This article is a re-edited version of an excerpt from "Accuracy verification of offshore wind conditions using the mesoscale meteorological model WRF and CFD model at the Mutsu Ogawara site," presented at the 2021rd Wind Energy Utilization Symposium held in November 11.
1. In the beginning
Doppler LIDAR (hereafter referred to as LIDAR) is a type of remote sensing device that can remotely measure wind direction and speed based on the speed of movement of minute particles by emitting laser light and receiving the reflected light from them. Since LIDAR can observe wind conditions in the sky above the wind turbine hub height and over the ocean, it is expected to be used in wind power development.
Due to the observation principle, LIDAR has the characteristic that wind condition data is missing in clear air with few particles in the atmosphere, or in weather conditions such as snowfall or fog. If it were possible to supplement such missing data using the results of simulations based on the numerical model used in wind condition surveys, this would be a very useful method. Therefore, in this paper, as an initial analysis for the interpolation of missing data from LIDAR observations, we use the results of simulations based on the numerical model used in wind condition surveys, conducted by the NEDO "Fixed-bottom Offshore Wind Farm Development Support Project (Establishment of Offshore Wind Condition Survey Methods)".[1]Using the results of research at Nomutsu Ogawara Port, we attempted to verify the accuracy of wind condition simulations using a numerical model.
2. Target area and weather observation
For the analysis in this study, we used wind speed and direction at 1 to 2 m above sea level, observed by a vertical Doppler lidar at two locations, St. A1 on land and St. B on the ocean (Figure 50). The wind direction at this site can be broadly divided into the ocean sector (250-0°; sea breeze) and the land sector (180-180°; land breeze), with the coastline stretching from north to south as the boundary.[2]In addition, air temperature (at two altitudes) and sea surface temperature were measured at St.B, and based on these observations, an analysis was also conducted dividing the atmospheric stability into two cases, stable and unstable (Table 2).
Figure 1. Overview of observation points and observation elements
Table 1. Atmospheric stability classification at St.B
3. Numerical model
In this study, we used the meso-scale meteorological model WRF[3]and the CFD model MASCOT[4]The offshore wind conditions at St.B were estimated using the numerical models. Table 2 shows the calculation settings for each model, and Table 3 shows four calculation patterns based on the input method of observed values.
Table 2. Numerical model calculation settings (Top: mesoscale meteorological model WRF, Bottom: CFD model MASCOT)
Table 3. Calculation patterns of the numerical model implemented in this study based on the input of observed values
WRF calculations do not require observed values within the analysis domain, but it is possible to correct the calculated values using observed values after calculation. Therefore, in this study, in addition to the uncorrected WRF estimated values (WRF-Raw), we attempted a method (WRF-VecC) to correct WRF-Raw with wind observation values at St. A1.6, an onshore site about 1 km away.
On the other hand, MASCOT calculations use wind observation values from one location as input values. Assuming that offshore wind conditions are estimated from the coast, in this study, St. A1 was used as the input value for wind conditions to estimate St. B. In addition, to investigate the effect of the input altitude on the estimated values, two types of input values were used: an altitude of 1 m (MAS-59 m) assuming a wind mast, and an altitude of 059 m (MAS-120 m) equivalent to the wind turbine hub.
4. Overview of wind observation results
The vertical wind speed profile (Figure 1) and the average wind speed for the period at St. B (2 m height) observed at offshore St. B and onshore St. A120 are shown in Table 4.
Figure 2. Observed wind speed vertical profiles on land (St. A) and offshore (St. B)
Left → Right: All samples, stable, unstable
Top→Bottom: All wind directions, land breeze, sea breeze
The stability classification was based on ocean observations.
Table 4. Average wind speed and number of samples for each period, classified by atmospheric stability and wind direction (St.B, 120m height)
While the wind speeds at the two points match well above an altitude of 150m, there is a difference in wind speed at the lower layers, and the closer you get to the land (sea) surface, the more the wind speed over land is < the ocean. This difference in wind speed at the lower layers is thought to be caused by the difference in surface roughness between the land and ocean surfaces. This is particularly noticeable during land breezes when the atmospheric stability over the ocean is unstable. When winds blow from the land to the ocean, if the atmospheric stability becomes unstable due to the supply of heat from the ocean, vertical mixing will strengthen, resulting in stronger winds.
Looking at the St.B observation values (stable) during sea breezes, the vertical shear of the wind speed is large up to about 70 m height, but becomes smaller at higher altitudes (about 70 to 180 m height), resulting in an S-shaped distorted profile. Therefore, under these conditions, it is difficult to estimate the wind speed vertical profile at the wind receiving height (about 120 m height) of offshore wind turbines located higher up in the sky from lower altitudes.
5. Wind condition estimation results using a numerical model
5.1 Vertical profile estimates from each numerical model
The vertical profile of wind speed estimated by each numerical model (Figure 3) and the frequency of wind direction occurrence (Figure 4) are shown. Taking into account the effects of topography, etc., these results only cover sea breezes.
The vertical wind shear estimated by the two WRF cases matches the observed trends well for both WRF-Raw and WRF-VecC during unstable conditions, while WRF-VecC shows a trend more similar to the observed values during stable conditions. This result suggests that using WRF in combination with observed values can estimate wind conditions with higher accuracy than using WRF alone. On the other hand, the shear for the two MASCOT cases is uniform regardless of atmospheric stability. This means that MASCOT cannot reproduce the vertical shear caused by atmospheric stability because it does not take into account thermodynamic effects.
Looking at the estimated wind direction occurrence frequency (Figure 4), we can see that WRF reproduces the vertical wind direction (vertical wind direction difference) with good accuracy. MASCOT reproduces the wind direction occurrence frequency at the input altitude with good accuracy, but shows the same occurrence frequency at all altitudes and is unable to represent the wind direction perturbations. These wind direction perturbations seen near the ground surface are caused not only by thermodynamic effects but also by the Coriolis force, and the characteristics of the wind direction perturbations cannot be seen in MASCOT, which does not take these meteorological effects into account.
Figure 3. Estimation results of each numerical model at St. B (vertical wind speed profile during sea breeze)
Left → Right: Observation, WRF-Raw, WRF-VecC, MAS-059m, MAS-120m
Figure 4. Estimation results of each numerical model at St. B (frequency of wind direction occurrence during stable conditions)
Left → Right: Observation, WRF-Raw, WRF-VecC, MAS-059m, MAS-120m
5.2 Wind speed estimation results of each numerical model (St.B 120m)
Figure 5 shows the average error Bias [%] of each numerical model estimate for the period-averaged wind speed at 120 m height at St.B.
Fig. 5. Average error bias [%] of mean wind speed for each numerical model (St.B 120m height)
Left to right: All wind directions, land breeze, sea breeze. The red numbers indicate the average error for all samples.
The WRF-Raw estimate has a bias of -0.4% for all samples, accurately estimating wind conditions at St.B throughout the period. This value was obtained by NeoWins at the Mutsu Ogawara Port site.[5]Results verified under the same conditions as the WRF calculation (+12.9%)[6]This is because the tendency for WRF wind speed to be overestimated near the coastline was significantly reduced by changing the surface roughness of the WRF calculation in this study to a value larger than the default value.[10]It is considered.
For MAS-120m, the error is small (-0.8% for all samples) for all wind direction sectors and atmospheric stability conditions. On the other hand, for MAS-059m, the bias shows a large negative value (-14.4%) for all samples, and this underestimation is mainly due to the tendency in the land sector. This is thought to be due to the fact that the vertical shear at the input point is not fully reproduced in the model, and such errors may occur due to the difference between the estimated altitude and the input altitude of MASCOT.
6. Summary
- Comparing wind observations over the ocean and on land, the wind speed over the ocean increases the closer you get to the surface (sea surface), and this is especially noticeable when the atmosphere is unstable.
- When using the bias of the period-averaged wind speed at an altitude of 120 m as an index, the accuracy of the WRF estimated value (uncorrected) and the MASCOT calculated value using land observation values at an altitude of 120 m as input was comparable, with both being within ±1%.
- For MASCOT, we found that the estimation accuracy depends heavily on the observation altitude of the input values. In order to minimize the influence of differences in the frequency of wind direction occurrence due to altitude and errors in vertical shear, it is important to use input values observed at altitudes close to the estimated altitude.
(Written by Mizuki Konagaya)
References
[1] Osawa, Shimada, Ogaki, Iwashita, Konagaya, Araki, and Imamura, 2020: NEDO Fixed-Bottom Offshore Wind Farm Development Support Project (Establishment of Offshore Wind Condition Survey Methods). Proceedings of the 42nd Wind Energy Utilization Symposium.
[2] Konagaya M., Ohsawa T., Inoue, T., Mito, T., Kato, H., Kawamoto, K., 2021: SOLA, Land–sea contrast of nearshore wind conditions: Case study in Mutsu-Ogawara , SOLA, 17, 225-228.
[3] Skamarock, WC, Klemp, JB, Dudhia, J., Gill, DO, Barker, DM, Wang, W., Powers, JG, 2008: A description of the advanced research WRF version 3. Tech. Note TN- 475+STR, 1−96.
[4] Ishihara, Yamaguchi, Fujino, and Hibi, 2002: Development of a nonlinear wind forecast model, MASCOT, and its applications. Proceedings of the 24th Wind Energy Utilization Symposium.
[5] NEDO, Offshore Wind Condition Map NeoWins, URL: https://appwdc1.infoc.nedo.go.jp/Nedo_Webgis/top.html (Accessed: February 2022, 2).
[6] Konagaya, Osawa, Mito, Kato, Inoue, Kawamoto, and Watanabe, 2018: Comparison of offshore wind resource estimates in coastal waters using multiple wind resource simulations. Proceedings of the 40th Wind Energy Utilization Symposium.
[7] Misaki, T., 2020: A study on improving the accuracy of coastal wind speeds simulated. Doctoral Dissertation, Graduate School of Maritime Sciences, Kobe University.