the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Aug 2024
Modelling Frontal Low-Level Jets and Associated Extreme Wind Power Ramps over the North Sea
Abstract. The increasing global demand for wind power underscores the importance of understanding and characterizing extreme ramp events, which are significant fluctuations in wind power generation over short periods, that pose challenges for grid integration. This study focuses on modeling frontal low-level jets (FLLJs) and associated extreme ramp-down events, particularly their impact on wind power production at Belgium offshore wind farms. Using the Weather Research and Forecasting (WRF) model, we analyzed five cases of extreme wind power ramp down events, including in-depth analysis of two cases and generalization of three additional cases. We assessed the sensitivity of various model configurations, including initial and boundary condition (IC/BC) datasets (ERA5 and CERRA), the activation of Fitch wind farm parameterization (WFP), planetary boundary layer (PBL) schemes, and single versus nested-domain configuration. Our findings indicate that CERRA IC/BCs provide a superior representation of atmospheric flow compared to ERA5, resulting in more accurate predictions of ramp timing, intensity, and FLLJ characteristics. The WFP significantly impacts wind power output by modeling turbine interactions and wake effects, leading to slightly lower wind speeds. The scale-aware Shin and Hong PBL scheme yielded a stronger FLLJ core at higher altitudes with a more pronounced jet nose, although wind speeds below 200 m were lower compared to the Mellor-Yamada-Nakanishi-Niio 2.5 scheme. Single-domain configuration proved more effective in simulating wind power ramps, although higher core heights and higher wind speeds below 200 m, resulting in a diffused jet profile. Our analysis highlights that reliable simulation of extreme ramps associated with FLLJ using a single domain configuration could reduce computational costs. Further, the FLLJ and associated extreme ramps can be predicted one day in advance, offering substantial benefits for operational efficiency in wind energy management.
- Preprint
(16796 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2024-99', Anonymous Referee #1, 23 Sep 2024
The paper aims to show how well frontal low-level jet (FLLJ) and associated ramp-down events can be modeled with the WRF model depending on forcing data, 2 different PBL schemes, and using wind farm parameterization or not. The topic is highly relevant, and the study presents valuable findings and discussions. However, the study suffers from unfinished editing (see below). While it does show some compelling anecdotal evidence for using direct nesting from CERRA and wind farm parameterization for modeling these FLLJ ramp events, some of the conclusions and the generality of the study remain questionable. Overall, the study is valuable but would benefit from a revision to clear up many of the questions below and finalize the required formatting.
Formatting
Please ensure that you follow the formatting guidelines of the journal: https://www.wind-energy-science.net/submission.html#assets. Some examples of what is not following the guidelines:
- Figures 1 and 2 (and Figures A1, A2, and A3): add subplot letters (a), (b), and (c)
- The date and time should follow the format: 25 July 2007 (dd month yyyy), 15:17:02 (hh:mm:ss).
- …The abbreviation "Fig." should be used when it appears in running text and should be followed by a number unless it comes at the beginning of a sentence, e.g.: "The results are depicted in Fig. 5. Figure 9 reveals that...".
- Regarding the notation, if units of physical quantities are in the denominator, contain numbers, and are abbreviated, they must be formatted with negative exponents (e.g. 10 km h-1 instead of 10 km/h).
Specific comments and questions
- P4L119-120: It’s not clear to me what “it’s” refers to here
- P5L145: Why did you choose the two cases from the five as you did? What criteria did you use?
- P6L162: I assume the “forecast” experiment is using GFS? Perhaps it’s obvious, but consider stating it explicitly here already
- MYNN2.5: what “bl_mynn_*” settings were used? The defaults of WRF v4.4? MYNN2.5 can be quite different depending on this
- In a study focussing on capturing the timing of an event, it’s surprising to me that you don't consider the influence of data assimilation (except to say that is one reason for the accuracy of re-analysis datasets). Why did you not test data assimilation and/or discuss this in the paper?
- You use 51 levels, why not more? Are you sure it’s not sensitive to this?
- You used 6 hr of spin-up. Are you sure the model has enough time to develop these strong weather events? Perhaps the poor performance of some of the experiments is due to insufficient spin-up time. See e.g. Lui et al. (2023)
- Figures 4-10: Why not show the results from the datasets used for forcing data: ERA5, CERRA, and GFS? It would be more convincing to show that downscaling is needed if I could see the reference data as well.
- The range of values in the colormap makes some of the plots difficult to it difficult to interpret. For example, do you really need to include values up to 30 m/s in figures D1-4. In Fig. C1, the breaks in the colormap seem to be inconsistent (sometimes 2 m/s, sometimes 1 m/s)
- Do you use the ERA5 pressure-levels or model-levels?
- In P14L298-300: What exactly are you arguing here? that WRF cannot generate sharp gradients from coarse ERA5 boundary-condition data? You say CERRA constitutes better BCs, but it is itself based on ERA5.
- P23L423: Why are you surprised that the model captured the event? Please elaborate on why it’s surprising
- Figure 11: Unfortunately, only the CERRA-1d1kmMYFP was included here, it would have been more convincing to show that the trend from cases 1 and 2 continues here
- P27L463: Same as above, you say CERRA provides better BCs than ERA5 to capture timing and intensity, but CERRA is based on downscaling from ERA5, so perhaps the problem is not ERA5 BCs but the downscaling and e.g. lack of data assimilation?
Technical corrections
- See “formatting” part
- P13L275: time-serires -> time-series
References
Liu, Ying, Lu Zhuo, and Dawei Han. "Developing spin-up time framework for WRF extreme precipitation simulations." Journal of Hydrology 620 (2023): 129443.
Citation: https://doi.org/10.5194/wes-2024-99-RC1 -
RC2: 'Comment on wes-2024-99', Anonymous Referee #2, 01 Oct 2024
Review of “Modelling Frontal Low-Level Jets and Associated Extreme Wind Power Ramps over the North Sea” by Baki et. al.
This study addresses the importance of modeling frontal low-level jets (FLLJs) and their impact on extreme wind power ramp-down events in Belgium’s offshore wind farms using the WRF model. Five cases studies are analyzed, with a detailed examination of two and generalization of the three additional cases. The paper evaluates different model configurations, including IC/BC datasets, wind farm parameterizations, planetary boundary layer schemes, and domain configurations. The study concludes that the CERRA dataset provides better atmospheric flow representation, and the Shin and Hong PBL scheme yields a stronger FLLJ core at higher altitudes. The findings also suggest that single-domain configuration is more effective in simulating extreme wind ramps.
The study presents a comprehensive investigation of the impact of different modeling configurations in FLLJs and wind power ramps, which is scientifically significant for wind energy forecasting. The methodological approach is sound, and the use of multiple case studies adds robustness to the findings. The manuscript is generally well-organized, with a clear structure and logical flow. However, there are sections that could be benefit from improved clarity and additional explanations. I would recommend a major revision before its possible publication on WES.
Major comments.
It is unclear how the timing of a down-ramp is determined. While Equation 1 defines the intensity of the ramp across different scales, what thresholds are used to identify the ramp? Additionally, is any preprocessing (e.g., smoothing) necessary to avoid false alarms?
The introduction highlights that unforecasted wind power down-ramps can lead to significant profit losses for farm owners. What is the critical timeframe within which an effective plan can be made in advance? Does a one-day forecast fall within this time window?
When comparing the runs with and without WFS, does the way calculating wind power influence the results? For instance, the choice of the power curve.
I am not fully convinced that CERRA-1d1kmMYFP outperforms CERRA-2d1kmMYFP or other CERRA-forced runs in simulating wind speed and direction in Figure 5 and Figure 6. CERRA-1d1kmMYFP produces a much more gradual change in wind speed and direction. The second pick at around 09:30 in case one is not captured in CERRA-1d1kmMYFP. As for the timing of wind power down ramp, a larger offset is also observed in other cases such as case 3.
Specific comments:
Line 42-43. Hasn’t this weather event been forested by the weather forecast service? Any potential reason for the difficulty? – I mainly want to confirm the down ramp event was not forecasted due to inaccurate weather forecast.
Line 72-74. A synoptic-scale phenomenon can have features at different scales. The synoptic-scale feature of FLLJ is expected to be better modeled than its local-scale feature such as the timing of passing a wind farm and the intensity at a single location.
Line 145. What does the criteria refer to?
Figure 1 and others. Please add a label to each sub-panel.
Line 208-210. What is the main difference between the MYNN2.5 and SH schemes? A brief explanation should help readers understand the rationale behind selecting these schemes and the distinctions they can expect.
Line 217. It appears to be a typo, yet this sentence stands as its own paragraph.
Line 219-224. Please give a reference for the schemes listed here. Change meter to m and kilometers to km.
Line 249-250. Are the percentage numbers calculated using eq. 1?
Line 290-291. Please see my preview comment, could you add a label to each panel, so that it can be cited by labels.
Line 315-316. Does the second ramp read from Figure 5?
Line 421. Does the Elia day-ahead forecast incorporate outputs from numerical weather forecast models? While detailed information isn't available, a high-level introduction would be helpful. Out of curiosity, what does the GFS original time series look like? This insight could be useful in determining whether the GFS serves as a low-cost alternative to Elia and whether dynamic downscaling is necessary for informing farm owners about upcoming down-ramps.
Citation: https://doi.org/10.5194/wes-2024-99-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 208 | 40 | 198 | 446 | 9 | 7 |
- HTML: 208
- PDF: 40
- XML: 198
- Total: 446
- BibTeX: 9
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1