CN111488712B - Wind power generator power curve modeling method based on transfer learning - Google Patents

Wind power generator power curve modeling method based on transfer learning Download PDF

Info

Publication number
CN111488712B
CN111488712B CN202010286951.5A CN202010286951A CN111488712B CN 111488712 B CN111488712 B CN 111488712B CN 202010286951 A CN202010286951 A CN 202010286951A CN 111488712 B CN111488712 B CN 111488712B
Authority
CN
China
Prior art keywords
wind
wind power
generator
atypical
typical
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.)
Active
Application number
CN202010286951.5A
Other languages
Chinese (zh)
Other versions
CN111488712A (en
Inventor
张方红
刘冰冰
胡号朋
刘城
温树森
雍彬
温钊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CSIC Haizhuang Windpower Co Ltd
China State Shipbuilding Corp Ltd
Original Assignee
CSIC Haizhuang Windpower Co Ltd
China State Shipbuilding Corp Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CSIC Haizhuang Windpower Co Ltd, China State Shipbuilding Corp Ltd filed Critical CSIC Haizhuang Windpower Co Ltd
Priority to CN202010286951.5A priority Critical patent/CN111488712B/en
Publication of CN111488712A publication Critical patent/CN111488712A/en
Application granted granted Critical
Publication of CN111488712B publication Critical patent/CN111488712B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/06—Wind turbines or wind farms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/06—Power analysis or power optimisation
    • 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)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Wind Motors (AREA)

Abstract

The invention provides a wind power generator power curve modeling method based on transfer learning, which is characterized in that wind power generators in a wind field are preliminarily divided into typical wind power generators and atypical wind power generators, operation data of each typical wind power generator are collected and analyzed to obtain an effective data set, then a power curve of the effective data set is obtained through fitting regression, the effective data set of each typical wind power generator is marked as a source domain, the operation data of each atypical wind power generator is marked as a target domain, each target domain is based on all the source domains, a power curve model is established through transfer learning regression, and then a power curve of the atypical wind power generator corresponding to the target domain is obtained. The rapid modeling of the power curve of the atypical wind power generator is realized in a migration learning mode of the atypical wind power generator.

Description

Wind power generator power curve modeling method based on transfer learning
Technical Field
The invention relates to the technical field of wind power generation, in particular to a wind power generator power curve modeling method based on transfer learning.
Background
The power curve of the wind generating set is a curve describing the relationship between wind speed and unit output power, and is not only an important basis for designing a control system in the design of the wind generating set, but also an important index for evaluating the performance of the wind generator and evaluating the generated energy and the capacity of the wind generator. From a theoretical point of view, for a control system of a wind generating set, a power curve is an important parameter required by the design of a controller and is an important means for evaluating the quality of the control system. In wind turbine control systems, the power curve is an important feedback signal that directly affects the control strategy and the adjustment of parameters in the control process. Due to the difference of the operating environment, errors exist between the theoretical power model and the actual generated power data of the wind generating set. In order to improve the applicability of the power curve model, an actually measured power generation model is usually established based on actually measured data.
Due to the cost relation, a complete sensor system is configured for only a few units with typical characteristics in a general wind field, and a typical unit (flagship unit) with the typical characteristics is constructed. The set has more complete acquisition information, and can analyze and construct a more accurate power curve by utilizing more acquisition information (wind speed, wind direction, temperature, set load and the like). The atypical unit mainly establishes a power curve based on equivalent wind speed and equivalent power, so that the modeling precision is limited. Obviously, the modeling capability of the power curve of the flagship set can assist the atypical set to improve the modeling capability and accuracy. Therefore, the method is based on the idea of transfer learning, and a new wind driven generator set modeling framework is provided from the angle of modeling of a power curve of a full-field wind driven generator of a wind power plant.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a wind generator power curve modeling method based on transfer learning, which aims to solve the technical problem that the existing atypical set establishes a power curve based on equivalent wind speed and equivalent power, and the modeling precision is limited.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a wind power generator power curve modeling method based on transfer learning preliminarily divides wind power generators in a wind field into a typical wind power generator and an atypical wind power generator, and comprises the following steps:
s1, collecting the operation data of each typical wind driven generator, analyzing and processing the operation data to obtain an effective data set, and further fitting regression to obtain a power curve of the effective data set;
s2, marking the effective data set of each typical wind driven generator as a source domain, marking the operation data of each atypical wind driven generator as a target domain, and establishing a power curve model through transfer learning regression on each target domain based on all the source domains so as to obtain the power curve of the atypical wind driven generator corresponding to the target domain.
Optionally, in step S2, in the migration learning process, if the similarity between the target domain and all the source domains is low, the atypical wind turbine corresponding to the target domain is divided into new typical wind turbines, and a new source domain is further constructed.
Optionally, the typical wind power generator comprises a wind power generator positioned at the edge of a wind field, and the wind power generator at the windward side and the wind power generator at the air outlet side are specifically selected.
Optionally, the typical wind power generator further comprises wind power generators located in the wind field, at least one wind power generator is selected from each wind power generator, and any two selected wind power generators are not adjacent to each other.
According to the technical scheme, the invention has the beneficial effects that:
the invention provides a wind power generator power curve modeling method based on transfer learning, which is characterized in that wind power generators in a wind field are preliminarily divided into typical wind power generators and atypical wind power generators, operation data of each typical wind power generator are collected and analyzed to obtain an effective data set, then a regression is fitted to obtain a power curve of the effective data set, the effective data set of each typical wind power generator is marked as a source domain, the operation data of each atypical wind power generator is marked as a target domain, each target domain is based on all the source domains, a power curve model is established through transfer learning regression, and then a power curve of the atypical wind power generator corresponding to the target domain is obtained. And the rapid modeling of the power curve of the atypical wind power generator is realized in a mode of migration learning of the atypical wind power generator.
Drawings
In order to more clearly illustrate the detailed description of the invention or the technical solutions in the prior art, the drawings that are needed in the detailed description of the invention or the prior art will be briefly described below. Throughout the drawings, like elements or portions are generally identified by like reference numerals. In the drawings, elements or portions are not necessarily drawn to scale.
FIG. 1 is a flow chart of a method for modeling a power curve of a wind turbine based on transfer learning;
FIG. 2 is a diagram of a power curve modeling process of a wind turbine based on transfer regression learning;
FIG. 3 is a flow chart of a wind turbine power curve modeling based on transfer regression learning.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and therefore are only examples, and the protection scope of the present invention is not limited thereby.
It is to be noted that, unless otherwise specified, technical or scientific terms used herein shall have the ordinary meaning as understood by those skilled in the art to which the invention pertains.
Referring to fig. 1, according to the method for modeling a power curve of a wind turbine based on transfer learning of the present invention, wind turbines in a wind farm are initially classified into typical wind turbines and atypical wind turbines. The selection of a typical wind driven generator is mainly determined according to the distribution of wind resources in a wind field and the layout of the wind driven generator, and the specific principle is as follows:
A. for a wind turbine located at the edge of a wind farm: selecting a windward side wind driven generator and an air outlet side wind driven generator;
B. for a wind turbine located within a wind farm: at least one wind power generator is selected from each wind power generator, and any two selected wind power generators are not adjacent (including diagonal directions).
After the types of the wind driven generators are divided, the method is implemented according to the following steps:
s1, collecting the operation data of each typical wind driven generator, removing points, complementing points, filtering and the like to obtain an effective data set, and fitting regression to obtain a power curve of the effective data set;
s2, marking the effective data set of each typical wind driven generator as a source domain, namely, one typical wind driven generator corresponds to one source domain; marking the operation data of each atypical wind driven generator as a target domain, namely one atypical wind driven generator corresponds to one target domain; and each target domain is based on all the source domains, and a power curve model is established through transfer learning regression, so that a power curve of the atypical wind driven generator corresponding to the target domain is obtained.
In one embodiment, referring to FIGS. 2-3,
1) based on the idea of the marker post unit, n marker post units are constructed in the wind field, the positions and the working conditions of the marker post units are representative in the wind field, and the basic principle of selection is one to two wind driven generators of an upwind wind driven generator with main wind direction and each exhaust wind driven generator in the wind field.
2) And selecting the data of the n wind power generators, such as wind speed, wind direction, generated power and the like as a source domain. And (3) by utilizing a similarity principle k-Nearest neighbor algorithm, enabling the data to keep a local geometric structure of the data in the original feature space in the mapping subspace:
Figure BDA0002448882160000041
wherein omega lp Is a weight matrix, y l As target domain samples, y j Are source domain samples.
3) And establishing a power curve of the wind driven generator in the target domain by using a regression algorithm based on the data of each source domain and the target domain. Are respectively f 1 ,...,f m . The regression algorithm can adopt a support vector machine and other neural network methods. By means of a support vector machineFor example, the following steps are carried out:
f i (x)＝f i 1 (x)+…+f i l (x)…+f i d (x),i＝1,...,m
Figure BDA0002448882160000042
Figure BDA0002448882160000043
wherein x is a regression function independent variable and is the wind speed, wind direction, rotating speed and the like of the wind driven generator; x is the number of i A sample set consisting of a source domain i and a target domain.
Figure BDA0002448882160000044
And
Figure BDA0002448882160000045
are the l and p samples of the sample set.
Figure BDA0002448882160000046
For the kernel function, a gaussian kernel is generally used.
Figure BDA0002448882160000047
And
Figure BDA0002448882160000048
is a weight coefficient, b i Is a threshold variable.
4) To f 1 ,...,f m Combined construction to obtain f:
f＝w 1 f 1 +w 2 f 2 +…+w m f m
5) using target domain data to weight w 1 ,...,w m And optimizing distribution. The optimization adopts the least square principle of min | | f (x (t) -y (t)) | calculation of luminance 2
6) And in consideration of seasonal climate change of the wind power plant and drift of the performance of the wind driven generator, selecting the similarity source domain periodically, reconstructing the source domain, and constructing a new fitting function f of the target domain.
As a further improvement to the above scheme, in step S2, in the migration learning process, if the similarity between the target domain and all the source domains is low, the atypical wind turbine corresponding to the target domain is divided into new typical wind turbines, and a new source domain is further constructed, so as to implement automatic selection of the source domain wind turbines.
The method can realize the rapid modeling of the power curve of the atypical wind power generator based on the modes of transfer learning and adaptive learning. For the wind driven generator with an unsatisfactory obtained power curve, the considered characteristic of the source domain of the typical wind turbine is not satisfied, so that the typical wind turbine set needs to be further refined and updated, namely adaptive learning is realized. The method is based on the overall wind field, the batch property and the correlation of the wind turbine are fully considered, and the efficiency of the power curve modeling mode is higher than that of the existing wind turbine power curve modeling mode.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.

Claims (1)

1. A wind power generator power curve modeling method based on transfer learning is used for preliminarily dividing wind power generators in a wind field into a typical wind power generator and an atypical wind power generator, and is characterized by comprising the following steps:
s1, collecting the operation data of each typical wind driven generator, analyzing and processing the operation data to obtain an effective data set, and fitting regression to obtain a power curve of the effective data set;
s2, marking the effective data set of each typical wind driven generator as a source domain, marking the operation data of each atypical wind driven generator as a target domain, and establishing a power curve model through transfer learning regression on each target domain based on all the source domains so as to obtain the power curve of the atypical wind driven generator corresponding to the target domain;
wherein the typical wind power generator is divided into:
specifically, aiming at a wind driven generator positioned at the edge of a wind field, a wind driven generator on the windward side and a wind driven generator on the air outlet side are selected as typical wind driven generators;
specifically, at least one of wind power generators in each row of wind power generators is selected as a typical wind power generator aiming at the wind power generators in the wind field;
in the transfer learning process, if the similarity between the target domain and all the source domains is low, the atypical wind driven generator corresponding to the target domain is divided into new typical wind driven generators, and then a new source domain is constructed.
CN202010286951.5A 2020-04-13 2020-04-13 Wind power generator power curve modeling method based on transfer learning Active CN111488712B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010286951.5A CN111488712B (en) 2020-04-13 2020-04-13 Wind power generator power curve modeling method based on transfer learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010286951.5A CN111488712B (en) 2020-04-13 2020-04-13 Wind power generator power curve modeling method based on transfer learning

Publications (2)

Publication Number Publication Date
CN111488712A CN111488712A (en) 2020-08-04
CN111488712B true CN111488712B (en) 2022-08-26

Family

ID=71797963

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010286951.5A Active CN111488712B (en) 2020-04-13 2020-04-13 Wind power generator power curve modeling method based on transfer learning

Country Status (1)

Country Link
CN (1) CN111488712B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070629B (en) * 2020-09-10 2022-06-14 中国船舶重工集团海装风电股份有限公司 Performance evaluation method of wind power plant energy management system
CN114638312B (en) * 2022-03-24 2024-07-19 申能新能源(青海)有限公司 Wind power unsupervised learning classification algorithm

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268627A (en) * 2014-09-10 2015-01-07 天津大学 Short-term wind speed forecasting method based on deep neural network transfer model

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706335B (en) * 2009-11-11 2012-01-11 华南理工大学 Wind power forecasting method based on genetic algorithm optimization BP neural network
CN103218405A (en) * 2013-03-20 2013-07-24 西安电子科技大学 Method for integrating migration text classifications based on dimensionality reduction
CN104217247B (en) * 2013-05-29 2018-08-31 乌托巴斯洞察公司 Method and apparatus for the output power for predicting the wind turbine in wind field
CN106815773A (en) * 2015-12-02 2017-06-09 中国电力科学研究院 A kind of wind power method of evaluating characteristic
US10778194B2 (en) * 2017-08-28 2020-09-15 University Of Florida Research Foundation, Inc. Automatic composition of universal filters

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268627A (en) * 2014-09-10 2015-01-07 天津大学 Short-term wind speed forecasting method based on deep neural network transfer model

Also Published As

Publication number Publication date
CN111488712A (en) 2020-08-04

Similar Documents

Publication Publication Date Title
CN110837226B (en) Thermal power generating unit operation optimization method based on intelligent optimization algorithm and related device
CN109274121B (en) Wind power plant control parameter optimization method and system
CN111488712B (en) Wind power generator power curve modeling method based on transfer learning
CN113657662B (en) Downscaling wind power prediction method based on data fusion
CN103942736B (en) A kind of wind power plant multimachine equivalent modeling method
CN103489046A (en) Method for predicting wind power plant short-term power
EP3741991A1 (en) Method for dynamic real-time optimization of the performance of a wind park and wind park
CN114091265A (en) Wind power plant layout optimization method and system based on local search strategy
KR20230076695A (en) Solar power generation forcasting device and method reflecting solar periodic time variable
CN115358606A (en) Method and system for evaluating energy efficiency of in-service wind power plant under gentle terrain
CN110991701A (en) Wind power plant fan wind speed prediction method and system based on data fusion
Kanchana et al. PV Power Forecasting with Holt-Winters Method
CN111178601B (en) Wind turbine generator power prediction method based on meteorological data post-processing
CN114429233A (en) Cluster algorithm-based station-level optimal power control method and system
JP2019091335A (en) Power generation amount prediction device, power generation amount prediction system, power generation amount prediction method, and power generation amount prediction program
CN110296055B (en) Wind direction prediction associated seed unit screening method
CN110188939B (en) Wind power prediction method, system, equipment and storage medium of wind power plant
CN113688581B (en) Method, device, electronic equipment and medium for optimally controlling active output of wind power plant
KR102439311B1 (en) Coordinated optimization method for optimiztion of wind farm using sparsified wake digraph and apparatus performing the same
CN115146742A (en) Offshore wind farm unit grouping and flagship machine selection method suitable for farm group control
Amaro e Silva et al. A regime-based approach for integrating wind information in spatio-temporal solar forecasting models
CN110991083B (en) Photovoltaic power station model determining method, device, equipment and storage medium
Preethi et al. Predicting the wind turbine power generation based on weather conditions
Pranupa et al. Wind farm layout optimization approach using bio-inspired meta-heuristic algorithm to minimize wake effect
Liu et al. Data-driven parameter calibration in wake models

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
TA01 Transfer of patent application right

Effective date of registration: 20210917

Address after: 401122 No. 30 Jinyu Avenue, Jingkaiyuan, North New District of Chongqing

Applicant after: CSIC HAIZHUANG WINDPOWER Co.,Ltd.

Applicant after: China Shipbuilding Group Co.,Ltd.

Address before: 401122 No. 30 Jinyu Avenue, Jingkaiyuan, North New District of Chongqing

Applicant before: CSIC HAIZHUANG WINDPOWER Co.,Ltd.

TA01 Transfer of patent application right
GR01 Patent grant
GR01 Patent grant