The present invention relates to a method and apparatus for detecting, verifying and/or adjusting the orientation of the blades of a wind turbine.

Over recent decades, wind turbines have become an increasingly important source of renewable energy. A typical wind turbine comprises a tower having a nacelle at its upper end. The nacelle houses an electrical generator, gearing and control systems and has mounted to it a rotor that carries a plurality of rotor blades (typically three). The nacelle is able to rotate relative to the tower so that the blades are directed towards the wind (i.e. so that the plane in which they rotate is perpendicular to the wind direction).

It will be appreciated that accurate alignment of the rotor blades with respect to the wind direction is important to ensure efficient operation of the wind turbine. The variation from the optimum orientation is known as yaw misalignment. Yaw misalignment of wind turbines has proved to be one of the major challenges for the entire wind industry. The impact of misalignment is lower production, reduced turbine lifetime and poor performance.

Over the past decade or so, there has been increasing interest in offshore wind turbines. These have significant advantages over land-based turbines.

Generally the wind is stronger and more reliable offshore (in this regard it is relevant that the power obtainable is proportional to the cube of the wind speed) and there may be fewer environmental and aesthetic objections (particularly when they are located some distance offshore). However, locating wind turbines offshore provides additional challenges in relation to yaw misalignment.

Most offshore wind turbines are similar to those used on land, in that they comprise towers that have foundations in the sea bed. However, in recent years floating offshore wind turbines have been developed. These can be located in deep water and hence further out to sea where the wind is stronger and where they may be entirely out of sight from land.

Although wind turbines may be used singly they are generally provided in groups known as farms. It would obviously be expected that the blades of all wind turbines within a given farm would be orientated in substantially the same direction.

Offshore wind turbines, whether floating or anchored, pose additional difficulties in ensuring the optimum orientation of their rotor blades since there is no fixed datum point because the turbines and any other floating structures are mobile to some extent on their moorings.

A number of methods are presently used to determine and control the orientation of wind turbine blades.

One approach is to provide instrumentation on top of the turbine structure, in particular, to provide anemometers on the nacelle to provide a measurement of the wind direction (and strength) relative to the nacelle. However, there are a number of limitations to this approach. The measurement of a post-rotor anemometer has errors influenced by the rotating blades because the anemometer is located in the turbulent air behind the blade disc. Also, nacelle anemometers are not calibrated. There can be significant differences between individual turbines, and even between anemometers on a single turbine.

Another system uses LiDAR, either based on a turbine or measured from a boat. LiDAR is a surveying method that measures distance to a target by illuminating that target with a pulsed laser light and measuring the reflected pulses with a sensor. Differences in laser return times and wavelengths can then be used to make digital 3D-representations of the target. However, the existing LiDAR technology can only measure one turbine at a time. This makes it impossible to compare turbine performance accurately. It is also relatively costly.

A still further technique is known as iSpin. This involves the use of three ultrasonic sensors mounted on the spinner of the rotor. (The spinner is the roughly cone-shaped cover that is provided at the centre of the rotor.) This can be effective but requires the use of expensive instrumentation which must be provided on the turbine itself.

Because of these drawbacks, at present wind turbines are generally not optimally aligned. It is estimated that if all wind turbines can be kept properly aligned, production could be increased by 2% (about £8 million per annum per typical wind farm).

Accordingly, there is a need for a method of measuring the orientation of a wind turbine rotor that addresses at least some of the above drawbacks and, in particular, is suitable for use with offshore wind turbine farms.

According to the invention there is provided a method of determining the orientation of the rotor blade disc of a wind turbine comprising the steps of:

obtaining an image of the wind turbine; identifying from the image the rotor blade disc of the wind turbine; and determining the orientation of the rotor blade disc from the image.

Thus, by means of the invention, a convenient method of determining the orientation of the rotor blade disc of a wind turbine is provided. Whilst it may be used for a solitary turbine, it is particularly advantageous for use with a plurality of turbines and preferably a farm of wind turbines since preferably the entire farm may be included within the image.

The image may be any suitable image from which the rotor blade disc may be resolved and its orientation determined relative to the frame of reference of the image - e.g. relative to the known orientation of the imaging system itself or to some external datum. (Where a plurality of wind turbines are imaged, the relative orientation of their blade discs may be determined.). Such images may most conveniently be provided from an elevated location, for example from an aerial craft (whether a conventional aircraft, drone/UAV, etc.), from a satellite, or even a suitable fixed platform. Such images may conveniently comprise a plurality of wind turbines, for example an entire wind farm. An elevated image of a wind farm may be termed a synoptic image because it provides information about the overall state of the orientation of the wind turbines within the farm.

The invention is therefore based upon the recognition by the inventors that suitable elevated images may be used for this purpose and, in particular, that it is both possible and convenient to extract the blade disc orientation from them. In particular, the recognition that such information may be extracted from satellite images is particularly significant as this facilitates the determination of blade orientation in remote or difficult-to-access locations.

Moreover, it is a process that may readily be automated, and so the invention may provide automatic way of measuring turbine blade position (i.e. yaw data) in a systematic and consistent way.

The yaw measurements may be integrated with additional data, including but not limited to: meteorological, oceanographic, operations and maintenance, and production data.

As noted above, the method of the invention is useful for measuring a plurality of wind turbines and it is preferred that the satellite image covers an entire farm or several wind farms. Thus, all turbines in a given wind farm may be measured at the same instant in time, which allows the measurements to be compared to each other. The invention may therefore provide some or all of the following advantages:-

1. Consistent performance monitoring - it enables monitoring of the efficiency of both single wind turbines and the entire wind farm.

2. Higher production and income - it facilitates an increase in the productivity of wind turbines because the method identifies and helps operators to adjust previously undetected yaw misalignment.

3. Lowers the loads on the turbine and extends the lifetime of turbine - it allows the owner/operator to increase the life of turbines by reducing loads.

4. Lowers the HSE risk - it removes risk of unnecessary inspection of a turbine. At present, turbines are inspected randomly so the invention saves both time and resources for unnecessary campaigns on correctly aligned turbines

5. Lowers the environmental footprint of wind turbine operation by reducing the number of trips to the offshore windfarm.

The invention may be performed manually. In other words, an engineer may obtain suitable image data, e.g. from a satellite, from which he or she may identify wind turbine(s) and measure the orientations of their blades. However, it is preferred that the process be automated by use of computing apparatus. The term “computer" is used herein to refer to any suitable system which, as is notoriously well known, will be provided with a processor, storage, memory, input and output means.

Accordingly, the method preferably includes the step of inputting image data into the computer. The image data may comprise optical images, synthetic aperture radar (SAR), etc. in the case of satellite image data, the data may be downloaded from any provider of such data. Alternatively, it may be obtained by means of aerial photography from a drone/UAV or (manned) aircraft.

Although an image may be sufficiently clear to use in its“raw” state, preferably it is pre-processed to remove noise and artefacts. It may also be convenient to normalise the image to a given size and scale.

The, or each, wind turbine may then be identified in the image. This may be done by means of any known pattern/image recognition algorithm such as the known types machine-learning based algorithms used for facial recognition. The algorithm may be trained to identify wind turbines from such images. Additionally or alternatively, a human operator may identify wind turbines from within the image, or he/she may verify those identified by the algorithm. Having identified the/each wind turbine, the blade disc may then be identified from within the (sub) image of the wind turbine in the case of satellite images, which are substantially from above, this will appear as a line (depending of course on the relative location of the satellite - if it is further from the overhead position, the disc will appear more elliptical). Having identified the disc, its orientation may then be determined - i e the angle of the line” with respect to north or some other predefined direction. Optionally, the diameter of the blade disc (i.e. the length of the“line”) may also be determined.

Where the image comprises a plurality of wind turbines (e.g. an entire farm), the orientation of the rotor blade disc of each of the plurality of wind turbines is preferably determined. The orientation of the rotor blade discs may then be compared to determine if one of the wind turbines is significantly misaligned.

Where the method is performed manually, this may be by inspection.

However, preferably the comparison comprises determining the average (e.g. the mean) orientation and determining whether any of the blade disc orientations deviates from the average by more than a predetermined threshold amount. The threshold may be a predetermined angle (e.g. more than 3 degrees, 5 degrees, 8 degrees, etc.) or more than a given fraction of the standard deviation of the angles. The latter arrangement has the advantage of accounting for the fact that sometimes conditions may be turbulent and so all of the wind turbines may“see” a different wind direction.

Having identified a misaligned wind turbine, the method preferably further comprises the step of performing remedial action in respect of a wind turbine identified as misaligned. A list of underperforming turbines and list of turbines with deviations may be combined to produce a list of turbines that are misaligned. The list will in turn be an O&M plan for the wind farm’s maintenance. A preferred application of the invention is to provide a decision-based tool to link to a revenue model for planning O&M activities so as to increase revenue and extend lifetime.

Satellite imagery may also be used to produce information on instantaneous wind direction and speed measurements throughout the wind farm co-incident with the turbine yaw measurements.

The invention may be performed using historic data and may do so using several sets of data over time to identify consistent misalignment issues.

Alternatively or additionally, it may be done using (substantially) real time image data - i.e. using the most recent available image data. As discussed above, the invention is preferably performed using a computer and therefore, viewed from another aspect, the invention provides a computing apparatus configured to perform the method of any preceding claim.

The invention also extends to a wind turbine (and indeed to a farm of wind turbines) controlled by means of the method described above.

Likewise, from another aspect it provides a software product containing code which when run on a computer enables it to perform the method described above.

In each case, the computer preferably performs the method according to the preferred features of the invention set out above.

Certain embodiments of the present invention will now be described, by way of example only, and with reference to the accompanying drawings in which:

Figure 1 is a flow chart showing a method of obtaining wind turbine rotor axis and length data according to an embodiment of the invention;

Figure 2 is an optical satellite image showing turbine axis measurement by means of the embodiment; and

Figure 3 is a synthetic aperture radar satellite image showing turbine axis measurement by means of a variant embodiment.

With reference to Figure 1 , a flow chart showing the overall method of the embodiment is provided.

The described embodiment is performed on conventional computing apparatus (not shown) using satellite image data. Starting at box 1 , the method begins with the receipt of suitable satellite data. This may be an optical image or other types of image, such as synthetic aperture radar (SAR). The images used may be obtained by downloading them from a commercial provider of such data. They may be substantially in real time or historic. The image data is input into the system for processing by means of an algorithm embodied in software running on the computer.

The next stage (box 2) is for the image to be pre-processed, to normalise the image to a standard size and format, filter out artefacts and noise, etc.

The pre-processed images are then processed at box 3 to identify images of wind turbines within them. This may be done by means of known image recognition algorithms, such as those based on machine learning systems. Additionally or alternatively, a human operator may identify and/or verify the wind turbines in the image. The number and location within the image of each wind turbine is stored. At box 4, for the first of the identified wind turbines, the long axis of the wind turbine - i.e. the plane of the blade disc - is identified from within the image portion identified as showing a wind turbine. Since the wind turbines are being viewed substantially from above, this appears as a line in the image.

The next step - shown in box 5 - is for the orientation of the disc (i.e. the orientation of the line relative to north in the image) is determined and stored. In addition, the diameter of the disc (i.e. length of that line) is determined and stored. The latter step is performed by scaling the length of the line in the image based upon the known scale of the image.

The algorithm provides a loop whereby the steps of boxes 4 and 5 are repeated for each turbine identified. Thus, box 6 provides a counter and determines whether there are further wind turbines to measure (“Yes”) or whether the last wind turbine has been measured (“No”).

Once the last wind turbine has been measured, the data is output into a table whereby for each turbine, its measured blade disc diameter (“blade length”) and orientation (“angle”) are provided.

Figure 2 shows an example of a portion of an optical satellite image 10 showing a wind turbine 11. The identified blade disc has been superimposed on the image as a line“a”.

Figure 3 shows a corresponding view of an SAR image. In this case, the identified blade disc is line“ ”.

It will be noted that in either case the length and angular orientation of the lines a and b may readily be determined.

Once the data has been obtained, it may be studied to identify if any of the wind turbines appears to be significantly misaligned. This is done on the basis that the wind direction should be substantially identical across a relatively small area such as a wind farm and therefore a wind turbine whose orientation is an“outlier” is likely to be misaligned. Thus, the average of the orientation angles is determined and the deviation from that average of each wind turbine is determined. A threshold is set - either in terms of a predetermined angle or a predetermined percentage of the standard deviation - any wind turbine having an angle which exceeds that threshold is identified as misaligned.

As noted above, the output of the system is in the form of a table. Table 1 below shows an example of data obtained using an optical satellite image in respect of a wind turbine farm comprising seven turbines.

Table 1

Here it may be seen that turbine 4 is an“outlier" in that its angle (125 degrees) is significantly different form all of the others, which lie in the 132 to 140 degree range.

The corresponding table for an SAR image (taken at a different time) is shown in Table 2 below.

In this case, turbine 4 is again shown as an outlier since at 187 degrees, its blade disc orientation differs significantly from the others, which are all in the 182 to 185 degree range. Having identified a turbine as being misaligned, remedial action can then be taken to determine and address the cause of the problem.