CN108468622A - Wind turbines blade root load method of estimation based on extreme learning machine - Google Patents
Wind turbines blade root load method of estimation based on extreme learning machine Download PDFInfo
- Publication number
- CN108468622A CN108468622A CN201810134688.0A CN201810134688A CN108468622A CN 108468622 A CN108468622 A CN 108468622A CN 201810134688 A CN201810134688 A CN 201810134688A CN 108468622 A CN108468622 A CN 108468622A
- Authority
- CN
- China
- Prior art keywords
- input
- learning machine
- extreme learning
- output
- hidden
- 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 15
- 238000013528 artificial neural network Methods 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 7
- 238000003062 neural network model Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 230000005284 excitation Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 2
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 230000003993 interaction Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 210000004218 nerve net Anatomy 0.000 claims 1
- 238000010008 shearing Methods 0.000 claims 1
- 230000005611 electricity Effects 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000011217 control strategy Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007246 mechanism Effects 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
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D80/00—Details, components or accessories not provided for in groups F03D1/00 - F03D17/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses the Wind turbines blade root load methods of estimation based on extreme learning machine, the advantages of making full use of wind-powered electricity generation data collecting system information resources and extreme learning machine Fast Learning, output variable is determined according to modeling demand, system input variable is determined using pivot analysis, Wind turbines blade root load is established by extreme learning machine and estimates model, and load estimated value under corresponding input state is calculated by model.It is not only simple with the progress Wind turbines modeling of extreme learning machine algorithm, but also fast convergence rate, unit performance and state can quickly identify and estimated, a kind of effective method is provided for wind turbine Modeling Research.
Description
Technical field
The present invention relates to wind-power electricity generation modeling estimation method and technology fields, and in particular to the wind turbine based on extreme learning machine
Group blade root load method of estimation.
Background technology
In order to utilize wind energy to greatest extent, the economic benefit and competitiveness of wind-powered electricity generation are improved, Wind turbines are to enlargement, light
Quantized directions develop.Since the randomness and wind of wind cut the influence of effect, tower shadow effect and turbulent flow so that act on wind wheel leaf
In load existence time on the components such as piece and pylon and inhomogeneities spatially, unbalanced load cause unit parts long
Time is vibrated, and the major fatigue of parts can be caused to damage, and then passing through control strategy reduces the dynamic load of unit, improves machine
Group reliability and service life are increasingly taken seriously.
Dynamic load control is primarily upon the load control of Wind turbines critical component and key position, blade and transmission chain
The load being subject to is larger, is the more fragile component of unit reliability.The blade root position load categories of wherein blade are more, most multiple
It is miscellaneous, influence big and be most vulnerable to tired damage, so being paid close attention to by researcher.
In dynamic load Study on Active Control Strategy, it is of crucial importance to establish accurate load model, due to blade root load
Complexity, close coupling, uncertain influence factor are more, major influence factors and the load such as the randomness and wind speed of wind speed, propeller pitch angle
Non-linear relation so that it is traditional based on internal mechanism analyze based on, by empirical equation and assume simplify premised on
Modelling by mechanism is difficult to meet the requirements.
Invention content
In view of the deficiencies of the prior art, the present invention provides the Wind turbines blade root load estimation sides based on extreme learning machine
Method.
Wind turbines blade root load method of estimation proposed by the present invention based on extreme learning machine, including estimation mode input
Output determines and estimation model extreme learning machine learns two parts；
Estimate the determination of mode input output variable：According to modeling demand, F is sheared to wave directionxAnd moment My, shimmy
Direction shears FyAnd moment MxAs estimation model output.Input variable determines：Wind speed size, propeller pitch angle, orientation are chosen first
Angle, 3 wind speed round, wind vector and inertial coodinate system axis 7 variables such as angle, take the data of above-mentioned 7 variables, form
Input matrix X (X1, X2 ..., X7) carries out pivot analysis to X, calculates the contribution rate of each pivot, the accumulation tribute of first four pivot
The rate of offering reaches 90% or more, therefore is 4 pivot X (X1, X2 ..., X4) by original 7 variable data dimensionality reductions, according to contribution
It can determine 4 major influence factors of blade root load：Wind speed v, propeller pitch angle β, azimuth angle theta and wind speed round ω, it is defeated as model
Enter variable.
Estimate that mode input exports normalized：To ensure the non-thread of extreme learning machine (ELM) neural network neuron
Property effect and faster pace of learning, avoid because net input absolute value it is excessive caused by neuron output saturation, should be neural by ELM
The input of network normalizes in a smaller numberical range；When ELM algorithms are returned for being fitted, generally by input and output value
Normalize to [0,1] section.Calculating is normalized to sample data according to normalization formula:
X in formulaiFor pending data, xpFor the data after normalized, xminAnd xmaxFor pending data minimum value and
Maximum value.
Extreme learning machine (ELM) is the solution neural networks with single hidden layer put forward by Nanyang Technolohy University professor Huang Guangbin
Algorithm.Equipped with N number of different sample (Xi,ti)∈Rn×Rm, wherein Xi=[xi1,xi2···xin]T, ti=[ti1,
ti2···tim]TI=1, N.For X input O outputs, there are the Single hidden layer feedforward neural networks of L hidden node can
To be expressed as：
Wherein g (x) is activation primitive；Wi=[wi1,wi2,···,win]TIt is the input power for connecting i-th of hidden node
Weight vector；βi=[βi1,βi2,···,βim]TIt is the output weight vectors of i-th of hidden node；biIt is i-th of hidden node
Threshold values.Activation primitive is the neural networks with single hidden layer of g (x), if approaching N number of sample (X with zero errori,ti), that is, to expire
Sufficient equationThere is βi,Wi,biMeet：
N number of equation of formula (3) can be expressed in matrix as：
H β=T (4)
Wherein H is the output matrix of hidden node, and β is output weight matrix, and T is desired output matrix
When the number of hidden nodes L is equal to sample number N, i.e. L=N, then matrix H, which is square formation, can ask its inverse matrix, single hidden layer feedforward
Neural network can approach sample value with zero error, however sample number N will be far longer than the number of hidden nodes L in many cases,
β may be not present in H non-square matrixs at this timei,Wi,bi(i=1, L, L) meets equation (4), at this moment it is desirable that findingMeet equation：
This is equivalent to minimize loss function：
Traditional Single hidden layer feedforward neural networks algorithm can solve this problem, but need during iteration
Adjust network parameter, the random initializtion input weight W in ELM algorithmsiWith hidden node threshold values bi, do not need iteration adjustment.One
Denier Wi,biIt determines, then the output matrix H of hidden node is now uniquely determined.Equation (1-4), which is equivalent to, at this time solves linear system H β
The least square solution of=TI.e.：
It can acquireWherein H+It is the Moore-Penrose generalized inverses of matrix H.
In conclusion specific ELM algorithms realization can be classified as following steps：
(1) training set ξ={ (x is providedi,ti)|xi∈Rn,ti∈Rm, i=1, L, N }, activation primitive g (x) and hidden node
Number L；
(2) random initializtion input weight WiWith hidden node threshold values bi, i=1, L, L；
(3) hidden layer output matrix H is calculated；
(4) output weight beta=H is calculated+T。
Entire modeling process is made of the following steps：
The ELM neural network blade root load estimation model that Step1 to be established shears F to wave directionxAnd moment My, shimmy
Direction shears FyAnd moment MxIt is exported as model.Using pca method, wind speed v, propeller pitch angle β, azimuth angle theta and wind are determined
It is input variable that wheel speed ω, which is used as,.The input node number of i.e. built ELM neural network models is determined as 4, output node
Number is 4；Inputoutput data is randomly selected from experiment gathered data, is pressed (1- α)：α (test scale factor) percentage
It is divided into training data and test data, determines excitation function G, excitation function can choose sin, sig, hardlmi function, in ELM
Random initializtion input weight W in algorithmiWith hidden node threshold values bi, do not need iteration adjustment, it is only necessary to which hidden layer node is set
Number initial value LsWith end value Lf。
Step2：Determine the basic structure and parameter of neural network
Using ELM algorithm learning neural network parameters, from setting hidden layer node number initial value LsStart, then constantly increases
Hidden node is added to count to preset maximum value Lf, but hidden layer section number maximum value LfGenerally less than training data number, training and survey
Try ELM networks under different hidden nodes, calculate training and test root-mean-square error, to training and test root-mean-square error into
Row is added, and L values when the sum of root-mean-square error is minimum value are the hidden layer neuron number of the network.
Any given input condition is pressed by trained neural network model is input to after the normalization of (1) formula
(4) corresponding output variable x therein after the output T of calculating neural networkpBeing transformed into after quantities by (8) formula can estimate accordingly
Load value xi。
xi=xp*(xmax-xmin)+xmin (8)
X in formulaminAnd xmaxData minimum value and maximum value are managed for original place.
Description of the drawings
Fig. 1 is extreme learning machine neural network structure figure.
Specific implementation mode
Estimate the determination of mode input output variable：According to modeling demand, F is sheared to wave directionxAnd moment My, shimmy
Direction shears FyAnd moment MxAs estimation model output.Input variable determines：Wind speed size, propeller pitch angle, orientation are chosen first
Angle, 3 wind speed round, wind vector and inertial coodinate system axis 7 variables such as angle, take the data of above-mentioned 7 variables, form
Input matrix X (X1, X2 ..., X7) carries out pivot analysis to X, calculates the contribution rate of each pivot, the accumulation tribute of first four pivot
The rate of offering reaches 90% or more, therefore is 4 pivot X (X1, X2 ..., X4) by original 7 variable data dimensionality reductions, according to contribution
It can determine 4 major influence factors of blade root load：Wind speed v, propeller pitch angle β, azimuth angle theta and wind speed round ω, it is defeated as model
Enter variable.
Estimate that mode input exports normalized：To ensure the nonlinear interaction of ELM neural network neurons and very fast
Pace of learning, avoid because only input absolute value it is excessive caused by neuron output saturation, the input of ELM neural networks should be returned
One changes into a smaller numberical range；When ELM algorithms are returned for being fitted, generally normalize to input and output value [0,
1] section.Calculating is normalized to sample data according to normalization formula:
X in formulaiFor pending data, xpFor the data after normalized, xminAnd xmaxFor pending data minimum value and
Maximum value.
Entire modeling process is made of the following steps：
The ELM neural network blade root load estimation model that Step1 to be established shears F to wave directionxAnd moment My, shimmy
Direction shears FyAnd moment MxIt is exported as model.Using pca method, wind speed v, propeller pitch angle β, azimuth angle theta and wind are determined
It is input variable that wheel speed ω, which is used as,.The input node number of i.e. built ELM neural network models is determined as 4, output node
Number is 4；Inputoutput data is randomly selected from experiment gathered data, is pressed (1- α)：α (test scale factor) percentage
It is divided into training data and test data, takes α=10% here, determines that excitation function is sig functions, it is random first in ELM algorithms
Beginningization input weight WiWith hidden node threshold values bi, do not need iteration adjustment, setting hidden layer node initial value Ls=15, end value Lf
=300.
Step2：Determine the basic structure and parameter of neural network
Using ELM algorithm learning neural network parameters, from setting hidden layer node number initial value LsStart, then constantly increases
Hidden node is added to count to preset maximum value Lf, but hidden layer section number maximum value LfGenerally less than training data number, training and survey
Try ELM networks under different hidden nodes, calculate training and test root-mean-square error, to training and test root-mean-square error into
Row is added, and L values when the sum of root-mean-square error is minimum value are the hidden layer neuron number of the network.
Any given input condition is pressed by trained neural network model is input to after the normalization of (9) formula
(10) corresponding output variable x therein after the output T of calculating neural networkpIt can estimate phase after being transformed into quantities by (10) formula
The load value x answeredi。
xi=xp*(xmax-xmin)+xmin (10)
X in formulaminAnd xmaxData minimum value and maximum value are managed for original place.
Above-mentioned specific implementation is the preferable realization of the present invention, and certainly, the invention may also have other embodiments,
Without deviating from the spirit and substance of the present invention, those skilled in the art make various in accordance with the present invention
Corresponding change and deformation, but these corresponding change and deformations should all belong to the scope of the claims of the present invention.
Claims (1)
1. the Wind turbines blade root load method of estimation based on extreme learning machine, which is characterized in that defeated including estimation mode input
Go out to determine and estimation model extreme learning machine learns two parts；
Estimate the determination of mode input output variable：According to modeling demand, F is sheared to wave directionxAnd moment My, edgewise direction
Shear FyAnd moment MxAs estimation model output；Input variable determines：Wind speed size, propeller pitch angle, azimuth, wind are chosen first
7 variables such as angle of 3 wheel speed, wind vector and inertial coodinate system axis take the data of above-mentioned 7 variables, composition input
Matrix X (X1, X2 ..., X7) carries out pivot analysis to X, calculates the contribution rate of each pivot, the accumulation contribution rate of first four pivot
Reach 90% or more, therefore is 4 pivot X (X1, X2 ..., X4) by original 7 variable data dimensionality reductions, it can be true according to contribution
Determine 4 major influence factors of blade root load：Wind speed v, propeller pitch angle β, azimuth angle theta and wind speed round ω become as mode input
Amount；
Estimate that mode input exports normalized：For ensure extreme learning machine neural network neuron nonlinear interaction and compared with
Fast pace of learning avoids the output saturation of the neuron caused by net input absolute value is excessive, should be by extreme learning machine nerve net
The input of network normalizes in a smaller numberical range；It, generally will input when extreme learning machine algorithm is returned for being fitted
Output valve normalizes to [0,1] section；Calculating is normalized to sample data according to normalization formula:
X in formulaiFor pending data, xpFor the data after normalized, xminAnd xmaxFor pending data minimum value and maximum
Value；
Equipped with N number of different sample (Xi,ti)∈Rn×Rm, wherein Xi=[xi1,xi2…xin]T, ti=[ti1,ti2…tim]TI=
1,…,N；For X input O outputs, there are the Single hidden layer feedforward neural networks of L hidden node that can be expressed as：
Wherein g (x) is activation primitive；Wi=[wi1,wi2,…,win]TIt is the input weight vector for connecting i-th of hidden node；βi
=[βi1,βi2,…,βim]TIt is the output weight vectors of i-th of hidden node；biIt is the threshold values of i-th of hidden node；Activate letter
Number is the neural networks with single hidden layer of g (x), if approaching N number of sample (X with zero errori,ti), that is, to meet equationThere is βi,Wi,biMeet：
N number of equation of formula (3) can be expressed in matrix as：
H β=T (4)
Wherein H is the output matrix of hidden node, and β is output weight matrix, and T is desired output matrix
When the number of hidden nodes L is equal to sample number N, i.e. L=N, then matrix H, which is square formation, can ask its inverse matrix, single hidden layer feed forward neural
Network can approach sample value with zero error, however sample number N will be far longer than the number of hidden nodes L in many cases, at this time
β may be not present in H non-square matrixsi,Wi,bi(i=1, L, L) meets equation (4), at this moment it is desirable that findingMeet equation：
This is equivalent to minimize loss function：
Traditional Single hidden layer feedforward neural networks algorithm can solve this problem, but need to adjust during iteration
Network parameter, the random initializtion input weight W in extreme learning machine algorithmiWith hidden node threshold values bi, do not need iteration tune
It is whole；Once Wi,biIt determines, then the output matrix H of hidden node is now uniquely determined；Equation (1-4) is equivalent to the linear system of solution at this time
The least square solution of system H β=TI.e.：
It can acquireWherein H+It is the Moore-Penrose generalized inverses of matrix H.
In conclusion specific extreme learning machine algorithm realization can be classified as following steps：
(1) training set ξ={ (x is providedi,ti)|xi∈Rn,ti∈Rm, i=1, L, N }, activation primitive g (x) and the number of hidden nodes L；
(2) random initializtion input weight WiWith hidden node threshold values bi, i=1, L, L；
(3) hidden layer output matrix H is calculated；
(4) output weight beta=H is calculated+T；
Entire modeling process is made of the following steps：
The extreme learning machine neural network blade root load estimation model that Step1 to be established shears F to wave directionxAnd moment My, pendulum
Shake direction shearing FyAnd moment MxIt is exported as model；Using pca method, determine wind speed v, propeller pitch angle β, azimuth angle theta and
It is input variable that wind speed round ω, which is used as,；The input node number of i.e. built extreme learning machine neural network model is determined as 4,
Output node number is 4；Inputoutput data is randomly selected from experiment gathered data, is pressed (1- α)：α (test ratio because
Son) percentage is divided into training data and test data, determine that excitation function G, excitation function can choose sin, sig, hardlmi letter
Number, the random initializtion input weight W in extreme learning machine algorithmiWith hidden node threshold values bi, iteration adjustment is not needed, is only needed
Hidden layer node number initial value L is setsWith end value Lf；
Step2：Determine the basic structure and parameter of neural network
Limits of application learning machine algorithm learning neural network parameter, from setting hidden layer node number initial value LsStart, then constantly
Increase hidden node and counts to preset maximum value Lf, but hidden layer section number maximum value LfGenerally less than training data number, training and
The extreme learning machine network under different hidden nodes is tested, training and test root-mean-square error are calculated, it is equal to training and test
Square error is added, and L values when the sum of root-mean-square error is minimum value are the hidden layer neuron number of the network；
For any given input condition, it is input to trained neural network model after being normalized by (1) formula, based on (4)
Corresponding output variable x therein after the output T of calculation neural networkpBeing transformed into after quantities by (8) formula can estimate to carry accordingly
Charge values xi；
xi=xp*(xmax-xmin)+xmin (8)
X in formulaminAnd xmaxFor pending data minimum value and maximum value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810134688.0A CN108468622B (en) | 2018-02-09 | 2018-02-09 | Wind turbines blade root load estimation method based on extreme learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810134688.0A CN108468622B (en) | 2018-02-09 | 2018-02-09 | Wind turbines blade root load estimation method based on extreme learning machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108468622A true CN108468622A (en) | 2018-08-31 |
| CN108468622B CN108468622B (en) | 2019-10-11 |
Family
ID=63266378
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810134688.0A Active CN108468622B (en) | 2018-02-09 | 2018-02-09 | Wind turbines blade root load estimation method based on extreme learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108468622B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109488526A (en) * | 2018-11-23 | 2019-03-19 | 湖南工业大学 | Based on ratio-extreme learning machine stable state estimation variable pitch control method |
| CN110985286A (en) * | 2019-12-04 | 2020-04-10 | 浙江大学 | Novel wind turbine generator pitch angle control method based on ELM |
| CN113435595A (en) * | 2021-07-08 | 2021-09-24 | 南京理工大学 | Two-stage optimization method for extreme learning machine network parameters based on natural evolution strategy |
| CN114689237A (en) * | 2020-12-31 | 2022-07-01 | 新疆金风科技股份有限公司 | Load sensor calibration method and device and computer readable storage medium |
| CN116561638A (en) * | 2023-05-24 | 2023-08-08 | 南京电力设计研究院有限公司 | Protective pressing plate non-correspondence checking method based on neural network learning and state evaluation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105243259A (en) * | 2015-09-02 | 2016-01-13 | 上海大学 | Extreme learning machine based rapid prediction method for fluctuating wind speed |
| CN105569923A (en) * | 2016-01-13 | 2016-05-11 | 湖南世优电气股份有限公司 | Radar-assisted load optimizing control method for large wind turbine unit |
| WO2016099558A1 (en) * | 2014-12-19 | 2016-06-23 | Hewlett Packard Enterprise Development Lp | Automative system management |
| WO2016186694A1 (en) * | 2015-05-15 | 2016-11-24 | General Electric Company | Condition-based validation of performance updates |
| CN107229736A (en) * | 2017-06-14 | 2017-10-03 | 北京唐浩电力工程技术研究有限公司 | A kind of wind power plant wind information estimating and measuring method |
-
2018
- 2018-02-09 CN CN201810134688.0A patent/CN108468622B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016099558A1 (en) * | 2014-12-19 | 2016-06-23 | Hewlett Packard Enterprise Development Lp | Automative system management |
| WO2016186694A1 (en) * | 2015-05-15 | 2016-11-24 | General Electric Company | Condition-based validation of performance updates |
| CN105243259A (en) * | 2015-09-02 | 2016-01-13 | 上海大学 | Extreme learning machine based rapid prediction method for fluctuating wind speed |
| CN105569923A (en) * | 2016-01-13 | 2016-05-11 | 湖南世优电气股份有限公司 | Radar-assisted load optimizing control method for large wind turbine unit |
| CN107229736A (en) * | 2017-06-14 | 2017-10-03 | 北京唐浩电力工程技术研究有限公司 | A kind of wind power plant wind information estimating and measuring method |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109488526A (en) * | 2018-11-23 | 2019-03-19 | 湖南工业大学 | Based on ratio-extreme learning machine stable state estimation variable pitch control method |
| CN110985286A (en) * | 2019-12-04 | 2020-04-10 | 浙江大学 | Novel wind turbine generator pitch angle control method based on ELM |
| CN114689237A (en) * | 2020-12-31 | 2022-07-01 | 新疆金风科技股份有限公司 | Load sensor calibration method and device and computer readable storage medium |
| WO2022142149A1 (en) * | 2020-12-31 | 2022-07-07 | 新疆金风科技股份有限公司 | Load sensor calibration method and apparatus, and computer-readable storage medium |
| CN113435595A (en) * | 2021-07-08 | 2021-09-24 | 南京理工大学 | Two-stage optimization method for extreme learning machine network parameters based on natural evolution strategy |
| CN113435595B (en) * | 2021-07-08 | 2024-02-06 | 南京理工大学 | Two-stage optimization method for network parameters of extreme learning machine based on natural evolution strategy |
| CN116561638A (en) * | 2023-05-24 | 2023-08-08 | 南京电力设计研究院有限公司 | Protective pressing plate non-correspondence checking method based on neural network learning and state evaluation |
| CN116561638B (en) * | 2023-05-24 | 2024-05-31 | 南京电力设计研究院有限公司 | Protective pressing plate non-correspondence checking method based on neural network learning and state evaluation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108468622B (en) | 2019-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108468622B (en) | Wind turbines blade root load estimation method based on extreme learning machine | |
| Qais et al. | Enhanced whale optimization algorithm for maximum power point tracking of variable-speed wind generators | |
| Abdelbaky et al. | Design and implementation of partial offline fuzzy model-predictive pitch controller for large-scale wind-turbines | |
| Bottasso et al. | Aero-servo-elastic modeling and control of wind turbines using finite-element multibody procedures | |
| Sierra-García et al. | Performance analysis of a wind turbine pitch neurocontroller with unsupervised learning | |
| CN104595106B (en) | Wind-power generating variable pitch control method based on intensified learning compensation | |
| WO2021073090A1 (en) | Real-time robust variable-pitch wind turbine generator control system and method employing reinforcement learning | |
| CN105649877B (en) | A kind of ant colony PID independent pitch control methods of large-scale wind electricity unit | |
| Chen et al. | Reinforcement-based robust variable pitch control of wind turbines | |
| CN106126906A (en) | Short-term wind speed forecasting method based on C C Yu ELM | |
| Jia et al. | Combining LIDAR and LADRC for intelligent pitch control of wind turbines | |
| CN102900603B (en) | Variable pitch controller design method based on finite time non-crisp/guaranteed-cost stable wind turbine generator set | |
| Araghi et al. | Enhancing the net energy of wind turbine using wind prediction and economic NMPC with high-accuracy nonlinear WT models | |
| CN115689375A (en) | Virtual power plant operation control method, device, equipment and medium | |
| Abbas et al. | Aero‐servo‐elastic co‐optimization of large wind turbine blades with distributed aerodynamic control devices | |
| Tao et al. | On comparing six optimization algorithms for network-based wind speed forecasting | |
| Ayoubi et al. | Intelligent control of a large variable speed wind turbine | |
| Yao et al. | Optimized active power dispatching of wind farms considering data-driven fatigue load suppression | |
| Merz et al. | A hierarchical supervisory wind power plant controller | |
| Lakhal et al. | Fuzzy logic control strategy for tracking the maximum power point of a horizontal axis wind turbine | |
| Li et al. | Wind power forecasting based on time series and neural network | |
| CN102900605B (en) | Wind turbine generator set variable pitch controller design method based on finite time stabilization | |
| CN102900613B (en) | Wind turbine generator set variable pitch controller design method based on finite time robustness or guaranteed cost stabilization | |
| Zhang et al. | Data-driven wind farm Volt/Var control based on deep reinforcement learning | |
| CN102900606B (en) | Wind turbine generator set variable pitch controller design method based on finite time guaranteed cost stabilization |
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 | ||
| GR01 | Patent grant |