WO2017211367A1

WO2017211367A1 - Adaptive control of a wind turbine by detecting a change in performance

Info

Publication number: WO2017211367A1
Application number: PCT/DK2017/050184
Authority: WO
Inventors: Thomas S. Bjertrup Nielsen; Amit Kalyani; Vimal Kumar MANI; Karthik KRISHNAN JAMUNA
Original assignee: Vestas Wind Systems A/S
Priority date: 2016-06-07
Filing date: 2017-06-06
Publication date: 2017-12-14

Abstract

The disclosure provides a method of controlling a wind turbine, wherein the method comprises: a) measuring and recording power output values from a sensor, and measured values from at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor for the wind turbine, during operation of the turbine with a reference pitch angle and/or rotor speed regulation curve for a time period t1; b) changing the pitch angle and/or rotor speed regulation curve(s); c) operating the wind turbine with the curve(s) from step b) during a time period t1 and receiving measured values from a wind turbine power output sensor and at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor; d) optionally repeating steps b) and c) one or more times; e) comparing a power output of the wind turbine obtained at step a) with a power output obtained in step c) during operation of the wind turbine under one or more changed pitch angle and/or rotor speed regulation curve(s); f) determining whether a power output obtained at step c) is higher than a power output obtained at step a); and g) optionally, changing the pitch angle and/or rotor speed reference curve(s) for the wind turbine to a curve or curves identified in step f) that results in higher power output than the power output measured in step a). A computer programme implementing the method, a controller configured to execute the method and a wind turbine comprising the controller are also provided.

Description

Adaptive control of a wind turbine by detecting a change in performance

Field of the invention

The present invention relates to methods and systems for controlling wind turbines and, more specifically, to methods and systems for detecting a change in the performance of a turbine.

Background of the invention

Generally speaking, wind turbines used for power generation convert the kinetic energy of wind into electrical energy. Due to the growing need for alternative sources of energy that do not rely on fossil fuels, wind turbines are increasingly used for providing energy into the electrical grid. Wind turbines used for electrical power generation typically include a rotor with a plurality of blades (e.g. three) attached to a nacelle located at the top of a tower, and coupled to a generator that converts the rotational energy of the rotor into electrical energy.

Wind turbines are typically controlled using complex models and methods relying on a plurality of parameters and measurements, related to the ambient conditions and the turbine itself, in order to optimise the production of energy and the lifetime of the turbine. Generally, the optimal operating parameters are determined under test conditions and are then imported into the control software for the turbine. Small differences in e.g. blade twist distributions, pitch angle markings in the blade roots, blade torsional stiffness, blade outer surface shape, blade surface roughness and level of seasonal rain can result in different models or parameters being optimal for different turbines. Even for the same wind turbine, the same operating conditions may not be optimal throughout the operation of the turbine, as e.g. events such as wear of the blades (e.g. leading edge erosion), build-up of dirt, insects, ice or any other irregularities on the blade surface may change the aerodynamics of one or more blades (lift and drag as a function of angle of attack) and impact the behaviour of the turbine, such that the initial model and/or parameters may no longer be optimal. This can have a significant impact on the efficiency and profitability of a wind turbine.

Document EP1959130 A2 discloses a method for optimising the operation of a wind turbine based on establishing a relation between a measured response variable and a control parameter, taking into account one or more ambient condition measured variable. The approach intends to adjust controller settings taking into account ambient conditions.

Accordingly, there is a need for new methods to adapt operating settings of a turbine in use, preferably without requiring a control or knowledge of the ambient conditions, and to detect

l events that impact the performance of a wind turbine, in particular events associated with the blades.

Summary of the invention

The invention generally relates to control systems, methods and computer programs for operation of wind turbines, and in particular, to control of wind turbines by adapting operating conditions for improved power production of a wind turbine. The invention can be performed at any desired interval during the operation of the wind turbine to check whether the current control parameters for the wind turbine are still capable of providing a desirable or optimal power output, or whether a new / different set of control parameters might provide improved performance, such as improved or increased power output. The methods of the invention desirably have minimum disruptive effect on the operation of the wind turbine and, therefore, may be used relatively frequently without adversely affecting the expected power output. Having performed a method of the invention it may be confirmed that the current set of wind turbine control parameters are optimal and/or suitable, meaning that no change is necessary to the operational controls; alternatively, it may be found that one or more different sets of control parameters provide an improved performance (such as increased power output), in which case the control parameters of the wind turbine may be changed to one of those sets of different control parameters (typically the set that provides the highest power output of those tested). Physical or technical faults or irregularities in the wind turbine leading to reduced power output may also be detected by way of the invention.

Thus, the invention also relates to the detection of loss in performance due to the presence of surface irregularities on the blades. The invention is particularly advantageous for monitoring either gradual or sudden changes in performance due to the presence of surface irregularities on the blade, and issuing warning messages or triggering appropriate actions. Loss of performance may cause operating parameters to be changed as part of the wind turbine adaptive operation of the invention and these changes can be used as an indication of a loss of performance.

According to a first aspect of the invention, there is provided a controller for a wind turbine comprising: a processor; an input/output interface; and a memory including instructions that, when executed by the processor, cause the processor to: a) receive via the input/output interface measured values of the wind turbine output power and measured values of at least one of a rotor speed and pitch angle, during operation of the turbine with a reference pitch angle and/or rotor speed reference value for a time period t1 , the reference pitch angle being based on a pitch angle reference curve and/or the rotor speed reference value being based on a rotor speed regulation curve; b) change the pitch angle and/or rotor speed value; c) operate the wind turbine with the changed pitch angle and/or rotor speed value for the time period t1 and receive measured values of the wind turbine output power and at least one of a rotor speed and pitch angle position; d) optionally repeat steps b) and c); e) compare a power output of the wind turbine obtained at step a) with a power output obtained in step c) during operation of the wind turbine under one or more changed pitch angle(s) and/or rotor speed value(s); and f) determine whether a power output obtained at step c) is higher than the power output obtained at step a). In some embodiments, the processor is additionally configured to g) change the pitch angle and/or rotor speed value(s) for the wind turbine reference curves to the value(s) identified in step f) that results in higher power output than the power output measured in step a).

Comparing the power output in step e) may comprise comparing measurements acquired over a predetermined range of wind speed, rotor speed and/or power output. In some embodiments, the predetermined range of wind speed may be below a rated wind speed of the wind turbine, below a rated power of a turbine, or below a rated rotor speed of a turbine. In some embodiments, the predetermined range of rotor speed may be below the rated rotor speed. In some embodiments, the predetermined range of rotor speed may be below the rated rotor speed and above the minimum rotor speed of the turbine.

Steps a) to c) may be repeated multiple times to obtain multiple points and comparing the power outputs at step e) may comprise comparing statistical estimates or cumulated values over the distribution of points. The changed pitch angle and/or rotor speed regulation curves of step b) may be chosen from a collection of pre-computed curves or may be obtained by applying pre-defined variations around the reference curve(s). In some embodiments, pre-defined variations around a pitch angle regulation curve may comprise one or more of the pitch angle offset of ±0.5°, ±1.0°, ±1.5°, ±2.0°; and/or pre-defined variations around a rotor speed regulation curve comprise one or more of the tip speed ratio offsets of ±0.2, ±0.4, ±0.6, ±0.8, or ±1.0.

Step g) may be performed if a power output obtained at step c) is higher than the power output obtained at step a) by at least a difference threshold value For example, a difference in accumulated power of at least 0.1 %, at least 0.2%, at least 0.5% or at least 1 % may be used as a threshold. In some embodiments, differences in accumulated power between 0.1 % and 3% may be considered sufficient. In some embodiments, step g) may be performed if the pitch offset value resulting in higher power output is within pre-defined limits with respect to noise and/or if it results in loads that are within the design loads of the turbine components. In some embodiments, a calibration run may be performed to determine one or more of: a number of times that a measurement cycle (steps a) to c)) is repeated, the period of time t1 , the number of settings tested (step d)), and/or the difference threshold value at step f) for step g) to be performed. In some embodiments, the period of time t1 may be between 5 seconds and 10 minutes. In preferred embodiments, the period of time t1 may be between 10 and 40 seconds.

Measured values may be separated into wind speed bins according to the measured or estimated wind speed at the time the data point is acquired. In some embodiments, the estimated wind speed associated with a measurement may be derived from power output and/or rotor speed measurements. In some embodiments, the measured average or accumulated power over t1 may be used to estimate the wind speed based on an expected relationship between wind speed and power output for a turbine. In some embodiments, measured values may be separated into groups depending on the average rotor speed and/or accumulated or average power output measured or estimated over t1 in relation to the minimum rotor speed, rated rotor speed and nominal power output of a turbine.

In some embodiments, the width of wind speed bins may be predetermined. In some embodiments, the width of wind speed bins may be constant over the range of wind speeds. In preferred embodiments, the wind speed bins may be about 0.5 m/s wide.

Alternatively, the width of wind speed bins may vary over the range of wind speeds. For example, the wind speed bins may be between 0.1 and 2 m/s wide

It will be appreciated that unless otherwise required the steps / actions of the controller may, in accordance with embodiments of the invention, be performed in any appropriate order. Advantageously, at least steps a) to g) may be performed at predefined intervals or events during operation of the wind turbine. For example, at least steps a) to d) may be performed at regular predetermined intervals during the operation of the wind turbine.

Steps e) to f) (and optional step g) if performed) may be performed at regular predetermined intervals during operation of the wind turbine. In some embodiments, steps a) to f) (and optional step g) if performed) may be performed at regular predetermined intervals during operation of the wind turbine. In some embodiments, steps a) or e) to f) (or g) may be performed every quarter year, e.g. at the start of every season, every 6 months, at the start of summer / winter, after a defined number of days, or every month. In some embodiments, steps a) to f) (and optional step g) if performed) may be performed on request by a user.

It may be desirable for steps a) to f) to be automatically performed following an event, instead or in addition to the steps being performed following a schedule and/or upon request. For example, the event may be selected from any one or more of the following: the occurrence of rain; a defined number of days without rain; a defined change in

environmental conditions; and a defined change in average temperature. In some embodiments, the pitch angle regulation curves may be different for each blade of a wind turbine, and steps a) to f) may be performed for each blade of a wind turbine

individually. In some embodiments, the rotor speed regulation curves may be different for each wind direction sector. In such embodiments, steps a) to f) may be performed separately for each wind sector.

In some embodiments, alone or in combination with any other embodiments, the reference regulation curve(s) may differ during operation of the wind turbine according to the period of the day. Advantageously, there may be two or more periods of the day and the period of the day may be recorded during steps a) to c); and steps e) to f) (and optional step g), if performed) may be performed separately for power output measurements obtained during each of the two or more periods of the day. For example, the two or more periods of the day may comprise a daylight period and a night-time period.

The controller in any embodiment of this aspect of the invention may additionally be configured to: h) produce an output when the difference between a power output of step a) and a power output of step c) is larger than a difference threshold value and/or when the difference between the changed reference curve(s) at step g) and any previous reference curve is larger than a difference threshold value. In some embodiments, step h) may be performed if there is an offset of at least about 5% between a changed reference curve and any previous reference curve. In some embodiments, step h) may be performed if the accumulated power for one or more wind speed bins with a given offset changes by at least 2%. In some embodiments, steps a) to g) may be repeated at different times throughout the operation of the turbine, and the rate of change of the highest power output at step f) or of the changed reference curve(s) at step g) over time may be calculated, and an output may be produced if the rate of change exceeds a rate threshold value. In some embodiments, an output may be produced if the accumulated power for any or all wind speed bins for the same offset decreases by at least 2% between two repetitions of the test. In some embodiments, an output may be produced if the difference between accumulated power for each or any wind speed bin for any pitch offset changes significantly between two repetitions of the test. For example, an output may be produced when the difference in accumulated power in any or all bins between two pitch offsets varies by more than 5% between two repetitions of the test. In some embodiments, an output may be produced if the absolute difference in accumulated power for any or all wind speed bins for the same pitch offset increases by more than a threshold and continues to increase over time. For example, an output may be produced if between tests performed every y days, the absolute difference in accumulated power for any or all wind speed bins for the same pitch offset increases by more than x%. In some embodiments, x% may be 2%. In some embodiments, y days may equal to approx. 6 months. For example, if a test is performed every half year, and the absolute difference in accumulated power between the same pitch offsets between consecutive half-yearly tests increases by more than 2%, an output may be produced. Producing an output may comprise transmitting a warning message to a user via an input/output interface. In addition or in the alternative, producing an output may comprise producing a signal to a blade cleaning system or a blade defrosting system to commence or schedule a cleaning or defrosting cycle, respectively, of the wind turbine blades. In some embodiments, the processor may additionally perform the step of: f) obtaining additional data relating to wind turbine operating parameters and/or environmental conditions proximate the wind turbine. For example, the additional data may comprise one or more of temperature data, rainfall data, time since last maintenance, time to next

maintenance, age of the wind turbine blades, expected wear of the blades, expected or recorded seasonal events such as insect density, environmental conditions leading to the presence of salt on the blades, etc.

In some embodiments, comparing the power output in step e) comprises comparing measurements acquired over a range of wind speed between about 66% and 99% of the rated wind speed of the wind turbine. In some embodiments, assessment of the difference between a power output of step a) and a power output of step c) and/or the difference between the changed reference curve(s) at step g) and any previous reference curve may comprise comparing measurements acquired over a range of wind speed above about 66% of the rated wind speed, for example, between about 66% and 99% of the rated wind speed of the wind turbine, or between about 66% and 100% of the rated wind speed. In some embodiments, measurements acquired in the operating region where the rotor is operated at its rated speed and the power output is below the nominal power of the turbine may be used. In accordance with a second aspect of the invention, there is provided a method of controlling a wind turbine, the method comprising: a) measuring and recording power output values from a sensor, and measured values from at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor for the wind turbine, during operation of the turbine with a reference pitch angle and/or rotor speed regulation curve for a time period t1 ; b) changing the pitch angle and/or rotor speed regulation curve(s); c) operating the wind turbine with the curve(s) from step b) during a time period t1 and receiving measured values from a wind turbine power output sensor and at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor; d) optionally repeating steps b) and c) one or more times; e) comparing a power output of the wind turbine obtained at step a) with a power output obtained in step c) during operation of the wind turbine under one or more changed pitch angle and/or rotor speed regulation curve(s); and f) determining whether a power output obtained at step c) is higher than a power output obtained at step a). In some embodiments, the method may additionally comprise optional step g) changing the pitch angle and/or rotor speed reference curve(s) for the wind turbine to a curve or curves identified in step f) that results in higher power output than the power output measured in step a).

Embodiments of this second aspect may comprise any of the additional features described above in relation to the first aspect.

According to a third aspect of the invention, there is provided a computer programme for a controller of a wind turbine, that when executed by the controller causes the controller to perform any of the methods of the invention as described herein, for example the methods as described in connection with the second aspect of the invention, optionally comprising any of the features described in relation to the first or second aspect of the invention.

According to a fourth aspect of the invention, there is provided a wind turbine comprising any embodiments of the controller described herein, for example the controller described in connection with the first or second aspect of the invention.

In any aspects of the invention as described above, the power output may be a

measurement directly obtained from a power output sensor or may be indirectly obtained from another measured quantity. In particular, the power output measurement may be obtained from various indirect measurements as an example a main shaft torsion or torque measurement; a gearbox shaft torsion or torque measurement; blade strain or load measurements; strain measurements on the blade bearings, blade bolts or hub; blade surface pressure measurements; blade tip mean deflection in edgewise direction; generator current; transformer power, transformer current, transformer power, transformer current and tower lateral moment. Brief description of the drawings

One or more embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

Figure 1 illustrates a typical model for the control of a wind turbine at different load / wind speed conditions (I, II, III and IV).

Figure 2 illustrates the effect of blade surface irregularities on the power output and the pitch angle regulation process for a wind turbine.

Figure 3 is a flowchart illustrating a method of operating a wind turbine according to a general embodiment of the invention.

Figure 4 illustrates schematically a wind turbine according to embodiments of the invention. Figure 5 is a flowchart illustrating a method of adapting operating settings of a wind turbine according to embodiments of the invention.

Figure 6 illustrates an example of a test of operating settings according to embodiments of the invention.

Figure 7 provides an example of the effect on power output of dirt accumulation on the blades of a wind turbine, as well as the effect of dirt accumulation on the pitch angle regulation process.

Figure 8 shows an example of settings that may be tested according to embodiments of the invention.

Description of embodiments

Although the invention will be described by way of examples, it will be appreciated by a person skilled in the art that the invention could be modified to take many alternative forms without departing from the spirit and scope of the invention as defined in the appended claims.

Control of wind turbines

Referring to Figure 1 , an example of wind turbine control will be explained. Figure 1 illustrates example reference curves of rotor speed ω, power output P and blades pitch Θ as a function of the wind speed V. Looking first at the Power curve, as the wind speed increases above a minimum or "cut-in" speed (not shown), the rotor starts to rotate and the turbine starts to produce electrical power (load / wind speed condition zone I). The power P generated typically increases with wind speed in the "partial load" region (zones 1, 11 and III), until the wind speed reaches a nominal wind speed (also called "rated speed") for the wind turbine (zone IV), at which the turbine generates its nominal power (also called "rated power"). In this zone (zone IV), the turbine functions at "full load".

Each wind turbine may be controlled by a local or remote controller that may regulate the rotor speed, blade pitch angle and, hence, power output of the turbine. The purpose of the controller is to operate the turbine according to the programmed objectives. Optimizing of power is not generally true. In the full load for example power is far from optimized. The turbine controller typically comprises a pitch controller that adjusts the pitch of the rotor blades in order to adjust the amount of wind power captured by the rotor, and/or a rotor speed controller. In the partial load region, the controller may aim to regulate these functions (pitch and rotor speed) in order to operate in conditions as close as possible to the optimal C_p, so as to maximise power capture.

Referring now to the rotor speed ω and blade pitch Θ curves in Figure 1 , in zone I the rotor functions at its minimum speed, and the controller will therefore adjust the pitch angle Θ in order to try to achieve the above objective of maximising C_p. In zone II, the speed of the wind is sufficient for the rotor speed to increase beyond its minimum speed. In this zone, the pitch angle will be kept constant and the rotor speed will be increased proportionally to the wind speed, in order to maintain a ratio of rotor speed to wind speed at which C_p is optimal. As mentioned above, for a given wind speed, the maximum power coefficient C_p ^* can be achieved with a particular blade pitch setting and rotational speed. In other words, for a given pitch angle, each turbine will have a curve of C_p as a function of the tip speed ratio λ (which relates the rotor speed to the wind speed), and this curve will have a single maximum. Conversely, for a given rotor speed, there will be a curve of C_p as a function of wind speed, which will also have a positive optimum. Therefore, as the wind speed increases in zone II, the pitch angle is maintained constant and the rotor speed is adjusted proportionally to the wind speed in order to maintain λ at the value that corresponds to the maximum achievable power coefficient C_p ^* . When the rotor speed reaches its maximum value or "rated speed", it cannot be increased anymore (zone III), and the pitch angle will be adjusted to stay at the maximum power coefficient C_p ^* , as long as the wind speed stays under the rated wind speed. When the wind speed increases beyond the rated wind speed, the turbine operates at maximum power and therefore the pitch angle will be adjusted to use a suboptimal C_p with the objective to maintain (i.e. not exceed) the nominal power production. As used herein, a pitch angle reference curve refers to any collection of parameters that may be used for setting the pitch angle of a wind turbine depending on the wind speed according to a regulation model for a wind turbine. This may be e.g. in the form of a curve or mathematical relationship between a regulated pitch angle and a wind speed, a set of logical rules, a combination of values for specific wind speed ranges, etc. Similarly, as used herein, a rotor speed reference curve refers to any collection of parameters that may be used for setting the rotor speed (or tip speed ratio) depending on the wind speed. This may be in the form of a curve, a function, a set of logical rules, a combination of values for specific wind speed ranges, etc. As used herein, any reference to a value, setting or curve for a rotor speed may be used interchangeably to refer to the corresponding tip speed ratio and vice versa. As used herein, the terms "rotor speed values" and "rotor speed measurements" may also refer to corresponding generator speeds. As used herein, the term "wind speed" may refer to either a measurement from a sensor or a wind speed estimate indirectly obtained from other measurements, such as a power output measurement / estimate and/or a rotor speed measurement. For example, a wind speed estimate may be indirectly obtained from a power output value based on the knowledge of the expected power curve of the wind turbine. As such, although the terms "wind speed", "wind speed ranges" and "wind speed bins" are used throughout this application for ease of understanding, the person skilled in the art would understand that no direct wind speed measurement is necessary for the performance of the invention. Wind speed estimates and ranges of wind speeds associated with a power output measurement may be derived from knowledge of the optimal power in different operating regions of a wind turbine, such as in regions defined by the minimum rotor speed, rated rotor speed and nominal power (see also explanations in relation to Figure 1 above).

As used herein, the "measured power output" or "power output" of a wind turbine may be a directly measured quantity (e.g. from a power output sensor) or may be indirectly obtained from another measured quantity. In particular, an improved regulation of pitch angle and/or rotor speed of a turbine may result in an increase of the lift versus drag force on the blades of the turbine, resulting in increased blade moment / torque and hence increase in the cumulative blade moments and main shaft torque. As this will result in an improved power production, any of these values may be used as an indication of a change in power output. For example, blade load sensors in a turbine may be used to obtain a measurement of blade torque, and this may be used as an indication of a change in power output as different reference curves are used to operate the turbine. In some embodiments, a "measured power output" or "power output" may be derived from any of the following measurements, alone or in combination: a main shaft torsion or torque; gearbox shaft torsion or torque; blade strain measurements (as an indication blade torque) such as from strain gauges, optical fibres, blade load sensors, etc.; strain measurements on the blade bearings, blade bolts or hub (as an indication of increased torque); blade surface pressure measurements (from which lift force and pressure drag may be determined in one or more blade cross sections, based on which an indication of torque change can be obtained); blade tip mean deflection in edgewise direction; generator current; and tower lateral moment from e.g. load sensors on tower moments, tower top acceleration or tower top lateral deflections (as the change in main shaft torque is transferred at least partly to the tower top). As the person skilled in the art would understand, any value that is proportional to the power output of the turbine may be used in the context of the invention to compare settings, as an indication of the relative power output associated with the settings tested. Factors that affect the behaviour and performance of a turbine

Wind turbines are normally operated to produce the maximum amount of electrical power possible according to a predicted behaviour of the turbine as a function of wind speed based e.g. on a theoretical model parameterised for a given turbine, as explained above. Such models may be used to derive a collection of parameters that may be used for setting the pitch angle and/or rotor speed of a wind turbine depending on the wind speed, and operate the turbine accordingly. In order to adjust the pitch and/or rotor speed as described above, a wind turbine comprises means to obtain a measurement or an estimate of the wind speed. This can be in the form of an anemometer, or it can be estimated from another

measurement, such as e.g. the power output (based on a relationship between wind speed and expected power output) and/or the rotor speed (based on a relationship between rotor speed and expected power).

Evidently, optimal regulation of wind turbine operation relies on an appropriate model of the behaviour described above, i.e. on using the appropriate collection of parameters for setting the pitch angle and/or rotor speed. In particular, the model may need to be changed or adapted for a particular turbine or location, or when an (expected or unexpected) event happens that modifies the properties of the turbine. For example, anything that would change the aerodynamics of the blade may result in a departure from the expected / predicted behaviour of the turbine. Events that commonly occur and that affect the performance of the blades of a wind turbine may include the presence of dirt, insects, salt etc. on the blades; the presence of ice on the blades; and/or leading edge erosion of the blade surface as the blade ages (e.g. due to wear and impact events). All of these events will result in a gradual or sudden increase in friction on the blade, thereby increasing drag and reducing the amount of power produced under given conditions. Additionally, turbines may also differ in terms of blade twist distributions, pitch angle markings in the blade roots, blade torsional stiffness and blade surface roughness. As a consequence, operating parameters that are a priori optimal may not prove to be optimal for a particular turbine or in particular conditions, or may no longer be optimal as properties of the turbine or of the environment vary throughout the use of the turbine.

Additionally, a change in the expected behaviour of the turbine can be used as an indicator that one or all of the blades of the wind turbine are exhibiting reduced performance.

Figure 2 illustrates an example of the effect of e.g. blade surface irregularities on the power output and the pitch angle regulation process for a wind turbine, when the wind turbine is controlled as described above. Looking at the pitch curves on Figure 2, it can be seen that the optimal pitch curve as a function of wind speed is influenced by the presence of surface irregularities, such that two different curves may be optimal when the blades are regular (expected / reference curve) or when irregularities are present on the surface of the blade (modified / irregular blades curve). In particular, the pitch angle that results in a given power output, see the settings indicated with filled circles on Figure 2 (or conversely the power output at a given pitch angle), and the pitch angle that is optimal at a given wind speed may vary. This may be particularly severe at wind speeds where the turbine approaches nominal power (knee of the power curve). As a consequence, operating the turbine with the modified pitch-angle regulation curve may result in a higher power production than operating the wind turbine with the expected / reference pitch angle regulation curve, at any given wind speed. Therefore, according to the methods of the invention, it is possible to detect a difference in power output when operating a turbine using alternative regulation curves from the current reference curve(s), and optionally to change the current reference curve to one that resulted in an improved power production. This may be achieved solely based on a measurement of the wind speed and power output of the turbine, both of which are typically already monitored in turbines used for power generation. Looking at the power curves (solid and long dashed lines), it can be seen that the presence of irregularities on the surface of the blades may result in a change in the parameters of the curve of power output achievable as a function of wind speed. In particular, a reduction in power output between an expected ("clean blades" / reference) curve and a modified curve (in the case where surface irregularities are present) may be observed at any given wind speed in the partial load region. Therefore, according to the methods of the invention, it is additionally possible to detect the effect or and, hence, the presence of blade surface irregularities on some or all of the blades of a wind turbine based on a difference in the power output achievable at a given wind speed and/or based on a difference between the regulation curve that results in the achievable power output.

Indeed, comparing the power generation, at a given wind speed, for normal blades

(expected value) with the power generation actually achieved (i.e. the vertical difference between the two horizontal dashed lines in Figure 2) indicates that the blades are currently less / under-performing. In other embodiments, alone or in combination with the above, a turbine may be identified as having reduced performance if the wind speed at which nominal power is reached differs from the expected rated wind speed (horizontal difference between the two vertical dashed lines in Figure 3). Adaptive control of wind turbines

Adaptive control of wind turbines relies on testing different models, e.g. different pitch and/or rotor speed reference curves and determining if a new setting results in better power production.

Figure 3 shows a flowchart of a method of operating a wind turbine according to a general embodiment of the invention. At step 400, a wind turbine is operated based on a reference pitch angle curve and/or rotor speed regulation curve during a period of time t1 , and the measured output power, rotor speed and/or pitch angle during t1 are recorded. At step 410, the pitch angle and/or rotor speed values are changed. At step 420, the turbine is operated for a time period t1 with the new settings and the measured output power, rotor speed and/or pitch angle are recorded. In some embodiments, steps 410 and 420 may be repeated a number of times with different values. Advantageously, in some embodiments, alone or in addition with the above embodiments, steps 400 to 420 may be repeated multiple times in order to collect multiple data points using the reference value(s) and each of the one or more changed value(s). At step 430, the power output recorded at the one or more instances of steps 400 and 420 are compared in order to identify any changed setting that results in a higher power output. In one embodiment the accumulated mean power during t1 recorded over multiple instances of steps 400 and 420 are used to compare power outputs for the different changed settings. Optionally, an additional step 440 may be performed where the reference values are changed to a value identified in step 430 as resulting in a higher power output. Any or all previously used reference values, and/or any or all previously tested changed values may be recorded and used, for example, to monitor the change in achievable power output and/or the change in optimal settings throughout the use of the turbine. In some embodiments, the test may be repeated a number of times. In some embodiments, the test may be repeated until a statistical estimate of the difference in power production may be obtained. In some embodiments, a setting may be considered to result in a difference in power output (i.e. may be considered as "significant") when the power output difference (or statistical estimate thereof) exceeds a threshold. In some embodiments, a difference between values that exceeds 0.2% may be considered significant. In some embodiments, a difference exceeding 0.5, 1 , or 2% may be considered significant. In some embodiments, a threshold may be predetermined, such as based on a known/quantified uncertainty around the measurements. In some embodiments, the uncertainty around the measurements may be quantified empirically in a calibrating period prior to the normal use of the turbine. In some embodiments, a threshold may be dynamically adjusted based on the statistical uncertainty around measurements corresponding to the same conditions. For example, a threshold may be determined based on a number of standard deviations of the distribution of power outputs measured at the same wind speed and pitch/rotor speed settings.

In some embodiments, the wind speed may be used to determine if power output data collected over the testing periods for compared settings is comparable. In some

embodiments, data may only be compared if the wind speed (e.g. average wind speed) over the testing time period for compared settings is comparable. In embodiments, in order to facilitate the acquisition and processing of data, measurements may be aggregated over wind speed bins, such as wind speed bins of 0.1 m/s, 0.5 m/s, 1 m/s, 1.5 m/s or 2 m/s width. Preferably, wind speed bins of 0.5 m/s may be used. For example, measurements acquired using a particular pitch angle and/or rotor speed in a wind speed bin may be compared to measurements using a different pitch angle and/or rotor speed over the same bin.

As the person skilled in the art would understand, the width of the bins represents a trade-off between the precision of estimates obtained and the accuracy of estimates due to the availability of measurement points falling within a bin. As such, the width of the bins may vary along the range of wind speeds observed for a particular turbine at a particular site, such that e.g. bins may be narrower around wind speeds that are frequently observed. In such embodiments, bin width may depend on the density of measurements along the wind speed axis, i.e. bins may be narrower in regions of more frequent wind speeds. Wind speed bins that correspond to areas where large differences are expected, such as e.g. around the knee of the power curve, as described above, may also be narrower as less measurements would be required to be able to conclude that a difference in power output is significant. The appropriate width of bins may be determined prior to putting the turbine into use (e.g. based on the expected behaviour of the turbine and/or expected environmental conditions on a site), or may be adjusted in use. For example, the appropriate width may be adapted through a learning/adaptive process throughout use, or may be changed depending on the time of the year and corresponding changes in expected environmental conditions, such as depending on the frequency of wind speeds falling in each bin.

Advantageously, the power output may be monitored below the rated wind speed of a turbine, in particular in areas where comparatively large differences may be observed or expected. For example, the power output may be monitored between 4 and 12 m/s, between 4 and 10.5 m/s, between 4 and 9 m/s, between 5.5 and 12 m/s, between 5.5 and 10.5 m/s, between 5.5 and 9 m/s. As the person skilled in the art would understand, these values may depend on the wind turbine, site and settings and any appropriate region around the nominal (rated) wind speed may be used, such as for example any region within 10, 15, 20, 25, 30, 35, or 40% of the rated wind speed. In some embodiments, the power output may be monitored at wind speeds where the rotor functions at minimum speed (e.g. region I in Figure 1), and/or at wind speeds where the rotor speed is between minimum speed and rated rotor speed (e.g. region II in Figure 1), and/or at wind speeds where the rotor is operated at rated speed and the power output is below nominal power (e.g. region III in Figure 1). For example, measurements corresponding to wind speeds above about 4 m/s and below about 10.5 m/s may be used. In some embodiments, regions where the rotor is operated at rated speed are not used, for example, because of load / noise restrictions. In such embodiments, measurements corresponding to wind speeds below about 9 m/s may be used. In some advantageous embodiments, for example, when tip speed ratio reference curves are tested, measurements corresponding to wind speeds where the rotor is operated between minimum and rated speed may be used (e.g. region II in Figure 1 above, where the optimal pitch angle may be expected to be constant). In such embodiments, measurements corresponding to wind speeds between 5.5 and 9 m/s may be used. In some embodiments, the settings (pitch angle and/or rotor speed) tested are selected from a user-input list of settings, such as from a user input list of pitch angle regulation and/or rotor speed regulation curves. In some embodiments, the settings tested are selected as the closest settings to the currently used ones, from a predetermined list of settings. In some embodiments, the settings to be tested are calculated on the fly. In some embodiments, the settings to be tested are calculated based on small differences around the currently used settings. In some embodiments, the settings to be tested are selected based on any known method to sample a parameter space. In some embodiments, two alternative settings are tested for a difference in power output compared to the situation where the turbine is operated with the reference value. In embodiments, the settings are on either side of the currently used reference setting, e.g. the pitch angle regulation curves located just below and above the currently used curve at the current wind speed, or the pitch and/or rotor speed that differ (i.e. are larger and smaller, respectively) by a set value from the reference value. In some embodiments, settings comprise different pitch angle regulation curves, different rotor speed regulation curves, or both (alone or in combinations). In some embodiments, settings may comprise slight pitch offsets in particular tip speed ratio intervals, such as ±0.5 degree, ±1 degree, ±1.5 degrees, ±2 degrees, etc. In some embodiments, settings may comprise slight tip speed ratio offsets, such as ±0.2, ±0.4, ±0.6, ±0.8, ±1.0, etc. In some embodiments, settings may comprise slight modifications of the reference rotor speed regulation or pitch angle regulation curves in some areas of the curves, instead or in combination with the use of a different curve. In some embodiments, multiple settings, such as 2, 3, 4, 5, 6, 7, 8, 9 or 10 settings may be tested in a test run.

In some embodiments, a test may be automatically triggered, for example at regular intervals. In embodiments, tests may be triggered every 6 months, every season, every 4 months, every three months, every 2 months or every month. In some embodiments a test may be triggered manually by an operator. In some embodiments, the test interval may be determined depending on the site. For examples, the testing interval may be higher on sites with larger variations in blade dirt or insect build-up, on sites with larger seasonal variations, or on sites with more severe occurrence of leading edge erosion (e.g. offshore). In embodiments, the interval between tests is a parameter set by an operator. In some embodiments, the interval between tests may be varied throughout the year, such as e.g. to increase in periods of more intense dirt/insects build-up. In some embodiments, a test may be automatically triggered by an event. In some embodiments, an event may be a heavy rainfall. In some embodiments, an event may be a predetermined period of time without rainfall. For example, a test may be triggered after a period of 5, 10 or 15 days without rain. In some embodiments, a test may be performed on a wind turbine that is part of a group of wind turbines, e.g. a test may be performed for all or some of the turbines present on a site. In embodiments, the results of the test may be extrapolated to all turbines in the group. In embodiments, some or all turbines in a group may be tested but the list of settings tested may vary depending on the group of turbines (e.g. location on a farm). In some

embodiments, separate tests may be run for multiple wind sectors. In some embodiments, different preferred settings may be determined depending on the wind sector. In some embodiments, data from tests may be separated depending on the wind sector, for example based on wind sectors of 15 degrees (wherein e.g. Sector 1 : 0 - 15 deg., where 0 deg. is North; Sector 2: 15 - 30 deg., etc.). In some embodiments, different rotor speed regulation values may be found for each wind sector. In some embodiments, wind sectors

corresponding to the most frequent wind directions may be prioritised for testing. For example, knowledge of the wind rose (i.e. main wind directions and percentage of time in each wind direction sector) may be used to prioritise the wind sectors to test.

In some embodiments, separate tests may be performed for each blade of a turbine. In such embodiments, tests may be run for each blade in succession, by keeping the settings for all other blades constant and testing new settings for blade 1 only, then blade 2, etc. In such embodiments, different settings, e.g. different pitch angle regulation curves may be selected for each blade.

In some embodiments, separate tests may be performed for different time periods or day periods, such as e.g. for night and day conditions. In such embodiments, different settings may be used for different time periods, e.g. for night and day time. Such embodiments may be particularly advantageous on sites where larger differences in wind shear are observed in night and day time or where larger differences in temperatures are observed in night and day time. Time/day periods may be pre-defined, such as e.g. between 8pm and 8am, and 8am and 8pm (or 7am-7pm, 6am-6pm, 9pm-6am, etc. based on e.g. location and season), or may be defined based on temperature or light measurements. For example, tests may be assigned to a "day test" or a "night test" depending on whether the test was performed during a period of relative high temperature/light levels or not.

Figure 4 shows schematically a typical wind turbine for use in the context of the disclosure. A wind turbine 10 may comprise a rotor 12 coupled to a generator 14 providing power to the grid 16. A pitch controller 18 may control the pitch of the blades by providing a pitch control signal 20 to a pitch control mechanism in hub 22. The pitch controller 18 may regulate the pitch of the blades based on a difference between a pitch position signal 24 measured by a pitch position sensor 26 and a pitch position command signal 28 provided by a turbine controller 30. The pitch controller 18 may determine a pitch control signal 20 based on a Proportional-lntegral-Derivative (PID), Proportional-Integral (PI) or any other type of control loop feedback mechanism known to the person skilled in the art. Pitch controls that do not rely on feedback signals from the position sensor 26 (i.e. open loop controller) may alternatively be used. As the person skilled in the art would understand, the pitch controller 26 may be designed as a separate controller or may be comprised in a turbine controller 30 that directly provides a pitch control signal 20 to a pitch control mechanism. As the person skilled in the art would understand, although a single pitch controller 18, pitch command signal 28, position signal 24 and sensor 26 are described, multiple controllers, sensors and signals may be provided to control the pitch angle of individual blades.

Similar considerations apply to a rotor speed controller 44 that may control the rotational speed of the rotor by providing a speed control signal 46 to a rotor speed control mechanism in rotor 12. Likewise, a rotor speed controller 44 may regulate rotor speed based on a difference between a rotor speed signal 48 from a rotor speed sensor 50 and a rotor speed command signal 52 provided by the turbine controller 30. A similar set up may be used for a yaw controller that regulates the orientation of the turbine in relation to the direction of the wind as determined by a wind direction sensor (not shown). Open loop controllers are also usable for both the rotor speed controller 44 and the yaw controller. As the person skilled in the art would understand, the precise implementation of all auxiliary controllers and sensors does not significantly impact the methods described in the present disclosure.

The turbine controller 30 may include a processor 32, a memory 34, and an input/output interface 36. The processor 32 may include one or more processing circuits, and the memory 34 may comprise one or more memory devices, as known to the person skilled in the art. The input/output interface 36 operatively couples the processor 32 to other components such as sensors, other controllers etc. The coupling may be wired or wireless, such as using a wireless network protocol as known in the art. The input/output interface may also couple the processor to a user interface 38. The user interface 38 may include input devices and controls, such as a keyboard, keypad, buttons, or any other device capable of accepting instructions from a user and transmitting it to the processor 32, as well as screens, displays or any other device capable of communicating information to a user. The turbine 10 may also be equipped with a series of sensors, such as a wind sensor 40, a power output sensor 42, a blade strain sensor 54 etc. Additional sensors may be present, as known in the art, such as a temperature sensor, a rain sensor, a main shaft torsion sensor, blade surface pressure sensor, blade bearing, bolts or hub strain sensors, etc. (not shown). In operation, the controller 30 may determine a pitch command signal 28 for the pitch controller 18 and/or a rotor speed command signal 52 for the rotor speed controller 44, based on a wind speed measurement or estimate (e.g. a signal from the wind sensor 40 or a wind speed estimate calculated based on signals from e.g. the power output sensor 42, rotor speed sensor 50 and/or pitch angle sensor 26), and a control method for optimal power output regulation as explained above, stored in the memory 34 of the controller. A controller 30 may request data from the various sensors, and in particular the power output sensor 42 and wind speed sensor 40, and store this data in the memory 34.

Figure 5 displays a flowchart of a method of adaptive control of wind turbines according to embodiments of the invention in which multiple settings (e.g. pitch angle regulation or rotor speed regulation curves) may be tested for power output differences using repeated cycles of short period operation with each test setting. At step 600, the controller 30 selects a series of N settings Xi to x_N (where N≥ 2 including any reference setting and comparative settings) from a list of available settings. N is a parameter that may be specified by a user and the list of settings may be specified by a user prior to the start of the test. Alternatively, candidate settings may be calculated on the fly rather than selected from a predetermined list. At step 602, cycle c is in progress and the controller selects setting Xj from the list of N settings. At step 604, the controller 30 executes instructions to operate the turbine 10 with setting x, for a time period of t1.

At step 606, the controller 30 records in memory 34 a power production signal e.g. provided by the power output sensor 42 or estimated from other sensor measurements such as from the blade strain sensor 54 (see above) over the period of operation of the turbine 10 at setting x, for cycle c. The signal may comprise e.g. a continuous or almost continuous signal of instantaneous power production over t1 or an accumulated power production (over t1 or since recording, from which a signal over t1 can be trivially derived based on a measurement at the start of t1). Any metric derived from the above signal may also be used, such as e.g. any statistical estimate of the instantaneous power production over t1 (e.g. mean, median, etc.). Optionally, data from additional sensors such as data from the pitch position sensor 26, the rotor speed sensor 50, or the wind sensor 40 may be recorded at step 606. The controller may instead or in addition calculate a wind speed estimate based on the power output, pitch and rotor speed signals, as explained above. It will be appreciated that the period of time t1 may not necessarily be the same for every cycle of the method and/or for every setting x,. For example, the period of time t1 may be a minimum period of time. In embodiments where t1 is different between different power output measurements, power output values may be normalised.

At step 608, the controller 30 checks whether all settings Xi to x_N have been used in cycle c. In the negative, the controller repeats steps 602 to 608 for a new setting x. In the affirmative, the controller 30 updates the cycle counter to c+1 and repeats steps 602 to 608 for all N settings in the list of settings. Optionally, the controller 30 may implement a break of t2 seconds between consecutive settings (where t2 may be a default period or a parameter set by a user), during which the power output measurements may not be taken into account, such as to allow for the settings to be changed, for the new conditions (power production in response to new setting, readings of sensors etc.) to stabilise, etc.

At step 610, the controller checks whether a sufficient amount of cycles c have been completed (i.e. the number of cycles has reached or exceeded a threshold c_m, where the number of cycles c_th≥ 1 can be a default parameter or can be set by a user). In some embodiments, a user may be able to specify, instead of an amount of cycles, an amount of time since the beginning of the process (which can be converted into an amount of cycles based on the parameters t1 and t2). In some embodiments, the minimum amount of cycles may be determined dynamically based on a minimum number of data points, a statistical metric of variability, etc. (see further below). In some embodiments, a user may be able to manually interrupt the process at any point (i.e. at any cycle).

At step 612, the controller accumulates the power values stored at step 606 over all c cycles separately for each setting Xi to x_N. In some embodiments, the controller may perform step 612 at the end of each day, or after a given number of cycles, then the controller may resume measurement for another period or set of cycles (i.e. the cycle counter may be reset and the method may start again from step 102). As the person skilled in the art would understand, other metrics derived from the power output data may be used, such as e.g. the average of the power values. The average of power values is directly proportional to the sum of the power production values, with an identical proportionality factor for all settings, thereby making the comparison of averages equivalent to the comparison of cumulated power. As further described below, the controller 30 may take the additional data from sensors that may have been recorded at step 606 to separate the data into different sets. For example, the controller 30 may classify the data into wind speed bins according to a wind speed value recorded at step 606. Other criteria to include or exclude individual data points or groups of data points may also be specified e.g. by a user or built into the instructions executed by the controller, such as e.g. a criterion on data outliers etc. In some embodiments, a cycle where any of the measurements falls outside of a pre-determined region may be disregarded. In some embodiments, mean power measurements that differ by more than a defined threshold from other measurements for the same setting in the same wind speed bin may be disregarded, for example in order to exclude outliers that may be the result of unknown errors.

At step 614, the controller 30 compares the accumulated (and optionally filtered or classified) values obtained for the different settings. For example, the controller 30 may evaluate the difference in accumulated power output between a test setting and a currently used or default setting. The controller 30 then may then optionally decide at step 616 whether any difference is significant (see further below). At step 618 the controller 30 may optionally operate the turbine 10 on a new setting that was found at steps 614 or 616 to result in a power production improvement. In some embodiments, some steps of the above method may be performed by an external controller, which exchanges information with the turbine controller 30. For example, the controller may instead or in addition to recording data in memory 34 at step 606,

communicate the data to an external computing device or a user via the input/output interface 36. In such embodiments, any of steps 610, 612, 614 and 616 may also be performed by a separate computing device (or a user, for steps 610 and 616).

In some embodiments, the settings Xi to x_N may be selected at step 602 in consecutive order from a list of possible settings. In such embodiments, checking whether all settings were tested in a cycle (step 608) may simply involve setting a counter to c=c+1 after x_N was used, and restarting the process for the next cycle from the start of the Xi to x_N list. In some embodiments, the settings Xi to x_N may be randomly selected at step 602 from a list of possible settings by a process of sampling without replacement. In these embodiments, checking at step 608 whether all settings were tested may involve checking that the list of settings not used yet in the current cycle is empty, or that the list of settings already used is of length N. Advantageously, such embodiments may control for dependencies between measurements compared to settings where a fixed testing order is used.

In embodiments, the length of the time period t1 may be such that the ambient conditions can reasonably be expected to be stable over the time period (and so should the power production), such that a large number of measurement periods can be obtained over a short period of time, and/or such that the length of time is sufficient to obtain a representative measurement. In some embodiments, these assumptions may be verified before using one or more data points, such as using the power output data and optionally additional data from sensors recorded at step 606. In some embodiments, data points that do not comply with these assumptions may be filtered out. In some embodiments, t1 is between 10 seconds and a minute, between 10s and 2 minutes, between 10s and 5 minutes. In some embodiments, t1 is about 10 seconds, about 15 seconds, about 20 seconds, about 25s, about 30s, about 35s, about 40s, about 45s, about 50s, about 55s, about a minute, about 75s, about 90s, about 2 minutes. In some embodiments, t1 is below 15 minutes, below 10 minutes, below 5 minutes, below 2 minutes, below a minute. In some embodiments, the length of the time period t2 may be set to the shortest period allowing for a change of settings and stabilisation of operating parameters following the change. In some embodiments, t2 may depend on the settings to be tested. In some embodiments, t2 may be shorter than t1. In some

embodiments, t2 may be about 2 seconds, about 5 seconds, about 10 seconds, about 15 s, about 20s, about 25s, about 30s. In some embodiments, t2 may be under a minute, under 2 minutes, under 5 minutes. In some embodiments, t1 and/or t2 may be automatically set to default values. In some embodiments, t1 and/or t2 may be specified by a user.

In some embodiments, the data may be separated into wind speed bins at step 606, i.e. the controller may only record a wind speed bin rather than a wind speed value. In some embodiments, the data may be binned at step 612 based on wind speed measurements recorded at step 106. In some embodiments, the data may not be separated as a function of wind speed, and the cumulated power production over all wind speeds is compared. In such embodiments, the power production during each period t1 may be the only measurement required to perform the method of the disclosure. In some embodiments, the decision at step 616 may be based on data from all wind speeds. In some embodiments, the decision at step 616 may be based on data from some wind speed bins. In some embodiments, the decision at step 616 may be based on data from bins where the (observed or expected) difference in power production between settings is the largest. In some embodiments, the decision at step 616 may be based on the wind speed bins that are expected to be the most frequent at a particular site. In some

embodiments, the wind speeds taken into account for the determination of performance improvement may be specified by a user. Such embodiments may be useful when some or all of the settings tested aim to produce a benefit in particular wind speed regions.

In embodiments, any or all of t1 , t2, the number of cycles, and the significance threshold may be determined based on a calibration run. In some embodiments, a calibration run may comprise running the method with a series of identical settings, i.e. x^.., x_N: x =x₂=x_N- In some embodiments, a calibration run may comprise running the method until the standard deviation of differences between accumulated power per day falls below an acceptable threshold. The standard deviation in a calibration run is expected to decrease as the number of data points increases because with identical settings there should not be any difference in power output, and variations due to uneven environmental conditions and measurement error should average out. Typically (see below), after a few days, the standard deviation may drop below a level that is sufficient to detect small differences in power production, and the decrease in standard deviation as additional data is collected decreases slowly, such that there is limited benefit in pursuing the test any further. In some

embodiments, a calibration run may be performed prior to using the method to compare different settings. In some embodiments, a calibration run is performed simultaneously with a testing run, by including a series of identical control settings as part of Xi , . . , x_N.

Detection of less performing blades

In some embodiments, the invention also provides a method of detecting less performing blades, based on any of the embodiments of the method of operating a wind turbine described above. As mentioned above, the power output achievable with the methods of adaptive control of a wind turbine described may be influenced by a series of events affecting e.g. the aerodynamics of the blades etc. Therefore, comparing the highest power output obtained at steps 430, 614 between a latest and previous run of the adaptive control method may indicate that the blades are currently less / under-performing. In some embodiments, the highest power output and/or the power output difference obtained at steps 430, 614 may be recorded and an output may be produced when the power output difference calculated at steps 430, 614 is above a threshold and/or the highest power output obtained at step 400/420, 612 is lower than a previously recorded power output. As the person skilled in the art would understand, comparing the highest power outputs obtained at different instances of step 400/420, 612 may in fact encompass comparing some of the values obtained at these instances. In particular, the highest power output obtained at a given wind speed / in a given wind speed bin or cumulated over some or all wind speeds may be compared, or conversely the wind speed/wind speed bin in which a given power output (e.g. the rated power of the turbine) was obtained. Additionally, if the controller implementing the adaptive operating process described above has to change the reference control curves more often or to a larger extend than expected in normal use of the turbine, then this may provide an indication that the blades are underperforming. Therefore, in some embodiments, an output may instead or in addition be produced when a changed value in steps 440, 618 differs from any previous reference value by a given threshold. For example, any of the difference between a previous reference and changed reference pitch angles or rotor speeds at a given wind speed or in a given wind speed bin, or the difference between a previous reference and changed pitch angles or rotor speeds resulting in a given power output may be used as an indication of less performing blades.

In some embodiments, the detection of less performing blades may be based on the change in value of any of the performance parameters, where performance parameters may comprise a power output (e.g. highest power output at a given wind speed / in a given wind bin / highest cumulated power output, or wind speed / wind speed bin at which the rated power was obtained) , and/or a regulation value such as pitch angle or wind speed (e.g. pitch angle or rotor speed at a given wind speed or in a given wind speed bin, pitch angle or rotor speed resulting in a given power output) described herein, for example, over a measurement period. A measured change in the appropriate parameter values over time may then be used as an indication of less performing blades. In particular, any combination of the wind speed at which nominal power is reached, the highest power output obtained at a given wind speed, the changed pitch angle and/or rotor speed resulting in a given power output, and the changed pitch angle and or rotor speed at a given wind speed may be used.

In some embodiments, the detection of less performing blades may be based on a sudden change in the value of any of the parameters disclosed herein. For example, a sudden change may be defined as a "significant" difference between statistical estimates over a sample of the n most recent qualifying instances as compared to the previous n qualifying instances. Measurement instances may qualify, e.g. by virtue of falling in the same wind speed bin. For example, the highest power output obtained in a given wind speed bin over the n most recent instances of steps 430, 614 may constitute a set of qualifying instances. A sudden change may also be based on a significant difference between individual values of the parameters described, where the individual values may be consecutive or individual values within a given window. In some embodiments, a sudden change may be described as an individual value that is considered to be significantly different from the distribution of all or some of the values previously measured. In particular, a value may be compared to a distribution of values previously measured over a specified amount of time (or a number of values). In some embodiments, the slope of a curve (or gradient) of monitored parameters over time may be measured and used to detect a sudden degradation in performance.

As the person skilled in the art would understand, any combination of the above is also envisaged in the present disclosure. For example any combination of the power output, rotor speed and/or pitch angle measurements may be used to identify a loss of performance, as well as any combination of the behaviour of these measurements, as described above. In particular, any combination of wind speed at which nominal power is reached, power output at a given wind speed, pitch angle resulting in a given power output, and optimal pitch angle at a given wind speed may be used.

A difference between parameter values (or statistical estimates of values) may be considered as "significant", as used herein, when it meets or exceeds a threshold. That is, once a threshold has been met an appropriate action may be taken. In some embodiments, a difference between parameter values that is equal to or exceeds 1 % may be considered to be significant, i.e. to meet a threshold. Alternatively, a significant difference may be equal to or exceeds 2%. In some embodiments, a difference that equals or exceeds 3, 4, 5, 10, 15 or 20% may be considered significant. The size of difference that makes a particular change "significant" may also depend on the parameter under consideration. For example, a significant difference in optimal pitch angle may be set at a lower level than a significant difference in wind speed at optimal power and so on.

As the person skilled in the art would understand, any of the considerations mentioned above in relation to determining a significant difference between power outputs obtained with multiple settings may also apply to the determination of a significant difference between highest power outputs obtained over multiple runs of the adaptive control method.

In some embodiments, one or more of the parameters described herein may be monitored in performance regions / under conditions within which comparatively large differences may be observed when blade surface irregularities are present. For example, performance regions around the knee of the power curve, e.g. at wind speeds close to nominal speed, are particularly advantageous. In some embodiments, measurements corresponding to wind speeds in the regions described above may be used.

In some embodiments, the ambient temperature may be monitored and the information may be combined with the change in performance information to determine a likely cause of loss of performance. For example, a sudden loss of performance or sudden change in any parameter in combination with a measured temperature below 0°C may indicate the presence of ice on the blades. Alternatively, a sudden loss of performance or change in any parameter at temperatures above freezing may be the result of a swarm of insects passing through the site and having stuck to the surface of the blade. A gradual loss of performance may be indicative of dirt or salt accumulation on the blades. Similar considerations would apply if the cause of the performance loss is blade leading edge erosion, although the gradual decrease in performance might be expected to occur at a different speed.

In some embodiments, additional data may also be used to discriminate between the possible causes of performance loss, such as user input data or sensor data. Such data may include information about the last time the blades were cleaned, or the occurrence of rainfall and the impact on the observed performance of the wind turbine. Indeed, significant levels of rainfall may at least partially clean the blade surface of irregularities such as dirt, insects or salt. Therefore, a change in performance that has occurred since the last rainfall may be attributed to such causes, whereas a change in performance that is unaffected by the occurrence of rainfall may be more likely to be attributable to e.g. blade leading edge erosion. In some embodiments, the likely cause of loss of performance may be predicted based on such data. In some embodiments, a warning message may be produced upon detection of a loss of performance. The warning message may include information as to the likely cause of the loss of performance, as described herein. In some embodiments, the detection of a loss of performance may trigger an action. For example, a sudden change as described above associated with temperatures below 0°C may trigger a defrosting mechanism as known in the art, such as heating systems for the blades. Alternatively, a detected change might cause a scheduling of a defrosting mechanism, or a message to be sent or displayed to an operator to clean the blades or to check the blades for ice build-up. By contrast, a sudden change as described above at temperatures above freezing may trigger an automatic cleaning of the blades, as known in the art.

In some embodiments the detection method above may be integrated as part of a process to monitor the functioning of a wind turbine and to determine time to maintenance. For example, a gradual loss of performance may be included as a parameter in scheduling a next maintenance of the turbine or for scheduling a blade inspection (such as e.g. done by a person in a lift, using a ground based camera, or using a camera on a drone), or even a change of the blades.

While the above has been explained in relation to the presence of irregularities on the blade, the person skilled in the art would understand that any event that may affect the

performance of a wind turbine may be detected with the methods disclosed herein. Such events may include changes in site conditions over time (e.g. increased vegetation, presence of buildings or other wind turbines, etc.), slippage in the blade pitch system causing pitch angles offsets, etc.

The invention is further illustrated by the following non-limiting examples.

Examples Example 1

Figure 6 illustrates a situation in which three different pitch angle regulation curves are tested. In this case, starting from the currently used pitch angle regulation curve (continuous line), two alternative curves are tested. At the current tip speed ratio of 9 (i.e. maintaining the rotor speed constant if the wind speed is constant, or adapting it to stay at this tip speed ratio), the three pitch angle regulation curves imply a different pitch control signal. The turbine can therefore be operated successively with the three pitch angle settings marked by stars, for a period of time t1. After the period of time, the power output is compared between the three periods (or multiple replications of the three periods) and the turbine operated with the pitch angle regulation curve that resulted in the best power production.

Example 2

Figure 7 illustrates an example of the effect of an accumulation of dirt on the achievable power output and optimal pitch setting at various wind speeds. Accumulation of dirt on the surface of the blades causes the power curve to shift towards the right, as seen in the example depicted in Figure 9, and modifies the pitch at which optimal power production is achieved for each wind speed. In particular, monitoring the optimal pitch setting within a zone (see dashed box on Figure 7) between 12 and 15 m/s (i.e. approaching rated wind speed) allows the detection of large differences in optimal pitch setting. Therefore, recordings of pitch settings within this zone at t=1 when the blades were clean may be compared to recordings of pitch settings in this zone at t=2. For example, assuming that all wind speeds are equally frequent within this bin, the average optimal pitch setting observed with clean blades for all measurements falling within the bin would be -0.5 degrees, whereas the corresponding average for dirty blades would be ~-0.9 degrees (see the stars and dashed-dotted lines on Figure 7). The average power output over that zone would also be reduced with the accumulation of dirt on the blades. Conversely, as can be seen from the curves on Figure 7, the power output observed over data points collected at a given pitch setting would be different for clean and dirty blades, and so would the optimal pitch setting for a given power output. Example 3

Figure 8 shows an example of settings that may be tested using the methods and apparatus of the invention. A currently used reference curve (toggle 3) for the pitch angle regulation of a turbine may be compared against e.g. four other candidate curves obtained by applying an offset of ±0.5 degrees to the currently used regulation curve. The offsets are applied to the region of the curves that are situated below the rated regions of the turbine (right hand side of the graph, see also explanations in relation to Figure 1 , zones I and II). The curve marked "toggle 1" may correspond to the originally used regulation curve, and the currently used curve (toggle 3 would have been found using the same process in a previous instance of the testing procedure).

Claims

1. A controller for a wind turbine comprising:

a processor;

an input/output interface; and

a memory including instructions that, when executed by the processor, cause the processor to:

a) receive via the input/output interface measured values of the wind turbine output power and measured values of at least one of a rotor speed and pitch angle, during operation of the turbine with a reference pitch angle and/or rotor speed reference value for a time period t1 , the reference pitch angle being based on a pitch angle reference curve and/or the rotor speed reference value being based on a rotor speed regulation curve;

b) change the pitch angle and/or rotor speed value;

c) operate the wind turbine with the changed pitch angle and/or rotor speed value for the time period t1 and receive measured values of the wind turbine output power and at least one of a rotor speed and pitch angle position;

d) optionally repeat steps b) and c);

e) compare a power output of the wind turbine obtained at step a) with a power output obtained in step c) during operation of the wind turbine under one or more changed pitch angle(s) and/or rotor speed value(s);

f) determine whether a power output obtained at step c) is higher than the power output obtained at step a); and

g) optionally, change the pitch angle and/or rotor speed value(s) for the wind turbine reference curves to the value(s) identified in step f) that results in higher power output than the power output measured in step a).

2. The controller of claim 1 , wherein comparing the power output in step e) comprises comparing measurements acquired over a predetermined range of wind speeds.

3. The controller of claim 2, wherein the predetermined range of wind speed is below a rated wind speed of the wind turbine.

4. The controller of any preceding claim, wherein steps a) to c) are repeated multiple times to obtain multiple points and comparing the power outputs at step e) comprises comparing statistical estimates or cumulated values over the distribution of points.

5. The controller of any preceding claim, wherein the changed pitch angle and/or rotor speed regulation curves of step b) are chosen from a collection of pre-computed curves or are obtained by applying pre-defined variations around the reference curve(s).

6. The controller of any preceding claim, wherein measured values are separated into wind speed bins according to the measured or estimated wind speed at the time the data point is acquired.

7. The controller of any preceding claim, wherein at least steps a) to g) are performed at predefined intervals or events during operation of the wind turbine.

8. The controller of any preceding claim, wherein the pitch angle regulation curves can be different for each blade of a wind turbine, and steps a) to f) can be performed for each blade of a wind turbine individually.

9. The controller of any preceding claim, wherein the reference regulation curve(s) may differ during operation of the wind turbine according to the period of the day, wherein there are two or more periods of the day and wherein the period of the day is recorded during steps a) to c); and wherein steps e) to g) are performed separately for power output measurements obtained during each of the two or more periods of the day.

10. The controller of any preceding claim, wherein the processor is additionally configured to:

h) produce an output when the difference between a power output of step a) and a power output of step c) is larger than a difference threshold value and/or when the difference between the changed reference curve(s) at step g) and any previous reference curve is larger than a difference threshold value.

11. The controller of claim 10, wherein producing an output comprises transmitting a warning message to a user via an input/output interface; and/or producing a signal to a blade cleaning system or a blade defrosting system to commence or schedule a cleaning or defrosting cycle, respectively, of the wind turbine blades.

12. The controller of any of claims 10 or 1 1 , wherein the processor additionally performs the step of:

f ) obtaining additional data relating to wind turbine operating parameters and/or environmental conditions proximate the wind turbine.

13. The controller of any of claims 10 to 12, wherein steps a) to g) are repeated at different times throughout the operation of the turbine, and the rate of change of the highest power output at step f) or of the changed reference curve(s) at step g) over time is calculated, and an output is produced if the rate of change exceeds a rate threshold value.

14. The controller of any preceding claim, wherein the power output comprises measurements from a power output sensor and/or values indirectly obtained from other measured quantities.

15. The controller of claim 14, wherein the power output is indirectly obtained from values derived from main shaft torsion measurements, main shaft torque measurements, gearbox shaft torsion measurements, gearbox shaft torque measurements, blade strain measurements, blade load measurements, strain measurements on the blade bearings, strain measurements on the blade bolts, strain measurements on the hub, blade surface pressure measurements, blade tip mean deflection in edgewise direction, generator current, transformer power, transformer current, and/or tower lateral moment.

16. A method of controlling a wind turbine, the method comprising:

a) measuring and recording power output values from a sensor, and measured values from at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor for the wind turbine, during operation of the turbine with a reference pitch angle and/or rotor speed regulation curve for a time period t1 ; b) changing the pitch angle and/or rotor speed regulation curve(s);

c) operating the wind turbine with the curve(s) from step b) during a time period t1 and receiving measured values from a wind turbine power output sensor and at least one of a rotor speed sensor, rotor speed controller, pitch controller or pitch position sensor;

d) optionally repeating steps b) and c) one or more times;

e) comparing a power output of the wind turbine obtained at step a) with a power output obtained in step c) during operation of the wind turbine under one or more changed pitch angle and/or rotor speed regulation curve(s);

f) determining whether a power output obtained at step c) is higher than a

power output obtained at step a); and

g) optionally, changing the pitch angle and/or rotor speed reference curve(s) for the wind turbine to a curve or curves identified in step f) that results in higher power output than the power output measured in step a).

17. A computer programme for a controller of a wind turbine, that when executed by the controller causes the controller to perform the method of claim 16.

18. A wind turbine comprising the controller of any of claims 1 to 15.