METHOD AND AN APPARATUS FOR COMPUTER-IMPLEMENTED MONITORING OF A WIND TURBINE

Abstract

A method for monitoring a wind turbine including: i) obtaining, from a data storage, a plurality of sets of measurement data of at least two measurement variables, the measurement variables being measurement variables of the wind turbine, acquired by first sensors, and/or the environment of the wind turbine, acquired by second sensors, and the measurement data of a respective set of measurement data being acquired at a same time point in the past; ii) processing the measurement data of the at least two measurement variables by creating an image suitable for visualization; iii) determining a deviation type from a predetermined operation of the wind turbine by processing the image by a trained data-driven model configured as a convolutional neural network, where the image is fed as a digital input to the trained data-driven model and the trained data-driven model provides the deviation type as a digital output.

Claims

1. A method for computer-implemented monitoring of a wind turbine comprising an upper section on top of a tower, the upper section being pivotable around a vertical yaw axis and having a nacelle and a rotor with rotor blades, the rotor being attached to the nacelle and the rotor blades being rotatable by wind around a substantially horizontal rotor axis, the method comprising: i) obtaining, from a data storage, a plurality of sets of measurement data of at least two measurement variables, the at least two measurement variables being measurement variables of the wind turbine, acquired by one or more first sensors, and/or an environment of the wind turbine, acquired by one or more second sensors, and the measurement data of a respective set of measurement data being acquired at a same time point in the past; ii) processing the plurality of sets of measurement data of the at least two measurement variables by creating an image suitable for visualization; and iii) determining a deviation type from a predetermined operation of the wind turbine by processing the image by a trained data-driven model configured as a convolutional neural network, wherein the image is fed as a digital input to the trained data-driven model and the trained data-driven model provides the deviation type as a digital output.

2. The method according to claim 1, wherein an information based on the deviation type is output via a user interface.

3. The method according to claim 1, wherein control commands are generated for the wind turbine.

4. The method according to claim 1, wherein transforming the plurality of sets of measurement data of the at least two measurement variables into the image comprises adding a reference graph characterizing and/or visualizing a predetermined operation of the wind turbine.

5. The method according to claim 1, wherein the plurality of sets of measurement data of the at least two measurement variables processed for transformation into the image is dependent on the failure type and greater than 1000.

6. The method according to claim 1, wherein the plurality of sets of measurement data of the at least two measurement variables is filtered to exclude periods of maintenance and/or downtimes.

7. An apparatus for computer-implemented monitoring of a wind turbine comprising an upper section on top of a tower, the upper section being pivotable around a vertical yaw axis and having a nacelle and a rotor with rotor blades, the rotor being attached to the nacelle and the rotor blades being rotatable by wind around a substantially horizontal rotor axis, the apparatus comprising: a processing unit configured to perform the following steps: i) obtaining, from a data storage, a plurality of sets of measurement data of at least two measurement variables, the at least two measurement variables being measurement variables of the wind turbine, acquired by one or more first sensors, and/or an environment of the wind turbine, acquired by one or more second sensors, and the measurement data of a respective set of measurement data being acquired at a same time point in the past; ii) processing the plurality of sets of measurement data of the at least two measurement variables by creating an image; iii) determining a deviation type from a predetermined operation of the wind turbine by processing the image by a trained data driven model configured as a convolutional neural network, wherein the image is fed as a digital input to the trained data driven model and the trained data driven model provides the deviation type as a digital output.

8. The apparatus according to claim 7, wherein the apparatus is configured to perform a method for computer-implemented monitoring of a wind turbine.

9. A wind turbine comprising an upper section on top of a tower, the upper section being pivotable around a vertical yaw axis and having a nacelle and a rotor with rotor blades, the rotor being attached to the nacelle and the rotor blades being rotatable by wind around a substantially horizontal rotor axis, wherein the wind turbine comprises the apparatus according to claim 7.

10. A computer program product, comprising a computer readable hardware storage device having computer readable program code stored therein, said program code executable by a processor of a computer system to implement the method according to claim 1 when the program code is executed on a computer.

Description

BRIEF DESCRIPTION

[0024] Some of the embodiments will be described in detail, with reference to the following figures, wherein like designations denote like members, wherein:

[0025] FIG. 1 is a schematic illustration of a wind turbine comprising a controller for performing an embodiment of the invention;

[0026] FIG. 2 is a two-dimensional diagram visualizing a plurality of sets of measurement data of a wind turbine together with a reference graph characterizing a predetermined operation of the wind turbine, wherein the wind turbine operates within design operation;

[0027] FIG. 3 is another two-dimensional diagram visualizing a plurality of sets of measurement data of a wind turbine together with a reference graph characterizing a predetermined operation of the wind turbine, wherein the measurement data comprises a number of deviations from design operation; and

[0028] FIG. 4 is another two-dimensional diagram visualizing a plurality of sets of measurement data of a wind turbine without a reference graph, wherein the measurement data comprises a number of deviations from design operation.

DETAILED DESCRIPTION

[0029] FIG. 1 shows in a lower part a wind turbine 10. The wind turbine 10 may be part of a wind farm or may be a single turbine. The method described herein may be applied to a single wind turbine and to wind turbines in wind farms as well.

[0030] The wind turbine 10 is shown in a plane view from above. A 3D coordinate system CS for indicating the spatial arrangement of the wind turbine 10 is part of FIG. 1. The vertical direction is indicated by the z-axis of the coordinate system CS whereas the directions parallel to the horizontal directions are indicated by the x-axis and y-axis of the coordinate system CS. The wind direction is along the y-axis of the coordinate system CS.

[0031] The wind turbine 10 comprises an upper section 11 being located on top of a tower (not shown) which extends in the vertical z-direction. The upper section comprises a nacelle 12 accommodating an electric generator for generating electricity. Furthermore, the upper section 11 comprises a rotor 13 having three rotor blades with an angle of 120° therebetween where FIG. 1 only shows two of those blades. The rotor 13 is rotated around a substantially horizontal rotor axis HA by wind resulting in the generation of electricity by the generator within the nacelle 12. The upper section 11 of the turbine 10 can be pivoted around the vertical yaw axis VA.

[0032] The turbine 10 is equipped with a number of not shown other sensors 14 (being a first sensor) for determining operating parameters of the turbine such as rotor values (rotor speed, rotor acceleration, rotor azimuth), blade pitch values (pitch angle, pitch speed), blade root moment, produced power and/or torque. Besides the sensor 14, a second sensor 15 may be installed at or nearby the wind turbine 10. The sensor 15 which can consist of a number of different sensors is adapted to determine environmental parameters, such as wind speed, wind direction, turbulence intensity, air density, outside temperature and so on. In addition, further sensors may be provided to determine further information, such as outside temperature, air pressure and so on.

[0033] The wind turbine 10 is equipped with a controller which aims to keep the wind turbine within design operation. The method as described in the following provides an easy method to detect deviations from design operation. To do so, the wind turbine 10 acquires, by means of the sensors 14, 15, measurement data MD of two or more measurement variables Var1, Var2 and stores them in a database. The acquisition of the measurement data of the different measurement variables takes place periodically, such as every minute, every ten minutes or every 15 minutes. If the processor and the sensors 14, 15 are adapted to acquire the measurement data of the measurement variables in a shorter time interval, filtering of the measurement data of a respective measurement variable may be made and the filtered measurement data may be stored in the data storage DB. As measurement variables Var1, Var2, for example, produced power, torque, rotor speed and so on are acquired. However, in the database any kind of measurement data may be stored.

[0034] Measurement data (values) of different measurement variables Var1, Var2 which are acquired at the same time are assigned with a same timestamp and denoted as set of measurement data MD. The data storage DB may be a database consisting of table having a plurality of columns (consisting of the timestamp, the number of measurement variables Var1, Var2, . . . ) where each line represents a set of measurement data MD being acquired at the same time.

[0035] To detect a deviation from a design operation, i.e. a predetermined operation of the wind turbine as guaranteed by the manufacturer of the wind turbine, a plurality of sets of measurement data MD consisting of the at least two measurement variables Var1, Var2 is obtained from the data storage DB. The sets of measurement data MD are transferred by a suitable communication link to a controller 100. The controller 100 may be a controller of the wind turbine 10 or an external computer for supervising operation of the wind turbine 10. The controller 100 comprises a processing unit PU with a transformation unit TRF to transform the plurality of sets of measurement data MD into an image IM and for implementing a trained data-driven model MO receiving a respective image IM as a digital input and providing a deviation type DT (DT1, DT2, DT3, . . . ) as a digital output.

[0036] The trained data-driven model MO is based on a convolutional neural network having been learnt beforehand by training data. The training data comprise a plurality of images IM together with the information of none, one or more deviation types DT1, DT2, DT3 occurring in the respective image IM. Convolutional neural networks are well-known from the prior art and particularly suitable for processing digital images. Convolutional neural network comprise convolutional layers followed by pooling layers as well as fully connected layers in order to determine at least one property of the respective image where the property according to embodiments of the invention is one or more deviation types DT1, DT2, DT3, . . . .

[0037] FIGS. 2 to 4 show different embodiments of a multidimensional diagram in which a plurality of sets of measurement data MD of two (FIGS. 2 and 3) or three (FIG. 4) has been transformed into an image IM.

[0038] In examples of FIGS. 2 and 3, variable Var1 corresponds to, for example, wind speed. Variable Var2 corresponds to, for example, produced electricity. Each point in the diagram depicted in grey color represents a set of measurement data MD, i.e. a pair of values of corresponding wind speed and produced electricity. Together with the sets of measurement data MD consisting of, in particular, more than 1000 sets of measurement data MD having been acquired in the past over a time period, for example, a month, a reference graph REG indicating a warranted power from the manufacturer of the turbine is shown.

[0039] As can be seen from FIG. 2, the plurality of sets of measurement data MD and the reference graph REG correspond to each other, i.e. the wind turbine 10 operates within design operation. In the embodiment of FIG. 3 three different deviation types DT1, DT2, DT3 indicated by a respective ellipse are shown in the image. Generally, a power curve can have several deviations such as speed calibration errors, power boost, unexplained curtailment (denoted in FIG. 3 as deviation type DT1), faulty anemometer (denoted in FIG. 3 as deviation type DT2), faulty cluster of points (denoted in FIG. 3 as deviation type DT3), no data, and so on.

[0040] In the images of FIGS. 2 and 3, the background containing neither measurement data MD nor the reference graph RFG is depicted with no color, i.e. in black or white. In this way, the reference graph REG is distinguishing from the measurement data MD.

[0041] In the example of FIG. 4, another image resulting from a transformation of a plurality of sets of measurement data MD of three measurement variables Var1, Var2, Var3 is illustrated. Measurement variable Var1 corresponds to a rotor speed and measurement variable Var2 corresponds to the active power or torque of the wind turbine. The third measurement variable Var3 corresponds to the ambient temperature and may be visualized in the image IM by color coding, i.e. the point of values resulting from measurement variables Var1 (rotor speed), Var2 (active power or torque) has the color according to the ambient temperature (Var3) as indicated by the color gradient on the right hand side. In case that a further measurement variable is to be visualized in the image IM, the size of the measurement data points may be chosen in different sizes. By way of example only, two deviation types DT4 and DT5 in the form of vertical series of points are illustrated in FIG. 4. They may be unexpected deviations and require further analysis by a human operator.

[0042] In the embodiment of FIG. 1, the deviation type DT (as a general denotation for the different deviation type DT1, . . . , DT5) produced as an output of the model MO results in an output on a user interface UI which is shown only schematically. The user interface UI comprises a display. The user interface provides information for a human operator. The output based on the deviation type DT may be the deviation type itself so that the operator is informed about the kind of deviation. Alternatively or additionally, the output may be a warning in case that the deviation type of a specific type which might cause damage to the wind turbine.

[0043] The deviation type DT determined by the model MO may also result in control commands which are provided to the wind turbine 10 in order to adjust, for example, the yaw angle, or to shut the wind turbine down. In this case, the controller 100 enables an automatic adjustment or shutdown of the wind turbine to avoid further damage.

[0044] Embodiments of the invention as described in the foregoing have several advantages. Particularly, an easy and straightforward method in order to detect deviations from design operations is provided. To do so, measurement data collected in the past is processed and transformed into an image in order to determine the deviation type via a suitably trained data-driven model configured as a convolutional neural network. The formalization of a specific failure type is simpler than classical engineering techniques. The process is fast as a domain expert only needs to classify training images correctly. The method provides consistent results as the same image will lead to the same prediction.

[0045] Although the present invention has been disclosed in the form of preferred embodiments and variations thereon, it will be understood that numerous additional modifications and variations could be made thereto without departing from the scope of the invention.

[0046] For the sake of clarity, it is to be understood that the use of “a” or “an” throughout this application does not exclude a plurality, and “comprising” does not exclude other steps or elements.