CN111859789B - Method for identifying trail of wind driven generator - Google Patents
Method for identifying trail of wind driven generator Download PDFInfo
- Publication number
- CN111859789B CN111859789B CN202010644875.0A CN202010644875A CN111859789B CN 111859789 B CN111859789 B CN 111859789B CN 202010644875 A CN202010644875 A CN 202010644875A CN 111859789 B CN111859789 B CN 111859789B
- Authority
- CN
- China
- Prior art keywords
- data
- flow field
- driven generator
- wind driven
- invariant
- 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
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
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- 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
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- 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)
- Aerodynamic Tests, Hydrodynamic Tests, Wind Tunnels, And Water Tanks (AREA)
Abstract
The invention provides a trail identification method of a wind driven generator, which comprises the steps of firstly collecting all flow field data of a wind driven generator to be tested, then carrying out derivation calculation on the flow field data through a difference method to obtain a velocity gradient tensor, and then calculating flow invariant data on each point of a flow field; analyzing the known data and dividing the collected data into strong turbulence and weak turbulence according to a predetermined standard; using flow invariant data as input quantity, using strong turbulence and weak turbulence data as learning objects, and generating an identifier by machine learning algorithm software; inputting the invariant data of the flow field to be identified into an identifier, and then drawing a data area conforming to the strong turbulence according to a preset standard, namely obtaining a wake area of the wind driven generator generating the current flow field to be identified. The invention uses a big data analysis method in modern computer science, only needs to provide sufficient data without adding other human interference, and can ensure the objectivity of the calculation result.
Description
Technical Field
The invention relates to the field of wind power generation, in particular to an identification method for identifying a trail of any wind power generator by establishing a learning model through known data.
Background
The traditional wind driven generator wake identification usually adopts a mode of 'one physical quantity + one threshold value', for example, for a time-average wake field, a center position can be determined through a maximum point of a speed loss, a wake range can be determined through the speed loss being greater than a certain threshold value, for an instantaneous speed field, the wind driven generator wake can be identified through the vorticity of a flow field and a certain threshold value, and the method has the following two disadvantages:
1. the univariate recognizer has poor robustness, the threshold value needs to be adjusted manually for the trail recognition of different flow parameters, however, the selection of the threshold value is seriously dependent on the subjective cognition of people, and therefore, the situation can not be reflected objectively.
2. Turbulence is a multi-scale physical phenomenon, and a variable can only reflect the flow characteristics under a certain characteristic scale, so that the flow state cannot be comprehensively identified, and the wake of the wind driven generator cannot be accurately identified.
Disclosure of Invention
The invention aims to provide an identification method for establishing a learning model through known data so as to identify the trail of any wind driven generator.
Specifically, the invention provides a trail identification method of a wind driven generator, which comprises the following steps:
step 100, collecting all flow field data in a region affected by a wake at the downstream of a tested wind driven generator, then carrying out derivation calculation on all the collected flow field data through a difference method to obtain a velocity gradient tensor, and then calculating flow invariant data on each point of a flow field;
step 400, inputting the invariant data of the flow field to be identified into an identifier, namely automatically outputting the flow state of each point in the flow field, and then drawing a data region conforming to strong turbulence according to a preset standard, namely obtaining a wake region of the wind driven generator generating the current flow field to be identified.
In an embodiment of the present invention, each of the flow field point data information in all the flow field data includes spatial coordinate information and time information.
In one embodiment of the present invention, all the flow field data is obtained by field measurement or laboratory measurement or numerical simulation.
In one embodiment of the invention, the flow invariants include at least turbulent impulse energy, pseudo-vortex energy, and vortex tensile strength.
In one embodiment of the present invention, the analysis method for analyzing the data with known flow state in all the collected flow field data is a big data analysis method.
In one embodiment of the present invention, said intensity dependent on the yaw influence of the wind turbine is: in the process that the orientation of the wind driven generator is changed along with the wind direction at any time under the influence of the wind direction in space, the change of a wake region of the wind driven generator is compared with that of a peripheral non-wake region in the moving process.
In one embodiment of the present invention, the invariant data required to identify the flow field in step 400 includes invariant data required to identify all known data acquired in the flow field.
In one embodiment of the present invention, the invariant data in the invariant data of the flow field to be identified is calculated by the difference method in step 100.
In an embodiment of the invention, the machine learning algorithm adopts a Python version of an XGBoost mathematical library, the XGBoost mathematical library is called through Python language and a binary tree-based classification model is selected, the invariant is used as model input, the XGBoost algorithm automatically adjusts adjustable parameters existing in the binary tree-based classification model, and finally the difference between the result given by the classification model and the strong turbulence and the weak turbulence is minimized, and then the binary tree-based classification model which is trained to meet the requirement is packaged into an executable file to form the recognizer.
In one embodiment of the present invention, the invariant data is: a physical quantity which can physically maintain objectivity without changing with the selection of the coordinate system.
The invention uses a big data analysis method in modern computer science, only needs to provide sufficient data without adding other human interference, and can ensure the objectivity of the calculation result. The method has the advantages that the multiple variables forming the flow invariants are utilized, the flow state on different characteristic scales of each variable reaction really reflects the multi-scale characteristics of the turbulent flow, each variable has definite physical significance, and different characteristics of the turbulent flow, including unsteady state, vortex structure, vortex stretching and the like, can be reflected, so that the identification of the finally obtained wake is more accurate.
Drawings
FIG. 1 is a flow diagram of a method for trail identification in accordance with an embodiment of the present invention;
FIG. 2 is a schematic view of a wind turbine flow field according to an embodiment of the present invention.
Detailed Description
The method is based on the modern computer data analysis technology and identifies the trail area of the wind driven generator, has more accurate and objective advantages compared with the traditional method, is vital to realizing advanced fan control and optimizing the arrangement of the wind driven generator, and is a key technology for improving the generating efficiency of the wind power plant.
According to the scheme, the big data analysis method is applied to the field of wind power, and the value of the invariant is equal to the color of the picture in the image recognition technology by combining the image recognition technology and the physical recognition process, so that the physical phenomenon is recognized. Compared with the image recognition technology, the physical phenomenon recognition processed by the scheme is more complex, each pixel point in the image can be expressed only by RGB (red, green and blue) three primary colors (equivalent to three variables), and the physical phenomenon of processing the trail can objectively reflect the characteristics of the trail by more variables.
In the prior art, when trail analysis is processed, analysis is generally performed by calibrating a threshold value of each variable, and the calibration process can only adopt a trial and error method, namely, the identified trail region is close to the cognition of a researcher as much as possible by repeatedly changing the threshold value. The comprehensive calibration of the threshold values of a plurality of variables is not realized in the existing method even if the cognition of a researcher is not considered to be doped with subjective factors, because the threshold value setting of each variable needs to be calibrated when the plurality of variables are considered comprehensively, and because orthogonality does not exist among the variables, the threshold values are mutually influenced, and the value of the physical quantity is continuously changed, theoretically, the selectable threshold value is infinite, so when the number of the variables is more than 3, the multivariate analysis is almost impossible by using the existing identification method, the existing technology adopts 2 variables to analyze the trail, and the accuracy of the finally obtained trail data is far lower than that of the processing change data of the scheme.
In addition, the adopted machine learning belongs to a developing technology, and the XGboost mathematical algorithm adopted by the scheme is a learning supervision algorithm which is only applied to the object segmentation problem in image recognition and is not widely applied in the industry. The method uses the supervised learning characteristic of the XGboost, and well solves the problems of objectivity and accuracy in identifying the trail of the wind driven generator.
The detailed structure and implementation process of the present invention are described in detail by the following embodiments and the accompanying drawings.
As shown in fig. 1, a method for identifying a wind turbine wake in an embodiment of the present invention includes the steps of:
step 100, collecting all flow field data in a region affected by a trail at the downstream of a wind driven generator to be tested, then carrying out derivation calculation on all the collected flow field data through a difference method to obtain a velocity gradient tensor, and then calculating flow invariant data on each point of a flow field;
in this step, all the flow field data may be obtained by field measurement (for example, light Detection and Ranging), or laboratory measurement (for example, PIV (Particle Image metrology)), or may be numerical simulation.
The collected flow field data downstream of the wind turbine is specifically denoted as u i (x i T), where x i Indicating the position of the measurement point in space, subscript i =1,2,3 indicating three directions of the three-dimensional space, and t indicating time.
Data u of convection field i (x i T) calculating the derivative by a difference method to obtain a velocity gradient tensor
Then calculating the flow invariants at each flow field point, including turbulent flow pulse energy k = u i u i 2, pseudo-vortex energy omega i ω i (ii)/2. Vortex tensile Strength S ij ω j Etc., wherein>
And &>
Respectively representing vorticity and strain rate tensor, ∈ ijk Representing the third order column of the vicat-odd vicat symbol.
the strength according to the influence of the yaw of the wind turbine is as follows: in the process that the orientation of the wind driven generator is changed along with the wind direction at any time under the influence of the wind direction in space, the wake area of the wind driven generator is compared with the change of the peripheral non-wake area in the moving process.
Among all the collected flow field data, the analysis method used for analyzing the data whose flow state is known is a big data analysis method, and the specific analysis is described in the following steps.
As shown in fig. 2, since various flow field data are known, data whose flow state is known can be marked. In the figure, the solid line is the lateral boundary region of the wake, and the state of the clamped region is necessarily strong turbulence due to the influence of the wake of the wind turbine, so that a label value of L =1 can be given, while the region far from the wind turbine (the region surrounded by the dashed line box) is necessarily a weak turbulence or non-turbulence region, so that a label value of L =0 is given. The pre-calibration criterion is divided according to the influence of the wind turbine yaw, and can be determined according to the analysis requirement of the actual wake.
the machine learning algorithm adopts a Python version of the XGboost mathematical library, the XGboost mathematical library is called through Python language, a binary tree-based classification model is selected, the invariant obtained in the previous step is used as model input, the XGboost algorithm automatically adjusts adjustable parameters in the binary tree-based classification model, finally, the difference value between the result given by the classification model and the strong turbulence and the weak turbulence obtained through actual measurement is minimized, and then the binary tree-based classification model which meets the training requirement is packaged into an executable file to form the recognizer.
The XGboost method is a big data analysis method, but the objectivity of wake identification cannot be guaranteed if the method is directly applied to original measured data (such as speed and pressure), because the flow speed is not invariant, the size of the flow speed depends on the selection of a coordinate system, the wind direction of an actual wind power station changes at any time, the direction of a wind driven generator also changes along with the wind direction, and the coordinate system of flow field measurement also changes continuously, so the scheme uses invariant data as input to identify the wake.
The invariant data physically refers to a physical quantity which does not change with the selection of the coordinate system, and has physical objectivity.
Step 400, inputting the invariant data of the flow field to be identified into an identifier, namely automatically outputting the flow state of each point in the flow field, and then drawing a data region conforming to strong turbulence according to a preset standard, namely obtaining a wake region of the wind driven generator generating the current flow field to be identified.
The invariant data needed to identify the flow field refers to: invariant data identifying all known data acquired in the flow field is required. And the invariant data is calculated by the difference method in step 100.
The identifier can be applied to the calculation of the trail of any wind driven generator after being obtained, and the identifier can automatically output the flowing state of each point in the flow field, namely the label value L, only by taking the invariant of each point in the corresponding flow field as input * . Tag value L * The flow field area of =1 is the wake area of the wind turbine.
The invention uses a big data analysis method in modern computer science, only needs to provide sufficient data without adding other human interference, and can ensure the objectivity of the calculation result. The method has the advantages that the multiple variables forming the flow invariants are utilized, the flow state on different characteristic scales of each variable reaction really reflects the multi-scale characteristics of the turbulent flow, each variable has definite physical significance, and different characteristics of the turbulent flow, including unsteady state, vortex structure, vortex stretching and the like, can be reflected, so that the identification of the finally obtained wake is more accurate.
Thus, it should be appreciated by those skilled in the art that while a number of exemplary embodiments of the invention have been illustrated and described in detail herein, many other variations or modifications consistent with the principles of the invention may be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Accordingly, the scope of the invention should be understood and interpreted to cover all such other variations or modifications.
Claims (10)
1. A method for identifying a trail of a wind driven generator is characterized by comprising the following steps:
step 100, collecting all flow field data in a region affected by a trail at the downstream of a wind driven generator to be tested, then carrying out derivation calculation on all the collected flow field data through a difference method to obtain a velocity gradient tensor, and then calculating flowing invariant data on each point of a flow field through the difference method;
step 200, analyzing the data with known flow state in all the collected flow field data according to the strength influenced by the yaw of the wind driven generator, and dividing the collected data into strong turbulence and weak turbulence according to a preset standard;
step 300, using flow invariant data as input quantity, using strong turbulence and weak turbulence data as learning objects, and generating an identifier through machine learning algorithm software;
step 400, inputting the invariant data of the flow field to be identified into an identifier, namely automatically outputting the flow state of each point in the flow field, then drawing a data region conforming to strong turbulence according to a preset standard, namely obtaining a wake region of the wind driven generator generating the current flow field to be identified,
wherein the predetermined criteria is divided according to the magnitude of the wind turbine yaw effect.
2. The trail identification method according to claim 1,
and each flow field point data information in all the flow field data comprises space coordinate information and time information.
3. The trail identification method according to claim 1,
all the flow field data are obtained by field measurement or laboratory measurement or numerical simulation.
4. The trail identification method according to claim 1,
the flow invariants include at least turbulent pulsating energy, pseudo-vortex energy and vortex tensile strength.
5. The trail identification method according to claim 1,
the analysis method for analyzing the data with known flow state in all the collected flow field data is a big data analysis method.
6. The trail identification method according to claim 1,
the strength influenced by the yaw of the wind driven generator is as follows: in the process that the orientation of the wind driven generator is changed along with the wind direction at any time under the influence of the wind direction in space, the wake area of the wind driven generator is compared with the change of the peripheral non-wake area in the moving process.
7. The trail identification method according to claim 1,
the invariant data needed to identify the flow field in step 400 includes invariant data of all known data acquired in the flow field.
8. The trail identification method according to claim 7,
the invariant data in the invariant data of the flow field to be identified is calculated by the difference method in step 100.
9. The trail identification method according to claim 1,
the machine learning algorithm adopts a Python version of the XGboost mathematical library, the XGboost mathematical library is called through a Python language, a binary tree-based classification model is selected, the invariant is used as model input, the XGboost algorithm automatically adjusts adjustable parameters in the binary tree-based classification model, finally the difference between the result given by the classification model and the strong turbulence and weak turbulence is minimized, and then the training required binary tree-based classification model is packaged into an executable file to form the recognizer.
10. The trail identification method according to claim 1,
the invariant data refers to: a physical quantity which can physically maintain objectivity without changing with the selection of the coordinate system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010644875.0A CN111859789B (en) | 2020-07-07 | 2020-07-07 | Method for identifying trail of wind driven generator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010644875.0A CN111859789B (en) | 2020-07-07 | 2020-07-07 | Method for identifying trail of wind driven generator |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111859789A CN111859789A (en) | 2020-10-30 |
| CN111859789B true CN111859789B (en) | 2023-04-14 |
Family
ID=73152385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010644875.0A Active CN111859789B (en) | 2020-07-07 | 2020-07-07 | Method for identifying trail of wind driven generator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111859789B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105041572A (en) * | 2014-04-29 | 2015-11-11 | 通用电气公司 | Systems and methods for optimizing operation of a wind farm |
| CN109885845A (en) * | 2017-12-06 | 2019-06-14 | 北京普华亿能风电技术有限公司 | A kind of wake analysis method and system of wind-driven generator |
| CN110543649A (en) * | 2018-05-29 | 2019-12-06 | 北京金风科创风电设备有限公司 | fan arrangement method and device based on rapid evaluation fluid model and wake flow model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6963291B2 (en) * | 2002-05-17 | 2005-11-08 | The Board Of Trustees Of The Leland Stanford Junior University | Dynamic wake prediction and visualization with uncertainty analysis |
-
2020
- 2020-07-07 CN CN202010644875.0A patent/CN111859789B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105041572A (en) * | 2014-04-29 | 2015-11-11 | 通用电气公司 | Systems and methods for optimizing operation of a wind farm |
| CN109885845A (en) * | 2017-12-06 | 2019-06-14 | 北京普华亿能风电技术有限公司 | A kind of wake analysis method and system of wind-driven generator |
| CN110543649A (en) * | 2018-05-29 | 2019-12-06 | 北京金风科创风电设备有限公司 | fan arrangement method and device based on rapid evaluation fluid model and wake flow model |
Non-Patent Citations (7)
| Title |
|---|
| Georgios Deskos 等.Turbulence-resolving simulations of wind turbine wakes.《Renewable Energy》.2019,第第134卷卷第989-1002页. * |
| 何伟 ; 田德 ; 邓英 ; 汪宁渤 ; .风力发电机组旋转湍流风场数值模拟.中国电机工程学报.2013,(第11期),全文. * |
| 曹娜 ; 于群 ; 王伟胜 ; 石立宝 ; .风电场尾流效应模型研究.太阳能学报.2016,(第01期),全文. * |
| 杨晓雷.单一水平轴风电机组尾迹的模拟方法与流动机理研究综述.《力学学报》.2021,第第53卷卷(第第53卷期),第3169-3178页. * |
| 杨祥生.风力机尾流效应数值模拟研究.《中国博士学位论文全文数据库（电子期刊）》.2017,(第undefined期),C042-28. * |
| 章书成 ; 东雪青 ; 汪建文 ; 高志鹰 ; 陈天阁 ; 杨超 ; 管飞 ; .风力机尾迹速度场脉动特性的实验研究.工程热物理学报.2016,(第07期),全文. * |
| 郭静婷 ; .用CFD方法模拟计算风力机尾流的研究.可再生能源.2010,(第04期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111859789A (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021155650A1 (en) | Image recognition model training method and apparatus, computer system, and storage medium | |
| CN108182433A (en) | A kind of meter reading recognition methods and system | |
| CN109508740B (en) | Object hardness identification method based on Gaussian mixed noise production confrontation network | |
| KR20100065677A (en) | Method for detection of multi moving objects in the high resolution image sequences and system thereof | |
| CN109284779A (en) | Object detection method based on deep full convolution network | |
| CN113259331A (en) | Unknown abnormal flow online detection method and system based on incremental learning | |
| CN112365497A (en) | High-speed target detection method and system based on Trident Net and Cascade-RCNN structures | |
| CN112069787A (en) | Log parameter anomaly detection method based on word embedding | |
| CN111553373B (en) | CNN + SVM-based pressure bubble image recognition algorithm | |
| CN114503131A (en) | Search device, search method, search program, and learning model search system | |
| CN111859789B (en) | Method for identifying trail of wind driven generator | |
| CN206897873U (en) | A kind of image procossing and detecting system based on detection product performance | |
| CN115493740B (en) | Method and system for measuring pressure pulsation of internal flow passage of water turbine by using machine vision | |
| CN116542254A (en) | Wind tunnel test data anomaly decision method and device, electronic equipment and storage medium | |
| CN109977797A (en) | The optimization method of single order object detector based on sequence loss function | |
| Zhang et al. | Digital instruments recognition based on PCA-BP neural network | |
| Lan et al. | PCB defect detection algorithm of improved YOLOv8 | |
| CN109800854A (en) | A kind of Hydrophobicity of Composite Insulator grade determination method based on probabilistic neural network | |
| CN115100451A (en) | Data expansion method for monitoring oil leakage of hydraulic pump | |
| CN114818945A (en) | Small sample image classification method and device integrating category adaptive metric learning | |
| CN113838008A (en) | Abnormal cell detection method based on attention-drawing mechanism | |
| CN113537339A (en) | Method and system for identifying symbiotic or associated minerals based on multi-label image classification | |
| CN112255141A (en) | Thermal imaging gas monitoring system | |
| JP2020144619A (en) | Abnormality detecting device and abnormality detecting method | |
| CN117798654B (en) | Intelligent adjusting system for center of steam turbine shafting |
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 |