WO2019178974A1 - 风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质 - Google Patents

风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质 Download PDF

Info

Publication number
WO2019178974A1
WO2019178974A1 PCT/CN2018/093013 CN2018093013W WO2019178974A1 WO 2019178974 A1 WO2019178974 A1 WO 2019178974A1 CN 2018093013 W CN2018093013 W CN 2018093013W WO 2019178974 A1 WO2019178974 A1 WO 2019178974A1
Authority
WO
WIPO (PCT)
Prior art keywords
blade
wind speed
fatigue
data
vibration data
Prior art date
Application number
PCT/CN2018/093013
Other languages
English (en)
French (fr)
Inventor
郝吉芳
贾志强
刘芳
Original Assignee
北京金风慧能技术有限公司
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 北京金风慧能技术有限公司 filed Critical 北京金风慧能技术有限公司
Priority to AU2018298513A priority Critical patent/AU2018298513B2/en
Priority to EP18829711.3A priority patent/EP3564526B1/en
Publication of WO2019178974A1 publication Critical patent/WO2019178974A1/zh

Links

Images

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
    • F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
    • F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
    • F05B2260/00—Function
    • F05B2260/80—Diagnostics
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
    • F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
    • F05B2270/00—Control
    • F05B2270/30—Control parameters, e.g. input parameters
    • F05B2270/32—Wind speeds
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
    • F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
    • F05B2270/00—Control
    • F05B2270/30—Control parameters, e.g. input parameters
    • F05B2270/332—Maximum loads or fatigue criteria
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
    • F05B—INDEXING SCHEME RELATING TO WIND, SPRING, WEIGHT, INERTIA OR LIKE MOTORS, TO MACHINES OR ENGINES FOR LIQUIDS COVERED BY SUBCLASSES F03B, F03D AND F03G
    • F05B2270/00—Control
    • F05B2270/30—Control parameters, e.g. input parameters
    • F05B2270/334—Vibration measurements
    • 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

Definitions

  • the present disclosure relates to fault monitoring techniques for wind turbine blades, and in particular to fatigue damage monitoring methods for wind turbine blades, electronic devices, and storage media.
  • the main reasons for blade failure are as follows: 1. There are defects in design. The current design may lack considerations such as wind speed, wind direction, swirl, counter-lift, chattering or counterweight; second, the manufacturing quality is not good, and Transportation, installation, commissioning and other links may cause artificial quality problems. Third, the external environment is complex and variable. After long-term periodic and non-periodic movement of the blade, the material may change inside, causing microscopic damage, and then microscopic damage is indexed. The magnitude increases and eventually visible damage can occur. The current blade anomaly monitoring technology is not applicable in many practical situations, and it is impossible to accurately monitor blade anomalies.
  • the present disclosure provides a fatigue damage monitoring method for a wind turbine blade.
  • the method may include: acquiring real-time vibration data and wind speed data of the unit; extracting operating characteristics and environmental characteristics of the blade according to the vibration data, the wind speed data, and a word frequency-inverse document frequency (TF-IDF) algorithm; a failure mechanism that combines the operational characteristics and the environmental characteristics to obtain a non-periodic load characteristic of the blade; and determines an accumulated fatigue value of the blade over time according to the non-periodic load characteristic and the PM fatigue theory.
  • TF-IDF word frequency-inverse document frequency
  • the present disclosure also provides an electronic device.
  • the electronic device can include a memory for storing a computer program, and a processor for executing a computer program stored in the memory to implement the above method of the present disclosure.
  • the present disclosure also provides a computer readable storage medium having stored thereon a computer program that, when executed, implements the above-described methods of the present disclosure.
  • fatigue damage monitoring of wind turbine blades is performed by combining non-periodic load characteristics with PM fatigue theory, taking into account aperiodic loads that are the root cause of blade failure, This allows for accurate monitoring of blade anomalies.
  • FIG. 1 is a flowchart of a fatigue damage monitoring method of a wind turbine blade according to a first embodiment of the present disclosure
  • FIG. 2 is a flow chart of a fatigue damage monitoring method of a wind turbine blade according to a second embodiment of the present disclosure
  • FIG. 3 schematically illustrates a fatigue curve obtained by a fatigue damage monitoring method of a wind turbine blade according to a second embodiment of the present disclosure
  • FIG. 4 schematically illustrates a fatigue curve obtained by a fatigue damage monitoring method of a wind turbine blade according to a second embodiment of the present disclosure
  • FIG. 5 is a flowchart of a fatigue damage monitoring method of a wind turbine blade according to a third embodiment of the present disclosure
  • FIG. 6 schematically illustrates a load peak ratio curve involved in a fatigue damage monitoring method of a wind turbine blade according to the present disclosure.
  • FIG. 1 schematically illustrates a flow of a fatigue damage monitoring method of a wind turbine blade according to a first embodiment of the present disclosure.
  • real-time vibration data and wind speed data of the unit can be acquired.
  • real-time vibration data and wind speed data of the unit can be obtained by collecting data monitored by the unit sensors.
  • the operating characteristics and environmental characteristics of the blade may be extracted according to the vibration data, the wind speed data, and the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm.
  • the operational characteristics may be characterized by the vibration data, which may be characterized by the wind speed data.
  • the operational characteristics and the environmental features may be combined according to a failure mechanism of the blade to obtain a non-periodic load characteristic of the blade.
  • the blade failure mechanism refers to the long-term periodic and non-periodic motion of the blade, especially the non-periodic motion that may generate a large load on the blade (for example, a small wind speed but a large vibration, turbulence, etc.).
  • the interior of the blade material may change, causing microscopic damage. Over time, the microscopic damage increases in order of magnitude, and eventually visible damage, such as cracks or cracks.
  • the accumulated fatigue value of the blade over time may be determined based on the aperiodic load characteristics and the PM fatigue theory.
  • the accumulated fatigue value or its varying characteristics may characterize the extent of blade fatigue damage. For example, when the accumulated fatigue value exceeds the preset first threshold, the degree of fatigue damage of the blade is large, and the blade of the unit has a large risk of cracking; or, when the rate of change of the accumulated fatigue value exceeds a preset second threshold At the time, it indicates that the degree of fatigue damage of the blade is large, and the blade of the unit has a large risk of cracking.
  • the above threshold may be set or adjusted according to actual conditions or experience. It is not difficult to understand that in practical applications, the degree of fatigue damage of the blade can also be judged based on other characteristics of the accumulated fatigue value.
  • the TF-IDF algorithm is a commonly used statistical method for natural language processing and is commonly used to assess how important a word is to a document set or a document in a corpus. The importance of a word increases as it appears in a document, but at the same time it decreases as its frequency in the corpus increases.
  • the term frequency (TF) refers to the frequency at which a given word appears in the document.
  • the word frequency is usually normalized in some way to prevent it from being biased towards long documents.
  • Inverse Document Frequency is a measure of the universal importance of a word. The IDF of a word can be obtained by dividing the total number of documents by the number of documents containing the word and then taking the logarithm of the ratio.
  • the TF-IDF algorithm can filter out common words and retain important words.
  • the TF-IDF algorithm can be expressed by the following formula (1):
  • W ij represents the TF-IDF weight of the word j in the document i
  • f ij represents the normalized word frequency of the word j in the document i
  • N represents the total number of documents in the corpus
  • n j represents the vocabulary containing the word j The number of documents.
  • the TF-IDF algorithm described above may be used to extract the operating characteristics and environmental characteristics of the blade based on the vibration data and the wind speed data.
  • the vibration data and the wind speed data can be used as words in the TF-IDF algorithm.
  • the measured wind speed may have different values, for example, any value between 1 m/s and 20 m/s.
  • the wind speed data measured in one year can be divided into multiple documents, and the TF-IDF weight of 1 m/s in a document and the TF-IDF weight of 2 m/s in a document are calculated according to the TF-IDF algorithm. So on and so forth.
  • vibration data can be similarly processed. Thereby, the operating characteristics and environmental characteristics of the blade can be extracted.
  • the fatigue damage monitoring of the wind turbine blade takes into consideration the aperiodic load which is the root cause of the blade failure, so that the blade abnormality can be accurately monitored.
  • FIG. 2 schematically illustrates a flow of a fatigue damage monitoring method of a wind turbine blade according to a second embodiment of the present disclosure.
  • the steps 201 to 204 are the same as the steps 101 to 104 described above with reference to FIG. 1 , and details are not described herein again.
  • a fatigue curve can be plotted based on the accumulated fatigue values of the blade over time.
  • FIGS. 3 and 4 respectively illustrate different fatigue curves obtained by the fatigue damage monitoring method of the wind turbine blade according to the second embodiment of the present disclosure, wherein the abscissa represents time (for convenience of drawing, The time is represented by a numerical value, and the ordinate represents the fatigue value.
  • the different curves in the same graph represent the fatigue curves of different units in the same wind field.
  • an early warning may be issued when the morphological characteristics of the plotted fatigue curve meet the preset conditions.
  • the performance of the fatigue curve of the failed blade can be pre-summarized based on a large number of fatigue curves.
  • the first case may be a surge in fatigue value represented by the highest fatigue curve as shown in FIG. 3
  • the second case may be a sharp change in the slope of the highest fatigue curve as shown in FIG.
  • the fatigue curve drawn in step 205 exhibits, for example, the first case or the second case described above, it can be determined that the blade of the corresponding unit has a large risk of cracking, and an early warning is required to prompt the operation and maintenance engineer to pay attention.
  • the performance of the fatigue curve of the above-mentioned failed blade is merely exemplary, and may be set or adjusted according to actual conditions or requirements in practical applications. Of course, in practical applications, it is also possible to determine whether there is a large risk of cracking in the blades of the corresponding unit according to other morphological characteristics of the fatigue curve, which will not be enumerated here.
  • the fatigue curve can be drawn based on the accumulated fatigue value of the blade over time, and an early warning is issued when the morphological characteristic of the fatigue curve satisfies the preset condition. In this way, by observing the fatigue curve drawn, it is possible to determine whether there is a large risk of cracking in the blades of the corresponding unit, so that the blade anomaly can be monitored intuitively and efficiently.
  • FIG. 5 schematically illustrates a flow of a fatigue damage monitoring method of a wind turbine blade according to a third embodiment of the present disclosure.
  • step 501 real-time vibration data and wind speed data of the unit can be acquired.
  • abnormal data processing may be performed on the acquired vibration data and wind speed data to obtain vibration data and wind speed data of the rejected abnormal data samples.
  • the vibration data and the wind speed data of the rejected abnormal data sample may be discretized to obtain discrete vibration data and wind speed data.
  • the acquired raw vibration data and wind speed data are generally continuous values, these continuous values can be discretized for subsequent feature extraction based on the amount of data, the amount of calculation, and the degree of precision. A better way to think about it.
  • the discretization processing method can be adjusted according to specific conditions.
  • operational characteristics and environmental characteristics of the blade may be extracted based on the discrete vibration data and wind speed data and the TF-IDF algorithm.
  • the operational characteristics may be characterized by the vibration data, which may be characterized by the wind speed data.
  • Steps 505 and 506 are the same as steps 103 and 104 described above with reference to FIG. 1, and are not described herein again.
  • the abnormal data samples can be eliminated, and the vibration data and the wind speed data of the abnormal data samples have been discretized. In this way, the accuracy of the data can be ensured, and the calculation efficiency can be improved, so that the blade anomaly can be accurately and efficiently monitored.
  • the real-time vibration data of the unit may include a first nacelle acceleration parallel to the unit generator bearing direction and a second nacelle acceleration perpendicular to the unit generator bearing direction.
  • the step 102, 202 or 504 of extracting the running features and the environmental features of the blade may include: performing window segmentation on the vibration data and the wind speed data in a preset time period; The first cabin acceleration, the second cabin acceleration and the wind speed are applied to the TF-IDF algorithm to obtain the TF-IDF value of the corresponding window; according to the TF-IDF value of each window, the first cabin acceleration in the corresponding window is calculated respectively.
  • the foregoing preset time period can be determined according to actual conditions, and the amount of data is too small to recognize the fatigue condition of the blade. If the above window is too thick, it is not conducive to the feature extraction, and if the rules are too large, the calculation amount is large, the speed is slow, and the calculation cost is high.
  • the specific window segmentation method can be determined according to the actual situation, and will not be described in detail here.
  • step 103, 203 or 505 of obtaining a non-periodic load characteristic of the blade those skilled in the art will appreciate that the greater the load peak ratio of the blade, the greater the aperiodic load the blade may be subjected to.
  • 6 schematically illustrates a load peak ratio curve involved in a fatigue damage monitoring method of a wind turbine blade according to the present disclosure, wherein the abscissa represents time (for convenience of drawing, time is represented by a numerical value), and the ordinate represents The peak load ratio of the blades.
  • the blade can bear the load when the load peak ratio is less than or equal to the preset threshold, and the load peak ratio greater than the preset threshold can be used as the non-periodic load characteristic that affects the blade life.
  • the preset threshold may be determined according to actual conditions, and is not specifically limited herein.
  • the step 103, 203 or 505 of obtaining the aperiodic load characteristic of the blade may include: obtaining a first load peak ratio according to a vector distance of the first cabin acceleration and a vector distance of the wind speed; according to the second cabin The vector distance of the acceleration vector distance from the wind speed is obtained, and the second load peak ratio is obtained; the larger value of the first load peak ratio and the second load peak ratio is selected; when the selected value is greater than the preset threshold, the selected value is determined. It is the non-periodic load characteristic of the blade.
  • the step 104, 204 or 506 of determining the accumulated fatigue value of the blade over time may include obtaining fatigue at the current applied load stress level according to the non-periodic load characteristic of the blade and the constant associated with the blade material. Lifetime; according to the fatigue life and PM fatigue theory, the accumulated fatigue value of the blade with time is obtained.
  • the fatigue life at the current applied load stress level can be calculated using the following formula (2):
  • N i represents the fatigue life at the current applied load stress level
  • S i represents the current aperiodic load characteristic of the blade
  • ⁇ represents a constant associated with the blade material
  • N i the fatigue life at the current applied load stress level
  • n the amount of the variable amplitude load
  • the fatigue life can be combined with the PM fatigue theory, and the formula (2) can be substituted into the formula (3), and the calculation formula (4) of the accumulated fatigue value caused by the non-periodic load of the blade can be obtained:
  • F represents the accumulated fatigue value caused by the non-periodic load of the blade
  • pr i and S i are the same, both indicating the current aperiodic load characteristics of the blade.
  • the non-periodic load characteristic can be obtained according to the load peak ratio of the wind turbine blade, and the non-periodic load feature and the PM fatigue theory are combined to perform the fatigue damage monitoring of the blade.
  • the aperiodic load characteristics determined from the load peak ratio are more accurate, allowing for more accurate monitoring of blade anomalies.
  • embodiments of the present disclosure may also provide an electronic device.
  • the electronic device can include a memory for storing a computer program, a processor for executing a computer program stored in the memory to implement a method in accordance with any of the above-described embodiments of the present disclosure.
  • embodiments of the present disclosure may also provide a computer readable storage medium having stored thereon a computer program that, when executed, may implement a method in accordance with any of the above-described embodiments of the present disclosure.
  • the computer readable storage medium can be, for example, a magnetic disk, an optical disk, a flash memory, or the like.

Landscapes

  • Engineering & Computer Science (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)
  • Wind Motors (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

一种风力发电机叶片的疲劳损伤监测方法，该方法包括：获取机组的实时振动数据和风速数据；根据振动数据、风速数据及TF-IDF算法，提取叶片的运行特征和环境特征；根据叶片的失效机理，对运行特征和环境特征进行组合，以得到叶片的非周期性载荷特征；以及根据非周期性载荷特征和PM疲劳理论，确定叶片随时间的累加疲劳值。同时还公开了用于实现上述方法的电子设备和存储介质。

Description

风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质 技术领域
本公开涉及风力发电机叶片的故障监测技术，具体地涉及风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质。
背景技术
近几年风力发电发展非常快，随之而来的是风电设施的损坏问题，这些问题中叶片开裂损坏是重中之重。
叶片失效的原因主要有以下几点：一、设计方面存在缺陷，目前的设计可能对风速、风向、旋流、逆扬、振颤或配重等缺乏考虑；二、制造质量不精良，以及在运输、安装、调试等环节可能人为造成的质量问题；三、外界环境复杂多变，叶片经过长期的周期性和非周期性运动，其材料内部可能发生变化，产生微观损伤，然后微观损伤以指数量级增大，最终可能产生可见的损伤。目前的叶片异常监测技术在很多实际场合并不适用，无法对叶片异常进行准确的监测。
发明内容
本公开提供一种风力发电机叶片的疲劳损伤监测方法。该方法可以包括：获取机组的实时振动数据和风速数据；根据所述振动数据、所述风速数据及词频-逆文档频率(TF-IDF)算法，提取叶片的运行特征和环境特征；根据叶片的失效机理，对所述运行特征和所述环境特征进行组合，以得到叶片的非周期性载荷特征；以及根据所述非周期性载荷特征和PM疲劳理论，确定叶片随时间的累加疲劳值。
本公开还提供一种电子设备。该电子设备可以包括：存储器，用于存储计算机程序；以及处理器，用于执行所述存储器中存储的计算机程序，以实现本公开的上述方法。
本公开还提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被执行时实现本公开的上述方法。
在本公开的各种实施例中，通过将非周期性载荷特征和PM疲劳理论相结合来进行风力发电机叶片的疲劳损伤监测，其中考虑了作为导致叶片失效的根本原因的非周期性载荷，从而可以对叶片异常进行准确的监测。
附图说明
本公开的上述和/或附加的方面和优点从下面结合附图对实施例的描述中将变得明显和容易理解，在附图中：
图1为根据本公开的第一实施例的风力发电机叶片的疲劳损伤监测方法的流程图；
图2为根据本公开的第二实施例的风力发电机叶片的疲劳损伤监测方法的流程图；
图3示意性地例示了根据本公开的第二实施例的风力发电机叶片的疲劳损伤监测方法所得到的疲劳曲线；
图4示意性地例示了根据本公开的第二实施例的风力发电机叶片的疲劳损伤监测方法所得到的疲劳曲线；
图5为根据本公开的第三实施例的风力发电机叶片的疲劳损伤监测方法的流程图；
图6示意性地例示了根据本公开的风力发电机叶片的疲劳损伤监测方法所涉及的载荷峰值比曲线。
具体实施方式
下面详细描述本公开的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的，仅用于解释本公开，而不应被视为对本公开的限制。
图1示意性地例示了根据本公开的第一实施例的风力发电机叶片的疲劳损伤监测方法的流程。
在步骤101，可以获取机组的实时振动数据和风速数据。具体地，可以通过采集由机组传感器监测的数据而获取机组的实时振动数据和风速数据。
在步骤102，可以根据所述振动数据、所述风速数据及词频-逆文档频率(Term Frequency–Inverse Document Frequency，TF-IDF)算法，提取叶片的运行特征和环境特征。所述运行特征可以由所述振动数据表征，所述环境特征可以由所述风速数据表征。
在步骤103，可以根据叶片的失效机理，对所述运行特征和所述环境特征进行组合，以得到叶片的非周期性载荷特征。叶片的失效机理是指，叶片经过长期的周期性和非周期性运动，特别是可能对叶片产生较大载荷的非周期性运动(例如，风速较小但振动较大的运动、湍流等)，叶片材料内部可能发生变化，产生微观损伤，久而久之，微观损伤以指数量级增大，最终可能产生可见的损伤，例如裂纹或开裂。
在步骤104，可以根据所述非周期性载荷特征和PM疲劳理论，确定叶片随时间的累加疲劳值。所述累加疲劳值或其变化特征可以表征叶片疲劳损伤的程度。例如，当累加疲劳值超过预设的第一阈值时，表示叶片疲劳损伤的程度较大，机组的叶片存在较大的开裂风险；或者，当累加疲劳值的变化率超过预设的第二阈值时，表示叶片疲劳损伤的程度较大，机组的叶片存在较大的开裂风险。这里需要说明的是，在实际应用中，可以根据实际情况或经验总结对上述阈值进行设定或调整。不难理解，在实际应用中，也可以根据累加疲劳值的其他特性来判断叶片的疲劳损伤程度。
为便于理解，下面简要说明TF-IDF算法。
TF-IDF算法是一种用于自然语言处理的常用统计方法，通常用以评估词语对于一个文档集或一个语料库中的一个文档的重要程度。一个词语的重要性会随着它在一个文档中出现的频率增加而上升，但同时会随着它在语料库中出现的频率增加而下降。在一个给定的文档里，词频(Term Frequency，TF)指的是某一个给定的词语在该文档中出现的频率。词频通常会以某种形式被归一化，以防止它偏向长的文档。逆文档频率(Inverse Document Frequency，IDF)是某一词语的普遍重要性的度量。 某一词语的IDF可以由总文档数目除以包含该词语的文档的数目再将得到的比值取对数而得到。某一词语在某一文档内的高词频以及该词语在整个语料库中的低文档频率可以产生高权重的TF-IDF。因此，TF-IDF算法可以过滤掉常见的词语，保留重要的词语。通常，TF-IDF算法可以用如下公式(1)来表示：
Figure PCTCN2018093013-appb-000001
其中，W ij表示词j在文档i中的TF-IDF权重，f ij表示词j在文档i中的归一化的词频，N表示语料库中的文档总数，n j表示语料库中包含词j的文档的数目。
在上述步骤102中，可以根据振动数据和风速数据，采用上述TF-IDF算法来提取叶片的运行特征和环境特征。具体地，可以将振动数据和风速数据作为TF-IDF算法中的词语。所测得的风速可能有不同的数值，例如，可能是1m/s～20m/s之间的任何数值。可以将一年中测得的风速数据分成多个文档，根据TF-IDF算法来计算1m/s在某个文档中的TF-IDF权重，2m/s在某个文档中的TF-IDF权重，依此类推。同理，可以对振动数据进行类似处理。由此可以提取叶片的运行特征和环境特征。
如上所述，在第一实施例中，对风力发电机叶片的疲劳损伤监测考虑了作为导致叶片失效的根本原因的非周期性载荷，从而可以对叶片异常进行准确的监测。
图2示意性地例示了根据本公开的第二实施例的风力发电机叶片的疲劳损伤监测方法的流程。
步骤201～204分别与上面参照图1描述的步骤101～104相同，在此不再赘述。
在步骤205，可以根据叶片随时间的累加疲劳值，绘制疲劳曲线。为便于理解，图3和图4分别例示了根据本公开的第二实施例的风力发电机叶片的疲劳损伤监测方法所得到的不同的疲劳曲线，其中，横坐标表示时间(为了方便作图，时间用数值来表示)，纵坐标表示疲劳值，同一曲线图中不同的曲线表示同一风场中不同机组的疲劳曲线。
在步骤206，当所绘制的疲劳曲线的形态特征满足预设的条件时，可以发出预警。
在实际应用中，可以根据大量的疲劳曲线，预先归纳出失效叶片的疲劳曲线的表现情形。例如，第一种情形可以是如图3中所示的最高的疲劳曲线表示的疲劳值激增，第二种情形可以是如图4中所示的最高的疲劳曲线的斜率急剧变化。
当在步骤205绘制的疲劳曲线表现出例如上述第一种情形或第二种情形时，即可判定相应机组的叶片存在较大的开裂风险，需要发布预警，以便提示运维工程师予以关注。
需要说明的是，上述失效叶片的疲劳曲线的表现情形仅仅是示例性的，在实际应用中可以根据实际情况或需求进行设定或调整。当然，在实际应用中，也可以根据疲劳曲线的其他形态特征，判定相应机组的叶片是否存在较大的开裂风险，此处不再列举。
如上所述，在第二实施例中，可以根据叶片随时间的累加疲劳值来绘制疲劳曲线，并在疲劳曲线的形态特征满足预设的条件时发出预警。这样，通过观测所绘制的疲劳曲线，即可判定相应机组的叶片是否存在较大的开裂风险，从而可以对叶片异常进行直观、高效的监测。
图5示意性地例示了根据本公开的第三实施例的风力发电机叶片的疲劳损伤监测方法的流程。
在步骤501，可以获取机组的实时振动数据和风速数据。
在步骤502，可以对所获取的振动数据和风速数据进行异常值处理，得到已剔除异常数据样本的振动数据和风速数据。
可选地，在步骤503，可以对已剔除异常数据样本的振动数据和风速数据进行离散化处理，得到离散的振动数据和风速数据。在实际应用中，由于所获取的原始振动数据和风速数据一般是连续型数值，可以对这些连续型数值进行离散化处理以便随后执行特征提取，这是基于对数据量、计算量和精确程度的综合考量而得出的较佳方式。当然，在实际应用中，可以根据具体情况调整离散化处理方式。
在步骤504，可以根据所述离散的振动数据和风速数据及TF-IDF算 法，提取叶片的运行特征和环境特征。所述运行特征可以由所述振动数据表征，所述环境特征可以由所述风速数据表征。
步骤505和506分别与上面参照图1描述的步骤103和104相同，在此不再赘述。
如上所述，在第三实施例中，可以剔除异常数据样本，并对已剔除异常数据样本的振动数据和风速数据进行离散化处理。这样，既可以保证数据的准确性，又可以提高计算效率，从而可以对叶片异常进行准确、高效的监测。
此外，在本公开的各种实施例中，机组的实时振动数据可以包括平行于机组发电机轴承方向的第一机舱加速度和垂直于机组发电机轴承方向的第二机舱加速度。在这种情况下，提取叶片的运行特征和环境特征的步骤102、202或504可以包括：对在预设时间段的所述振动数据和所述风速数据进行窗口切分；将每个窗口中的第一机舱加速度、第二机舱加速度和风速应用于TF-IDF算法，得到相应窗口的TF-IDF值；根据每个窗口的TF-IDF值，分别计算相应窗口中的第一机舱加速度、第二机舱加速度和风速相对于全部窗口的矢量距离，其中，第一机舱加速度的矢量距离、第二机舱加速度的矢量距离对应所述运行特征，风速的矢量距离对应所述环境特征。
可以理解，上述预设时间段可以根据实际情况确定，避免数据量太少而无法识别叶片的疲劳状况。上述窗口切分若太粗则不利于特征的提取，而若太细则会导致计算量较大、速度较慢、计算成本较高。具体的窗口切分方式可以根据实际情况确定，这里不做详述。
此外，关于得到叶片的非周期性载荷特征的步骤103、203或505，本领域技术人员可以理解，叶片的载荷峰值比越大，叶片承受的非周期性载荷可能就越大。图6示意性地例示了根据本公开的风力发电机叶片的疲劳损伤监测方法所涉及的载荷峰值比曲线，其中，横坐标表示时间(为了方便作图，时间用数值来表示)，纵坐标表示叶片的载荷峰值比。
在实际应用中，基于叶片的设计和生产，可以认为在载荷峰值比小于等于预设阈值时叶片能够承受载荷，而可以把大于预设阈值的载荷峰值比 作为影响叶片寿命的非周期性载荷特征。该预设阈值可以根据实际情况确定，这里不做具体限定。因此，在实际应用中，得到叶片的非周期性载荷特征的步骤103、203或505可以包括：根据第一机舱加速度的矢量距离与风速的矢量距离，得到第一载荷峰值比；根据第二机舱加速度的矢量距离与风速的矢量距离，得到第二载荷峰值比；选取第一载荷峰值比和第二载荷峰值比中较大的值；当所选取的值大于预设阈值时，确定所选取的值为叶片的非周期性载荷特征。
此外，在实际应用中，确定叶片随时间的累加疲劳值的步骤104、204或506可以包括：根据叶片的非周期性载荷特征和与叶片材料相关的常数，得到当前作用载荷应力水平下的疲劳寿命；根据所述疲劳寿命和PM疲劳理论，得到叶片随时间的累加疲劳值。
具体地，可以采用如下公式(2)计算当前作用载荷应力水平下的疲劳寿命：
Figure PCTCN2018093013-appb-000002
其中，N i表示当前作用载荷应力水平下的疲劳寿命，S i表示叶片的当前非周期性载荷特征，α表示与叶片材料相关的常数。
根据PM理论，非周期性载荷的每一次作用对构件造成的损伤可以线性叠加，如下面公式(3)所表示的：
Figure PCTCN2018093013-appb-000003
其中，D表示构件的损伤程度，N i表示当前作用载荷应力水平下的疲劳寿命，n表示变幅载荷的数量。
可以将所述疲劳寿命与PM疲劳理论结合，将公式(2)代入公式(3)，整理得到叶片的非周期性载荷导致的累加疲劳值的计算公式(4)：
Figure PCTCN2018093013-appb-000004
其中，F表示叶片的非周期性载荷导致的累加疲劳值，pr i与S i指代相同，均表示叶片的当前非周期性载荷特征。
随着时间的推移，非周期性载荷数量可能不断增加，叶片的累加疲劳值也会相应增加，因此可以得到随时间变化的疲劳曲线，如图3、图4所示。
由此可见，根据本公开的实施例，可以依据风力发电机叶片的载荷峰值比得到非周期性载荷特征，再将非周期性载荷特征和PM疲劳理论相结合来进行叶片的疲劳损伤监测。通常，根据载荷峰值比确定的非周期性载荷特征更准确，从而可以对叶片异常进行更准确的监测。
此外，本公开的实施例还可以提供一种电子设备。该电子设备可以包括：存储器，用于存储计算机程序；处理器，用于执行所述存储器中存储的计算机程序，以实现根据本公开的上述任一实施例的方法。
此外，本公开的实施例还可以提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被执行时可以实现根据本公开的上述任一实施例的方法。该计算机可读存储介质可以是例如磁盘、光盘、闪存等。
需要说明的是，这里所描述的本公开的各实施例仅是示例性的，而非限制性的。对于本技术领域的普通技术人员来说，在不脱离本公开的原理的前提下，可以做出若干变型和润饰，这些变型和润饰也应被视为在本公开的范围内。

Claims (10)

  1. 一种风力发电机叶片的疲劳损伤监测方法，包括以下步骤：
    获取机组的实时振动数据和风速数据；
    根据所述振动数据、所述风速数据及词频-逆文档频率(TF-IDF)算法，提取叶片的运行特征和环境特征；
    根据叶片的失效机理，对所述运行特征和所述环境特征进行组合，以得到叶片的非周期性载荷特征；以及
    根据所述非周期性载荷特征和PM疲劳理论，确定叶片随时间的累加疲劳值。
  2. 根据权利要求1的方法，其中，所述振动数据包括平行于机组发电机轴承方向的第一机舱加速度和垂直于机组发电机轴承方向的第二机舱加速度；
    所述提取叶片的运行特征和环境特征的步骤包括：
    对在预设时间段的所述振动数据和所述风速数据进行窗口切分；
    将每个窗口中的第一机舱加速度、第二机舱加速度和风速应用于TF-IDF算法，得到相应窗口的TF-IDF值；
    根据每个窗口的TF-IDF值，分别计算相应窗口中的第一机舱加速度、第二机舱加速度和风速相对于全部窗口的矢量距离，其中，第一机舱加速度的所述矢量距离、第二机舱加速度的所述矢量距离对应所述运行特征，风速的所述矢量距离对应所述环境特征。
  3. 根据权利要求2的方法，其中，所述得到叶片的非周期性载荷特征的步骤包括：
    根据第一机舱加速度的所述矢量距离与风速的所述矢量距离，得到第一载荷峰值比；
    根据第二机舱加速度的所述矢量距离与风速的所述矢量距离，得到第二载荷峰值比；
    选取第一载荷峰值比和第二载荷峰值比中较大的值；
    当所选取的值大于预设阈值时，确定所选取的值为叶片的非周期性载 荷特征。
  4. 根据权利要求1至3中任一项的方法，其中，所述确定叶片随时间的累加疲劳值的步骤包括：
    根据所述非周期性载荷特征和与叶片材料相关的常数，得到当前作用载荷应力水平下的疲劳寿命；
    根据所述疲劳寿命和PM疲劳理论，得到叶片随时间的累加疲劳值。
  5. 根据权利要求1至4中任一项的方法，还包括：
    根据叶片随时间的累加疲劳值，绘制疲劳曲线；
    当疲劳曲线的形态特征满足预设的条件时，发出预警。
  6. 根据权利要求1至5中任一项的方法，还包括：
    在所述提取叶片的运行特征和环境特征的步骤之前，对所获取的振动数据和风速数据进行异常值处理，得到已剔除异常数据样本的振动数据和风速数据。
  7. 根据权利要求6的方法，其中，
    对已剔除异常数据样本的振动数据和风速数据进行离散化处理，得到离散的振动数据和风速数据；以及
    所述提取叶片的运行特征和环境特征的步骤包括：根据所述离散的振动数据和风速数据及TF-IDF算法，提取叶片的运行特征和环境特征。
  8. 根据权利要求1至7中任一项的方法，其中，通过采集由机组传感器监测的数据而获取所述机组的实时振动数据和风速数据。
  9. 一种电子设备，包括：
    存储器，用于存储计算机程序；以及
    处理器，用于执行所述存储器中存储的计算机程序，以实现根据权利要求1至8中任一项的方法。
  10. 一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被执行时实现根据权利要求1至8中任一项的方法。
PCT/CN2018/093013 2018-03-23 2018-06-27 风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质 WO2019178974A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
AU2018298513A AU2018298513B2 (en) 2018-03-23 2018-06-27 Method, electronic device and storage media for monitoring a fatigue damage of a blade of a wind turbine unit
EP18829711.3A EP3564526B1 (en) 2018-03-23 2018-06-27 Method, electronic device and storage media for monitoring a fatigue damage of a blade of a wind turbine unit

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810247182.0 2018-03-23
CN201810247182.0A CN110296053B (zh) 2018-03-23 2018-03-23 风力发电机叶片的疲劳损伤监测方法和装置

Publications (1)

Publication Number Publication Date
WO2019178974A1 true WO2019178974A1 (zh) 2019-09-26

Family

ID=67437022

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/093013 WO2019178974A1 (zh) 2018-03-23 2018-06-27 风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质

Country Status (4)

Country Link
EP (1) EP3564526B1 (zh)
CN (1) CN110296053B (zh)
AU (1) AU2018298513B2 (zh)
WO (1) WO2019178974A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111368459A (zh) * 2020-03-25 2020-07-03 河北振创电子科技有限公司 风力发电支撑塔筒振动疲劳评估方法、装置、介质及终端
CN113357098A (zh) * 2021-05-31 2021-09-07 西安热工研究院有限公司 一种风机变桨子系统故障预警方法
CN116292150A (zh) * 2023-05-23 2023-06-23 三峡智控科技有限公司 一种基于变桨电机转矩异常监测的叶片失效保护方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112855457B (zh) * 2019-11-12 2022-08-12 江苏金风科技有限公司 失速监测系统、方法及叶片
CN112611584B (zh) * 2020-05-18 2023-06-02 北京金风慧能技术有限公司 风力发电机组的疲劳失效检测方法、装置、设备及介质
CN112796957B (zh) * 2021-03-26 2022-08-26 厦门理工学院 一种风机叶片的检测方法和装置以及设备

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4525633A (en) * 1982-09-28 1985-06-25 Grumman Aerospace Corporation Wind turbine maximum power tracking device
WO2010069926A2 (en) * 2008-12-17 2010-06-24 Vestas Wind Systems A/S Fairing for wind turbine blade
CN102968669A (zh) * 2011-08-31 2013-03-13 富士通株式会社 对负荷进行预测的方法和装置
CN105508147A (zh) * 2015-12-28 2016-04-20 山东理工大学 风电叶片单点疲劳加载试验弯矩匹配方法
CN105938481A (zh) * 2016-04-07 2016-09-14 北京航空航天大学 一种城市多模式文本数据的异常检测方法
CN106384163A (zh) * 2016-09-08 2017-02-08 国网能源研究院 一种新能源发电技术成熟度预测方法及装置
CN106762454A (zh) * 2016-12-23 2017-05-31 上海致远绿色能源股份有限公司 一种风力发电机叶片的疲劳测试方法
CN106846170A (zh) * 2016-12-22 2017-06-13 国网上海市电力公司 一种发电机组跳闸监测方法及其监测装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2530302B1 (en) * 2011-06-03 2014-04-30 Wilic S.Àr.L Wind turbine and control method for controlling the same
IL216118A0 (en) * 2011-11-03 2012-02-29 Google Inc Customer support solution recommendation system
CN104515677A (zh) * 2015-01-12 2015-04-15 华北电力大学 风电机组叶片故障诊断与状态监控系统
CN107356384B (zh) * 2017-07-26 2019-07-02 安徽容知日新科技股份有限公司 一种监测风力发电设备中叶片的状态的方法、计算设备和系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4525633A (en) * 1982-09-28 1985-06-25 Grumman Aerospace Corporation Wind turbine maximum power tracking device
WO2010069926A2 (en) * 2008-12-17 2010-06-24 Vestas Wind Systems A/S Fairing for wind turbine blade
CN102968669A (zh) * 2011-08-31 2013-03-13 富士通株式会社 对负荷进行预测的方法和装置
CN105508147A (zh) * 2015-12-28 2016-04-20 山东理工大学 风电叶片单点疲劳加载试验弯矩匹配方法
CN105938481A (zh) * 2016-04-07 2016-09-14 北京航空航天大学 一种城市多模式文本数据的异常检测方法
CN106384163A (zh) * 2016-09-08 2017-02-08 国网能源研究院 一种新能源发电技术成熟度预测方法及装置
CN106846170A (zh) * 2016-12-22 2017-06-13 国网上海市电力公司 一种发电机组跳闸监测方法及其监测装置
CN106762454A (zh) * 2016-12-23 2017-05-31 上海致远绿色能源股份有限公司 一种风力发电机叶片的疲劳测试方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3564526A4 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111368459A (zh) * 2020-03-25 2020-07-03 河北振创电子科技有限公司 风力发电支撑塔筒振动疲劳评估方法、装置、介质及终端
CN111368459B (zh) * 2020-03-25 2023-08-01 河北振创电子科技有限公司 风力发电支撑塔筒振动疲劳评估方法、装置、介质及终端
CN113357098A (zh) * 2021-05-31 2021-09-07 西安热工研究院有限公司 一种风机变桨子系统故障预警方法
CN116292150A (zh) * 2023-05-23 2023-06-23 三峡智控科技有限公司 一种基于变桨电机转矩异常监测的叶片失效保护方法
CN116292150B (zh) * 2023-05-23 2023-08-04 三峡智控科技有限公司 一种基于变桨电机转矩异常监测的叶片失效保护方法

Also Published As

Publication number Publication date
EP3564526A4 (en) 2021-04-21
AU2018298513B2 (en) 2020-07-02
EP3564526A1 (en) 2019-11-06
AU2018298513A1 (en) 2019-10-10
EP3564526B1 (en) 2022-09-14
EP3564526A8 (en) 2020-01-01
CN110296053B (zh) 2020-06-05
CN110296053A (zh) 2019-10-01

Similar Documents

Publication Publication Date Title
WO2019178974A1 (zh) 风力发电机叶片的疲劳损伤监测方法、电子设备和存储介质
US11835030B2 (en) Methods and systems for predicting risk of observable damage in wind turbine gearbox components
JP4995134B2 (ja) 風車の監視装置及び方法並びにプログラム
JP6334130B2 (ja) ブレードの健康状態を監視するための方法およびシステム
WO2011024304A1 (ja) 風車の監視装置及び方法並びにプログラム
Butler et al. A feasibility study into prognostics for the main bearing of a wind turbine
CN103925155A (zh) 一种风电机组输出功率异常的自适应检测方法
EP3642481B1 (en) A method for determining wind turbine blade edgewise load recurrence
CN112459971A (zh) 风力发电机组塔筒的异常振动检测方法
Zhang et al. Detection of emerging faults on industrial gas turbines using extended Gaussian mixture models
Wu et al. Variable forgetting factor identification algorithm for fault diagnosis of wind turbines
Joshuva et al. Split-point and attribute-reduced classifier approach for fault diagnosis of wind turbine blade through vibration signals
WO2019135747A1 (en) Probabilistic life evaluation algorithm for gas turbine engine components
CN112597607A (zh) 风力发电机高速轴承的预测性维修方法及系统
CN116500390A (zh) 一种基于分解产物的绝缘纸老化检测方法及相关装置
CN116032016A (zh) 智慧风电场运行状态在线监测预警系统及预警方法
CN114254904B (zh) 一种风电机组机舱运行健康度评价方法及装置
CN113052716B (zh) 风力发电机组主轴承的异常预警方法及装置
CN112699598A (zh) 齿轮箱油温异常智能诊断方法及装置
CN116664096B (zh) 基于联邦学习的风电螺栓数据处理方法和装置
Joshuva et al. Research on hyper pipes and voting feature intervals classifier for condition monitoring of wind turbine blades using vibration signals
Francisco de Santos et al. Farm-wide fatigue loads estimation: a data-driven approach
CN117128143A (zh) 一种叶片故障检测方法和相关装置
Karlsson Indirect detection of ice on wind turbine rotor
Zhang et al. Prediction of pick wear degradation based on Wiener process

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2018829711

Country of ref document: EP

Effective date: 20190123

ENP Entry into the national phase

Ref document number: 2018829711

Country of ref document: EP

Effective date: 20190115

ENP Entry into the national phase

Ref document number: 2018298513

Country of ref document: AU

Date of ref document: 20180627

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE