WO2018121667A1

WO2018121667A1 - A method of and system for empirically evaluating wind turbine generator operation

Info

Publication number: WO2018121667A1
Application number: PCT/CN2017/119356
Authority: WO
Inventors: Matthew SUMMERS
Original assignee: Envision Energy (Jiangsu) Co., Ltd.
Priority date: 2016-12-30
Filing date: 2017-12-28
Publication date: 2018-07-05
Also published as: CN110139983A; CN110139983B

Abstract

Disclosed are a method and a system of evaluating operation (100) of at least one wind turbine generator (WTG) (1) in a set of wind turbine generators (WTGs) controlled by at least one controller (20) operating the WTG based on at least one set of settings (30). The method may comprise the following acts: monitoring operation (110) of at least one WTG; monitoring a set of settings (120) used by the controller (20) operating the at least one WTG; detecting (13) a change in the settings (132) from the monitored set of settings and recording a time-of-change (135) of the detected change in the settings; and determining an interrelationship (140) between change in the operation (134) and change in the settings before and after the detected time-of-change.

Description

A Method of and System for Empirically Evaluating Wind Turbine Generator Operation

Field of the Invention

The present invention relates to a method of evaluating operation of at least one wind turbine generator (WTG) in a set of wind turbine generators (WTGs) controlled by at least one controller operating the WTG based on at least one set of settings. The method may comprise one or more of the following acts.

Background of the Invention

Modern wind turbine generators (WTGs) use variable pitch and torque settings to control the net velocity of the air relative to the rotating blade section chord lines. This behaviour is governed by control software. Theoretical models like blade element momentum theory (BEM) may be used to estimate optimal settings for these controls.

Relying on BEM will require understanding of limitations and implementing a BEM model as for example ” Aerodynamics of Wind Turbines” 2nd Ed by Martin OL Hansen.

Furthermore, reliance on BEM models will require a prior knowledge about a particular wind turbine generator.

US 2016/0084233 discloses some relevant art by describing a method and system for evaluating performance of one or more wind turbines after actively changing a setting, such as software upgrades, control upgrades, hardware upgrades, etc., on the wind turbine; see paragraphs. However, the disclosure assumes precise a-priory information about changing a setting.

Limitations and expanded usage are to be overcome.

Object of the Invention

It is an object to provide a simple procedure or system that can be applied to wind turbine generators without having prior knowledge about the configuration.

One object is to be able to detect changes in operation of a wind turbine generator. One object is to quantify changes in operation and impact on performance of a wind turbine generator. One object is to determine optimal parameters to be used to control or operate a wind turbine.

It is also an object to be able to achieve one or more of the above objects without or with only partial prior information or knowledge about a wind turbine generator.

Description of the Invention

An objective is achieved by a method of evaluating operation of at least one wind turbine generator (WTG) in a set of wind turbine generators (WTGs) controlled by at least one controller operating the WTG based on at least one set of settings. The method may comprise one or more of the following acts.

There may be an act of monitoring operation of at least one WTG.

There may be an act of monitoring a set of settings used by the controller operating the at least one WTG.

There may be an act of detecting a change in the settings and recording a time-of-change.

There may be an act of determining an interrelationship between change in the operation and change in the settings before and after the detected time-of-change.

Thereby is achieved valuable operational information without actually requiring direct access to the controller or at least only having partial access to the controller.

This allows operators or owners of a WTG to obtain real operational data or information about a particular WTG. The required method or actions further provide a foundation for inspection in an easy or simpler manner.

Thus, what is disclosed allows for monitoring, even as an additional layer on established wind turbine generators or as a third party, wind turbine generators and flagging or informing users or operators when controllers change; and even quantifying effects of such changes.

The method can furthermore easily be applied to wind turbine generators without the difficulties of reprogramming the controller or even “looking” at the controller.

In particular, the method or actions do not rely on more or less complicated engineering, aerodynamic or physics models or assumptions to be applied or computed to evaluate operation.

This will provide detection of changes that can be reported as part of surveillance and reporting.

A further advantage is that the actions provide measures that quantify performance of a WTG.

Furthermore, the results form a basis for determining optimal parameters or settings that can be used to feed to the controller.

Monitoring settings may involve monitoring output from the SCADA data. Monitoring settings may also include monitoring environmental conditions within a WGT or ambient conditions such as meteorological conditions.

In cases there may be direct access to controller data or parameters or settings may be inferred.

For detecting changes, one WTG may be enough. In other cases, detecting changes may involve applying the method to more WTGs. There may be a set of WTGs and quantifying the impact of the detected changes may rely on evaluating multiple WTGs. For example, when comparing the power between WTGs.

In another embodiment, multiple WTGs may be used to detect the changes.

The time-of-change, or point-of-change, may be identified by the implemented so-called “change point” algorithms. Some algorithms are readily available from software libraries and a person skilled in the art will be able to implement and modify an algorithm.

Thus, when applied to a horizontal axis, a wind turbine generator (WTG) has an apparatus for actively rotating (pitching) its blades about an axis and controlling the rotational speed of the rotor by adjusting the torque of a generator.

The method allows maximizing the power output by the generator while staying within certain load and noise limits. The controls which satisfy such goals may depend on the local conditions (e.g. shear, veer, and air density) along with the condition of the blades (dirt and/or leading edge erosion) . Therefore, it is very difficult to determine the optimal parameters for a given WTG at a given point in time. Such difficulties are minimized or overcome by the outlined methods.

For example, the control parameters may be changed when a software update is applied to a wind turbine. Additionally, a given version of the control software may have several modes of operation including some modes of operation which further limit noise or power. In some cases, operators are not aware of these changes in controls. Even when they are, operators may not be aware of the impact of these changes. When a software update e.g. changes the torque which reduces power output, the operator may wish to revert the software or to update it in order to have a new torque setting.

Applying the above methodology when several WTGs are updated with different values, it may even be possible to estimate optimal settings by constructing a regression function of performance (or loads or noise) versus control parameter or control setting values. This may be accomplished by comparing performance metrics, load sensors, and/or acoustic measurements before and after said changes occurred.

In an aspect, the interrelationship is determined purely empirically.

In an aspect, monitoring a set of settings is the output SCADA data.

In an aspect, determining an interrelationship is performed by use of a regression model between the change in operation and the change in settings.

Regression model may be used to perform a regression between values like power and pitch or power and torque. A regression may also use rpm and pitch, rpm and torque, or others.

In principle, the regression model may also use more than two variables in the model, e.g. torque as a function of rotor RPM and air density.

A piecewise linear regression may be used.

In an aspect, there may be a further act of determining multiple interrelationships amongst change in the operation and change in the settings.

In an aspect, determining multiple interrelationships is based on clustering techniques.

Applying clustering techniques may be used to determine different relationships. Subsequently, the data can be assessed when it changed from one relationship to another.

One usable clustering technique may be implemented based on the disclosure by Arias-Castro, E. and Chen, G. and Lerman, G. (2011) , "Spectral clustering based on local linear approximations. " , Electronic Journal of Statistics, 5: 1537–1587.

An alternative clustering approach may be to implement a method known as DBSCAN as described by Ester, Martin; Kriegel, Hans-Peter; Sander,

Xu, Xiaowei (1996) . Simoudis, Evangelos; Han, Jiawei; Fayyad, Usama M., eds. A density-based algorithm for discovering clusters in large spatial databases with noise. Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD-96) . AAAI Press. pp. 226–231.

A different approach may be to use a multivariable change-point algorithm to perform the action to reduce or eliminate a need for additional regression or clustering analysis.

In an aspect, the act of monitoring operation includes monitoring one or more of operational conditions such as power output or a power output performance measure. There may be an act of monitoring load measures. There may be an act of monitoring noise measures.

We monitor the SCADA data (e.g. relationship between torque and power) in order to detect when the control schedule has changed. Once we determine that the control schedule has changed, separate algorithms are employed to determine the impact of that control schedule on power performance, loads, or noise.

In an aspect, the act of monitoring the settings includes monitoring one, more, or all of pitch settings or torque settings. Generally the monitoring involves monitoring settings available as SCADA data.

In an aspect, the act of monitoring the settings includes monitoring one or more ambient conditions.

In an aspect, the interrelationship is determined empirically between monitored pitch and torque from SCADA data and power performance.

The empirical relationship between torque (pitch) and other variables like power or generator speed is first identified and a regression model of e.g. torque vs. power and density is created from the SCADA data.

As new data is collected, it is compared to this regression model and a change-point algorithm is used to detect if/when the torque schedule changes significantly.

The software then compares the performance (loads or noise) before the change to that after the change.

Once enough time has passed after the change, a statistically significant conclusion can be made and shown to the user.

After a change-point, the regression model is reconstructed only using data subsequent to that time and the process is repeated if/when another change-point is detected.

The system for detecting and evaluating pitch and torque changes in horizontal axis wind turbine generator comprising:

Detection of change in pitch and torque behaviour by analysing SCADA data and automatically report them to a software user.

Quantify the power performance, loads and/or acoustic effects of such changes in software and automatically report such information to a user.

Compute optimal parameters for controllers and report this with evidence to a user.

An object of the invention is achieved by a method of optimising performance of at least one wind turbine generator (WTG) in a set of wind turbine generators (WTGs) controlled by at least one controller operating the WTG based on at least one set of settings. The method may comprise one or more of the following acts.

There may be an act of evaluating operation according to any one or more of the methods or acts disclosed, and thereby determining interrelationship between change in the operation and change in the settings.

There may be an act of finding optimised settings with an optimal operation.

In an aspect, there may further be an act of passing optimised settings (1010) to at least one WTG in a set of WTGs controlled by at least one controller (20) operating the WTG.

The disclosed acts may be implemented in a computer program product comprising instruction to perform one or more actions of evaluating operation.

An object of the invention is achieved by a wind turbine operation evaluating system for empirically evaluating operation of a WTG controlled by a controller, the system comprising an operating monitoring system configured to monitor operation of at least one WTG and a setting monitoring system configured to monitor a set of settings used by the controller of the WTG. Furthermore, there may be a computer configured to detecting a change in the settings and recording a time-of-change and determining interrelationship between change in the operation and change in the settings before and after the detected time-of-change.

Example:

The method disclosed may be applied to real data and to provide for AEP (annual expected power) increase in wind farms, both by changing torque and pitch. The empirical relationship between torque or pitch and other variables like power or generator speed is first identified.

This may be performed by a regression model of e.g. torque vs. power and density and is created from the SCADA data. As new data is collected, it may be compared to this regression model and a change-point algorithm is used to detect if/when the torque schedule changes significantly. The software then compares the performance, loads or noise before the change to that after the change.

Once enough time has passed after the change, a statistically significant conclusion can be made and shown to an operator.

There are a number of ways that this can be done, e.g. using the prior performance metrics to evaluate changes in power performance before/after the change-point.

Confidence intervals may be applied to help determine when enough data has been collected to provide a meaningful conclusion about the impact of the changes.

Finding change-points and using a regression model as described above is just one implementation method.

Alternative or additional methods based on clustering techniques, like spectral clustering or DBSCAN, may be used to detect clusters in e.g. the space of torque and power.

If a time interval is associated with a given cluster, then a change has occurred before (after) the first (last) point in that cluster.

Changes may be applied to many WTGs in a wind farm, each by a different amount. For example, a change is made to wind farm A, such that the torque is reduced for many WTGs, but each by a different amount. Power performance may then be observed to increase for each WTG as a result of or after the torque was reduced. Improvements may be found to be nearly proportional to the %decrease in torque as shown as will be exemplified in details.

Example of optimising:

In an example of optimising performance, the following may be performed.

One way may be to apply a physics based trend matched to the empirical data. For example, the power performance as a function of pitch in particular ambient conditions should follow a (physics-based) curve Physics as a function of (pitch; p1, p2, ...) with parameters p1, p2, then the empirical available by monitoring of say different pitch settings allows to find values for p1, p2, ... and then the best or optimised values can be found by a lookup of the value of pitch (within this condition) that maximizes power performance.

The curve Physics as a function (Pitch; p1, p2, ...) may change over time, and therefore this optimal value may change too (e.g. due to blade contamination) , and therefore this is continuously monitored and updated.

Another way may be to use sufficient empirical data to locate e.g. max performance settings.

Description of the Drawing

Fig. 1 illustrates a wind turbine operation evaluating system and a WTG;

Fig. 2 illustrates monitoring empirical data, SCADA data, and may be used to evaluate operation;

Fig. 3 illustrates evaluating operation where the set of setting includes torque settings;

Fig. 4 illustrates performance ratios as a function of the torque ratio for wind farm having a set of WTGs;

Fig. 5 method of evaluating operation of at least one wind turbine generator (WTG) in a set of wind turbine generators WTGs controlled by at least one controller operating the WTG based on at least one set of settings;

Fig. 6 illustrates an implementation where determining an interrelationship is performed by a model that may be implemented as instructions in a computer;

Fig. 7 illustrates a further implementation of the method of evaluating operation as illustrated in figure 5;

Fig. 8 illustrates in view of figures 6 and 7, the act of determining the interrelationship performed by a model;

Fig. 9 illustrates in view of the previously disclosed implementation specific types of data monitored;

Fig. 10 illustrates a method of optimising performance of at least one wind turbine generator (WTG) in a set of wind turbine generators (WTGs) controlled by at least one controller operating the WTG based on at least one set of settings 31 as illustrated in figure 1;

Fig. 11 illustrates a flow diagram of a method of evaluating operation using elements or actions disclosed.

Detailed Description of the Invention

Item	No
Wind turbine generator, WTG	1
Tower	2
Nacelle	3
Rotor	4
Blade	5
Evaluating system	10
Controller	20
Setting monitoring system	30
Set of Settings	31
SCADA	40
Pitch setting	42
Torque setting	44
Operating monitoring system	50
Set of operating data	51
Power output performance measure	52
Load measure	54
Ambient conditions	60
Evaluating operation	100

Monitoring operation	110
Monitoring a set of settings	120
Detecting a change	130
Change in the settings	132
Change in the operation	134
Time-of-change	135
Determining an interrelationship	140
Determining multiple interrelationships	142
Interrelationship	145
Reset Determining interrelationship	150
Model/computer program product	200
Regression model	210
Clustering techniques	250
Triggering Alarm/Notify	300
Method of optimising performance	1000
Finding optimised setting	1010
Optimal operation	1020
Computer	2000

Figure 1 illustrates a generic wind turbine generator, a WTG, 1 having a tower 2 supporting a nacelle 3 with a rotor 4 having blades 5. The WTG 1 represents a typical wind turbine generator referred to in the following.

The WTG 1 may have a controller 20 for controlling the WTG 1. The controller 20 will generally be configured to communicate externally. Various operational data will generally be available based on monitoring systems. There may be a Supervisory control and data acquisition (SCADA) implementation.

Figure 1 also illustrates a wind turbine operation evaluating system 10 for empirically evaluating operation 100, as will be described in the subsequent figures, of a generic WTG 1 controlled by a controller 20, the evaluation system 10 comprises an operating monitoring system 50 configured to monitor a set of operating data 51 representing operation of at least one WTG 1.

The system 10 has a setting monitoring system 30 configured to monitor a set of settings 31 used by the controller 20 of the WTG 1. Generally, there may be a SCADA system providing access to SCADA data 40.

There is a computer 2000 configured to detecting a change 130 in the settings 30 and recording a time-of-change 135 and to determining an interrelationship 140 between change in the operation and change in the settings before and after the detected time-of-change as will be described in the following.

The computer is configured to execute a computer program product 200 comprising instruction to cause the computer 2000 to perform actions or instructions 100 as will be illustrated in the subsequent figures.

Figure 2 illustrates a result of the methods and systems disclosed along with definitions. The results are obtained based on empirical data from a WTG 1 as seen in figure 1 and by methods that will be described in the following. However, for illustrative purposes and definitions, showing the result may guide the understanding. It is also to be appreciated that although a person skilled in the art may know some of the mathematical algorithms, then simply trying to look for the features in the data and actually finding useful information in non-trivial.

Figure 2 specifically illustrates how monitoring 120 empirical data, SCADA data 40, over time can be applied to find different pitch schedules by use of the disclosed method or system.

A set of operating data 51, including power output performance 52, is provided by monitoring. The pitch setting 42 is provided in a set of settings 31. The data are obtained from the SCADA data 40 of a WTG 1.

Four plots or curves are shown; each corresponds to a different period of time. The data began Jan 2 2016 (1/2/2016) with the solid curve shown; and subsequent time periods are shown with data staring at indicated dates (7/5/2016; 7/8/2016; and 8/5/2016) .

On July 5 (7/5/2016) there was a sudden change, which is detected 130 as a change in settings 132A, to that next curve (7/5/2016) where the pitch was always -2 deg, which curve represents the interrelationship 140A between pitch and power. And thus, the change in power as a function, quantified, or mapped relationship, of the change in pitch settings is established or determined.

A few days later (7/8/2016) , the pitch schedule changed, which is detected 130 as a change in settings 132B, to that third curve (7/8/2016) , which starts at power=300, pitch=0 to about pitch=2deg at 1100kW and showing a change in operation 134 and which curve represents the interrelationship 140B between pitch and power.

Finally, on Aug 5 (8/5/2016) , the pitch schedule changed, which is detected 130 as a change in settings 132C, to that last dotted curve in which the pitch was always 2deg, which curve represents the interrelationship 140C between pitch and power.

The figure illustrates a robust method of change detection method for finding the pitch changes in the SCADA data 40. The actual data are points that vaguely scatter about the curves determined to represent the interrelationship after a time-of-change.

Additionally, for each of those periods of time monitoring of other metrics, e.g. power perf, loads, noise, is performed to enable information about which pitch schedule is best or worst according to different criteria.

Figure 3 illustrates evaluating operation where the set of setting 31 includes torque settings 44 and the set of operating data 51 includes power performance 52.

The graph shows the different torque schedules detected by the algorithm or method disclosed. In this case the values are scaled, which mean that there are no units. The real data are points vaguely scattered in the regions.

The data started December 31, 2015 (12/31/2015) and the grey area with black border lines is where 90%of the data points were during that time. On July 26, 2016 (7/26/2016) , the torque schedule changed (lower torque) such that 90 %of the values where within the region bounded by the dashed lines. The change is detected 130 as a change in settings 132. The curve region bounded by the dashed lines represents the interrelationship 140C between torque and power.

Although not directly seen from the figures or curves, some processing schemes may be applied. The method partitions data into bins of the conditional variable (s) . Within each bin regression is performed online as data arrives. Once there is statistical confidence in the regression within a given bin, it is used to estimate an error associated with subsequent points that arrive in that bin. If the errors become consistently large, a CUSUM is computed on the errors within all bins to determine if/when a change-point has occurred. If a change-point is detected, all bin data are reset and new regression models are created from subsequent data within each bin. Errors are computed again once enough subsequent data has arrived to have statistical confidence in the new regressions models.

The CUSUM method mentioned above may be implemented based on the disclosure by Grigg; Farewell, VT; Spiegelhalter, DJ; et al. (2003) . "The Use of Risk-Adjusted CUSUM and RSPRT Charts for Monitoring in Medical Contexts" . Statistical Methods in Medical Research. 12 (2) : 147–170.

The binning process helps the process, although not necessary, but binning may improve when searching for change points –many times there are conditions where the pitch (torque) is higher, but other conditions where it is lower, and others where it is the same after a change. Hence, if errors are analysed over all regions (without binning) , it will have large variation and will be difficult or impossible to decipher changes.

The changes were detected as disclosed. The figure illustrates a robust method of change detection method for finding the torque changes in the SCADA data

For each of those periods of time monitoring of other metrics (power perf, loads, noise) may be performed to determine which torque schedule is best or worst according to different criteria.

A person skilled in the art might appreciate that clustering techniques might also have been applied here.

Figure 4 illustrates a performance ratio (from summer to the rest of the year) as a function of the torque ratio (reduced amount to initial torque) for wind farm having a set of WTGs. The larger the torque reduction (further left) , the higher the performance (the lower the performance drop) .

The figure is based on methods and change of settings discussed in figure 3 for one WTG and here applied to WTGs in a farm.

The methods described earlier detected that several WTGs in this farm had significant torque schedule changes in the summer months.

There is a dot in this plot for each WTG that had such a detected change. The plot here shows that the power performance impact of the torque changes.

The plot shows that when the torque was reduced a lot (K _q ratio small, further to the left in the plot) , the performance penalty was less (higher in the plot) compared to WTGs where the torque dropped less (higher K _q ratio, further right in the plot) .

Hence, the interrelationship 140 is established, which quantifies that the bigger the torque schedule change (the more torque was dropped) , the better the power performance.

Performance ratio values shown in the plot: A performance ratio value here is the mean of the ratios of the active power (during summer months) to active power in ” similar conditions” in other times of the year, for one WTG. With “similar conditions” is understood similar wind speed, wind direction, turbulence, and/or wind shear. There is one value/point for each WTG.

As such, the performance ratio is a quantified measure of a change in operation 134.

K _q-ratio values shown in the plot: A K _q-ratio here is the mean of the ratios of the torque in the summer months, after the torque change, to torque during other parts of the year, for one WTG and after accounting for the density effect. There is one value for each WTG.

As such, the K _q-ratio is a quantified measure of a change in operation 132.

The interrelationship 140 between the performance and torque is determined by the curve with indicated uncertainty bands as thinner lines forming an envelope about the thicker curve.

Figure 5 illustrates, with reference to a general wind turbine generator 1 as previously described, a method of evaluating operation 100 of at least one wind turbine generator (WTG) 1 in a set of wind turbine generators (WTGs) controlled by at least one controller 20 operating the WTG based on at least one set of settings 30. The method 100 comprises one or more of the following acts.

There is an act of monitoring operation 110 of at least one WTG 1.

There is an act of monitoring a set of settings 120 used by the controller 20 operating the at least one WTG 11.

There is an act of detecting 130 a change in the settings 132 and recording a time-of-change 32.

There is an act of determining an interrelationship 140 between change in the operation 134 and change in the settings 132 before and after the detected time-of-change 32.

The act of determining the interrelationship 140 is determined purely empirically.

Monitoring operation 110 is illustrated to be based on a set of operating data 51.

Monitoring a set of settings 120 is illustrated to be based on is the output SCADA data 40.

Figure 6 illustrates an implementation where determining an interrelationship 140 is performed by a model 200 that may be implemented as instructions in a computer. The model 200 may be implemented as a regression model 210 to determine the interrelationship 140 between the change in operation 134 and the change in settings 132.

Figure 7 illustrates a further implementation of the method of evaluating operation 100 as illustrated in figure 5. The method 100 further involves an act of determining multiple interrelationships 145 amongst change in the operation 134 and change in the settings 132.

Figure 8 illustrates in view of figures 6 and 7, the act of determining the interrelationship140 performed by a model 200 that may be implemented as instructions in a computer, where the act determining multiple interrelationships 145 is based on clustering techniques 250 implemented as a model 200.

Figure 9 illustrates in view of the previously disclosed implementation specific types of data monitored.

The act of monitoring operation 110 may include monitoring a set of operating data, one 51 or more of power outputs or power output performance measure 52, load measures 54, and/or noise measures 56.

The act of monitoring the settings 120 may include monitoring one, more, or all of pitch settings 42 and/or torque settings 44. The settings may be from SCADA data 40.

The act of monitoring the settings 120 may also include monitoring one or more ambient conditions 60, such as wind direction, temperature, humidity, or similar meteorological conditions.

In a particular combination of determining an interrelationship 140, the monitored data may be empirically monitored pitch 42 and torque 44 from SCADA 40 data regressed against power performance 50.

Figure 10 illustrates a method of optimising performance 1000 of at least one wind turbine generator (WTG) 1 in a set of wind turbine generators (WTGs) controlled by at least one controller 20 operating the WTG based on at least one set of settings 31 as illustrated in figure 1. The method comprising acts of evaluating operation 100 according to any one or more of claim 1 to 10 and determining interrelationship 140 between change in the operation 134 and change in the settings 132 and finding optimised settings 1010 with an optimal operation 1020.

There may be an act of passing optimised settings 1010 to at least one WTG in a set of WTGs controlled by at least one controller 20 operating the WTG, such as illustrated in figure 1.

Figure 11 illustrates a flow diagram of a method of evaluating operation 100 using elements or actions disclosed.

New data or incoming data from a wind farm SCADA system is monitored 110, 120. A change in settings 132 or an update in estimated pitch/torque schedule is performed. If no detection of a change 130 is observed, then more data is monitored. If a detection of change 130 is observed, then a time-of-change 135 is recorded. An alarm or notification to a user is triggered 300 and there is an act of estimating or determining an interrelationship 140. The evaluation of operation 100 may be annual estimated power (AEP) , loads, or noise. Finally, there is an act of resetting 150 so that the estimation only applies data after the time-of-change 135.

Claims

A method of evaluating operation (100) of at least one wind turbine generator (WTG) (1) in a set of wind turbine generators (WTGs) controlled by at least one controller (20) operating the WTG based on at least one set of settings (30) , the method comprising acts of:

- monitoring operation (110) of at least one WTG (1) ,

- monitoring a set of settings (120) used by the controller (20) operating the at least one WTG (1) ,

- detecting (130) a change in the settings (132) from the monitored set of settings (120) and recording a time-of-change (135) of the detected change in the settings (132) , and

- determining an interrelationship (140) between change in the operation (134) and change in the settings (132) before and after the detected time-of-change (32) .
The method according to claim 1, wherein the interrelationship (140) is determined purely empirically based on monitored operation (110) and the monitored set of settings (120) .
The method according to claim 1 or 2, wherein monitoring a set of settings (120) is monitoring the output SCADA data (40) .
The method according to any one or more of claim 1 to 3, wherein determining an interrelationship (140) is performed by use of a regression model (210) relating at least two variables of the WTG, the two variables representing the change in operation (134) and the change in settings (132) .
The method according to any one or more of claim 1 to 4, further involves an act of determining multiple interrelationships (145) amongst change in the operation (134) and change in the settings (132) .
The method according to claim 6, wherein the act determining multiple interrelationships (145) based on clustering techniques (250) .
The method according to any one or more of claim 1 to 6, wherein the act of monitoring operation (110) includes monitoring one or more of:

- power output (50) or power output performance measure (52) ,

- load measures (54) , and/or

- noise measures (56) .
The method according to any one or more of claim 1 to 7, wherein the act of monitoring the settings (120) includes monitoring one, more, or all of:

- pitch settings (42) ,

- torque settings (44) .
The method according to any one or more of claim 1 to 8, wherein the act of monitoring the settings (120) includes monitoring one or more ambient conditions (60) .
The method according to claim 1, wherein the interrelationship (144) is determined empirically between monitored pitch (42) from SCADA (40) data and power performance (50) or torque (44) from SCADA (40) data and power performance (50) .
A method of optimising performance (1000) of at least one wind turbine generator (WTG) (1) in a set of wind turbine generators (WTGs) controlled by at least one controller (20) operating the WTG based on at least one set of settings (31) , the method comprising acts of:

- evaluating operation (100) according to any one or more of claim 1 to 10 and determining interrelationship (140) between change in the operation (134) and change in the settings (132) ,

- finding optimised settings (1010) with an optimal operation (1020) .
The method according to claim 11, further comprising an act of passing optimised settings (1010) to at least one WTG in a set of WTGs controlled by at least one controller (20) operating the WTG.
A computer program product (200) comprising instruction to perform one or more actions of evaluating operation (100) according to the methods of one or more of claim 1 to 12.
A wind turbine operation evaluating system (10) for empirically evaluating operation (100) of a WTG controlled by a controller (20) , the system comprising:

- a operating monitoring system (50) configured to monitor operation of at least one WTG (1) ,

- a setting monitoring system (30) configured to monitor a set of settings (31) used by the controller (20) of the WTG,

- a computer (2000) configured to

detecting a change (130) in the settings (30) from the monitored set of settings (120) and recording a time-of-change (135) of the detected change in the settings (132) ;

determining interrelationship (140) between change in the operation (134) and change in the settings (132) before and after the detected time-of-change (135) .
A computer program product (200) comprising instruction to cause the computer (2000) of claim 14 to execute the methods or actions of one or more of claim 1 to 13.