Modifying control strategies for controlling wind turbines using load probabilities and design load limits

Abstract

The present disclosure relates to controlling operation of a wind turbine. A first plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during a first time period are determined, and then a load probability signature is determined based on a statistical analysis of a distribution of the first plurality of extreme load measurements. A control strategy for controlling operation of the wind turbine is then modified based at least in part on a comparison of the load probability signature to a design load limit, and the wind turbine is subsequently controlled during a second time period according to the modified control strategy.

Description

Technical Field

[0001] The invention relates to a method and a control system for controlling the operation of a wind turbine. Background Art

[0002] Typically, the purpose of operating a wind turbine or a wind turbine park is to obtain maximum return on the capital invested therein, and therefore the wind turbine control system is configured to maximize the output power, i.e., to operate the wind turbine to capture the maximum power available in the wind with due regard to keeping the wind turbine within operating limits.

[0003] During operation, wind turbines experience changing weather conditions, and the control system of the wind turbine is designed to take these changes into account. To this end, existing wind turbines may include a monitoring system that monitors operating and weather conditions and controls the operation of the wind turbine to maximize power output while remaining within safe operating limits.

[0004] However, the inventors have recognized that the safe operating limits are based on design assumptions about safe operation that may or may not match the actual safe operating limits of each installed wind turbine. Therefore, there is a need for an improved method of controlling wind turbine operation that better takes into account the true safe operating limits of the wind turbine. Summary of the invention

[0005] A first aspect of the present disclosure provides a method of controlling operation of a wind turbine, the method comprising: controlling operation of the wind turbine according to a control strategy during a first time period; obtaining a first set of multiple extreme load measurements indicating extreme loads experienced by at least a portion of the wind turbine during the first time period; determining a load probability signature based on a distribution of at least the first set of multiple extreme load measurements by statistical analysis; comparing the load probability signature to a design load limit; modifying the control strategy based at least in part on the comparison of the load probability signature to the design load limit; and controlling operation of the wind turbine according to the modified control strategy during a second time period. By modifying the control strategy in this manner, the actual extreme loads experienced by the wind turbine during the first time period can be used to safely reduce a safety margin employed by the control strategy. By reducing the safety margin, it should be possible to achieve a larger control range, which will enable the wind turbine to be driven more vigorously and generate more power.

[0006] The load probability feature may include a maximum extreme load value indicating a value of an extreme load corresponding to a threshold percentile of a distribution of the plurality of extreme load measurements.

[0007] The comparison of the load probability characteristic to the design load limit may include the difference between the maximum extreme load value and the design load limit.

[0008] Modifying the control strategy may also be based at least in part on the magnitude and / or sign of the difference between the maximum extreme load value and the design load limit.

[0009] The design load limit may be based at least in part on a material failure load value, a siting uncertainty value, and a simulated safety margin.

[0010] The plurality of extreme load measurements may include a plurality of extreme load estimates, and wherein the method further comprises: receiving a plurality of operating status measurements of the wind turbine during the first time period; and determining the plurality of extreme load measurements based at least in part on the plurality of operating status measurements.

[0011] The method may also include obtaining a second plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during the second time period; determining an updated load probability signature based on a distribution of the first plurality of extreme load measurements and the second plurality of extreme load measurements; comparing the updated load probability signature to a design load limit; further modifying the control strategy based at least in part on the comparison of the updated load probability signature to the design load limit; and controlling operation of the wind turbine using the further modified control strategy during a third time period.

[0012] A control strategy and a modified control strategy are defined in a model predictive control (MPC) routine. Modifying the control strategy based at least in part on a comparison of the load probability signature to the design load limit may include adjusting at least one cost weight within the MPC routine based at least in part on the comparison of the load probability signature to the design load limit. Additionally or alternatively, modifying the control strategy based at least in part on a comparison of the load probability signature to the design load limit may include adjusting at least one constraint of the MPC routine based at least in part on the comparison of the load probability signature to the design load limit.

[0013] A distribution of at least the first plurality of extreme loads may be determined based on the first plurality of extreme loads and the simulated load distribution.

[0014] Statistical analysis may also include Bayesian analysis.

[0015] A second aspect of the present disclosure provides a computer program, which is configured to perform the method according to the first aspect when executed on a processor of an electronic device.

[0016] A third aspect of the present disclosure provides a control system for a wind turbine, the control system comprising: a controller, the controller being configured to: use a control strategy to control the operation of the wind turbine during a first time period; obtain a first set of multiple extreme load measurement values ​​indicating the extreme load borne by at least a portion of the wind turbine during the first time period; determine a load probability characteristic based on the distribution of the first set of multiple extreme load measurement values ​​through statistical analysis; compare the load probability characteristic with a design load limit; modify the control strategy based at least in part on the comparison of the load probability characteristic with the design load limit; and use the modified control strategy to control the operation of the wind turbine during a second time period.

[0017] The controller may also be configured to: obtain a second plurality of extreme load measurements indicating extreme loads experienced by at least a portion of the wind turbine during the second time period; determine an updated load probability signature based on a distribution of the first plurality of extreme load measurements and the second plurality of extreme load measurements; compare the updated load probability signature to a design load limit; further modify the control strategy based at least in part on the comparison of the updated load probability signature to the design load limit; and control operation of the wind turbine using the further modified control strategy during a third time period.

[0018] A fourth aspect of the present disclosure provides a wind turbine comprising the control system of the third aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Embodiments of the present invention will now be described with reference to the accompanying drawings, in which:

[0020] Figure 1 An example of a wind turbine is shown;

[0021] Figure 2 An embodiment of a control system together with elements of a wind turbine is shown;

[0022] Figure 3 Shows Figure 2 Details of controllers in the control system;

[0023] Figure 4 Shown by Figure 3 a graphical representation of a distribution of a first plurality of extreme load measurements determined by the distribution controller;

[0024] Figure 5 Shows Figure 3 An example initial control strategy for a controller of

[0025] Figure 6 Shows Figure 3 An example of a modified control strategy for a controller of; and

[0026] Figure 7 A flow chart illustrating elements of the present disclosure is shown. DETAILED DESCRIPTION

[0027] The control strategy used to control the operation of a wind turbine is typically initially set up based on simulated turbine loads, with safety built in to account for siting uncertainties and simulated safety margins. However, the inventors have realized that built-in safety often results in wind turbines being operated too conservatively, which can result in wasted power generation opportunities. Therefore, they have designed a technique to measure / estimate the extreme loads experienced over a period of time and perform statistical analysis on the distribution of the extreme loads to identify hidden, unnecessary safety margins built into the control strategy. The control strategy can then be modified to reduce this hidden safety margin, which can allow for greater control flexibility and increased power generation from the wind turbine without compromising safety.

[0028] Figure 1 An example of a wind turbine 1 is shown in a schematic perspective view. The wind turbine 1 includes a tower 2, a nacelle 3 disposed at the apex of the tower, and a rotor 4 operably coupled to a generator housed inside the nacelle 3. In addition to the generator, the nacelle also houses various components required to convert wind energy into electrical energy and various components required to operate, control and optimize the performance of the wind turbine 1. The rotor 4 of the wind turbine includes a central hub 5 and a plurality of blades 6 protruding outwardly from the central hub 5. In the embodiment shown, the rotor 4 includes three blades 6, but the number may vary. In addition, the wind turbine includes a control system. The control system may be placed inside the nacelle or distributed at multiple locations inside the turbine and communicated.

[0029] The wind turbine 1 may be included in a collection of other wind turbines belonging to a wind power plant, also referred to as a wind farm or wind park, which serves as a power plant connected to an electrical grid via transmission lines. An electrical grid generally consists of a network of power stations, transmission lines and substations that transmit electricity via the network of transmission lines to loads in the form of end users and other customers of the power company.

[0030] Figure 2 An embodiment of a control system 20 together with elements of a wind turbine is schematically shown. The wind turbine comprises rotor blades 21, which are mechanically connected to a generator 22 via a gearbox 23. The electricity generated by the generator 22 is injected into a grid 24 via an electrical converter 25. The generator 22 may be a doubly fed induction generator, but other generator types may also be used. Furthermore, a gearbox may not be required.

[0031] The control system comprises a plurality of elements, including at least one controller 200 having a processor and a memory, such that the processor is capable of performing computational tasks based on instructions stored in the memory. Typically, the wind turbine controller 200 ensures that the wind turbine generates the requested power output level in operation. This can be achieved by adjusting the pitch angle and / or the power extraction of the converter. To this end, the control system 20 comprises: a pitch system comprising a pitch controller 27 using a pitch reference 28; and a power system comprising a power controller 29 using a power reference 26. The wind turbine rotor comprises rotor blades that can be pitched by a pitch mechanism. The rotor may comprise a common pitch system that adjusts all pitch angles of all rotor blades simultaneously, and may further comprise individual pitch systems capable of individually pitching the rotor blades.

[0032] Figure 3 Schematically shown are details of a controller 200 according to one embodiment. The controller 200 comprises a model predictive control (MPC) unit 310 , an adapter unit 320 and a load estimator 330 .

[0033] The MPC unit 310 is configured to determine control signals (such as signals 26 and 28) for controlling the operation of the wind turbine (such as blade pitch and / or power extraction of the converter). The MPC unit 310 determines the control signals based at least in part on an MPC routine that utilizes a cost function that includes costs associated with one or more types of loads on the wind turbine (e.g., loads on one or more different components / assemblies of the wind turbine) and / or one or more measurements of the operation of the wind turbine (e.g., operating state, which will be described in more detail later). A skilled person will readily understand how the MPC routine can be used to determine the control signals. However, by way of brief explanation, MPC is a multivariable control algorithm that uses an optimization cost function J at a fallback prediction level to calculate an optimal control action.

[0034] The optimization cost function can be given by:

[0035] J=min(λ1*Load-Power)

[0036] In this example, "load" is a load cost associated with one or more types of loads of the wind turbine (e.g., load on the tower and / or load on the blades, etc.). Additionally or alternatively, "load" may be a load cost associated with one or more operational measurements of the wind turbine. λ1 is a cost weight corresponding to "load". "Power" is the electrical power output of the wind turbine. Thus, by minimizing λ1*Load-Power, the cost function may be optimized to achieve optimal power output for a particular load level on the wind turbine.

[0037] It is easy to understand that in equivalent implementations, the optimization benefit function can be used for exactly the same results. The equivalent optimization benefit function can be given by the following formula:

[0038] B=max(Power-λ1*Load)

[0039] Therefore, it will be understood that the term "cost function" used throughout this disclosure encompasses the equivalent "benefit function".

[0040] Figure 3 The MPC unit 310 in is configured to receive an operating state signal from the wind turbine, which includes one or more measured values ​​of one or more operating states of the wind turbine. However, the MPC unit 310 may additionally or alternatively be configured to receive at least one load measurement 335 from the load estimator 330. The operating state signal may include, for example, one or more measured values ​​of pitch value, rotor angular velocity, tower top position, power output, and derivatives of these parameters. The measurements in the operating state signal may be based on sensor readings of sensors arranged to measure sensor data related to physical state values ​​of the wind turbine. The MPC unit 310 then determines the control signal using the received operating state (and / or measured values ​​of load) and the MPC optimization routine.

[0041] Initially, the MPC routine may be configured with an initial control strategy based on design assumptions. The initial control strategy may be defined in the MPC routine by various parameters set in the MPC routine, such as cost weights and / or constraints. Figure 4 Explain the design assumptions used to set the initial control strategy.

[0042] Figure 4 A representation of the load on a wind turbine (y-axis) is shown relative to the operating time of the wind turbine (x-axis). Before installing a wind turbine, a simulation of the load distribution over time can be determined. Figure 4This is denoted as the “simulated load distribution” in the MPC routine. This represents how a particular load (e.g., tower load or blade load, etc.) varies over time. When the failure threshold percentile of the distribution is at or above the design load limit, then it is predicted that the wind turbine component may fail. For example, when the failure threshold percentile is 90%, if the first 10% or more of the distribution is above the design load limit, then it is predicted that the component may fail. If only the first 9% or only the first 5%, etc. of the distribution is above the design load limit, then it is predicted that the component may not fail. Therefore, during the design phase, the MPC routine can be configured to define an initial control strategy that is expected to result in a simulated load distribution that is in a safe position relative to the design load limit (i.e., such that the failure threshold percentile of the simulated distribution is at or below the design load limit). However, because the consequences of turbine failure due to overloading of the wind turbine are severe, siting uncertainties (e.g., uncertainties in the environmental conditions to which the wind turbine will be subjected, such as wind field and / or climate uncertainties) and simulation-based safety margins are also built into the initial control strategy to help ensure that, after installation, the wind turbine is not accidentally placed at risk due to an overly aggressive initial control strategy.

[0043] return Figure 3 , the load estimator 330 is configured to receive an operating status signal from the wind turbine. Then, at least one load measurement 335 related to one or more components of the wind turbine is determined and provided to the adapter unit 320 and the MPC unit 310. For example, the at least one load measurement 335 may be related to the load on the tower or the load on the blade, etc. The at least one load measurement 335 may be determined regularly (e.g., every second, or every three seconds, or every ten seconds, etc.) based on the current operating status signal. It will be appreciated that in an alternative, the operating status may include direct measurements of one or more loads on the wind turbine, in which case the controller 200 may not include the load estimator 330, and the adapter unit 320 may directly receive at least one load measurement in the operating status of the wind turbine.

[0044] In the controller 200 of the present disclosure, the adapter unit 320 has been implemented and configured to determine an extreme load measurement value for each type of load measurement value received from the load estimator 330. For example, the load measurement value 335 may include a measurement value of the load on the turbine tower, and the adapter unit 320 may receive a new load measurement value 335 every second. The adapter unit 320 may determine an extreme load measurement value indicating the extreme load experienced by the turbine tower based on the load measurements 335 received within the specific amount of time (such as a 10-minute period). In this example, the extreme load measurement value is the maximum load within the specific time period. It will be understood that the 10-minute period is only a non-limiting example, and different time lengths may be used alternatively. The adapter unit 320 continues to determine extreme load measurements over a first time period (such as a month, or three months, or six months, or a year, etc.). Therefore, the adapter unit 320 will determine a first set of multiple extreme load measurements over the first time period—in this specific example, one extreme load measurement value is determined every 10 minutes.

[0045] exist Figure 4 The distribution of these extreme load measurements can be seen in Figure 1. In particular, after the wind turbine is installed, more and more extreme load measurements are determined over time. It can be seen that over the course of the first time period, as more and more extreme load measurements are added to the distribution, the variance within the distribution decreases and the average level gradually tends to be lower than that of the simulated distribution. This is because there is no siting uncertainty or aeroelastic model uncertainty added to the distribution of the first plurality of extreme load measurements, since they are based on actual field measurements.

[0046] At the end of the first time period, a maximum extreme load value of the distribution of the first plurality of extreme load measurements is determined. This is the value of the extreme load that corresponds to a failure threshold percentile (e.g., the 90% percentile) of the distribution. The inventors have recognized that when performing a statistical analysis of actual extreme load measurements, the maximum extreme load value is often well below the design load limit, which represents a hidden, unnecessary safety built into the control strategy of the MPC unit 310. This means that the MPC unit 310 can safely allow larger extreme loads without risking safety, which will provide the MPC unit 310 with a larger control range and potentially allow the wind turbine to generate more power.

[0047] Thus, the adapter unit 320 generates adapted MPC settings 325 to modify the control strategy of the MPC unit 310 based at least in part on the comparison of the maximum extreme load value to the design load limit. The adapted MPC settings 325 are based at least in part on the magnitude and / or sign of the difference between the design load limit and the maximum extreme load value (e.g., if the maximum extreme load value is much lower than the design load limit, the adapted MPC settings 325 may be set to make a larger change to the MPC control strategy, whereas if the difference is small, the adapted MPC settings 325 may be set to make a smaller change to the MPC control strategy).

[0048] This statistical analysis for determining the distribution from which the maximum extreme load value is determined may also be based on a simulated load distribution and may utilize a Bayesian analysis. For example, the simulated load distribution may represent "prior knowledge" in the Bayesian analysis and the distribution may be effectively modified by performing a Bayesian process using the first plurality of extreme load measurements. In this sense, the distribution determined based on the simulated load distribution and the first plurality of extreme loads is a "posterior possible distribution". The maximum extreme load value may then be determined from this distribution.

[0049] After modifying the control strategy, the wind turbine will then be operated using the modified control strategy, which will result in future extreme loads being distributed in Figure 4 The shift upwards on the diagram shown here reduces the gap between the maximum extreme load value and the design load limit and enables greater power output without compromising safety.

[0050] Optionally, after the control strategy has been modified, the adapter unit 320 may determine a second plurality of extreme load measurements during a second time period following the first time period. An updated maximum extreme load value may then be determined based on a distribution of the first plurality of extreme load measurements and the second plurality of extreme load measurements (e.g., the maximum extreme load value may be found by finding a load value corresponding to 90% of the distribution of the first plurality of extreme load measurements and the second plurality of extreme load measurements). Optionally, the distribution may be determined using a Bayesian analysis, wherein the distribution prior to the start of the second time period represents "prior knowledge" in the Bayesian analysis (as previously described, the distribution itself may have been determined earlier using a simulated load distribution and a Bayesian analysis of the first plurality of extreme loads), and the second plurality of extreme loads is used to modify the distribution to obtain a "later possible distribution" using which an updated maximum extreme load value is determined. The control strategy may then be further modified based at least in part on a comparison of the updated maximum extreme load value with the design load limit, and subsequent control of the wind turbine may be based on the further modified control strategy. In this way, the control strategy can be gradually refined over time as the wind turbine ages and more extreme load data becomes available.

[0051] The first time period and the second time period, as well as any other time periods during which extreme load measurements are determined and used in the statistical analysis to determine further updated maximum extreme load values, may be any suitable time periods. For example, they may be of the order of seconds, minutes, hours, days, months or years, so that the control strategy may be updated and refined at any suitable frequency.

[0052] Figure 5 An example graphical representation is shown of how the control strategy of the MPC unit 310 may be modified. In this example, the extreme load measurements relate to tower loads, and the MPC unit 310 takes tower deflection into account when controlling the wind turbine, but the same principles may be equally applied to loads on any other component of the wind turbine (e.g., loads on blades). Figure 5 The x-axis of represents the tower deflection (eg, in millimeters, centimeters, meters, etc.). The y-axis represents the cost penalty applied to the tower load cost in the MPC routine.

[0053] The MPC TdMax limit represents an MPC constraint. It can be seen that when the tower deflection Td is below this limit, no cost penalty is applied. Once Td exceeds this limit, a cost penalty is applied to the tower load cost in the MPC routine, with the expected result that the control signal generated by the MPC unit 310 is changed to avoid the tower load reaching dangerous levels.

[0054] In the initial control strategy determined based on the simulated distribution of extreme loads, siting uncertainty, and simulated safety margin, MPC TdMax is set to the “initial MPC TdMax limit”. It can be seen that the initial MPC TdMax limit is set to:

[0055] Initial MPC TdMax limit = material failure - site selection uncertainty - safety margin (simulation)

[0056] Herein, material failure refers to the amount of tower deflection at which material failure is expected to occur.

[0057] As explained previously, siting uncertainties can be reduced or eliminated by comparison of the maximum extreme load values ​​with the design load limits. Furthermore, if necessary, the size of the safety margin can be adjusted and the effect on the actual tower deflection observed to ensure that the safety margin is still large enough to allow safe operation of the wind turbine. Thus, modification of the control strategy can involve moving the MPC TdMax limit so that the modified control strategy has a constraint of the "modified MPC TdMax limit".

[0058] Figure 6 An example graphical representation of the modified control strategy of the MPC unit 310 is shown. The modified MPC TdMax limit is set to:

[0059] Modified MPC TdMax limit = Material failure - Safety margin (modified)

[0060] Therefore, it can be seen that since there is no longer any siting uncertainty, and since the safety margin can be set against actual measurements of extreme loads (rather than simulations), the MPC TdMax limit constraint can be moved further along the x-axis to allow for larger tower deflections before cost penalties begin to apply. This means that after modifying the control strategy, Figure 4 The maximum extreme load value in the distribution of medium extreme load measurements should begin to shift upward in the y-axis direction over time, thereby reducing the amount of hidden safety. Using the modified control strategy, the MPC unit 310 is able to drive the wind turbine more strongly and output more power while still operating safely.

[0061] Figure 7 Elements of a method of controlling a wind turbine to perform various embodiments of the present disclosure are shown. The elements of the method may be implemented by software as a computer program product, hardware, or a combination of software and hardware. The software may include computer readable code that, when executed on a processor of any electronic device, may execute Figure 7The software may be stored on any suitable computer readable medium, such as a non-transitory computer readable medium (such as read-only memory, random access memory, CD-ROM, DVD, Blu-ray, tape, hard drive, solid state drive and optical drive). Typically, the software is stored and executed by the wind turbine control system.

[0062] In step S710 , the control system 20 controls the operation of the wind turbine according to a control strategy (eg, an initial control strategy).

[0063] In step S720 , a plurality of extreme load measurements (eg, a first plurality of extreme load measurements) are obtained within a time period (eg, a first time period).

[0064] In step S730 , a probability feature (eg, a maximum extreme load value) is determined based on the distribution of the plurality of extreme load measurement values.

[0065] In step S740, the determined load probability signature is compared to the design load limit, and in step S750, the control strategy is modified based on the comparison. The method then returns to step S710 to control the wind turbine based on the control strategy, except that the control strategy has now been modified in step S750.

[0066] It will be appreciated that this cycle may be repeated multiple times over the lifetime of the wind turbine (e.g., if the operational lifetime of the wind turbine is 20 years, the control strategy may be modified according to the method every minute, or every hour, or every day, or every week, or every month, or every 3 months, or every 6 months, or every year, or every two years, etc., over the lifetime of the wind turbine). The time period between each modification may be equal or may vary over the lifetime of the wind turbine. Thus, the control strategy may be gradually refined to obtain the optimum power output of the wind turbine while still operating within safety limits.

[0067] Although the invention has been described above with reference to one or more preferred embodiments, it will be appreciated that various changes or modifications may be made without departing from the scope of the invention as defined in the appended claims.

[0068] For example, although the above disclosure has focused specifically on a first plurality of extreme load measurements relating to a time period immediately following installation of the wind turbine and modification of the control strategy to an initial design stage control strategy, the first plurality of extreme load measurements may relate to extreme load measurements obtained at any time period during the life of the wind turbine. In this case, the modification to the control strategy is a modification to any control strategy currently in place (the currently in place control strategy may be the initial design stage control strategy or some other control strategy that has been previously modified).

[0069] Although one specific example of control strategy modification is given with reference to changing constraints within the MPC routine, it will be appreciated that the control strategy may be modified in any other suitable manner based on a comparison of the maximum extreme load values ​​and the design load limits, such as by modifying one or more cost weights within the MPC routine.

[0070] Furthermore, although in the above description the maximum extreme load value is determined based on the distribution of extreme load measurements and then compared to the design load limit, in alternatives any other statistical analysis may be performed on the distribution of extreme load measurements to determine any other suitable load probability feature. For example, a different threshold percentile may be used, or a probability feature related to the mean, variance or median of the distribution, etc. may be determined. This may then be compared to the design load limit as appropriate. Therefore, it may also be appreciated that although the comparison of the load probability feature to the design load limit is generally described above as finding the difference between the two, the comparison may alternatively take any other suitable form depending on the characteristics of the load probability feature.

[0071] In the above description, the adapter unit determines the plurality of extreme load measurements based on the one or more load measurements 335. However, in an alternative, the adapter unit 320 may simply receive the plurality of extreme load measurements, for example from a load estimator (which may be configured to determine the extreme load measurements and provide them to the adapter unit 320), or from any other suitable unit or entity. Thus, the adapter unit 320 is configured to obtain the plurality of extreme load measurements by determining the plurality of extreme load measurements itself or by receiving the plurality of extreme load measurements from another unit / entity.

Claims

1. A method of controlling operation of a wind turbine, the method comprising: controlling operation of the wind turbine according to the control strategy during a first time period; obtaining a first plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during said first period of time; determining a load probability signature based on a distribution of at least the first plurality of extreme load measurements by statistical analysis; comparing the load probability characteristic with a design load limit; modifying the control strategy based at least in part on a comparison of the load probability signature to the design load limit; controlling operation of the wind turbine according to the modified control strategy during a second time period; obtaining a second plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during the second period of time; determining, by statistical analysis, an updated load probability signature based on the distribution of at least the first plurality of extreme load measurements and the second plurality of extreme load measurements; comparing the updated load probability characteristics with the design load limits; further modifying the control strategy based at least in part on a comparison of the updated load probability signature to the design load limits; and The operation of the wind turbine is controlled using the further modified control strategy during a third time period.

2. The method according to claim 1, wherein: The load probability feature includes a maximum extreme load value indicating a value of an extreme load corresponding to a threshold percentile of a distribution of the plurality of extreme load measurements.

3. The method according to claim 2, wherein: The comparison of the load probability signature to the design load limit includes a difference between the maximum extreme load value and the design load limit.

4. The method according to claim 3, wherein: Modifying the control strategy is also based at least in part on a magnitude and / or a sign of a difference between the maximum extreme load value and the design load limit.

5. The method according to any one of claims 1 to 4, wherein: The design load limit is based at least in part on a material failure load value, a siting uncertainty value, and a simulated safety margin.

6. The method according to any one of claims 1 to 4, wherein: The plurality of extreme load measurements include a plurality of extreme load estimates, and wherein the method further comprises: receiving a plurality of operating status measurements of the wind turbine during the first time period; and The plurality of extreme load measurements are determined based at least in part on the plurality of operating condition measurements.

7. The method according to any one of claims 1 to 4, wherein: The control strategy and the modified control strategy are defined in a model predictive control (MPC) routine.

8. The method according to claim 7, wherein: The MPC routine has cost weights set therein, and modifying the control strategy based at least in part on the comparison of the load probability signature with the design load limit includes: adjusting at least one cost weight within the MPC routine based at least in part on the comparison of the load probability signature with the design load limit.

9. The method according to claim 7, wherein: The MPC routine has constraints set therein, and modifying the control strategy based at least in part on the comparison of the load probability signature with the design load limit includes adjusting at least one constraint of the MPC routine based at least in part on the comparison of the load probability signature with the design load limit.

10. The method according to any one of claims 1 to 4, wherein: A distribution of at least the first plurality of extreme loads is determined based on the first plurality of extreme loads and a simulated load distribution.

11. The method according to any one of claims 1 to 4, wherein: The statistical analysis also includes Bayesian analysis.

12. A computer program product comprising code which, when executed on a processor, is configured to cause the processor to perform the method according to any one of the preceding claims.

13. A control system for a wind turbine, comprising: A controller, the controller being configured to: controlling operation of the wind turbine using the control strategy during a first time period; obtaining a first plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during said first period of time; determining a load probability characteristic based on the distribution of the first plurality of extreme load measurements by statistical analysis; comparing the load probability characteristics to design load limits; modifying the control strategy based at least in part on a comparison of the load probability signature to the design load limit; controlling operation of the wind turbine using the modified control strategy during a second time period; obtaining a second plurality of extreme load measurements indicative of extreme loads experienced by at least a portion of the wind turbine during the second period of time; determining an updated load probability signature based on the distribution of the first plurality of extreme load measurements and the second plurality of extreme load measurements; comparing the updated load probability characteristics with the design load limits; further modifying the control strategy based at least in part on a comparison of the updated load probability signature to the design load limits; and The operation of the wind turbine is controlled using the further modified control strategy during a third time period.

14. A wind turbine comprising a control system according to claim 13.

Images