KR20220132847A - Prediction Method and Systems of Wind Turbine Load and Life-time Based On Vibration Data - Google Patents

Abstract

The present invention relates to a method and a system for predicting the load and life-time of a wind turbine based on vibration, which predict the load and life-time of the wind turbine from vibration data. The present invention is to provide a load and life-time prediction system which can predict a load and a life-time through a learning model for estimating a load using a vibration sensor which can be relatively easily installed, instead of a conventional strain sensor which is used for estimating a load but is sensitive to electromagnetic noise and difficult to install. In addition, it is possible to prepare for repair and replacement times of parts through load prediction and to facilitate the operation of wind turbines and wind power generators. The present invention includes a collection step for collecting vibration data and load data, a modeling step for assuming a correlation between the collected vibration data and load data, a training step for increasing accuracy by inputting continuous vibration data and load data into the assumed model, and a prediction step for predicting a load or life-time using the trained model and the vibration data.

Description

{Prediction Method and Systems of Wind Turbine Load and Life-time Based On Vibration Data}

The present invention relates to a method and system for predicting the load and lifespan of a wind turbine, and more particularly, to a method and system for predicting the load and lifespan by collecting various data of a wind turbine.

Recently, interest in global environmental problems due to global warming is increasing internationally. In addition, concerns about the safety of nuclear power plants have been amplified as radioactivity leaked due to the destruction of the Fukushima nuclear power plant in the Great East Japan Earthquake in March 2011. There are many issues such as controversy.

As a result, wind power generation and solar energy, which are new and renewable energies, are on the rise, and in recent years, research for the enlargement of wind turbines for large-scale energy production and the development of offshore wind power generation complexes are being actively conducted. However, overturning damage to wind turbine towers is also frequently occurring due to super-large typhoons and gusts caused by global warming. Moreover, in the case of offshore wind turbine towers located in the sea, such as offshore wind farms, unlike other thermal power or nuclear power generation facilities, a large number of facilities are widely distributed compared to power generation capacity, and when performing maintenance work, It is a facility that is not easy to manage because it is affected by accessibility parts, supply, supply, and the number of workers. In the event of an abnormality, access restrictions may occur depending on weather conditions and sea conditions, which may cause secondary damage as well as huge cost losses. A conventional monitoring system detects a change in state of a main part by detecting a change in state of the main part to provide an alarm and a warning.

Wind turbines must not undergo permanent deformation under limit loads or breakage under extreme loads. To verify this, a structural test is performed, and the structural stability of the wind turbine is verified by measuring the strain. At this time, the sensor used to measure the strain is a strain gauge. The signal measured by the sensor is transmitted through the lead wire, and the transmitted signal has a very low level and is amplified by the amplifier. Although the signal level is low, these signals have a problem of causing inaccuracy in the measurement signal due to the penetration of noise by the electrostatic field (electric field) and electromagnetic field (magnetic field) of the lead wire. In addition, there is a lot of wiring when configuring the measurement system, and it is installed in a narrow location, so it is difficult to continuously operate the measurement system.

The present invention has been devised to solve the above-described problems, and an object of the present invention is to predict or model data measured by a strain sensor from vibration data, which is relatively easy to install and maintain, To provide a method and system for estimating load and life. The condition monitoring system based on this monitors and diagnoses the condition of mechanical parts of the wind turbine.

In particular, offshore wind power generation has restrictions on access to wind power generators according to wind speed and wave height when a failure occurs, so it is essential to detect component failures early and establish a maintenance plan to prevent serious accidents in order to reduce maintenance costs.

The present invention relates to a collection step of collecting vibration data and load data, which are system responses, based on operating data, a modeling step assuming a correlation between the collected vibration data and load data after the collection step, and a modeling step, after the assumed model. It is characterized in that it comprises a learning step of learning to increase accuracy by injecting continuous vibration data and load data, and a prediction step of predicting load or life using the learned model and vibration data after the learning step.

In addition, the collecting step is characterized in that the measurement of environmental factors including temperature, atmospheric pressure and wind speed affecting the vibration.

In addition, the modeling step is characterized in that the vibration and load change due to the collected environmental factors are assumed.

In addition, after the prediction step, it characterized in that it further comprises an operation step of operating the wind power generator through the predicted load.

In addition, the operation step is characterized in that the maintenance plan by collecting vibration data from wind turbines installed in different locations.

In addition, the operation step is characterized in that the maintenance cycle and trend are analyzed through the accumulated failure-related data, and failure prediction due to fatigue is performed through the real-time fatigue analysis.

In addition, a vibration sensor for measuring the vibration of the wind turbine, a collection unit including a strain sensor for measuring the load of the wind turbine, the data collected in the collection unit is a predictive model and a learning module for estimating the load with the vibration data It is characterized in that it includes a main server that includes.

In addition, the collection unit is characterized in that it further comprises an anemometer, an atmospheric pressure gauge and a thermo-hygrometer that affects the vibration of the wind turbine.

In addition, the prediction model is characterized in that the load is predicted by correcting the vibration change according to wind speed, atmospheric pressure, and temperature and humidity.

In addition, the wind turbine is a generator that generates power through the rotation of the blades, a gearbox that increases the rotational speed of the blades to transmit power to the generator, and is positioned between the blades and the speedup and fixes the shaft at a fixed position, It includes a main bearing rotating while supporting a load, a coupling positioned between the gearbox and the generator and transmitting rotational power, a tower supporting the entire structure, and a substructure in the case of offshore wind power, wherein the vibration sensor is It is characterized in that it is attached to the component surface of the wind turbine.

According to the present invention, the load can be predicted using vibration data that is relatively easier to install than the strain sensor.

In addition, maintenance and life prediction are possible through predicted data.

In addition, it is easy to establish an operation plan for a wind farm in which a plurality of wind power generators are installed.

In addition, it can be applied not only to wind turbines but also to parts that generate vibration.

1 is a flowchart of the present invention;
2 is a block diagram of the present invention;
3 is a block diagram of the collection step of the present invention;
4 is a position diagram of the sensor of the present invention;
5 is an operational flowchart of the present invention;
6 is a wind turbine operation block diagram
7 is a wind turbine operation flow chart

As described above, in the conventional wind turbine load and life prediction, the load and lifespan are predicted by installing a strain gauge, but the strain gauge is sensitive to electromagnetic noise, so it is difficult to install. An object of the present invention is to estimate the load and lifespan of a wind turbine using a vibration sensor that is relatively easy to install.

Hereinafter, a vibration-based wind turbine load and life prediction method and system according to the present invention having the configuration as described above will be described in detail with reference to the accompanying drawings.

[1] Load prediction method of the present invention

First, Figure 1 is a flowchart of the present invention. Referring to FIG. 1 , the load and lifetime prediction of a wind turbine based on vibration are shown in stages. First, the collection step (S1) is a step of simultaneously measuring the system response vibration data and load data based on the actual operation data. The modeling step (S2) performed after the collection step (S1) is a step of creating a learning model through the correlation between the collected vibration data and the load data. After the modeling step (S2), the learning step (S3) is a step of completing the model by injecting continuous vibration data and load data. After the learning step (S3), the load and lifespan of the wind turbine are predicted through the prediction step (S4) of estimating the load and lifespan when the vibration data is injected into the learned model.

In this case, in the collection step (S1), vibration data is collected by a vibration sensor attached to the wind turbine, and the load data can be simultaneously measured by a factor affecting vibration using a strain gauge sensor. At this time, the measurement variable is for diagnosing the state of the wind turbine and may be determined by the user. Measured elements are values such as vibration, temperature, voltage, current, rotation speed, noise, atmospheric pressure, etc. measured from a plurality of sensors independently installed according to the parts and positions of the wind turbine, or input to the wind turbine including the wind turbine. Values such as wind speed, wind direction, etc. may also be included.

In the modeling step (S2), a predictive model is built for each part included in the wind turbine from the collected data. It is made based on the relationship between the vibration data and the load data through the data obtained in the collection step (S1). For example, when the vibration change according to the load change is collected and the load measured value is a1, and the vibration measured value measured when a1 is measured is b1, it is modeled to correspond to each other.

In the data learning step (S3), the model generated in the modeling step (S2) continuously acquires data collected from a plurality of parts to correct the relationship between vibration and load. In addition, the model is modified by additionally considering factors such as temperature, wind speed, noise, atmospheric pressure, and rotation speed that affect the measured vibration. For example, generator A installed in an environment with an average wind speed of 4 m/s had a vibration value of b1 when the load measurement value was a1, but generator B installed in an environment with an average wind speed of 6 m/s had a load measurement value. When this a1, the vibration value can be measured as b2. This is because, depending on the wind speed, the vibration generated from the tower of the generator may lead to the turbine, and the vibration sensor attached to each part may be affected. Like wind speed, factors such as rotation speed and noise act on the relationship between vibration and load.

In the prediction step (S4), vibration data is measured through the operation of the wind turbine and input to the prediction model, and the predicted load or lifespan of each part is calculated from the vibration data. For example, the vibration measurement value up to 100Hz is first processed and converted into a vibration index (RMS, vibration magnitude for each characteristic frequency), and it is classified into temperature, wind speed, atmospheric pressure, and rotation speed that affect vibration, and input this into the prediction model. to calculate the predicted load or life. In the prediction step, the state is determined by considering various measured values, rather than predicting the state using only the measured variables obtained for one part.

In order to predict the lifespan of a wind turbine that is already in operation, the accumulated load (Lc) from the past to the present is predicted by inputting historical vibration and operation data into the developed load prediction model. If the accumulated load on the wind turbine operated for nh hour and nm minute on nM month nd day of ny year ny is predicted, the average load in units of year/month/day/hour/minute is as follows. However, from the start of wind turbine operation to the date of cumulative load evaluation, the number of months is Mt, the number of days is Dt, the past hour is Ht, and the last minute is mt.

Average annual load (Ly)

= Lc / (ny + nM/12 + nd/365 + nh/365/24 + nm/365/24/60)

= Lc / Yt

Average load per month (LM) = Lc/Mt

Average daily load (Ld) = Lc/Dt

Average load per hour (Lh) = Lc / Ht

Average load in minutes (Lm) = Lc/mt

From each average load, the extra load Lr can be calculated as follows.

1) If the design life of a wind turbine is Y years, (Y-Yt) * Ly

2) If the design life of a wind turbine is M months, (M-Mt) * LM

3) If the design life of a wind turbine is D days, (D-Dt) * Ld

4) If the design life is considered as h hours, (h-Ht) * Lh

5) Considering the design life in m minutes, (m-mt) * Lm

After the excess load is secured, if the vibration data of the wind turbine is continuously acquired and the load is predicted, the remaining life expectancy is estimated as follows.

( Lr ?? Ln ) / Lr * 100

However, Ln is the cumulative load applied to the wind turbine after the extra load was estimated.

2 is a block diagram of the present invention. A vibration-based wind turbine load and life prediction system using a vibration-based wind turbine load and life prediction method will be described. Wind turbines include blades because they are affected by vibrations generated by blades, towers, and substructures. Referring to FIG. 2 , the wind power generator 100 includes a blade 120 rotating by wind, a turbine 130 generating energy by the blade 120 , and a control unit controlling the turbine 130 to rotate at a constant speed. 110 , and a communication module 140 capable of transmitting and receiving with the outside is included, and the tower 150 and the substructure 160 that are the base of the wind power generator 100 are included. The part additionally includes a collection unit 170 for collecting data required for modeling. The collection unit 170 includes a weather sensor 171 for collecting external environmental conditions, a strain sensor 172 and a vibration sensor 173 serving as modeling data, a noise sensor 174 and a temperature sensor 175 .

At this time, the collected data is stored in the main server 200 , and the main server 200 includes a predictive model 210 , a learning module 220 , and a database 230 . The data measured by the collecting unit 150 is accumulated in the main server 200 . The accumulated data estimates the load as vibration data by the prediction model 210 and the learning module 220 .

3 is a block diagram of a wind power generator of the present invention. Referring to FIG. 3 , the wind power generator 100 includes a blade 120 rotating by wind energy, a turbine 130 generating electricity through the blade, and a controller 110 controlling the angle of the blade and parts necessary for power generation. ), the communication module 140 connected to the central server, the tower 150 and the substructure 160 that are the base of the wind power generator 100 are included. Among them, the collecting unit 170 for predicting the load of the wind turbine is included, and the collecting unit 170 includes the weather sensor 171 for collecting external environmental conditions, the strain sensor 172 for measuring the load, and the present invention. A vibration sensor 173 serving as a base, a noise sensor 174 affecting the vibration of the wind turbine, an anemometer and a temperature sensor 175 are attached.

At this time, the collection unit 170 simultaneously measures the vibration data and the load data based on the actual operation data, and the load measurement may be measured using a strain gauge. In the collection unit 170, a sensor is additionally installed to measure environmental factors such as wind speed and temperature and humidity that affect the vibration of the wind turbine.

4 is a view showing the positions of the sensors for monitoring the vibration state of the wind turbine. Referring to FIG. 4 , the main bearing 131 and the speed increaser 132 connected to the blade 120 increase the rotation speed of the blade 120 so as to rotate to a rotation speed suitable for power generation of the generator 134 . carry out The gearbox 132 may use a gearbox used in a typical wind turbine. On the other hand, the gearbox 132 is configured as a medium speed gearbox that does not generate vibration by performing heat treatment on the gear and a gearbox that generates vibrations because heat treatment is not performed, so that the data of the wind turbine is collected separately. Energy is transmitted to the generator 134 by the coupling 133 connected to the speed increaser 132 . The generator 134 serves to convert the rotational force generated by the wind turbine into electrical energy. When energy is transferred from the blade 120 to the generator 134, vibration is generated in each component. In order to predict the load of each part, a number of sensors (S) are attached and are attached horizontally or vertically to measure the vibration. In order to monitor the vibration state, a plurality of acceleration sensors capable of detecting a change in speed and a laser tachometer capable of measuring the rotational speed of an object are installed. It collects vibration data through the attached sensor and makes a learning model to estimate the load and lifespan, and to help with the maintenance of the wind turbine.

Additionally, since the wind turbine is directly affected by the vibration of the blade 120 , when a sensor is attached to the blade 120 to generate abnormal vibration of the blade, it can be excluded from the vibration data of the wind turbine.

[2] Example of operation method using the present invention

5 is a flowchart including operating steps of the present invention. Referring to FIG. 5 , the load prediction of the wind turbine based on vibration is illustrated in stages. After the collection step (S1) of simultaneously measuring the system response vibration data and load data based on the actual operation data, the modeling step (S1) to create a learning model through the correlation between the collected vibration data and load data ( After the modeling step (S2), the learning step (S3) to complete the model by injecting continuous vibration data and load data after the modeling step (S2), after the learning step (S3), the load and lifespan are estimated when the vibration data is injected into the learned model The load and life of the wind turbine are predicted through the prediction step (S4). It includes an operation step (S5) of operating a wind turbine and a wind power plant based on the predicted load data.

In the operation stage, the wind turbine is evaluated based on the predicted load and lifespan. In a wind farm, the evaluation of the lifespan of an individual wind turbine is essential for the establishment of a maintenance and repowering plan through the operation of the wind turbine.

6 is a block diagram of a wind turbine operation according to the present invention. Referring to FIG. 4 , a plurality of wind turbines installed at different locations collects vibration data and load data from attached sensors, and accumulates a plurality of data in the main server through a communication module. The learning model learns the data accumulated in the main server, and when vibration data is input to the learned model, the load and lifespan are predicted. By synthesizing the predicted load data, it is helpful for wind turbine maintenance as well as wind turbine maintenance. At this time, the input vibration data may be injected vibration data in the database.

7 is a flowchart of a wind turbine operation step. Referring to FIG. 7 , the system operator monitors the state of the wind turbine through vibration, temperature, humidity, and fatigue sensors in the system to be analyzed, and collects and classifies failure-related data. Among the classified data, the accumulated data is used to analyze the maintenance cycle and trends, and the data to analyze the fatigue in real time is used to analyze the failure prediction according to temperature and humidity and vibration fatigue. It predicts the next failure and maintenance time and provides feedback to the system operator so that the operation plan can be efficiently performed.

At this time, the sensor is installed not only on the wind turbine but also on the rotor blade, tower, substructure, etc. to estimate the load and evaluate the lifespan of each structure.

The present invention is not limited to the above-described embodiments, and the scope of application is varied, and anyone with ordinary knowledge in the field to which the present invention pertains without departing from the gist of the present invention as claimed in the claims It goes without saying that various modifications are possible.

100: wind generator
110: control unit 120: blade 130: turbine
131: main bearing 132: gearbox 133: coupling
134: generator
140: communication module 150: tower 160: substructure
170: collection unit
171: weather sensor (wind speed/wind direction/atmospheric pressure/humidity) 172: strain sensor
173: vibration sensor 174: noise sensor 175: temperature sensor
220: main server
230: predictive model 240: learning module 250: database

Claims (10)

a collection step of collecting vibration data and load data that are system responses based on the driving data;
After the collection step, a modeling step of assuming a correlation between the collected vibration data and load data;
a learning step of learning to increase accuracy by continuously injecting vibration data and load data into the assumed model after the modeling step; and
After the learning step, a prediction step of predicting load and life using the learned model and vibration data;
Vibration-based wind turbine load and life prediction method comprising a.
The method of claim 1,
The collecting step is a vibration-based wind turbine load and life prediction method, characterized in that for measuring environmental factors including temperature, atmospheric pressure, and wind speed affecting the vibration.
3. The method of claim 2,
The modeling step is a vibration-based wind turbine load and life prediction method, characterized in that assuming vibration and load changes from the collected environmental factors.
The method of claim 1,
After the prediction step, the vibration-based wind turbine load and life prediction method, characterized in that it further comprises an operating step of operating the wind power generator through the predicted load.
5. The method of claim 4,
The operation step is a vibration-based wind turbine load and life prediction method, characterized in that the maintenance plan by collecting vibration data from wind turbines installed in different locations.
5. The method of claim 4,
The operation step is a vibration-based wind turbine load and lifespan prediction method, characterized in that the maintenance cycle and trend are analyzed through the accumulated failure-related data, and the failure prediction due to fatigue is performed through real-time fatigue analysis.
In the vibration-based wind turbine load and life prediction system using the vibration-based wind turbine load and life prediction method of claims 1 to 6,
a vibration sensor for measuring vibration of the wind turbine;
a collection unit including a strain sensor for measuring the load of the wind turbine; and
The data collected by the collecting unit includes a main server including a prediction model and a learning module for estimating load and lifespan with vibration data;
Vibration-based wind turbine load and life prediction system comprising a.
8. The method of claim 7,
The collection unit Vibration-based wind turbine load and life prediction system, characterized in that it further comprises an anemometer, an atmospheric pressure gauge, and a temperature-hygrometer that affects the vibration of the wind turbine.
9. The method of claim 8,
The prediction model is a vibration-based wind turbine load and life prediction system, characterized in that for predicting the load and life by correcting the vibration change according to wind speed, atmospheric pressure, and temperature and humidity.
8. The method of claim 7,
The wind turbine is a blade that converts wind into rotational force,
A generator that generates power through the rotation of the blade,
a speed increaser for transmitting power to the generator by increasing the rotational speed of the blade;
A main bearing that is positioned between the blade and the speed increaser and rotates while fixing the shaft at a fixed position and supporting the load, and
A coupling positioned between the gearbox and the generator to transmit rotational power,
In the case of the tower of the wind turbine and the offshore wind power, it includes a substructure,
The vibration-based wind turbine load and life prediction system, characterized in that the vibration sensor is vertically and horizontally attached to the surface of the components of the wind turbine.

Images