After that, the updated state of the system,/>The rewards for evaluating the quality of this state transition are shown in formulas (12) - (14). Next, for each subsystem formed by decoupling in the first part of wind farm cluster decoupling, a depth deterministic strategy gradient network (DDPG) is respectively built to make the decision of the markov process, DDPG is composed of an evaluation network and a strategy network, and each network contains a current network and a target network, as shown in fig. 3. Finally, the agent in DDPG is configured to determine the current system state/>, by sensing the current system state/>Policy enforcement action/>Thereby changing the environmental state and obtaining rewards/>As shown in fig. 4, the goal of the agent is to maximize the expected rewards, which represent all rewards that can be brought after performing an action, including instant rewards and future rewards at the current time, instant rewards/>As shown in formulas (12) - (14):
-----(12)；
-----(13)；
-----(14)；
wherein the reward function First part/>Representing the difference value of the wind power plant after power and fatigue normalization, the larger the rewards obtained by the intelligent agent, the more obvious the load reduction and synergy effects are represented. In addition, in order to ensure the safe operation of the intelligent agent, the risk of the yaw angle of the fan out of limit needs to be considered in the reward function, so that the reward function/>Second part/>Representing a penalty for dangerous actions by the agent. C p and C f are normalized weighting factors,/>For the power generated by the blower WT i at time t,/>For fatigue load of the blower WT j at time t,/>Penalty coefficient for yaw out-of-limit,/>And/>The upper and lower limits of the yaw angle set point for blower WT i, respectively.
To ensure that the intelligent agent can aim at different statesGive the optimal action/>The invention constructs a state-action cost function/>As shown in formula (15): /(I)-----(15)。
The function represents the in-stateIf according to policy/>Is required to perform the action/>Expected rewards that can be brought,/>Is a prize discount factor, indicating that the farther from the current time, the lower the weight of the prize, E () represents the mathematical expectation.
According to the above settings, to realize wake optimal control, it is necessary for the agent to learn the optimal strategy in the interaction with the environmentSo that the expected rewards can be maximized. The optimal strategy may be implemented by maximizing a state-action cost function, as shown in equation (16): /(I)-----(16)。
S3, DDPG training frames.
This section gives the training process of DDPG constructed, i.e., how to learn optimal strategies for the agents in DDPGThe overall training framework is shown in fig. 5.
(1) Parameters in the DDPG algorithm are first initialized. The current and target evaluation networks respectively adopt parameters asAnd/>Is constructed to estimate a state-action cost function; the current and target policy networks respectively adopt parameters as/>And/>Is constructed for learning action strategies/>And execute action/>. Before training starts,/>Employing random parameters and then let/>，/>Employing random parameters and then let/>。
(2) During the training process, the current strategy network is based onAnd (3) formulating yaw angle control strategies of all fans, wherein N is random noise. Then perform action/>Interacting with FAST.Farm of open source wind farm simulation software to generate tuple/>, wherein the tuple is generated by interactionAnd storing into an experience playback pool.
(3) M tuples are randomly sampled from the experience pool for parameter updates of the current policy network and the current evaluation network. The current strategy network is updated by a method of minimizing a loss function, and the current evaluation network is updated by a method of moving the action strategy to the gradient direction of the state-action cost function.
(4) And finally, according to the set time period, updating parameters of the target strategy network and the target evaluation network in a soft updating mode.
(5) When T reaches T max, a training period C is considered to be completed, and when C reaches C max, the pre-training process is ended, and the DDPG architecture trained at the moment is stored for online control. Here Tmax is set to 12h and cmax is set to 1000.
S4, online control.
Using a pre-trained DDPG architecture, accessing a wind power plant SCADA system to acquire wind power plant actual measurement data as a current system stateThen, utilizing the intelligent agent in DDPG to obtain the yaw value/>, of each fan at the current momentIn order to reduce yaw mechanism failure caused by frequent yaw mechanism actions, the yaw action threshold is designed as shown in the formula (17): /(I)-----(17)；
In the method, in the process of the invention,Representing the actual yaw angle of the current fan i, when the intelligent agent in DDPG gives a yaw instructionAnd/>When the difference is 0.0174 rad, the yaw mechanism can only act to adjust the yaw angle to/>。
In some embodiments, the wind farm cluster wake optimization system may include a plurality of functional modules comprised of computer program segments. The computer program of the individual program segments in the wind farm clustered wake optimization system may be stored in a memory of a computer device and executed by at least one processor to perform (see fig. 1 for details) the functions of wind farm clustered wake optimization.
In this embodiment, the wind farm cluster wake optimization system may be divided into a plurality of functional modules according to the functions performed by the wind farm cluster wake optimization system. The functional modules of the system may include: the system comprises a grouping decoupling module, a network construction module, a network training module and a yaw adjustment module. The module referred to in the present invention refers to a series of computer program segments capable of being executed by at least one processor and of performing a fixed function, stored in a memory. In the present embodiment, the functions of the respective modules will be described in detail in the following embodiments.
The grouping decoupling module is used for dividing the whole wind power plant into a plurality of subsystems according to the pneumatic coupling level between the fans;
The network construction module is used for constructing wake optimization into a Markov decision process and constructing a depth deterministic strategy gradient network for each subsystem respectively;
the network training module is used for training the depth deterministic strategy gradient network and acquiring the optimal fan yaw angle control strategy of each subsystem;
And the yaw adjustment module is used for adjusting the fan yaw angles of the corresponding subsystems based on the optimal fan yaw angle control strategy.
Alternatively, as an embodiment of the present invention, dividing the entire wind farm into a plurality of subsystems according to the level of aerodynamic coupling between fans, comprises:
Constructing a fan wake flow influence model;
Constructing a fan fatigue load quantization model;
constructing an influence factor of an upstream fan on a downstream fan based on a fan wake influence model;
converting the influence factors into influence weights of the upstream fans on the downstream fans;
Traversing all fan combinations to obtain influence weights among fans, and generating an undirected graph representing the pneumatic coupling degree among fans based on the influence weights.
Optionally, as an embodiment of the present invention, constructing a fan wake impact model includes:
Constructing a downstream wind speed value calculation model: ；
，/>，/>；
Wherein, The wind speed value at the downstream distance x and the radial distance r of the fan is represented, v is the wind speed of the inflow wind,/>Representing the width of the wake plane,/>Represents the diameter of a wind wheel of a fan,/>Represents the fan thrust coefficient,/>Represents the yaw angle of the fan,/>Is constant, the value of which depends on the roughness of the ground; /(I)
Distance of deviationThe calculation formula of (2) is as follows: /(I)；
Wherein,，/>For the initial wake angle,/>Is an uncertain model parameter that depends on the atmosphere.
Optionally, as an embodiment of the present invention, constructing a fan fatigue load quantization model includes:
；
Wherein, Fatigue for wind turbine power generation,/>Is fatigue caused by fluctuation of the ambient wind speed,/>And k w is a weight factor that is a weight factor,For/>Generating power of fan at moment,/>For rated power generation of fan,/>For the design life of the fan, r is the action coefficient of repairing and maintaining the fan, and is/(For standard turbulence used in fan design,/>For/>The fan is subjected to environmental turbulence at any time.
In an alternative embodiment, constructing an influence factor of an upstream fan to a downstream fan based on a fan wake influence model includes:
Setting downstream fan The coordinates at the hub are/>If it is located at the upstream fan/>Within the wake region of (2) to be satisfied/>；
Construct upstream fanFor downstream blower/>Influence factor/>The calculation formula is as follows:
；
Wherein, And/>For the upstream blower/>Setting range of yaw angle,/>Is the weight factor of different yaw angles,/>Indicating yaw angle as/>Downstream blower/>Wind speed value at the hub.
Optionally, as an embodiment of the present invention, converting the influence factor into an influence weight of the upstream fan on the downstream fan includes:
Upstream fan For point i, downstream blower/>For point j, upstream blower/>For downstream blower/>Influence factor/>Is the weight between point i and point j; /(I)
If the influence factor is，/>If the preset fixed parameter is set, the point i and the point j are not communicated any more, and the weight is set to 0.
Alternatively, as an embodiment of the present invention, constructing the wake optimization as a markov decision process and constructing a depth deterministic strategy gradient network for each subsystem separately includes:
Constructing wake optimization control problems as a data-driven Markov decision process, transitioning states of a system through tuples Description of the preferred embodiments wherein/>The system state variables at the time t comprise yaw angles of all fans, power generated by all fans and fatigue degree/>, of all fansEtc./>The control action variables taken for time t, including the yaw angle of each fan,Signifying taking the control action/>After that, the updated state of the system,/>Rewards for evaluating the state transition quality at this time;
Aiming at each subsystem formed by decoupling in the first part of wind power plant grouping decoupling, respectively constructing a depth deterministic strategy gradient network to make a decision of a Markov process, wherein the depth deterministic strategy gradient network consists of an evaluation network and a strategy network, and each network comprises a current network and a target network;
agent in depth deterministic policy gradient network by sensing current system state Policy-based implementation of actionsThereby changing the environmental state and obtaining rewards/>The goal of the agent is to maximize the expected rewards, which represent all rewards that can be brought after performing an action, including instant rewards and future rewards at the current time;
Wherein rewards are awarded Comprises:
；
wherein the reward function First part/>Representing the difference value of the wind power plant after power and fatigue normalization, wherein the larger the rewards obtained by the intelligent agent is, the more obvious the load reduction and synergy effects are represented; reward function/>Second part/>Representing a penalty to dangerous actions of the agent; c p and C f are normalized weighting factors,/>For the power generated by the blower WT i at time t,/>For fatigue load of the blower WT j at time t,/>Penalty coefficient for yaw out-of-limit,/>And/>The upper limit and the lower limit of the yaw angle of the fan WT i are respectively set;
Constructing a state-action cost function ，/>Representing the state/>If according to policy/>Is required to perform the action/>Expected rewards that can be brought,/>Is a rewards discount factor, which means that the farther from the current moment, the lower the weight of rewards, E () represents mathematical expectations;
the expression of the state-action cost function is: ；
enabling an agent to learn an optimal strategy in interaction with an environment Maximizing the expected rewards; the optimal strategy is implemented by maximizing a state-action cost function, expressed as: /(I)。
Optionally, as an embodiment of the present invention, training the depth deterministic strategy gradient network to obtain an optimal fan yaw control strategy for each subsystem includes:
initializing various parameters in a depth deterministic strategy gradient network;
in the training process, interacting with the simulated wind field environment, and storing tuples generated by interaction into an experience playback pool;
Based on the tuple in the experience playback pool, the value network parameters and the strategy network parameters of the depth deterministic strategy gradient network are iteratively updated by adopting a minimum loss function method until the set iteration times are reached.
Optionally, as an embodiment of the present invention, adjusting the fan yaw angle of the corresponding subsystem based on the optimal fan yaw angle control strategy includes:
The actual yaw angle of the fan is different from the optimal yaw angle of the fan, and a yaw value of the fan is obtained;
Setting a yaw action threshold value: ；
Representing the actual yaw angle of the current wind turbine i, when the optimum wind turbine yaw angle/> And/>When the difference is 0.0174 rad, the yaw mechanism can adjust the yaw angle to be/>。
Although the present invention has been described in detail by way of preferred embodiments with reference to the accompanying drawings, the present invention is not limited thereto. Various equivalent modifications and substitutions may be made in the embodiments of the present invention by those skilled in the art without departing from the spirit and scope of the present invention, and it is intended that all such modifications and substitutions be within the scope of the present invention/be within the scope of the present invention as defined by the appended claims.
Claims (10)
1. A method for optimizing a wind farm group wake, comprising:
Dividing the whole wind farm into a plurality of subsystems according to the pneumatic coupling level between fans;
constructing wake optimization as a Markov decision process, and constructing a depth deterministic strategy gradient network for each subsystem respectively;
performing offline pre-training on the depth deterministic strategy gradient network to obtain an optimal fan yaw angle control strategy of each subsystem;
And on-line adjustment of fan yaw angles of the corresponding subsystems is performed based on an optimal fan yaw angle control strategy.
2. The method of claim 1, wherein dividing the entire wind farm into a plurality of subsystems based on the level of aerodynamic coupling between fans, comprises:
Constructing a fan wake flow influence model;
Constructing a fan fatigue load quantization model;
constructing an influence factor of an upstream fan on a downstream fan based on a fan wake influence model;
converting the influence factors into influence weights of the upstream fans on the downstream fans;
Traversing all fan combinations to obtain influence weights among fans, and generating an undirected graph representing the pneumatic coupling degree among fans based on the influence weights.
3. The method of claim 2, wherein constructing a fan wake impact model comprises:
Constructing a downstream wind speed value calculation model: ；
，/>，/>；
Wherein, The wind speed value at the downstream distance x and the radial distance r of the fan is represented, v is the wind speed of the inflow wind,/>Representing the width of the wake plane,/>Represents the diameter of a wind wheel of a fan,/>Represents the fan thrust coefficient,/>Represents the yaw angle of the fan,/>Is constant, the value of which depends on the roughness of the ground;
Distance of deviation The calculation formula of (2) is as follows: /(I)；
Wherein,，/>For the initial wake angle,/>Is an uncertain model parameter that depends on the atmosphere.
4. The method of claim 2, wherein constructing a fan fatigue load quantification model comprises:
；
Wherein, Fatigue for wind turbine power generation,/>Is fatigue caused by fluctuation of the ambient wind speed,/>And k w is a weight factor,/>Is thatGenerating power of fan at moment,/>For rated power generation of fan,/>For the design life of the fan, r is the action coefficient of repairing and maintaining the fan, and is/(For standard turbulence used in fan design,/>For/>The fan is subjected to environmental turbulence at any time.
5. The method of claim 2, wherein constructing an upstream fan influence factor for a downstream fan based on a fan wake influence model comprises:
Setting downstream fan The coordinates at the hub are/>If it is located at the upstream fan/>In the wake region of (2) to satisfy；
Construct upstream fanFor downstream blower/>Influence factor/>The calculation formula is as follows:
；
Wherein, And/>For the upstream blower/>Upper and lower limits of yaw angle set point,/>Is the weight factor of different yaw angles,/>Indicating yaw angle as/>Downstream blower/>Wind speed value at the hub.
6. The method of claim 5, wherein converting the impact factor into an impact weight of an upstream fan on a downstream fan and implementing group decoupling between fans in a wind farm according to the weight comprises:
Upstream fan For point i, downstream blower/>For point j, upstream blower/>For downstream blower/>Is the influencing factor of (2)Is the weight between point i and point j;
If the influence factor is ，/>If the pneumatic coupling cut-off value is a preset pneumatic coupling cut-off value, the point i and the point j are considered to be not communicated any more, and the weight is set to be 0;
traversing all wind turbine generators in sequence, obtaining fan numbers and influence weights of the fan numbers in downstream wake areas, and constructing an undirected graph representing pneumatic coupling degree among the fans;
Traversing the undirected graph by breadth-first search, and solving the connected components to realize wind farm group decoupling.
7. The method of claim 1, wherein constructing the wake optimization as a markov decision process and constructing a depth deterministic strategy gradient network for each subsystem separately comprises:
Constructing wake optimization control problems as a data-driven Markov decision process, transitioning states of a system through tuples Description of the preferred embodiments wherein/>The system state variables at the time t comprise yaw angles of all fans, power generated by all fans and fatigue degree/>, of all fans，/>The control action variables adopted for the moment t comprise the yaw angle of each fan,/>Signifying taking the control action/>After that, the updated state of the system,/>Rewards for evaluating the state transition quality at this time;
Aiming at each subsystem formed by decoupling in the first part of wind power plant grouping decoupling, respectively constructing a depth deterministic strategy gradient network to make a decision of a Markov process, wherein the depth deterministic strategy gradient network consists of an evaluation network and a strategy network, and each network comprises a current network and a target network;
agent in depth deterministic policy gradient network by sensing current system state Policy enforcement action/>Thereby changing the environmental state and obtaining rewards/>The goal of the agent is to maximize the expected rewards, which represent all rewards that can be brought after performing an action, including instant rewards and future rewards at the current time;
Wherein rewards are awarded Comprises:
；
wherein the reward function First part/>Representing the difference value of the wind power plant after power and fatigue normalization, wherein the larger the rewards obtained by the intelligent agent is, the more obvious the load reduction and synergy effects are represented; reward function/>Second part/>Representing a penalty to dangerous actions of the agent; c p and C f are normalized weighting factors,/>For the power generated by the blower WT i at time t,/>For fatigue load of the blower WT j at time t,/>Penalty coefficient for yaw out-of-limit,/>And/>The upper limit and the lower limit of the yaw angle of the fan WT i are respectively set;
Constructing a state-action cost function ，/>Representing the state/>If according to policy/>Is required to perform the action/>Expected rewards that can be brought,/>Is a rewards discount factor, which means that the farther from the current moment, the lower the weight of rewards, E () represents mathematical expectations;
enabling an agent to learn an optimal strategy in interaction with an environment Maximizing the expected rewards; the optimal strategy is implemented by maximizing a state-action cost function, expressed as:。
8. The method of claim 1, wherein offline pre-training the depth deterministic strategy gradient network to obtain an optimal fan yaw control strategy for each subsystem comprises:
initializing various parameters in a depth deterministic strategy gradient network;
In the training process, interacting with a simulated wind farm environment FAST.Farm, and storing tuples generated by interaction into an experience playback pool;
based on the tuple in the experience playback pool, performing distributed priority playback, performing iterative updating on the evaluation network parameters of the depth deterministic strategy gradient network by adopting a minimized loss function method, and performing iterative updating on the strategy network parameters of the depth deterministic strategy gradient network by adopting a gradient direction moving method of enabling the action strategy to the state-action cost function until the set iterative times are reached.
9. The method of claim 1, wherein online adjusting the fan yaw angle of the respective subsystem based on the optimal fan yaw angle control strategy comprises:
accessing a wind power plant SCADA system to obtain corresponding measured data as a current subsystem state ；
Obtaining an optimal fan yaw angle of each fan at the current moment based on an optimal fan yaw angle control strategy;
The actual yaw angle of the fan is different from the optimal yaw angle of the fan, and a yaw value of the fan is obtained;
Setting a yaw action threshold value: ；
Representing the actual yaw angle of the current wind turbine i, when the optimum wind turbine yaw angle/> And/>When the difference is 0.0174 rad, the yaw mechanism can adjust the yaw angle to be/>Fatigue caused by frequent actions of the yaw mechanism is avoided.
10. A wind farm clustered wake optimization system, comprising:
the grouping decoupling module is used for dividing the whole wind power plant into a plurality of subsystems according to the pneumatic coupling level between the fans;
The network construction module is used for constructing wake optimization into a Markov decision process and constructing a depth deterministic strategy gradient network for each subsystem respectively;
The network training module is used for performing offline pre-training on the depth deterministic strategy gradient network and acquiring an optimal fan yaw angle control strategy of each subsystem;
And the yaw adjustment module is used for carrying out online adjustment on the fan yaw angles of the corresponding subsystems based on the optimal fan yaw angle control strategy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410391204.6A CN117967499B (en) | 2024-04-02 | 2024-04-02 | Wind power plant grouping wake optimization method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410391204.6A CN117967499B (en) | 2024-04-02 | 2024-04-02 | Wind power plant grouping wake optimization method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117967499A true CN117967499A (en) | 2024-05-03 |
| CN117967499B CN117967499B (en) | 2024-06-25 |
Family
ID=90851833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410391204.6A Active CN117967499B (en) | 2024-04-02 | 2024-04-02 | Wind power plant grouping wake optimization method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117967499B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150240789A1 (en) * | 2014-02-25 | 2015-08-27 | State Grid Corporation Of China | Method of analyzing wake flow of wind turbine based on multiple wake flow models |
| CN106203695A (en) * | 2016-07-07 | 2016-12-07 | 华北电力大学 | Optimization Scheduling in a kind of wind energy turbine set reducing wake effect |
| DE102016212364A1 (en) * | 2016-07-06 | 2018-01-11 | Universität Stuttgart | Follow-up flow diversion using feedback control to improve the output of wind farms |
| CN113541192A (en) * | 2021-07-27 | 2021-10-22 | 重庆大学 | Offshore wind farm reactive power-voltage coordination control method based on deep reinforcement learning |
| CN114421468A (en) * | 2022-01-26 | 2022-04-29 | 太原理工大学 | Primary frequency modulation capacity planning method considering wind power cluster shared energy storage joint participation |
| CN116221021A (en) * | 2023-04-10 | 2023-06-06 | 香港大学深圳研究院 | Wind farm collaborative yaw control method based on multilayer artificial intelligence system |
-
2024
- 2024-04-02 CN CN202410391204.6A patent/CN117967499B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150240789A1 (en) * | 2014-02-25 | 2015-08-27 | State Grid Corporation Of China | Method of analyzing wake flow of wind turbine based on multiple wake flow models |
| DE102016212364A1 (en) * | 2016-07-06 | 2018-01-11 | Universität Stuttgart | Follow-up flow diversion using feedback control to improve the output of wind farms |
| CN106203695A (en) * | 2016-07-07 | 2016-12-07 | 华北电力大学 | Optimization Scheduling in a kind of wind energy turbine set reducing wake effect |
| CN113541192A (en) * | 2021-07-27 | 2021-10-22 | 重庆大学 | Offshore wind farm reactive power-voltage coordination control method based on deep reinforcement learning |
| CN114421468A (en) * | 2022-01-26 | 2022-04-29 | 太原理工大学 | Primary frequency modulation capacity planning method considering wind power cluster shared energy storage joint participation |
| CN116221021A (en) * | 2023-04-10 | 2023-06-06 | 香港大学深圳研究院 | Wind farm collaborative yaw control method based on multilayer artificial intelligence system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117967499B (en) | 2024-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gebraad et al. | Wind plant power optimization through yaw control using a parametric model for wake effects—a CFD simulation study | |
| CN104632521B (en) | A kind of wind power optimization system and method based on drift correction | |
| Song et al. | Coordinated optimization on energy capture and torque fluctuation of wind turbines via variable weight NMPC with fuzzy regulator | |
| CN110009135B (en) | Wind power prediction method based on width learning | |
| Jia et al. | A reinforcement learning based blade twist angle distribution searching method for optimizing wind turbine energy power | |
| CN105649877A (en) | Ant colony PID individual pitch control method of large wind turbine unit | |
| CN114341488A (en) | Reinforcement learning based wind turbine yaw offset control | |
| CN105888971A (en) | Active load reducing control system and method for large wind turbine blade | |
| CN113139326A (en) | Wind power plant yaw cooperative control optimization method, electronic device and storage medium | |
| CN107045574A (en) | The low wind speed section effective wind speed method of estimation of wind power generating set based on SVR | |
| Shu et al. | Decentralised optimisation for large offshore wind farms using a sparsified wake directed graph | |
| CN115333168A (en) | Offshore wind farm field level control strategy based on distributed rolling optimization | |
| Ma et al. | An innovative aerodynamic design methodology of wind turbine blade models for wind tunnel real-time hybrid tests based on genetic algorithm | |
| CN117967499B (en) | Wind power plant grouping wake optimization method and system | |
| Song et al. | Application of surrogate-assisted global optimization algorithm with dimension-reduction in power optimization of floating offshore wind farm | |
| Yu et al. | Reinforcement learning-based multi-objective differential evolution for wind farm layout optimization | |
| CN117454939A (en) | Wind power prediction method for optimizing BP neural network based on NSABO-Gold algorithm | |
| CN113236487A (en) | Wind power plant noise control method, system, device and readable storage medium | |
| CN112052604A (en) | Method, system, equipment and readable medium for predicting equivalent fatigue load of fan | |
| CN116581827A (en) | Wind farm unit start-stop strategy under load limiting condition considering wake flow influence | |
| CN116306236A (en) | Wind power plant dynamic equivalent modeling method and system based on GRU-LSTM-FC combined network | |
| CN115796327A (en) | Wind power interval prediction method based on VMD (vertical vector decomposition) and IWOA-F-GRU (empirical mode decomposition) -based models | |
| CN115408958A (en) | Wind turbine simulation method and system based on grid-independent actuating line model | |
| CN111245008B (en) | Wind field cooperative control method and device | |
| Xu et al. | Dynamic safety control of offshore wind turbine based on model predictive control |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |