Method and apparatus for computer-implemented monitoring of one or more wind turbines in a wind farm

Abstract

The invention relates to a method for computer-implemented monitoring of one or more wind turbines (1) in a wind farm, each wind turbine (1) having a rotor (4), the rotor (4) having rotor blades (5), the rotor blades (5) being rotatable about a rotor axis (R) by wind, wherein during operation of the wind farm a process is performed once or several times for a respective rotor blade (5) of the one or more rotor blades (5) of a respective wind turbine (1) of the one or more wind turbines (1): i) obtaining a digital image (IM) of the respective rotor blade (5), the image (IM) being a current image taken by a camera (7) observing the respective rotor blade (5); ii) determining one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) by processing the image (IM) by a trained data-driven model (MO), wherein the image (IM) is fed as a digital input to the trained data-driven model (MO), and the trained data-driven model (MO) provides the one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) as a digital output.

Description

[0001] The present invention relates to a method and apparatus for computer-implemented monitoring of one or more wind turbines in a wind park.

[0002] There is a need to monitor the operating characteristics of rotor blades during operation of wind turbines in a wind farm. The operating characteristics give feedback about the forces acting on the wind turbine and enable an adaptation of the control strategy of the wind turbine in order to avoid damage to the rotor blades. It is known to measure the root bending moments acting on rotor blades based on strain gauges mounted inside the blades close to the blade root. Those strain gauges often fail and are difficult to replace due to their location at the blade root. Furthermore, the measurement accuracy of the strain gauges strongly depends on their mounting positioning.

[0003] It is an object of the present invention to provide a simple method for determining one or more operational characteristics of a rotor blade of a wind turbine during operation.

[0004] This object is solved by the independent patent claims. Preferred embodiments of the invention are defined in the dependent claims.

[0005] The invention provides a method for computer-implemented monitoring of one or more wind turbines in a wind farm. Each wind turbine has a rotor with rotor blades, which are rotatable about a rotor axis by wind. According to the method of the invention, a process is performed once or several times during operation of the wind farm. The method comprises the following steps i) and ii), which are performed for a respective rotor blade of the one or more rotor blades of a respective wind turbine of the one or more wind turbines being monitored.

[0006] In step i), a digital image of the respective rotor blade is obtained, which image is a current image taken by a camera observing the respective rotor blade. The term "current image" refers to an image taken during the current process being performed. The term "camera" refers to an imaging device operating in at least a part of the spectrum between ultraviolet light and infrared light. For example, the camera may operate in the visible light spectrum. In a preferred embodiment, the camera operates in the infrared spectrum of light, so that the image can also be obtained at night.

[0007] In step ii), one or more operational characteristics of the respective rotor blade are determined by processing the images by a trained data-driven model, wherein the images are fed as digital inputs to the trained data-driven model and the trained data-driven model provides the one or more operational characteristics of the respective rotor blade as digital outputs. The data-driven model is trained by training data, the training data comprising a plurality of images of the respective rotor blade taken by a camera and information about the one or more operational characteristics as captured in the respective images of the training data.

[0008] An advantage of the present invention is that the operational characteristics of the rotor blades can be determined in a simple and direct manner simply by taking images of the rotor blades and feeding the images to a trained data-driven model. Thus, there is no need to install error-prone sensors, such as strain gauges, in the wind turbine having the rotor blades to be monitored.

[0009] In the method according to the invention, any known data-driven model learned by machine learning can be used. In a particularly preferred embodiment, the trained data-driven model is a neural network, preferably a convolutional neural network that is particularly suitable for processing image data. However, other trained data-driven models can also be implemented in the method of the invention, such as models based on decision trees or support vector machines.

[0010] In a preferred embodiment of the invention, information based on one or more operating characteristics (i.e. based on the digital output of the data-driven model) is output via a user interface. For example, the one or more operating characteristics themselves may be output via the user interface. Alternatively or additionally, in case at least one of the operating characteristics does not meet predetermined criteria regarding normal operation of the wind turbine, a warning may be provided via the user interface. For example, in case the operating characteristics indicate that the load on the rotor blades is too large, a corresponding warning may be generated via the user interface so that a human operator is informed of the operating condition and appropriate countermeasures may be initiated. Preferably, the user interface comprises a visual user interface, but it may also comprise another type of user interface (e.g. an acoustic user interface).

[0011] In a particularly preferred embodiment, the one or more operating characteristics include at least one root bending moment, i.e. a bending moment acting on an attachment point on the rotor blade on the rotor hub. Preferably, the at least one root bending moment includes an edgewise root bending moment and / or a flapwise root bending moment. These moments are well known. The edgewise root bending moment acts on the respective rotor blade in the direction of rotation of the rotor, and the flapwise root bending moment is a bending moment acting on the respective rotor blade along the rotor axis.

[0012] In another embodiment, the one or more characteristics include a deflection of the respective rotor blade tip, i.e. the positioning of the rotor blade tip relative to the rotor blade root. Preferably, the wind speed is determined from the deflection, thereby providing the functionality of an anemometer. The wind speed may be output via a user interface in order to inform a human operator.

[0013] In another embodiment, the one or more operational characteristics include a twist of the respective rotor blade, i.e. a twist of a longitudinal axis of the respective rotor blade from a root to a tip of the rotor blade compared to an untwisted state, i.e. a state in which no wind acts on the respective rotor blade.

[0014] The camera that takes the image of the respective rotor blade can be mounted at different locations. In one embodiment, the camera is mounted on the nacelle of the respective wind turbine for which the process is performed. However, the camera can also be mounted on the (rotating) rotor hub of the respective wind turbine for which the process is performed. In this case, the camera rotates with the rotor and is used to monitor a specific rotor blade of the respective wind turbine.

[0015] In another embodiment, the camera is mounted on another wind turbine than the respective wind turbine for which the procedure is performed. In this case, the camera preferably views the respective rotor blade in a direction substantially perpendicular to the rotor axis of the respective wind turbine. In other words, the optical axis of the camera is substantially perpendicular to the rotor axis.

[0016] In a particularly preferred embodiment, the angular position of the respective rotor blade about the rotor axis at the time the image was taken is fed as an additional input to the data driven model. This enables detection of operating characteristics of the rotor blade for different angular positions. To implement this embodiment, the training data also includes the angular position of the respective rotor blade captured in the respective image of the training data.

[0017] In another preferred variant of the invention, in step i), several images of the respective rotor blade taken by the camera are available, wherein the image with the best image quality according to one or more quality criteria is selected as the image obtained in step i). Quality criteria for defining image quality are well known to the skilled person. For example, a quality criterion may refer to image clarity.

[0018] In another preferred embodiment of the invention, the process comprising steps i) and ii) is performed for several rotor blades of a respective wind turbine, wherein the identified rotor blades have one or more operating characteristics whose deviation from one or more operating characteristics of other rotor blades is above a predetermined threshold. Depending on the situation, the deviation may be defined differently. For example, a vector comprising the (preferably normalized) operating characteristics as entries may be used for the respective rotor blade, wherein the Euclidean distance of the vectors indicates the deviation.

[0019] In another embodiment, a gust of wind with a corresponding gust wind speed is detected based on one or more operating characteristics of at least one rotor blade of the respective wind turbine for which the process is performed. It is well known to those skilled in the art that there is a correlation between changes in operating characteristics and gusts of wind. In particular, a gust of wind will result in a high deflection of the tip of the respective rotor blade. The amount of this deflection is an indicator of the gust wind speed.

[0020] In a preferred variation of the above embodiment, the control of a wind turbine located downstream with respect to the wind direction of the respective wind turbine for which the process is performed is adapted depending on the gust wind speed. For example, it may be calculated when the gust arrives at the downstream wind turbine by using a known distance between the respective wind turbine and the downstream wind turbine, wherein the control of the downstream wind turbine is adapted when the gust arrives.

[0021] In addition to the above method, the present invention relates to a device for computer-implemented monitoring of one or more wind turbines in a wind farm, each wind turbine having a rotor with rotor blades rotatable about a rotor axis by wind, wherein the device comprises a processor configured to perform the method according to the invention or one or more preferred embodiments of the method according to the invention.

[0022] Furthermore, the present invention relates to a wind farm comprising a plurality of wind turbines, ie at least one wind turbine, wherein the wind farm comprises the above-described device according to the present invention for monitoring one or more wind turbines of the wind farm.

[0023] Furthermore, the invention relates to a computer program product with a program code stored on a non-transitory machine readable carrier for performing a method according to the invention or one or more preferred embodiments thereof, when the program code is executed on a computer.

[0024] Furthermore, the invention relates to a computer program with a program code for implementing the method according to the invention or one or more preferred embodiments thereof, when the program code is executed on a computer.

[0025] Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Figure 1 is a schematic diagram of a wind farm including a controller for executing an embodiment of the present invention.

[0026] Figure 1 A top plan view of a part of a wind farm comprising three wind turbines 1, 2 and 3 is shown, wherein the wind turbine 1 is monitored based on an embodiment of the invention. The entire wind farm comprises more than these three wind turbines 1 to 3. Figure 1 Also included is a 3D coordinate system CS indicating the spatial arrangement of the wind turbines 1, 2 and 3 of the wind farm. The z-axis of the coordinate system CS refers to the vertical direction, while the x-axis and the y-axis refer to the horizontal direction perpendicular to the vertical z-direction. The direction along the y-axis corresponds to the wind direction also indicated by a plurality of arrows P.

[0027] The construction of each of the wind turbines 1, 2, 3 is identical. That is, each wind turbine comprises a tower (not shown) with a nacelle 3 located on top of the tower. A rotor 4 is attached to the nacelle. The rotor 4 comprises a rotor hub 6 from which three rotor blades 5 extend, wherein the rotor blades 5 are located on the top of the tower. Figure 1 Only two of those rotor blades can be seen in FIG. 4 . The rotor is rotated by the wind about a rotor axis R. The rotor blades are spaced apart from each other at an angle of 120° about the rotor axis R. The nacelle 3 houses a generator (not shown) which converts the mechanical energy of the rotor 4 driven by the wind into electrical energy.

[0028] The method described herein enables monitoring of a turbine 1 by means of images IM captured by a camera 7. The camera is mounted on top of the nacelle of the wind turbine 3, positioned adjacent to the turbine 1 in the x-direction. The camera 7 is a side-view camera and captures images of the rotor blades of the wind turbine 1. Figure 1 In the scene of FIG. 1 , the camera 7 looks in a direction substantially perpendicular to the rotor axis R of the wind turbine 1. In other words, the optical axis of the camera 7 is substantially perpendicular to the rotor axis R of the wind turbine 1. Preferably, the camera 7 covers a wide angular range to cover the turbine 1 in many cases (for as many wind directions as possible). Furthermore, the camera is preferably not statically mounted on the wind turbine 3, but adapts itself so that an adjacent turbine 1 is located in its center.

[0029] The images IM of the camera 7 are used in order to determine one or more operational characteristics of the rotor blades of the wind turbine 1 for each 360° rotation of the rotor thereof. In the embodiment described herein, the operational characteristics when a particular rotor blade is in an upward direction along the z-axis of the coordinate system CS are determined by taking an image IM of the rotor blade by the camera 7 when the rotor blade is in an upward position. However, operational characteristics of more than one rotor blade or other rotor blades of the wind turbine 1 may also be derived by taking appropriate images of these rotor blades by the camera 7. Furthermore, the blade does not need to be in an upward position when taking the image of the respective rotor blade, but may have another angular position around the rotor axis R.

[0030] The corresponding images IM taken by the camera 7 are transferred via appropriate communication links to a controller 8 which is connected to all wind turbines of the wind park through several communication links and controls the operation of those wind turbines. The controller 8 comprises a processor PR implementing the trained data driven model MO.

[0031] The data driven model MO is learned by machine learning methods. In the embodiments described herein, the data driven model is a convolutional neural network, which is particularly suitable for processing images. However, other data driven models may be used, such as decision trees or support vector machines. The data driven model MO receives the corresponding image IM as digital input and generates a digital output in the form of an operating characteristic of the corresponding rotor blade of the wind turbine 1 as captured by the image IM.

[0032] In the embodiments described herein, the operating characteristics include the root bending moment BM in both the flap direction and the edge direction of the respective rotor blade. To generate this output, the data driven model MO is trained with training images, where the root bending moment in the edge direction and the flap direction for each training image is known. Those images are generated during the training phase by Figure 1 The camera 7 in the wind farm or by having Figure 1 The images are taken from a camera in another wind farm in the same arrangement as in FIG1 , wherein the images are taken from the rotor blades of a wind turbine of the same type as wind turbine 1. In order to generate accurate outputs through the trained data-driven model MO, the training phase has covered the entire wind speed range of wind turbine 1 from switch-in (i.e., when the wind turbine starts generating electricity) to switch-off (i.e., when the wind turbine is shut down due to excessive wind speed).

[0033] In addition to or as an alternative to the root bending moment BM, other operating characteristics of the respective rotor blade may be output by the trained data-driven model MO. Figure 1 , those operating characteristics may refer to a deflection DE of the tip end of the respective rotor blade, i.e. a positioning of the tip end of the rotor blade relative to the root end of the rotor blade. In another embodiment, the operating characteristic comprises a twist of the respective rotor blade, i.e. a twist of the longitudinal axis of the respective rotor blade from the root end to the tip end of the rotor blade compared to an untwisted state, i.e. a state in which no wind acts on the respective rotor blade.

[0034] In order to output the operating characteristics DE and TW, it is necessary to train the data-driven model MO or an additional data-driven model for these characteristics by means of appropriate training data. Those training data include images taken by the camera 7, and additional information for each image, where the turbine state captured in the image has blade deflection DE and / or twist TW.

[0035] exist Figure 1 In an embodiment, information based on the operating characteristics is output via a user interface UI to inform a human operator of those characteristics. The information based on the operating characteristics may be the characteristics themselves. Alternatively or additionally, the information may be a warning in the event that the operating characteristics exceed a predetermined threshold.

[0036] Furthermore, in a modified embodiment, the blade deflection DE is used to adapt the control strategy of a wind turbine 2 located downstream with respect to the wind direction of the wind turbine 1. To this end, a gust of wind (if present) is detected by analyzing the blade deflection DE. This is possible because the blade deflection is a direct measure of the gust of wind (i.e. a short-time deflection). In case a gust of wind is detected, the corresponding wind speed of the gust of wind can also be derived from the blade deflection. By using the wind speed of the gust of wind, the controller 8 determines when the gust of wind will arrive at the wind turbine 2. This determination is direct because the distance between the two turbines 1 and 2 in the y direction is stored in the controller 8, so that the arrival time of the gust of wind can be calculated by dividing this distance by the wind speed of the gust of wind. Based on the knowledge about the arrival time of the gust of wind, the control strategy of the wind turbine 2 is adapted so that a lower load acts on the turbine 2 when the gust of wind arrives in order to compensate for the gust of wind.

[0037] In another modified embodiment, at least one of the operating characteristics, and in particular the blade deflection and the blade twist, is determined for all three rotor blades of the wind turbine 1, and the values ​​of the operating characteristics of the different blades are compared with each other. By doing so, significantly different operating characteristics of one blade compared to the other blades can be detected. This detection can be an indication of a malfunction of the rotor blade with the significantly different operating characteristics. The information about the significantly different operating characteristics can be recorded for further analysis and / or a warning can be output via the user interface UI so that a human operator is informed of a possible malfunction of a rotor blade of the wind turbine 1.

[0038] exist Figure 1 In the embodiment of the invention, the blades of a wind turbine 1 are monitored by a camera 7 mounted on another wind turbine 2. However, the blades of a wind turbine 1 may also be monitored by a camera mounted on the wind turbine 1 itself. For example, the camera may be mounted on the nacelle 3 of the wind turbine 1. Preferably, the camera looks upwards in order to capture the rotor blades in an upward orientation. Furthermore, the camera may also be mounted at the rotor hub 6 of the wind turbine 1 rotating with the blades 5. In this case, the camera preferably looks upwards along the longitudinal axis of the blade to be monitored. If more than one blade is to be monitored, one camera per blade needs to be mounted at the rotor hub.

[0039] In another embodiment of the invention, several images are taken from the same rotor blade during one rotation of the rotor 4 of the wind turbine 1. Thereafter, the image with the best quality (e.g. best image clarity) according to a predetermined criterion is used as the image IM to be input to the data-driven model MO. This improves the accuracy of the operational characteristics determined by the data-driven model MO. A similar approach can be applied during the training phase. That is, several images of the corresponding blade and the information about the operational characteristics captured by the images can be taken during the training phase. The image with the best image quality is thereafter used as the training data set, while the other images are discarded.

[0040] In another preferred embodiment, the angular positioning of the rotor blades 5 about the rotor axis R of the wind turbine 1 is used as an additional input to the trained data driven model. In this case, the images used to train the data driven model also include information on the blade positioning present in the corresponding images of the training data.

[0041] The invention described above has several advantages. In particular, a simple and direct method is provided for detecting the operating characteristics of a rotor blade of a wind turbine. To this end, the operating characteristics are determined based on the images of a camera via a suitably trained data-driven model. Thus, no sensors, such as strain gauges, need to be installed to detect the operating characteristics of the rotor blade.

Claims

1. A method for computer-implemented monitoring of one or more wind turbines (1) in a wind farm, each wind turbine (1) having a rotor (4), the rotor (4) having rotor blades (5), the rotor blades (5) being rotatable about a rotor axis (R) by wind, wherein during operation of the wind farm a process is performed once or several times for a respective one of the one or more rotor blades (5) of a respective one of the one or more wind turbines (1): i) obtaining a digital image (IM) of the respective rotor blade (5), said image (IM) being a current image taken by a camera (7) observing the respective rotor blade (5), wherein said camera (7) is mounted on another wind turbine (3) than the respective wind turbine (1) for which said process is performed; ii) determining one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) by processing the image (IM) by a trained data driven model (MO), wherein the image (IM) is fed as digital input to the trained data driven model (MO) and the trained data driven model (MO) provides one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) as digital output.

2. The method of claim 1, wherein the trained data-driven model (MO) is a neural network.

3. Method according to claim 1 or 2, wherein information based on one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) is output via a user interface (UI).

4. The method according to claim 1 or 2, wherein: The one or more operational characteristics (BM, DE, TW) include one or more of the following characteristics: - at least one root bending moment (BM) acting on the respective rotor blade (5); - a deflection (DE) of the tip of the respective rotor blade (5), wherein the wind speed is determined by said deflection; - the twist (TW) of the respective rotor blade (5).

5. The method of claim 2, wherein the neural network is a convolutional neural network.

6. The method according to claim 4, wherein the at least one root bending moment (BM) is an edge-wise root bending moment and / or a flap-wise root bending moment.

7. The method according to claim 1 or 2, wherein the camera (7) observes the respective rotor blade (5) in a direction substantially perpendicular to a rotor axis (R) of the respective wind turbine (1).

8. The method according to claim 1 or 2, wherein: The angular position of the respective rotor blade (5) about the rotor axis (R) when the image (IM) is recorded is fed as an additional input to the data driven model (MO).

9. The method according to claim 1 or 2, wherein: In step i), several images (IM) of the respective rotor blade (5) taken by the camera (7) are processed, wherein the image (IM) having the best image quality according to one or more quality criteria is selected as the acquired image (IM).

10. The method according to claim 1 or 2, wherein: The process is performed for a number of rotor blades (5) of a respective wind turbine (1), wherein the identified rotor blades (5) have one or more operating characteristics (BM, DE, TW) which deviate from one or more operating characteristics (BM, DE, TW) of the other rotor blades (5) by above a predetermined threshold.

11. The method according to claim 1 or 2, wherein: A wind gust having a corresponding gust wind speed is detected based on one or more operating characteristics (BM, DE, TW) of at least one rotor blade (5) of the respective wind turbine (1) for which the process is performed.

12. The method according to claim 11, wherein: The control of a wind turbine (2) located downstream with respect to the wind direction relative to the respective turbine (1) for which the process is performed is adapted depending on the gust wind speed.

13. An apparatus for computer-implemented monitoring of one or more wind turbines (1) in a wind farm, each wind turbine (1) having a rotor (4) with rotor blades (5) rotatable about a rotor axis (R) by wind, wherein the apparatus (8) comprises a processor (PR) configured to perform, during operation of the wind farm, a process for a respective one of the one or more rotor blades (5) of a respective one of the one or more wind turbines (1) comprising the following steps one or more times: i) obtaining a digital image (IM) of the respective rotor blade (5), said image (IM) being a current image taken by a camera (7) observing the respective rotor blade (5), wherein said camera (7) is mounted on another wind turbine (3) than the respective wind turbine (1) for which said process is performed; ii) determining one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) by processing the image (IM) by a trained data driven model (MO), wherein the image (IM) is fed as digital input to the trained data driven model (MO) and the trained data driven model (MO) provides one or more operating characteristics (BM, DE, TW) of the respective rotor blade (5) as digital output.

14. The device according to claim 13, wherein: The device (8) is configured to carry out a method according to one of claims 2 to 12.

15. A wind farm comprising a plurality of wind turbines (1, 2, 3), wherein: The wind farm comprises an arrangement (8) according to claim 13 or 14.

16. A computer program product having a program code stored on a non-transitory machine-readable carrier for carrying out the method according to one of claims 1 to 12 when the program code is executed on a computer.

17. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions implement any one of methods 1 to 12 when executed by a processor.

Images