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

The invention refers to a method and an apparatus for comput er-implemented monitoring of one or more wind turbines in a wind farm.

Wind turbines of wind farms comprise an upper section with a rotor and a nacelle on the top of a tower, where the upper section can be rotated around a vertical yaw axis in order to vary the yaw angle of the respective turbine. The yaw angle of a wind turbine is usually adjusted such that the rotor of the wind turbine faces the wind. To do so, wind sensors (i.e. anemometers) are installed on the respective wind turbines to estimate the wind direction. The measurements provided by the wind sensors are often imprecise and, thus, result in a yaw misalignment of the corresponding wind turbine, i.e. the tur bine is not aligned with the wind direction. This yaw misa lignment causes power loss in the production of electric en ergy in the wind farm and shall be avoided.

It is an object of the invention to provide an easy method in order to detect yaw misalignments of wind turbines within a wind farm.

This object is solved by the independent patent claims. Pre ferred embodiments of the invention are defined in the de pendent claims.

The invention provides a method for computer-implemented mon itoring of one or more wind turbines in a wind farm. The wind farm comprises a first wind turbine and a second wind tur bine, i.e. a pair of first and second wind turbines. The method is applied to this pair and may also be applied to several of such pairs. Each of the first and second wind tur bines comprises 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 is at tached to the nacelle and the rotor blades are rotatable by wind around a horizontal rotor axis.

According to the method of the invention, the following steps i) and ii) are performed at each time point of one or more time points during the operation of the wind farm.

In step i), an image of the upper section of the first wind turbine is obtained. In the following, the term "image" re fers to a digital image. The term "obtaining an image" means that the image is received by a processor implementing the method of the invention. The digital image is a current image (i.e. an up-to-date image) taken by a camera installed at the upper section of the second wind turbine. The term "camera" refers to an imaging device working in at least a part of the spectrum between ultraviolet and infrared light. E.g., the camera may work in the spectrum of visible light. In a pre ferred embodiment, the camera works in the infrared spectrum of light so that images may also be obtained at night.

In step ii), a yaw misalignment angle between the first and second turbines is determined by processing the image by a trained data driven model, where the image is fed as a digi tal input to the trained data driven model and where the trained data driven model provides the yaw misalignment angle as a digital output. This yaw misalignment angle is the ob tuse angle between the rotor axis of the first wind turbine and the rotor axis of the second wind turbine.

The method of the invention provides an easy and straightfor ward method for determining yaw misalignments based on images of a camera. To do so, a trained data driven model is used. This model is trained by training data comprising a plurality of images of the upper section of a first wind turbine taken by a camera installed at the upper section of a second wind turbine together with the information about the yaw misalign- ment angle captured in the respective image of the training data .

Any known data driven model being learned by machine learning may be used in the method according to the invention. In a particularly preferred embodiment, the trained data driven model is a neural network, preferably a Convolutional Neural Network which is particularly suitable for processing image data. Nevertheless, other trained data driven models may also be implemented in the method of the invention, e.g. models based on decision trees or Support Vector Machines.

In a preferred embodiment of the invention, an information based on the yaw misalignment angle (i.e. based on the digi tal output of the data driven model) is output via a user in terface. E.g., the yaw misalignment angle itself may be out put via the user interface. Alternatively or additionally, a warning may be provided via the user interface in case that the yaw misalignment angle exceeds a predetermined threshold. Thus, a human operator is informed about an unacceptable yaw misalignment angle so that he can initiate appropriate coun ter measures by adjusting the yaw offset angles of the wind turbines in order to enhance the generated electric power of the wind farm. Preferably, the user interface comprises a visual user interface but it may also comprise a user inter face of another type (e.g. an acoustic user interface) .

In another particularly preferred embodiment, the method of the invention generates control commands for the first and second turbines. The control commands are such that the upper section of the first wind turbine and the upper section of the second wind turbine are rotated around their correspond ing yaw axes in opposite directions by half of the amount of the yaw misalignment angle so that the yaw misalignment angle becomes zero. This embodiment enables an automatic alignment of the first and second turbines to each other. This reduces a power loss due to yaw misalignment significantly, although it is not known which wind turbine is not correctly aligned with the wind direction.

In another embodiment, the camera taking the images is mount ed on the nacelle of the second wind turbine. This position provides a good view of the first turbine resulting in high- quality images taken by the camera.

In another preferred embodiment, the second wind turbine is located downstream with respect to the wind direction rela tive to the first wind turbine, where the image taken by the camera shows an end of the upper section of the first wind turbine opposite to an end of the upper section where the ro tor is attached to the nacelle of the first wind turbine. Those images enable a high precision in detecting the yaw misalignment angle. In an alternative embodiment, the camera may also be installed on an upstream first wind turbine in order to take images of the rotor side of a downstream second wind turbine.

Besides the above method, the invention refers to an appa ratus for computer-implemented monitoring of one or more wind turbines in a wind farm, where the apparatus is configured to perform the method according to the invention or one or more preferred embodiments of the method according to the inven tion .

Moreover, the invention refers to a computer program product with a program code, which is stored on a non-transitory ma chine-readable carrier, for carrying out the method according to the invention or one or more preferred embodiments thereof when the program code is executed on a computer.

Furthermore, the invention refers to a computer program with a program code for carrying out the method according to the invention or one or more preferred embodiments thereof when the program code is executed on a computer. An embodiment of the invention will now be described in de tail with respect to the accompanying drawing. Fig. 1 is a schematic illustration of a wind farm comprising a controller for performing an embodiment of the invention.

Fig. 1 shows in the lower part a section of a wind farm com prising a first wind turbine 1 and a second wind turbine 2. The wind farm has more than those two wind turbines. The method described herein is applied to the wind turbines 1 and 2. Nevertheless, the method may also be applied to other pairs of wind turbines being part of the wind farm.

The wind turbines are shown in plan view from above. A 3D co ordinate system CS for indicating the spatial arrangement of the wind turbines is part of Fig. 1. The vertical direction is indicated by the z-axis of the coordinate system CS where as the directions parallel to the horizontal direction are indicated by the x-axis and y-axis of the coordinate system CS . The wind direction is along the x-axis of the coordinate system CS .

Wind turbine 1 which is an upstream turbine with respect to the wind direction comprises an upper section 101 being lo cated on top of a tower (not shown) which extends in the ver tical z-direction. The upper section comprises a nacelle 102 accommodating an electric generator for generating electrici ty. Furthermore, the upper section 101 comprises a rotor 103 having three rotor blades with an angle of 120° therebetween where Fig. 1 only shows two of those blades. The rotor 103 is rotated around the horizontal rotor axis R1 by wind resulting in the generation of electricity by the generator within the nacelle 102. The upper section 101 of the turbine 1 can be pivoted around the vertical yaw axis VI.

Wind turbine 2 which is located downstream with respect to the wind direction has the same construction as wind turbine 1. I.e., wind turbine 2 comprises an upper section 201 locat ed at the top of a vertical tower, the upper section 201 com- prising a nacelle 202 accommodating a generator as well as a rotor 203 attached to the nacelle 202. Due to the rotation of the rotor by wind around the horizontal rotor axis R2, elec tricity is generated by the generator within the nacelle 202. Analogously to turbine 1, turbine 2 can be pivoted around a vertical yaw axis V2.

The yaw angles of the upper sections 101 and 201 around the respective yaw axes VI and V2 are adjusted such that the ro tor axes R1 and R2 correspond to the wind direction measured by respective wind sensors installed at both wind turbines 1 and 2. However, due to imprecise measurements of the wind sensors, a yaw misalignment may occur. This yaw misalignment is indicated by the yaw misalignment angle f in Fig. 1. This angle is the obtuse angle between the rotor axis R1 of the wind turbine 1 and the rotor axis R2 of the wind turbine 2.

Even small yaw misalignments lead to high losses of electric power produced by the wind farm. The method as described in the following provides an easy method to detect and correct those yaw misalignments. To do so, the downstream wind tur bine 2 comprises a camera 3 installed on the nacelle 201. The camera 3 is positioned such that it takes images IM of the rear side of the wind turbine 1. The camera is working in the visible light spectrum and also in the infrared light spec trum so that the method as described herein may also be per formed at night. The respective images IM of the camera 3 are transferred by a suitable communication link to a controller 4 of the wind farm. The controller comprises a processor PR implementing a trained data driven model MO receiving a re spective image IM as a digital input and providing a yaw mis alignment angle f as a digital output.

In the embodiment described herein, the trained data driven model MO is based on a Convolutional Neural Network having being learned beforehand by training data. The training data comprise a plurality of images of an upstream turbine taken by the camera of a downstream turbine together with the in- formation of the yaw misalignment angle occurring in the re spective image. Convolutional Neural Networks are well known from the prior art and are particularly suitable for pro cessing digital images. A Convolutional Neural Network com prises 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 ac cording to the invention is the yaw misalignment angle.

In the embodiment of Fig. 1, the yaw misalignment angle f produced as an output of the model MO results in an output on a user interface UI which is only shown schematically. Pref erably, the user interface UI comprises a display. The user interface provides information for a human operator. The out put based on the yaw misalignment angle may be the angle it self so that the operator is informed about the misalignment. Alternatively or additionally, the output may be a warning in case that the yaw misalignment angle exceeds a predetermined threshold .

In the embodiment of Fig. 1, the yaw misalignment angle f de termined by the model MO also results in control commands CO which are provided to the wind turbines 1 and 2 in order to adjust the yaw angles of the turbines. The control commands are such that the wind turbines 1 and 2 are rotated in oppo site directions around their respective yaw axis VI and V2 by an angle which is half of the yaw misalignment angle so that the yaw misalignment angle will become zero. Hence, the con troller 4 enables an automatic adjustment of the yaw angles to eliminate yaw misalignment between the wind turbines 1 and 2 resulting in a significantly reduced loss of generated electric power produced by the wind farm.

The invention as described in the foregoing has several ad vantages. Particularly, an easy and straightforward method in order to detect yaw misalignments between adjacent turbines is provided. To do so, the yaw misalignment is determined based on images of a camera via a suitably trained data driv- en model. In a preferred embodiment, the yaw angles of the wind turbines are also automatically controlled based on the yaw misalignment determined by the data driven model in order to reduce electric power losses.