Hyperparameter tuning framework for calibrating analytical wake models using SCADA data of an offshore wind farm

van Binsbergen, Diederik; Daems, Pieter-Jan; Verstraeten, Timothy; Nejad, Amir R.; Helsen, Jan

doi:https://doi.org/10.5194/wes-9-1507-2024

Articles | Volume 9, issue 7

https://doi.org/10.5194/wes-9-1507-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/wes-9-1507-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 9, issue 7

Research article

|

12 Jul 2024

Research article |

| 12 Jul 2024

Hyperparameter tuning framework for calibrating analytical wake models using SCADA data of an offshore wind farm

Diederik van Binsbergen, Pieter-Jan Daems, Timothy Verstraeten, Amir R. Nejad, and Jan Helsen

Download

Final revised paper (published on 12 Jul 2024)
Preprint (discussion started on 21 Aug 2023)

Interactive discussion

Status: closed

RC1: 'Comment on wes-2023-98', Robert Braunbehrens, 21 Sep 2023

The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-98/wes-2023-98-RC1-supplement.pdf

Citation: https://doi.org/10.5194/wes-2023-98-RC1
RC2: 'Comment on wes-2023-98', Anonymous Referee #2, 22 Sep 2023

The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-98/wes-2023-98-RC2-supplement.pdf

Citation: https://doi.org/10.5194/wes-2023-98-RC2
AC1: 'Comment on wes-2023-98', Diederik van Binsbergen, 26 Oct 2023

Dear reviewers,
We would like to thank you for your time providing thorough feedback and insightful comments and suggestions for our paper. Please find our answers to your comments and the corresponding updates to the paper in the attached file.
Thank you

Citation: https://doi.org/10.5194/wes-2023-98-AC1

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Diederik van Binsbergen on behalf of the Authors (26 Oct 2023) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (04 Dec 2023) by Julie Lundquist

RR by Anonymous Referee #2 (18 Dec 2023)

Suggestions for revision or reasons for rejection

Review 2

The authors responded satisfactorily to most of the comments and the changes to the manuscript improved it. Especially the inclusion of a map helped to follow the discussion. Unfortunately, two comments have not been addressed sufficiently in my opinion.

Comment 1

I was not satisfied with the authors’ response to the first general comment of Reviewer 2. The found relationship between the wake growth rate and the turbulence intensity differs from a body of literature that observed a stronger relationship from field experiments (e.g. Trabucchi et al., 2017), wind tunnel observations (Ishihara and Qian, 2018*), and large-eddy simulations (e.g. Niayifar and Porte-Agel, 2016). In my opinion, a discussion of what might contribute to the differences and mentioning possible limitations of the present study is required in the manuscript given the differences.

The authors replied to the general comment no. 1 of Reviewer 2 that their “decision to focus on the velocity deficit model and to exclude others, like the wake-added turbulence model, comes from the interdependencies these models show. Including both submodels could lead to a multimodal solution space, where certain parameters in one submodel influencing adjustments in another.” From this, I understand that the optimized parameters might change if those submodels are changed. Therefore, the found wake growth rates and the conclusion that “comparing the optimized parameters to the baseline reveals that the baseline parameters underestimate the wake effects, which subsequently leads to an overestimation of the expected yield” should be softened by mentioning this limitation in the paper explicitly in my opinion.

Regarding the implementation of the model that was optimized here, the reply also did not address my initial concern. It is true that nacelle mounted anemometers are not perfect, which can be seen from a comparisons to upstream looking lidars/upstream located met towers that have been published for some field campaigns. However, the power curve provided by the manufacturer is of unknown quality (or at least it is not provided in the manuscript) and the power curve might have been created for an undisturbed inflow (e.g. an inflow that can be described with a log- or power-law). It remains the questions if and how it can be applied to a waked or partially waked wind turbine in the interior of a wind farm. The reply of the authors did not rule out that the differences in the wake growth rates found here and those in other literature might be partly explained by such issues. Therefore, I am missing here as well that the authors mention such possible limitations of the their findings.

However, there are also possible reasons that can be mentioned why it might be different from the studies cited in the first paragraph above. They only used isolated wind turbines to determine the relationship between turbulence intensity and wake growth rate. It might be that the relationship is naturally different inside a wind farm, because the turbulence of the interior flow changes not only in its intensity, but also in scales. Some papers showed that the thrust coefficient of the wind turbine affects the wake recovery as well.

* Ishihara and Qian (2018) do not use a linear relationship, but the values for wake growth rates are larger still.

Comment 2

Section 2.4 / line 253-259: The filtering criterion is still mysterious to me and the clarification by the authors have not improved this. The authors say that they sometimes observe discrepancies between the active power and the power set point at wind speeds above rated wind speeds. Addressing this issue with threshold for the ratio of the wind speed variance and the wind direction variance is still not intuitively clear to me. I suspect that the criterion might work for the authors due to special circumstances at their experimental site, but I doubt whether it is an universal approach to detect this issue with the turbine operation (I wanted to test this on a set of SCADA data to verify, but I did not find the time in the end). Therefore, I believe the authors should add the following information to the paper:
(1) A summary of the impact of this filter criteria (How many data points does it remove from the total? Can it be shown that that it is successful at removing what the authors describe?).
(2) It should be clearly stated whether this filter criterion is based purely on a correlation or whether there is also a physical causation to support it (and in case of the former that it might not be a procedure that can be universally recommended to identify this state of turbine operation).

References

Trabucchi, D., Trujillo, J.-J., and Kühn, M. (2017). Nacelle-based lidar measurements for the calibration of a wake model at different offshore operating conditions. Energy Procedia 137, 77–88. doi: https: //doi.org/10.1016/j.egypro.2017.10.335. 14th Deep Sea Offshore Wind R&D Conference, EERA DeepWind’2017

Ishihara, T., & Qian, G. W. (2018). A new Gaussian-based analytical wake model for wind turbines considering ambient turbulence intensities and thrust coefficient effects. Journal of Wind Engineering and Industrial Aerodynamics, 177, 275-292.

Niayifar, A. and Porté-Agel, F., 2016. Analytical modeling of wind farms: A new approach for power prediction. doi:10.3390/en9090741. Energies, 9(9), p.741.

Hide

RR by Anonymous Referee #1 (22 Dec 2023)

ED: Reconsider after major revisions (29 Jan 2024) by Julie Lundquist

AR by Diederik van Binsbergen on behalf of the Authors (27 Feb 2024) Author's response Author's tracked changes Manuscript

ED: Publish subject to minor revisions (review by editor) (19 Apr 2024) by Julie Lundquist

AR by Diederik van Binsbergen on behalf of the Authors (22 Apr 2024) Author's response Author's tracked changes Manuscript

ED: Publish as is (02 May 2024) by Julie Lundquist

ED: Publish as is (09 May 2024) by Paul Fleming (Chief editor)

AR by Diederik van Binsbergen on behalf of the Authors (16 May 2024)

Short summary

Wind farm yield assessment often relies on analytical wake models. Calibrating these models can be challenging due to the stochastic nature of wind. We developed a calibration framework that performs a multi-phase optimization on the tuning parameters using time series SCADA data. This yields a parameter distribution that more accurately reflects reality than a single value. Results revealed notable variation in resultant parameter values, influenced by nearby wind farms and coastal effects.