Regional Study of Changes in Wind Power in the Indian Shelf Seas over the Last 40 Years
Abstract
:1. Introduction
2. Data and Methodology
2.1. Data
2.2. Methodology
3. Results and Discussion
3.1. Variations in Wind Power
3.2. Characteristics of Wind Power
3.3. Consistency of Wind Power
3.4. Exploitable Wind Speed
3.5. Directional Distribution of Wind Power
3.6. Long-Tterm Changes in Wind Power
4. Conclusions
Author Contributions
Funding
Acknowledgments
Conflicts of Interest
Notations and Abbreviations
AS | Arabian Sea |
BoB | Bay of Bengal |
ECMWF | European Centre for Medium-Range Weather Forecasts |
ENSO | El Niño–Southern Oscillation |
ERA5 | Fifth generation of ECMWF atmospheric reanalyses of the global climate |
JTWC | Website Joint Typhoon Warning Centre |
Mv | monthly variability index |
ONI | Oceanic Niño Index |
P | wind power density |
R | correlation coefficient |
Sv | variability index |
V | wind speed |
References
- Neill, S.P.; Hashemi, M.Z. Fundamentals of Ocean Renewable Energy; Academic Press: London, UK, 2018; p. 336. ISBN 9780128104484. [Google Scholar]
- Kåberger, T. Progress of renewable electricity replacing fossil fuels. Glob. Energy Interconnect. 2018, 1, 48–52. [Google Scholar] [CrossRef]
- GWEC. Annual Market Update; Global Wind Report; The Global Wind Energy Council (GWEC): Brussels, Belgium, 2018. [Google Scholar]
- GWEC. Global Wind Statistics 2016. 2017. Available online: www.gwec.net (accessed on 23 August 2019).
- Blaabjerg, F.; Ma, K. Wind Energy Systems. Proc. IEEE 2017, 105, 2116–2131. [Google Scholar] [CrossRef] [Green Version]
- Sheridan, B.; Baker, S.D.; Pearre, N.S.; Firestone, J.; Kempton, W. Calculating the offshore wind power resource: Robust assessment methods applied to the U.S. Atlantic Coast. Renew. Energy 2012, 43, 224–233. [Google Scholar] [CrossRef]
- Hasager, C.B.; Mouche, A.; Badger, M.; Bingöl, F.; Karagali, I.; Driesenaar, T.; Stoffelen, A.; Peña, A.; Longépé, N. Offshore wind climatology based on synergetic use of Envisat ASAR, ASCAT and QuikSCAT. Remote Sens. Environ. 2015, 156, 247–263. [Google Scholar] [CrossRef]
- Chang, R.; Zhu, R.; Badger, M.; Hasager, C.B.; Xing, X.; Jiang, Y. Offshore Wind Resources Assessment from Multiple Satellite Data and WRF Modeling over South China Sea. Remote Sens. 2015, 7, 467–487. [Google Scholar] [CrossRef] [Green Version]
- Guo, Q.; Xu, X.; Zhang, K.; Li, Z.; Huang, W.; Mansaray, L.R.; Liu, W.; Wang, X.; Gao, J.; Huang, J. Assessing Global Ocean Wind Energy Resources Using Multiple Satellite Data. Remote Sens. 2018, 10, 100. [Google Scholar] [CrossRef] [Green Version]
- Zheng, C.; Xiao, Z.; Yuehua, P.; Li, C.; Du, Z. Rezoning global offshore wind energy resources. Renew. Energy 2018, 129, 1–11. [Google Scholar] [CrossRef]
- Murthy, K.S.R.; Rahi, O.P. A comprehensive review of wind resource assessment. Renew. Sustain. Energy Rev. 2016, 72, 1320–1342. [Google Scholar] [CrossRef]
- Tank, V.; Bhutka, J.; Harinarayana, T. Wind Energy Generation and Assessment of Resources in India. J. Power Energy Eng. 2006, 4, 25–38. [Google Scholar] [CrossRef] [Green Version]
- Zheng, C.; Pan, J.; Li, J. Assessing the China Sea wind energy and wave energy resources from 1988 to 2009. Ocean Eng. 2013, 65, 39–48. [Google Scholar] [CrossRef]
- Olauson, J. ERA5: The new champion of wind power modelling? Renew. Energy 2018, 126, 322–331. [Google Scholar] [CrossRef] [Green Version]
- Hersbach, H.; Dee, D. ERA5 reanalysis is in production. ECMWF Newsl. 2016, 147, 7. [Google Scholar]
- Bosch, J.; Staffell, I.; Hawkes, A.D. Temporally explicit and spatially resolved global offshore wind energy potentials. Energy 2018, 163, 766–781. [Google Scholar] [CrossRef]
- Arent, D.; Sullivan, P.; Heimiller, D.; Lopez, A.; Eurek, E.; Badger, J.; Jørgensen, H.; Kelly, M. Improved Offshore Wind Resource Assessment in Global Climate Stabilization Scenarios. Technical Report: NREL/National Renewable Energy Laboratory (NREL), TP-6A20–55049; 2012; pp. 1–24. Available online: http://www.nrel.gov/docs/fy13osti/55049.pdf (accessed on 23 August 2019).
- Wind Europe. Offshore Wind in Europe. Key Trends and Statistics 2018; Technical Report; Wind Europe: Brussels, Belgium, 2018; Available online: https://windeurope.org (accessed on 3 December 2019).
- GWEC. Global Wind 2018 Report; Technical Report; Global Wind Energy Council: Brussels, Belgium, 2018; Available online: http://www.gwec.net (accessed on 3 December 2019).
- Sanjeev, G.; Paresh, C.D. Offshore wind power resource assessment using Oceansat-2 scatterometer data at a regional scale. Appl. Energy 2016, 176, 157–170. [Google Scholar] [CrossRef]
- Calif, R.; Schmitt, F. Modeling of atmospheric wind speed sequence using a lognormal continuous stochastic equation. J. Wind Eng. Ind. Aerodyn. 2012, 109, 1–8. [Google Scholar] [CrossRef]
- Ramon, J.; Lledo, L.; Torralba, V.; Soret, A.; Doblas-Reyes, F.J. What global reanalysis best represents near-surface winds? Q. J. R. Meteorol. Soc. 2019, 145, 3236–3251. [Google Scholar] [CrossRef] [Green Version]
- Findlater, J. A major low-level air current near the Indian Ocean during the northern summer. Quart. J. R. Meteorol. Soc. 1969, 95, 362–380. [Google Scholar] [CrossRef]
- Sirisha, P.; Remya, P.G.; Modi, A.; Tripathy, R.R.; Balakrishnan Nair, T.M.; Rao, B.V. Evaluation of the impact of high-resolution winds on the coastal waves. J. Earth Syst. Sci. 2019, 128, 226. [Google Scholar] [CrossRef] [Green Version]
- Chowdhury, S.; Zhang, J.; Messac, A.; Castillo, L. Unrestricted wind farm layout optimization (UWFLO): Investigating key factors influencing the maximum power generation. Renew. Energy 2012, 38, 16–30. [Google Scholar] [CrossRef]
- Aboobacker, V.M.; Shanas, P.R. The climatology of shamals in the Arabian Sea–Part 1: Surface winds. Int. J. Climatol. 2018, 38, 4405–4416. [Google Scholar] [CrossRef]
- Anoop, T.R.; Shanas, P.R.; Aboobacker, V.M.; Sanil Kumar, V.; Nair, L.S.; Prasad, R.; Reji, S. On the generation and propagation of Makran swells in the Arabian Sea. Int. J. Climatol. 2020, 40, 585–593. [Google Scholar] [CrossRef]
- Naseef, T.M.; Kumar, V.S. Climatology and trends of the Indian Ocean surface waves based on 39-year long ERA5 reanalysis data. Int. J. Climatol. 2020, 40, 979–1006. [Google Scholar] [CrossRef]
- Anoop, T.R.; Kumar, V.S.; Shanas, P.R.; Johnson, G. Surface wave climatology and its variability in the north Indian Ocean Based on ERA-interim reanalysis. J. Atmos. Ocean. Technol. 2015, 32, 1372–1385. [Google Scholar] [CrossRef] [Green Version]
- Shanas, P.R.; Kumar, V.S. Trends in surface wind speed and significant wave height as revealed by ERA-Interim wind wave hindcast in the Central Bay of Bengal. Int. J. Climatol. 2015, 35, 2654–2663. [Google Scholar] [CrossRef]
- George, V.; Kumar, V.S. Wind-wave measurements and modelling in the shallow semi-enclosed Palk Bay. Ocean Eng. 2019, 189, 106401. [Google Scholar] [CrossRef]
- Kumar, V.S.; Anoop, T.R. Spatial and temporal variations of wave height in shelf seas around India. Nat. Hazards 2015, 78, 1693–1706. [Google Scholar] [CrossRef]
- Torralba, V.; Doblas-Reyes, F.J.; Gonzalez-Reviriego, N. Uncertainty in recent near-surface wind speed trends: A global reanalysis intercomparison. Environ. Res. Lett. 2017, 12, 114019. [Google Scholar] [CrossRef]
- Klein, S.A.; Soden, B.J.; Lau, N.C. Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge. J. Clim. 1999, 12, 917–932. [Google Scholar] [CrossRef] [Green Version]
- Wyrtki, K. An Estimate of Equatorial Upwelling in the Pacific. J. Phys. Oceanogr. 1981, 11, 1205–1214. [Google Scholar] [CrossRef] [Green Version]
- Amrutha, M.M.; Kumar, V.S. Changes in Wave Energy in the Shelf Seas of India during the Last 40 Years Based on ERA5 Reanalysis Data. Energies 2020, 13, 115. [Google Scholar] [CrossRef] [Green Version]
Location | Geographic Position | Water Depth (m) | Annual Mean Wind Power Density (W/m2) | Occurrence of Exploitable Wind Energy (%) | |||||
---|---|---|---|---|---|---|---|---|---|
>50 W/m2 | >200 W/m2 | ||||||||
Range in Block | at Grid Point | 10 m | 110.8 m | 10 m | 110.8 m | 10 m | 110.8 m | ||
1 | 22.50° N; 68.25° E | 36–89 | 63 | 200.33 | 254.26 | 78 | 83 | 33 | 45 |
2 | 21.00° N; 69.50° E | 44–89 | 65 | 203.42 | 260.30 | 80 | 84 | 33 | 46 |
3 | 19.50° N; 71.25° E | 45–86 | 75 | 199.25 | 266.06 | 75 | 80 | 32 | 43 |
4 | 18.00° N; 72.00° E | 29–112 | 47 | 181.62 | 237.37 | 70 | 74 | 29 | 37 |
5 | 16.50° N; 72.75° E | 33–147 | 68 | 157.78 | 207.59 | 67 | 70 | 25 | 33 |
6 | 15.00° N; 73.50° E | 44–102 | 75 | 128.55 | 168.34 | 61 | 65 | 20 | 28 |
7 | 13.50° N; 73.50° E | 38–1171 | 80 | 131.12 | 172.81 | 58 | 62 | 21 | 27 |
8 | 12.00° N; 74.50° E | 66–1086 | 129 | 82.29 | 108.52 | 46 | 51 | 10 | 16 |
9 | 10.50° N; 75.50° E | 42–1171 | 46 | 81.86 | 111.01 | 49 | 55 | 10 | 18 |
10 | 09.00° N; 76.00° E | 13–362 | 90 | 86.10 | 117.34 | 47 | 51 | 13 | 21 |
11 | 07.75° N; 77.25° E | 47–90 | 61 | 299.02 | 443.64 | 80 | 81 | 58 | 64 |
12 | 08.25° N; 78.25° E | 5–1041 | 28 | 352.51 | 446.33 | 82 | 84 | 61 | 67 |
13 | 10.50° N; 80.25° E | 19–261 | 104 | 203.33 | 311.73 | 80 | 84 | 39 | 52 |
14 | 12.00° N; 80.00° E | 0–1187 | 27 | 107.32 | 174.00 | 66 | 76 | 15 | 32 |
15 | 13.50° N; 80.50° E | 2–1809 | 90 | 125.29 | 179.26 | 67 | 73 | 20 | 33 |
16 | 15.00° N; 80.25° E | 1–280 | 49 | 131.66 | 182.73 | 68 | 73 | 23 | 34 |
17 | 16.75° N; 82.50° E | 0–1113 | 99 | 150.62 | 209.72 | 67 | 73 | 26 | 38 |
18 | 18.50° N; 84.50° E | 0–813 | 41 | 100.11 | 163.58 | 54 | 65 | 15 | 30 |
19 | 20.00° N; 86.75° E | 6–995 | 58 | 182.57 | 227.63 | 65 | 69 | 32 | 38 |
20 | 21.00° N; 89.00° E | 30–700 | 75 | 165.93 | 196.96 | 61 | 64 | 27 | 32 |
Location | Maximum Wind Power (W/m2) | Date of Occurrence | Caused by |
---|---|---|---|
1 | 8431.35 | 20 May 1999 | Extremely severe cyclonic storm ARB 01 (02A) |
2 | 8246.88 | 08 Jun 1998 | Extremely Severe Cyclonic Storm ARB 02 (03A) |
3 | 5763.96 | 18 Jun1996 | Severe Cyclonic Storm ARB 01 (04A) |
4 | 5043.88 | 17 Jun 1996 | Severe Cyclonic Storm ARB 01 (04A) |
5 | 3597.72 | 23 Jul 1989 | During SW monsoon |
6 | 3534.93 | 10 Nov 2009 | Cyclonic Storm Phyan |
7 | 3015.67 | 29 May 2006 | During onset of SW monsoon |
8 | 3445.77 | 07 May 2004 | Cyclonic Storm ARB 01 |
9 | 2282.54 | 22 Jun 2007 | During SW monsoon |
10 | 2849.04 | 30 Nov 2017 | Very Severe Cyclonic Storm Ockhi |
11 | 2676.34 | 13 Nov 1992 | Severe Cyclonic Storm BOB 07 |
12 | 3513.90 | 13 Nov1992 | Severe Cyclonic Storm BOB 07 |
13 | 4075.57 | 03 Dec 1993 | Extremely Severe Cyclonic Storm BOB 02 |
14 | 4359.53 | 30 Dec 2011 | Very Severe Cyclonic Storm Thane |
15 | 8183.53 | 12 Dec 2016 | Very Severe Cyclonic Storm Vardah |
16 | 10,799.65 | 12 May 1979 | Cyclone One (1B) |
17 | 4414.27 | 12 Oct 2014 | Extremely Severe Cyclonic Storm Hudhud |
18 | 5177.32 | 12 Oct 2013 | Extremely Severe Cyclonic Storm Phailin |
19 | 8239.31 | 29 Oct 1999 | Super Cyclonic Storm BOB 06 (05B) |
20 | 11,081.02 | 25 May 2009 | Severe Cyclonic Storm Aila |
Location | Percentage of Occurrence | |||||||
---|---|---|---|---|---|---|---|---|
Wind Speed <3 m/s | Wind Speed <4 m/s | Wind Speed >6 m/s | Wind Speed >20 m/s | |||||
10 m | 110.8 m | 10 m | 110.8 m | 10 m | 110.8 m | 10 m | 110.8 m | |
1 | 7.05 | 6.09 | 17.27 | 13.76 | 48.12 | 59.04 | 0.02 | 0 |
2 | 7.27 | 5.99 | 15.93 | 12.60 | 49.76 | 61.17 | 0 | 0 |
3 | 9.17 | 7.78 | 19.95 | 16.18 | 45.98 | 56.56 | 0 | 0 |
4 | 12.61 | 11.34 | 24.60 | 21.48 | 41.84 | 49.85 | 0 | 0 |
5 | 15.22 | 13.83 | 28.02 | 25.06 | 38.23 | 45.72 | 0 | 0 |
6 | 18.81 | 17.26 | 33.50 | 30.16 | 32.27 | 39.45 | 0 | 0 |
7 | 21.87 | 19.98 | 36.95 | 33.58 | 31.47 | 37.74 | 0 | 0 |
8 | 29.61 | 26.97 | 47.62 | 43.21 | 19.53 | 26.26 | 0 | 0 |
9 | 27.53 | 24.91 | 44.82 | 39.85 | 20.83 | 28.82 | 0 | 0 |
10 | 33.39 | 31.12 | 47.89 | 44.48 | 23.62 | 30.87 | 0 | 0 |
11 | 10.75 | 10.15 | 17.66 | 16.48 | 66.91 | 70.58 | 0 | 0 |
12 | 9.08 | 8.30 | 15.01 | 13.64 | 70.00 | 73.81 | 0 | 0 |
13 | 6.84 | 5.54 | 15.87 | 12.48 | 54.12 | 64.85 | 0 | 0 |
14 | 12.78 | 9.68 | 27.25 | 12.48 | 29.12 | 47.44 | 0 | 0 |
15 | 13.72 | 11.57 | 26.95 | 19.69 | 35.79 | 46.92 | 0 | 0 |
16 | 13.98 | 11.87 | 26.93 | 23.31 | 37.29 | 47.40 | 0 | 0.01 |
17 | 15.10 | 12.84 | 27.59 | 22.53 | 40.12 | 50.75 | 0 | 0 |
18 | 24.62 | 18.45 | 40.66 | 30.50 | 26.54 | 41.68 | 0 | 0 |
19 | 17.62 | 15.81 | 29.96 | 26.77 | 42.81 | 48.23 | 0.08 | 0.01 |
20 | 21.42 | 19.53 | 34.73 | 31.76 | 38.10 | 42.79 | 0.01 | 0.01 |
Location | Different Percentile Wind Speed (m/s) at 10 m | Different Percentile Wind Speed (m/s) at 110.8 m | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
5 | 10 | 50 | 75 | 90 | 99 | 5 | 10 | 50 | 75 | 90 | 99 | |
1 | 2.7 | 3.4 | 5.9 | 7.5 | 9.2 | 12.1 | 2.8 | 3.6 | 6.6 | 8.3 | 9.9 | 10.8 |
2 | 2.6 | 3.4 | 6.0 | 7.4 | 9.2 | 12.2 | 2.8 | 3.7 | 6.6 | 8.3 | 9.9 | 10.9 |
3 | 2.4 | 3.1 | 5.8 | 7.4 | 9.3 | 12.3 | 2.5 | 3.3 | 6.4 | 8.3 | 10.2 | 11.3 |
4 | 2.0 | 2.7 | 5.5 | 7.2 | 9.0 | 12.2 | 2.1 | 2.8 | 6.0 | 7.9 | 10.0 | 11.2 |
5 | 1.8 | 2.5 | 5.3 | 6.9 | 8.6 | 11.6 | 1.8 | 2.6 | 5.7 | 7.6 | 9.5 | 10.7 |
6 | 1.6 | 2.2 | 4.9 | 6.5 | 8.0 | 10.8 | 1.6 | 2.3 | 5.3 | 7.1 | 8.8 | 9.9 |
7 | 1.4 | 2.0 | 4.8 | 6.5 | 8.2 | 11.2 | 1.5 | 2.1 | 5.1 | 7.1 | 9.1 | 10.3 |
8 | 1.2 | 1.7 | 4.1 | 5.6 | 6.9 | 9.7 | 1.2 | 1.8 | 4.4 | 6.1 | 7.6 | 8.6 |
9 | 1.2 | 1.8 | 4.3 | 5.8 | 7.0 | 9.4 | 1.3 | 1.8 | 4.6 | 6.3 | 7.8 | 8.7 |
10 | 1.0 | 1.5 | 4.1 | 5.9 | 7.2 | 9.3 | 1.1 | 1.5 | 4.4 | 6.5 | 8.0 | 8.9 |
11 | 2.0 | 2.9 | 7.5 | 9.1 | 10.2 | 11.9 | 2.1 | 3.0 | 8.4 | 10.4 | 11.8 | 12.5 |
12 | 2.2 | 3.2 | 7.7 | 9.3 | 10.4 | 11.9 | 2.3 | 3.3 | 8.6 | 10.5 | 11.8 | 12.4 |
13 | 2.6 | 3.3 | 6.4 | 8.1 | 9.5 | 11.5 | 2.6 | 3.5 | 7.0 | 9.0 | 10.6 | 11.5 |
14 | 2.0 | 2.7 | 5.1 | 6.2 | 7.3 | 9.0 | 2.3 | 3.0 | 5.9 | 7.3 | 8.6 | 9.4 |
15 | 1.9 | 2.6 | 5.3 | 6.6 | 7.7 | 9.7 | 2.1 | 2.8 | 5.8 | 7.4 | 8.7 | 9.5 |
16 | 1.9 | 2.6 | 5.3 | 6.7 | 7.9 | 9.7 | 2.0 | 2.8 | 5.8 | 7.5 | 8.9 | 9.6 |
17 | 1.8 | 2.5 | 5.4 | 6.9 | 8.3 | 11.0 | 1.9 | 2.7 | 6.1 | 7.8 | 9.3 | 10.2 |
18 | 1.3 | 1.9 | 4.6 | 6.1 | 7.4 | 9.7 | 1.5 | 2.1 | 5.4 | 7.2 | 8.7 | 9.6 |
19 | 1.6 | 2.3 | 5.5 | 7.5 | 9.1 | 11.9 | 1.7 | 2.4 | 5.9 | 8.1 | 9.8 | 10.8 |
20 | 1.4 | 2.0 | 5.1 | 7.1 | 8.9 | 12.1 | 1.5 | 2.2 | 5.8 | 8.7 | 11.3 | 12.9 |
Location | Optimum Direction (Deg) | 10 m | 110.8 m | ||
---|---|---|---|---|---|
Annual Mean Power (W/m2) | Time (%) | Annual Mean Power (W/m2) | Time (%) | ||
1 | 265 | 125.21 | 78.9 | 152.01 | 78.1 |
2 | 280 | 105.80 | 80.4 | 127.28 | 79.7 |
3 | 275 | 109.07 | 72.0 | 138.92 | 71.4 |
4 | 290 | 73.04 | 74.6 | 93.30 | 74.0 |
5 | 295 | 67.41 | 79.2 | 87.20 | 78.8 |
6 | 295 | 67.41 | 79.2 | 87.38 | 79.2 |
7 | 295 | 72.65 | 79.4 | 95.41 | 79.1 |
8 | 295 | 51.41 | 80.4 | 68.29 | 80.5 |
9 | 300 | 56.82 | 77.5 | 75.35 | 77.5 |
10 | 300 | 67.24 | 79.8 | 92.40 | 80.0 |
11 | 285 | 197.10 | 66.4 | 293.23 | 66.6 |
12 | 235 | 201.67 | 60.8 | 289.11 | 60.8 |
13 | 210 | 137.50 | 50.0 | 193.54 | 50.1 |
14 | 190 | 28.04 | 37.7 | 46.34 | 37.9 |
15 | 185 | 34.63 | 36.2 | 51.14 | 35.8 |
16 | 165 | 13.59 | 24.2 | 19.59 | 23.6 |
17 | 225 | 72.62 | 59.7 | 98.69 | 59.6 |
18 | 220 | 71.96 | 64.6 | 115.93 | 65.1 |
19 | 215 | 131.34 | 60.7 | 164.96 | 60.9 |
20 | 210 | 108.04 | 58.5 | 128.90 | 58.3 |
Location | Trend at 10 m (W/m2) | Trend at 110.8 m (W/m2) | ||||
---|---|---|---|---|---|---|
Mean | 95 Percentile | 99 Percentile | Mean | 95 Percentile | 99 Percentile | |
1 | −0.26 | −0.16 | −1.48 | −0.33 | −0.03 | −0.78 |
2 | −0.04 | 0.03 | −2.05 | 0.01 | −0.39 | −1.31 |
3 | 0.41 | 1.81 | 2.42 | 0.42 | 1.83 | 3.05 |
4 | 0.06 | 0.27 | 1.25 | 0.13 | 0.11 | 0.42 |
5 | −0.17 | −0.39 | −0.66 | −0.22 | −0.49 | −0.99 |
6 | −0.38 | −0.61 | −0.69 | −0.37 | −0.79 | −0.89 |
7 | −0.12 | −0.51 | −1.88 | −0.17 | −0.63 | −2.42 |
8 | −0.19 | −0.02 | −0.58 | −0.24 | 0.15 | −0.48 |
9 | −0.15 | −0.11 | 0.83 | −0.15 | −0.05 | 1.50 |
10 | 0.04 | 0.01 | 0.29 | 0.05 | −0.14 | 0.46 |
11 | −0.40 | −1.82 | −2.15 | −0.37 | −2.04 | −1.15 |
12 | 0.20 | −0.29 | −0.27 | 0.17 | −0.78 | 0.01 |
13 | 0.04 | −0.04 | 0.90 | 0.03 | 0.11 | 1.83 |
14 | −0.15 | −0.33 | −0.16 | −0.23 | −0.43 | −0.25 |
15 | −0.22 | −0.23 | 0.65 | −0.31 | −0.26 | 0.79 |
16 | −0.77 | −1.65 | −2.83 | −1.13 | −2.36 | −3.55 |
17 | −0.86 | −1.94 | −2.42 | −1.00 | −2.78 | −3.10 |
18 | −0.12 | −0.35 | 1.43 | −0.33 | −1.07 | 2.17 |
19 | −0.35 | −0.77 | 1.08 | −0.36 | −0.71 | 0.94 |
20 | −0.32 | 0.15 | −0.65 | −0.22 | 0.28 | −1.87 |
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Kumar, V.S.; Asok, A.B.; George, J.; Amrutha, M.M. Regional Study of Changes in Wind Power in the Indian Shelf Seas over the Last 40 Years. Energies 2020, 13, 2295. https://doi.org/10.3390/en13092295
Kumar VS, Asok AB, George J, Amrutha MM. Regional Study of Changes in Wind Power in the Indian Shelf Seas over the Last 40 Years. Energies. 2020; 13(9):2295. https://doi.org/10.3390/en13092295
Chicago/Turabian StyleKumar, V. Sanil, Aswathy B. Asok, Jesbin George, and M. M. Amrutha. 2020. "Regional Study of Changes in Wind Power in the Indian Shelf Seas over the Last 40 Years" Energies 13, no. 9: 2295. https://doi.org/10.3390/en13092295
APA StyleKumar, V. S., Asok, A. B., George, J., & Amrutha, M. M. (2020). Regional Study of Changes in Wind Power in the Indian Shelf Seas over the Last 40 Years. Energies, 13(9), 2295. https://doi.org/10.3390/en13092295