the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Impact of inflow conditions and turbine placement on the performance of offshore wind turbines exceeding 7 MW
Abstract. Accurately assessing wind turbine performance in large offshore wind farms requires a nuanced understanding of how inflow parameters—turbulence intensity (TI), wind shear, and wind veer—affect power production across different turbine rows. In this study, we analyze 13 months of 10-minute operational data from more than 40 high-capacity turbines in a North Sea offshore wind farm, complemented by nacelle-based LiDAR measurements of inflow. Our objectives are to (1) determine how power production differs between front, middle and rear sections of the farm under the influence of TI, shear, and veer, and (2) evaluate the effectiveness of International Electrotechnical Commission (IEC)–based normalization methods, including Rotor Equivalent Wind Speed (REWS) and turbulence corrections in the front row and inside a wind farm consisting of large-scale wind turbines.
The results indicate that the impact of wind shear and veer on power output is strongly dependent on the turbine location: free stream shear and veer correlate negatively with active power in the front row, yet show positive correlations in the mid and rear rows. In addition, the TI in the wake region has a distinct influence on power production—particularly at lower wind speeds—relative to the TI observed in the free-flow region. Finally, the rear section of the wind farm exhibits approximately 20 % lower variability in active power relative to the front section. These location-specific changes underscore the evolving nature of inflow conditions within large wind farms. Furthermore, IEC-based REWS do not fully capture the effects of shear and veer in a large scale wind turbines in an offshore environment. The findings highlight that turbines operating in non-free-flow conditions may require additional inflow-characterization parameters beyond standard IEC norms to achieve more accurate performance evaluations and enhance overall farm efficiency.
- Preprint
(1684 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2025-32', Anonymous Referee #1, 25 Mar 2025
-
RC2: 'Comment on wes-2025-32', Anonymous Referee #2, 18 May 2025
The manuscript entitled « Impact of inflow conditions and turbine placement on the performance of offshore wind turbines exceeding 7 MW” deals with the correlation between some parameters of the inflow conditions (veer, shear, turbulence intensity) and the power production at some sections of a wind farm. Some corrections as proposed by the IEC standards are applied and their impact on the results are discussed.
Unfortunately, (1) the poor description of the operational database, (2) the absence of proof of validity of the representativeness (as reference inflow measurement) of the nacelle lidar data located at a big distance from the wind farm of interest and, (3) the very low level of correlations (Pearson coefficient mainly lower than 0.3) between the inflow parameters and the power performance make the present manuscript not suitable for publication.
Major comments:
- LiDAR data measurement : (1) No detail on the measurement. Since the lidar is nacelle-mounted, one can assume that the measurement location is following the wind turbine orientation and so, is not always located at the same position. At which distance from the wind turbine is the measurement location? Is it out of the wind farm induction zone? How can you prove this?. (2) The inflow parameters are measured at 8km from the wind farm of interest. How do you prove that these measurements are representative of the inflow impacting the wind farm of interest? Considering the density of wind farms, there are probably strong blockage effects
- Turbine operational regime: were the 10-min periods used when not all wind turbines were under operation during the 10 minutes? If yes, it means that the level of interactions might be modified by the number of running wind turbines in the wind farm
- Wind sector selection: it is 105° wide. The wake interactions are very different depending on the wind direction. This wind sector range is too wide.
- Figure 3 is not cited in the manuscript. It is not clear how the rejected data had been determined.
- 4 to 9: the Pearson Correlation coefficients values are mainly between 0.3 and -0.3. This magnitude is generally considered as a proof of no correlation between two data sets. One would consider that the interpretation of the evolution of the Pearson coefficient with the wind speed, or the comparison between wind farm segments is fully invalid. If it is not the case, the authors need to explain why such low levels of correlations can be interpreted.
Minor comments:
- 2 and Line 91: 50% or 64% of the data are gathered in this wind sector range?
- Lines 87-88 : More details on the DB-scan algorithm
- Fig 3 : “blue” instead of “green”
- 3.3.2 Turbulence intensity correction and Equation 3 : it is not clear what the “simulated average power” refers to. Which simulation?
- Lines 167-168 : explain the Sliding Window Pearson Correlation approach
- 7 is not cited in the manuscript
Citation: https://doi.org/10.5194/wes-2025-32-RC2
-
RC2: 'Comment on wes-2025-32', Anonymous Referee #2, 18 May 2025
-
RC3: 'Comment on wes-2025-32', Anonymous Referee #3, 18 Jul 2025
### Review of WES-2025-32
**Impact of inflow conditions and turbine placement on the performance of offshore wind turbines exceeding 7 MW**
#### Authors
Konstantinos Vratsinis
Rebeca Marini
Pieter-jan Daems
Lukas Pauscher
Jeroen van Beeck
Jan Helsen
#### Overview:
#### Overview
This manuscript investigates the correlation between atmospheric inflow parameters—namely wind shear, veer, and turbulence intensity—and offshore wind turbine power production. The analysis is framed in the context of IEC-recommended normalization, with the goal of exploring performance dependencies across wind speed bins. The topic is clearly relevant for wind energy research and operations, particularly as offshore turbines exceed 7 MW in rated capacity and sit within more variable marine boundary layer conditions.However, it is not clear what novel contribution this manuscript makes to the field or the central hypothesis being investigated. IEC normalization approaches are not intended to serve as predictive models, but rather to establish standard design and compliance conditions. The fact that these normalized parameters correlate weakly with power production is not necessarily surprising, and the authors need to contextualize their work more rigorously. If the intent of the paper is to show that IEC-style normalization underrepresents atmospheric variability or leads to weak statistical connections to power, that point must be made explicit and defended with clear analysis. Even better would be to outline a meaningful improvement to the approach outlined in the IEC Standards.
In its present form, the paper suffers from several methodological and framing issues that limit its scientific clarity. Many of the assumptions are not justified, key steps in the analysis are missing or difficult to understand, and several potentially confounding factors are not adequately discussed.
#### Comments
- The nacelle-mounted lidar appears to be located much farther from the wind plant than indicated in the text. Figure 1 suggests a distance closer to 18–20 km rather than the 8 km noted. This large separation raises serious concerns about the representativeness of the inflow measurements used in the correlation analysis. The authors should justify the use of this lidar system for characterizing inflow to the plant or clearly state the limitations that this introduces.
- The manuscript lacks sufficient discussion of existing literature on nacelle-mounted lidar systems and their associated uncertainties, especially for continuous-wave systems used at long range. Important work should be cited and discussed to help the reader assess the quality and validity of the measurement inputs. Letizia et al., 2021, is a good place to begin.
- The authors assume that wind shear and veer estimates from the distant lidar are representative of conditions across the entire wind plant. This is a strong assumption, especially in the offshore context, where wind direction and shear can vary significantly across distances of several kilometers. The authors should validate this assumption using available SCADA data, such as nacelle wind speed or yaw measurements, or else explicitly acknowledge and try to quantify the uncertainty introduced by this assumption.
- The data selection process is not fully described. Specifically, it is not clear whether or how dynamic atmospheric events (e.g., wind speed ramps, gusts, frontal passages, etc.) are filtered out of the data set. These events can have a strong impact on the correlation strength between inflow and power production. The authors should consider filtering based on time series dynamics (e.g., Hamilton, 2020) to isolate quasi-stationary atmospheric periods. This is likely to produce stronger correlations between observed quantities.
- In Figure 1, several turbines within the wind plant are shown in gray and appear to be excluded from the analysis, but the reason for this exclusion is not stated. Similarly, if there are turbines located closer to the source of atmospheric measurements, perhaps they should be considered instead.
- The paper mentions the use of DB-SCAN for outlier detection but provides no mathematical description or reference. At a minimum, a citation should be added, along with a brief description of how the DB-SCAN parameters were selected. Moreover, it appears that the DB-SCAN filtering may bias the power deviation metric discussed in Section 3.2. By excluding data points that deviate from the nominal power curve, the authors may artificially reduce the observed variability. This potential bias should be acknowledged and discussed.
- The analysis sector from 180° to 285° includes turbines that are likely to be in the wake of upwind turbines. The authors should discuss the extent to which wake effects influence turbulence intensity estimates and correlation strength. It may be appropriate to apply a wake-added turbulence model or include a wake filter to avoid spurious correlations driven by wake-induced variability.
- The correlation analysis itself is not clearly explained. The use of a 0.75 m/s sliding window appears central to the results, but it is not described in sufficient detail. Does the window represent a binning interval for mean values, a pairwise filter on measurement differences, or something else? How are variables with different units (e.g., shear, veer, TI) handled in this scheme? Further, what range of wind speeds is included in the analysis, and how is the resolution of the sliding window chosen? The authors may wish to define the correlation structure in terms of normalized wind speed (e.g., $U/U_{rated}$) to provide more generalizable insight.
- Figures 4–11 present a variety of correlations between atmospheric parameters and power output or power deviation, but the organization of these figures is unclear. The alternation between absolute power and deviation metrics should be explained more explicitly. In most cases, the correlation coefficients are below 0.4. While low correlations are not inherently uninteresting, the authors should explore and discuss the potential reasons for these weak results. For instance, are the variables themselves weakly coupled, is the inflow poorly characterized, or are key intermediate variables missing from the analysis? The paper would benefit from a thoughtful discussion of these points, which could help inform future efforts to better predict wind turbine performance from inflow metrics.Citation: https://doi.org/10.5194/wes-2025-32-RC3
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
254 | 51 | 19 | 324 | 26 | 35 |
- HTML: 254
- PDF: 51
- XML: 19
- Total: 324
- BibTeX: 26
- EndNote: 35
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1