Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

ICEM 2017 June 27-29, 2017 @Bari, Italy

Accuracy of WRF simulations used for Japanese


offshore wind resource maps

Teruo Ohsawa, Hirokazu Uede, Takeshi Misaki, Mari Kato


Kobe University, Japan

Susumu Shimada
The National Institute of Advanced Industrial Science and Technology, Japan

1 /23
New offshore wind resource map in Japan
➢ Published by the New Energy and Industrial Technology Development
Organization (NEDO) on March 23rd, 2017.

NeoWins
http://app10.infoc.nedo.go.jp/Nedo_Webgis/top.html

2 /23
New offshore wind resource map in Japan
➢ Integrating information on wind resources with that on social and natural
environmental conditions

NeoWins
http://app10.infoc.nedo.go.jp/Nedo_Webgis/top.html

3 /23
Purpose of the study
➢ This NEDO project was conducted by 4 pertners.

National Institute of Kobe University Asia Air Survey Wind Energy Institute
Advanced Industrial Co., LTD. of Tokyo, Inc.
science and technology

➢ Kobe University’s missions:


Design of wind simulation for NeoWins and variation of the simulated winds

➢ In this presentation, we present


• Outline of the wind simulation in NeoWins
• Accuracy of the simulation based on validation at 41 offshore sites

4 /23
Outline of wind simulations for NeoWins
➢ Mesoscale model WRF was used to obtain wind information for NeoWins.
➢ Japanese waters within 30 km from the coast were divided into 160 areas.
➢ WRF simulations with 2.5km and 500m domains were carried out for each area.
➢ The simulations were conducted for three years (2009, 2012 and 2014).
➢ It took about a half year to complete all the simulations using super computers
with more than 4,000 CPU cores.

160 1º×1º areas


An example of WRF domains for an 1º×1º area
Domain 1 Domain 2
(100 x 100 2.5km-grids) (216 x 251 500m-grids)
Kitakyushu
北九州沖タワー

5 /23
Model configuration and input data

Model Advanced Research WRF (ARW) ver 3.6.1


Period 3 years (2009, 2012 and 2014)
Met: JMA Meso Analysis MANAL (3-hourly, 5 km×5 km)
Input data Soil: NCEP Final Analysis FNL (6-houlry, 1º×1º)
SST: AIST-Kobe Univ. MOSST (daily, 0.02º×0.02º)
Elevation: METI-NASA ASTER-GDEM (30 m)
Terrain data
Land use: MLIT National Land Numerical Information (100 m)
Domain 1: 2.5 km×2.5 km (100×100 grids)
Grids
Domain 2: 0.5 km×0.5 km (adjusted to the size of a 1º×1º area)
Vertical 40 levels (Surface to 100 hPa)
levels Lowest half levels: 23m, 73m, 130m, 199m, 287m, ・・・
Domain 1: Enabled (U, V, T, Q)
FDDA
Domain 2: Enabled (U, V, T, Q), excluding the interior of PBL
Shortwave process: Dudhia scheme
Longwave process: RRTM (Rapid Radiative Transfer Model) scheme
Cloud microphysics process: Ferrier (new Eta) scheme
Physics
PBL Process: Mellor-Yamada-Janjic (Eta operational) scheme
options
Surface layer process: Monin-Obukhov (Janjic Eta) scheme
Land-surface process: Noah land surface model scheme
Cumulus parameterization: None

6 /23
Annual mean wind speed from 160 domains

(made by WEIT)

7 /23
Method of validation
WRF-simulated winds are evaluated as follows.
➢ Hourly values of WRFOUT (max. 8,760 data for a year)
➢ Nearest grid value to an observation site
➢ Vertical interpolation (liner(U)–logarithmic(log Z))
➢ For 3 years (depending on data availability at an observation site)
➢ This presentation is focusing on annual and monthly wind speeds.

8 /23
In-situ observation sites for validation
1. 青森東岸沖(GPS波浪計)
2. 岩手北部沖(GPS波浪計) Red: Hub-height
3. 岩手中部沖(GPS波浪計)
4. 岩手南部沖(GPS波浪計) ( > 50m)
5. 宮城北部沖(GPS波浪計) measurements
6. 宮城中部沖(GPS波浪計)
7. 福島県沖(GPS波浪計)
8. 青森西岸沖(GPS波浪計)
9. 秋田県沖(GPS波浪計) Black: Lower-
10. 能代港風況観測タワー height (< 30m)
11. 山形県沖(GPS波浪計)
12. 波崎桟橋ライダー/桟橋上風速計 measurements
13. 銚子沖洋上風況観測タワー
14. 浦安沖
15. 千葉港波浪観測塔
16. 千葉港口第一号灯標
17. 川崎人工島
18. 平塚沖総合実験タワー
19. 静岡御前崎沖(GPS波浪計)
20. 伊勢湾湾奥
21. 伊勢湾湾央
22. 中山水道(2号灯標)
23. 伊勢湾湾口
24. 伊勢湾口沖(GPS波浪計)
25. 三重尾鷲沖(GPS波浪計)
26. 淀川河口
27. 大阪湾波浪観測塔
28. 阪南沖窪地
29. 関空MT
30. 田辺中島高潮観測塔
31. 和歌山南西沖(GPS波浪計)
32. 徳島海陽沖(GPS波浪計)
33. 黒潮牧場10号ブイ(室戸沖)
34. 黒潮牧場12号ブイ(高知沖)
35. 高知西部沖(GPS波浪計)
36. 黒潮牧場13号ブイ(足摺沖)
37. 北九州沖洋上風況観測タワー
38. 椛島沖浮体式洋上タワー
39. 沖の鳥島
40. JKEOブイ
41. KEOブイ

➢ In-situ measurements were collected from 41 offshore sites and used for validation.
9 /23
Validation at hub height - Kitakyushu -
NEDO met mast at Kitakyushu OBS WRF Bias
Year Month
Monthly mean @80m [m/s] [m/s] [m/s] [%]
2013 Jan 7.1 7.5 0.4 5.4
Feb 7.0 7.1 0.1 2.0
Mar 7.6 7.0 -0.6 -8.3
Apr 8.5 8.3 -0.3 -2.9
May 6.1 6.3 0.2 3.9
Jun 6.4 6.7 0.3 4.7
Jul 6.2 6.6 0.4 6.9
Aug 5.5 5.7 0.2 3.0
Annual bias: +0.3% Sep 6.9 5.9 -1.0 -14.8
Oct 7.1 7.5 0.4 5.7
Nov 8.2 8.3 0.1 1.0
Dec 8.8 8.8 0.0 0.0
Annual Mean 7.1 7.1 0.0 0.3

Wind rose @80m


Wind speed histogram @80m
(from NEDO HP)

Area 95

Data source: Offshore wind measurement data published on NEDO website.

10 /23
Validation at hub height - Choshi -
NEDO met mast at Choshi OBS WRF Bias
Year Month
Monthly mean @80m [m/s] [m/s] [m/s] [%]
2013 Jan 7.1 7.3 0.2 2.4
Feb 7.1 6.9 -0.2 -2.3
Mar 9.4 8.9 -0.6 -5.9
Apr 10.3 10.2 -0.1 -1.3
May 7.3 7.0 -0.3 -4.5
Jun 6.4 6.5 0.1 0.9
Jul 6.8 6.7 -0.1 -1.0
Aug 5.9 6.3 0.4 7.1
Annual bias: +1.2% Sep 7.4 7.8 0.4 5.0
Oct 8.8 8.7 -0.1 -1.0
Nov 7.1 7.5 0.4 5.1
Dec 6.2 7.1 0.9 15.2
Annual Mean 7.5 7.6 0.1 1.2

Wind rose @80m


Wind speed histogram @80m
(from NEDO HP)

Area 78

Data source: Offshore wind measurement data published on NEDO website.

11 /23
Validation at hub height - Noshiro -
Met mast at Noshiro Year Month
OBS WRF Bias
Monthly mean @50m [m/s] [m/s] [m/s] [%]
2014 Aug 6.1 5.8 -0.3 -4.1
Sep 6.0 5.8 -0.2 -3.1
Oct 7.0 7.1 0.1 1.7
Nov 8.6 8.3 -0.3 -3.9
Dec 11.8 11.2 -0.6 -7.8
2015 Jan 10.3 9.9 -0.4 -5.8
Feb 9.8 9.6 -0.2 -2.3

Annual bias:-2.9% Mar 9.0 8.8 -0.1 -1.9


Apr 6.5 6.4 -0.1 -1.8
May 6.1 5.8 -0.3 -3.3
Jun 6.1 5.7 -0.4 -5.4
Jul 4.8 5.1 0.3 3.4
Annual Mean 7.7 7.4 -0.2 -2.9

Wind rose @50m


Wind speed histogram @50m
(from proceedings of NEDO annual workshop 2015)

Area 38

Data source: Offshore wind measurement data in the NEDO project conducted by Obayashi Corporation and Omori Kensetsu KK.

12 /23
Validation at hub height - Hazaki -
LIDAR on pier of HORS Year Month
OBS WRF Bias
Monthly mean @87m [m/s] [m/s] [m/s] [%]
2015 Oct 7.3 7.5 0.2 2.1
Nov 7.4 7.4 0.0 0.4
Dec 6.4 6.5 0.2 2.6
2016 Jan 6.5 6.5 0.0 0.2
Feb 6.6 7.1 0.5 7.7
Mar 7.3 7.6 0.3 4.2
Hazaki Oceanographic Research Station Apr 7.5 7.9 0.3 4.0
(HORS) May 7.5 8.1 0.6 8.0
Jun 6.0 6.6 0.6 9.8
Annual bias: +4.5%
Jul 4.5 4.4 -0.1 -2.2
Doppler LIDAR Aug 7.6 8.1 0.5 6.3
Sep 6.0 6.3 0.3 5.3
Annual Mean 6.9 7.2 0.3 4.5

Wind rose @87m


Wind speed histogram @87m

Area 78
Data source: LIDAR measurement conducted in the NEDO project regarding offshore wind resource map.

13 /23
Annual bias at hub-height observation sites
Annual bias (estimation error in annual mean wind speed)
5.0 4.5
Annual bias (%)

2.5
1.2
0.3 Ave 0.8%
0.0
±5%
-2.5
-2.9
-5.0
Kitakyushu Choshi Noshiro Hazaki
(80m) (80m) (50m) (87m)

2013 2013 Aug/2014 – Jul/2015 Oct/2015 – Sep/2016


Doppler
LIDAR

(sited from proceedings of


(NEDO HP) (NEDO HP) NEDO annual workshop 2015)

➢ Annual biases are within ±5 %, which is the target accuracy of the project.
➢ The average is 0.8 % and three sites indicate a positive bias. → slightly
overestimation
14 /23
Monthly bias at hub-height observation sites
Monthly bias
20
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
15
Monthly bias (%)

10
5
0
-5
-10
-15
Kitakyushu Choshi Noshiro Hazaki
(80m) (80m) (50m) (87m)

RMS value of 12 monthly biases at each site


8
12 monthly biases (%)

5.9
RMS value of

6 5.1 5.3
4.1 Ave 5.1 %
4

0
Kitakyushu Choshi Noshiro Hazaki
(80m) (80m) (50m) (87m)

➢ Monthly bias is about ±5 % on average, though the max and min exceed ±15%.

15 /23
Annual bias at lower-height observation sites
Annual bias at 31 lower-height (<30m) observation sites
(averaged for 1 to 3 years with available data at each site)
15
12.1

10
6.9 6.9
5.6 6.2
5.1 4.9 5.2 5.3
5 3.7 3.6 3.1 4.2 4.3 3.6
3.5
Bias(%)

2.3 2.8 Ave.


2.1 1.8
1.4 1.3
0.7 0.5 0.7 0.8 +2.8%
0
-0.9-1.2
-2.1
-5 -3.4
-4.7

-10
1 2 3 4 5 6 7 8 9 11 12 14 15 16 18 19 23 24 25 26 27 28 29 30 31 32 33 34 35 36 39
Station number

➢ Minimum:-4.7% and Maximum: +12.1%


➢ 26 sites of 31 indicate a positive bias. → Overestimation
➢ Average for 31 sites is +2.8%. → cf. +0.8 % at hub-height observation sites

16 /23
Monthly bias at lower-height observation sites
RMS value of 12 monthly biases at each of 31 lower-height observation sites
(averaged for 1 to 3 years with available data at each site)
25
RMS value of 12 monthly biases(%)

20.5
19.8
20

14.9
15 13.8
13.1
11.8 11.3 11.7 12.0
11.1 10.8 11.1
9.9 9.9 Ave.
10 8.9 8.8 9.1 9.3 9.1
8.5
7.6
8.4
7.6
8.2
7.6 7.5
8.2 10.2%
7.4
5.5 5.6 6.0
5

0
1 2 3 4 5 6 7 8 9 11 12 14 15 16 18 19 23 24 25 26 27 28 29 30 31 32 33 34 35 36 39

Station number

➢ Minimum:5.6% and maximum: 20.5%


➢ Average for 31 sites: 10.2% → cf. 5.1 % at hub-height observation sites
➢ Indicating that the bias of WRF wind speed becomes smaller with height.

17 /23
Vertical variation of bias
Vertical profiles of annual mean wind speed Vertical profile of annual bias
100 220 220
Choshi met mast Hazaki (LIDAR)
Kitakyushu
80 200 200 Choshi
Hazaki
60 180 180
Height (m)

40 160 160

OBS
20 140 140
WRF

0 120 120
Decreasing

Height (m)
Height (m)

0 2 4 6 8 10
Wind Speed (m/s) with height
100
100 100
Kitakyushu met mast

80
80 80

60 60 60
Height (m)

40 40 40

OBS OBS
20
WRF
20 WRF 20

0 0 0
0 2 4 6 8 10 0 2 4 6 8 10
-5 0 5 10
Wind Speed (m/s) Wind Speed (m/s)
Bias (%)

➢ WRF tends to output a weaker vertical shear than the measured profile.
➢ The positive bias tends to decrease with height.
18 /23
Spatial variation of bias
Relation between annual mean wind speed and distance from coast
9

8
Wind speed (m/s)

5
OBS
4
WRF
3
0 5 10 15 20 25 30 35 40
Distance from coast (km)

Relation between bias and distance from coast


15
12
9
Bias (%)

6
3
0
-3
-6
0 5 10 15 20 25 30 35 40
Distance from coast (km)

➢ A bias of several % is found in nearshore waters within 30 km.


➢ The overestimation gradually decreases with the distance from coast.
19 /23
Relation between bias and fetch
Definition of fetch Number of samples in each bin of fetch

Fetch vs. bias (all plots)

➢ Wind direction is divided


into 8.
➢ The distance to the
nearest coast is defined
as fetch for the direction.

20 /23
Relation between bias and fetch
最低風速別
Fetch-bin averaged bias
10%

8%

6%

4%
≧0m/s
All data
2% ≧1m/s
Exc. <1m/s
Bias

0% ≧2m/s
Exc. <2m/s
0-10 10-20 20-30 30-40 40-60 60-100 100~ ≧3m/s
Exc. <3m/s
-2%
Exc. <4m/s
≧4m/s
-4%
≧5m/s
Exc. <5m/s
-6%

-8%

-10%
離岸距離

➢ Wind blowing over a shorter fetch (< 40 km) exhibits a positive bias.
➢ This indicates that the positive bias found in nearshore waters is attributed
to the winds blowing from land sectors.

21 /23
Conclusions
In this study, the accuracy of WRF simulations used for the Japanese
offshore wind resource map NeoWins, using measurements from 41
offshore observation sites. Main results are summarized as follows.
➢ WRF-simulated wind speed at a hub height has an annual bias within
±5 %, which is the target accuracy of the NEDO project.
➢ At a lower height (<30m) WRF has a positive bias of +2.8 % on
average in coastal waters, while it is +0.8 % at a hub height. This
indicates that the positive bias decreases with height.
➢ The large positive bias found in nearshore waters gradually decreases
with the distance from coast, vanishing at around 30 km.
➢ The positive bias is primarily caused by winds blowing from land,
though physical factors of this overestimation are still unknown.
Further investigation is necessary to find them to solve this problem.

22 /23
Acknowledgement
This work has been conducted under the NEDO project P07015
“Offshore Wind Condition Observation System Proving Research
(Offshore Wind Map)”.

Please visit here!


http://app10.infoc.nedo.go.jp/Nedo_Webgis/

23 /23

You might also like