the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Nov 2024
Analysis and calibration of optimal power balance rotor-effective wind speed estimation schemes for large-scale wind turbines
Abstract. Modern wind turbines' size growth creates challenges in their control system design, particularly due to greater wind variability across larger rotor areas. As modern turbine control systems rely on the availability of accurate wind speed information, the increasing unrepresentativeness of pointwise measurement devices, such as anemometers, necessitates the incorporation of more representative rotor-effective wind speed (REWS) estimation. Classical REWS estimators, based on static power relations, often fail to account for dynamic changes, leading to inaccurate estimation. To overcome these challenges, this paper introduces a power balance-based REWS estimation framework and splits the estimation problem into two modules: an aerodynamic power estimator and a wind speed estimate solver. Two possible aerodynamic power estimation techniques are discussed based on numerical derivative and state estimation. As state estimator calibration remained a challenge for varying wind turbine sizes, a gain-tailoring method for the performance calibration throughout a range of modern wind turbine sizes has been derived for the state estimation-based aerodynamic power estimator. Two types of wind speed estimate solvers are analyzed, namely the continuous and iterative single-step methods. From the two modules, the best-performing methods—the state-estimation aerodynamic power estimator and iterative single-step wind speed solver—are chosen to form the optimal power balance REWS estimator. The combined optimal estimator is validated through OpenFAST simulations of the NREL-5MW and IEA-22MW turbines and compared against a baseline method. The proposed method demonstrates good tracking of the REWS, better noise resilience, and convenient estimator gain calibration across different turbine sizes.
- Preprint
(2601 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2024-158', Anonymous Referee #1, 08 Dec 2024
General Comments
The paper is well written and makes valuable contributions which opens the path to better and more accurate rotor effective wind estimators. This is critical to achieve effective control and lowering of loads and thus cost on modern very large wind turbines.The work is based on a number of assumptions which significantly limits the direct applicability of the work. These assumptions are clearly stated, but the expected consequences and limitations that these assumptions entail are not treated. For the paper to give a accurate picture of the contribution made and where further work is needed, I believe that such assumptions should be discussed as part of the paper.
Specific Comments
What follows are comments on the assumptions
Assumption 1: No drivetrain loss
This assumption is obviously not realistic and will easily lead to a rotor power estimation error in the order of 10%. It would be beneficial to describe how this assumption could potential be lifted in order to show the feasibility using the proposed scheme in the presence of uncertain drivetrain losses.
Assumption 2: Only below rated operation with constant pitch angle
This assumption could potentially make the results of the paper unobtainable in real conditions. The state of the art wind turbine control uses pitch extensively in the below rated region. For e.g. optimal power tracking outside the optimal lambda region, thrust peak shaving, noise reduction control, fore-aft tower damping.
The REWS is also critical for turbine control in the above rated region as various load reduction control techniques rely on accurate wind speed information. Ensuring that the estimator can work in above rated is therefore also of high importance.
From experience designing and tuning wind speed estimators assuming constant pitch angles can lead to poorly tuned wind estimators where the pitch angle signal contaminates the REWS estimate leading to poor dynamic performance. The tuning method procedure derived in this paper might therefore not work if this assumption is lifted.
A discussion on the consequences of lifting this assumption, the feasibility of the presented method when this assumption is lifted and possible mitigations would make the applicability of the method much clearer.
Assumption 3: Accurate power coefficient
Assuming an accurate power coefficient is reasonable given clean blades, however the power coefficient model used in this paper is not suitable for modern very large wind turbine designs such as the IEA reference wind turbines used in this paper. The model neglects the substantial aerodynamic effects of rotor deflection which are especially prevalent in the below rated region.
The effect is not seen in the simulations as the ElastoDyn module is used for the blade beam models and this model does capture the torsion of the blades which is critical for the aerodynamic performance. Instead using the BeamDyn module would show the effect of blade deflection on the estimator performance.
Discussing this aspect would likely give a clear path for future work. The authors could even consider computing the Cp value for a few operating points of the IEA22MW turbine near the rated wind speed with and without blade deflection in order to discuss / assess the magnitude of the error induced
No comments on assumptions 4 and 5.
Technical Corrections
Seems that the Fast simulations are sometimes referred to as mid-fidelity, sometimes high-fidelity. e.g. line 85 and 95.
Line 116 It is claimed that the reference turbines represent industrial turbine designs which is not accurate since these turbines are designed by academia. If you compare the relations of the listed quantities across industrially designed turbines I expect you will find a high variance.
Line 240 Often the "noise" seen on the speed measurement are unmodelled dynamics and not Gaussian white noise. Some of these unmodelled dynamics will be present in the fast simulation. The filtering chosen thus often needs to consider the unmodelled dynamics rather than the actual noise variance
Line 340 The turbulence intensity used for the simulations is very low. What justified using such low values?
Figure 5: It would be beneficial to include the normalized power error and maybe even the cubic root of the normalized error as this should be proportional to the resulting error in estimated wind speed.
Line 574 As part of assessing the performance of the estimator it would be beneficial to have step wind or gust simulation cases where it is easy to see the transient response e.g. lag. Minimizing the lag is very important for the turbine control in gust like scenarios.
Figure 13: A zoom would nice as the both the actual REWS and the I&I REWS have higher frequency content and it is not possible to see if they are actually correlated
Figure 15: Very hard to read as the bars blue bars cover the green bars. Consider changing from bars to lines, such that both distributions can easily be seen,Citation: https://doi.org/10.5194/wes-2024-158-RC1 -
RC2: 'Comment on wes-2024-158', Anonymous Referee #2, 16 Dec 2024
The paper discusses the topic of estimating the wind speed for wind turbines, the topic is relevant to wind energy industry. The paper is clear and well written, and it add a contribution to the field of wind turbine wind speed estimation, the methods used to develop both the aerodynamic power estimator and wind speed estimate solver has been well implemented and described. Nevertheless, the assumptions of neglecting the power losses between the generator and the gearbox of the wind make the method developed not suitable for direct implementation, the efficiency of the drivetrain can be added to the problem in future study based on the outcomes of this work.
There are minor typos observed in the manuscript as follow:
1- Page 3 line 76 "...outlines the contributions of this work" a colon (:) could be added after 'work'.
2- Page 7 line 157, the expression at the end of the line exceeds the page margin.
3- The front size in figures [1, 4 - 6, 8, 9, 11 - 16] is small and could be enlarged.
Citation: https://doi.org/10.5194/wes-2024-158-RC2 -
RC3: 'Comment on wes-2024-158', Anonymous Referee #3, 20 Dec 2024
The paper addresses the topic of wind estimation, which is of great relevance in wind turbine engineering, and uses a systematic approach and well-known tools such as Kalman filters and Luenberger estimators. Besides, the method has been validated using aeroelastic simulations. However, some important issues should be addressed before the paper is published:
- The introduction states that many wind turbine control structures are based on wind estimators. However, this is far from true. Wind turbine controllers are mainly based on the feedback of generator speed and nacelle acceleration because wind measurements and estimates should not be trusted or have to great a delay. How does the proposed method change this scenario?
- What is the main purpose of the proposed wind estimate? If it is going to be used for control purposes, how does the estimator block improve performance? As the control action is going to be based on the generator speed either way, why is a two step procedure (estimator+controller) a better option than the standard feedback? How does the main purpose affect the requirements on the estimated signal?
- The estimator is based on the idea that both the inertia and the cP are known values. Is that a realistic approach? Have the authors performed a sensitivity analysis?
- The paper says that high power wind turbines show great wind speed variance across the rotor. How does it affect the estimation? Would it make sense to focus on blade effective wind speed?
- Section 4.1 proposes a first order filter to eliminate measurement noise. Could the time constant be chosen by characterizing the spectrum of the sensor noise and the generator speed? Maybe a frequency domain analysis could be a useful guide.
The grammar should be reviewed, specifically the use of the Saxon genitive for nonanimate subjects and the subject-verb agreement.
Citation: https://doi.org/10.5194/wes-2024-158-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 149 | 34 | 8 | 191 | 3 | 5 |
- HTML: 149
- PDF: 34
- XML: 8
- Total: 191
- BibTeX: 3
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1