assessment of wind energy potential over india using high

Assessment of wind energy potential over India usinghigh-resolution global reanalysis data

CHINA SATYANARAYANA GUBBALA1,* , VENKATA BHASKAR RAO DODLA

2

and SRINIVAS DESAMSETTI3

1Center for Atmospheric Science, Koneru Lakshmaiah Education Foundation, Green Fields, Vaddeswaram,Andhra Pradesh 522 502, India.2Department of Meteorology and Oceanography, Andhra University, Visakhapatnam, Andhra Pradesh 530 003,India.3National Centre for Medium Range Weather Forecasting, Ministry of Earth Sciences, A-50, Sector-62, Noida,Uttar Pradesh, India.

*Corresponding author. e-mail: [email protected]

MS received 27 February 2020; revised 20 October 2020; accepted 9 December 2020

An assessment of wind energy potential based onwind speed data over the Indian subcontinent has beenmadeusing high spatio-temporal resolution global reanalysis for the period from 1979 to 2018. Regions of high windspeed exceeding 4.5 m/s are identiBed over West Rajasthan, West Gujarat, Saurashtra and Kutch, CentralMaharashtra, Interior Karnataka, and Rayalaseema. Threshold wind speeds are noted to occur during thedaytime, and during the summermonths fromMay through September.Wind speeds and the spatial extent ofthreshold winds increase rapidly with height below 40 m and then gradually up to 100 m. The wind powerdensity is highestbetween50and80m,with thepotential highestoverGujarat,Kutch,and InteriorKarnatakaandmoderate over Saurashtra andRayalaseema.This study also notiBes that oAshorewind potential is higherthan over land, and most of the western parts of India are congenial for low wind farming. The present studyclearly delineates wind speed distributions and wind power productivity regions over the entire Indian sub-continent. The results would provide authentic wind speed and wind power potential information that wouldbe useful for the industries, government agencies, and industries concerning wind harness over India.

Keywords. Wind speed distribution; wind energy potential; Indian subcontinent; ERA global analysis.

1. Introduction

The world looms in energy crisis because of fast-depleting energy resources due to increasing energyusage from rapid industrialization. As naturalresources of energy are limited and require verylong periods for identiBcation and replenishment,the use of renewable energy gained importance andpriority at present. Renewable sources are hydro,wind, solar, geothermal, and biomass, whereas thenon-renewable sources are coal, natural gas, andcrude oil. Solar energy tops as it is abundantly

available all over and more so in the tropicaldomain. Wind energy is also eAective as themovement of air is common and exists everywhere.Globally, the renewable share of total Bnal energyconsumption (TFEC) by source, as per 2018estimations, was 79.9% of fossil fuels, 2.2% fromnuclear, 6.9% from traditional biomass, and11% from modern renewable sources. In India,comparatively, renewable sources account for8.97% (https://www.iea.org/fuels-and-technologies/renewables). Proportionately, out of the 11%contribution from renewable energy sources; 2.1%

J. Earth Syst. Sci. (2021) 130:64 � Indian Academy of Scienceshttps://doi.org/10.1007/s12040-021-01557-7 (0123456789().,-volV)(0123456789().,-volV)

comes from the wind; 3.6% from hydro; 4.3% fromsolar/biomass/geothermal heat together; and bio-fuels for transport account for 1% (REN21 2020).Apart, considering the electricity generation andconsumption, as of March 2020, renewable energysources contribute to 35.86% and 21.22% of India’sinstalled electricity generation capacity and totalelectricity consumption, respectively. Of the35.86% attributed to electricity generation fromrenewable sources, wind power account for 10%;large hydro concern 13%; 8% comes from biomassand the remaining from small hydro and waste,respectively (Central Electricity Authority 2020).In the scenario of depleting natural energyresources, renewable energy provides a potentialalternative and India stands Bfth to followDenmark, Germany, Spain, and the USA.Due to the growing importance and need, several

investigations on the identiBcation and resource-fulness of renewable energy sources were carriedout globally. In the current world scenario ofincreasing wind harnessing as an importantrenewable energy resource, analysis of low-levelwind speed and assessment of wind energy poten-tial over different world regions have beenattempted using different approaches. Estimationof wind power potential requires characterizationof the wind speed distribution at the turbine level,and measurements at the desired locations andheights may not always be possible due to logisticconstraints. The sources of wind data are throughconventional observations near the surface whichcorrespond to 10-m level and through the atmo-spheric model application. Both of these datasources have drawbacks, where observations havethe barriers of spatial and temporal resolution, andmodel-generated analysis data have boundedcomputational resources and the complexities ofimplied physics, dynamics, and numerical methods(Bhaskar Rao 2019). Due to the sparsity of obser-vations, statistical methods do not provide appro-priate estimations of wind at the requiredlocations, whereas atmospheric models providedynamically consistent wind Belds at differentvertical levels and at the desired spatial resolutionof the atmosphere which are more compatible.The most commonly used method hitherto has

been to analyze near-surface wind speed fromobservations, identify the best Bt, and then toassess the wind power density. All of these studieshave indicated Weibull distribution to be the mostsuitable as compared to Rayleigh and Gammadistributions to describe the wind speed variations.

Wind energy estimations were made using thisconcept for different regions of Japan (Dokur andKurban 2015); Iraq (Ahmed and Mahammed 2012;Ghitas et al. 2016); Nigeria (Himrib et al. 2008;Dikko and Yahaya 2012; Odo et al. 2012); Ethiopia(Wolde-Ghiorgis 1998); Spain (Carrillo et al. 2014);Egypt (Ghitas et al. 2016); Kenya (Maina et al.2016); Libya (El-Osta and Kalifa 2003); Greece(Fyrippis et al. 2010); Jordan (Al-Nhoud and Al-Smairan 2015); Syria (Al-Mohamad and Karmeh2003); Yemen (Mahyoub 2006); Tunisia (Elamouriand Amar 2008; Dahmouni et al. 2010); Turkey(Gokcek et al. 2007); Azores Islands (Rusu andSoares 2012); European coastal regions (Sempre-viva et al. 2008); California (Dvorak et al. 2010);Korean peninsula (Kim et al. 2011); South ChinaSea (Zhou et al. 2006; Zheng et al. 2013; Zheng andLi 2015); Ontario and Great Lakes (Ashtine et al.2016); and India (Kumar et al. 2017; Satya-narayana et al. 2019).In contrast, a few studies were attempted to

assess wind energy using numerical atmosphericmodels. The Department of Environment, Aus-tralia has produced wind maps at the 80-m level at8 km spatial resolution from 1995 to 2005 assimi-lating the observations into a wind mapping model(http://www.dpi.vic.gov.au/˙data/assets/pdf˙Ble/0006/38841/SKM-DPI-Renewable-Energy-Part2-v5˙Part1.pdf). Willow et al. (2014) have usedMERRA (Modern Era Retrospective-analysis forResearch and Analysis) data at 67 km (longitu-dinal) 9 50 km (latitudinal) resolution for theperiod from 1979 to 2009 to produce the wind Beldat the turbine heights of 50-, 80-, and 150-m overAustralia. In their study, the authors have com-puted wind power density, using median (in lieu ofthe mean) and coefBcient of variation (in lieu ofstandard deviation) and interpreted the results interms of abundance and variety. Xi et al. (2009)made an assessment of wind power on a globalscale using global wind analysis data from GEOS-5 DAS (Goddard Earth Observing System version5 Data Assimilation System), available at 67 9 50km2 horizontal resolution and three lowest levelsof 71, 201, and 332 m. These data were used toestimate the wind speed and wind power at 10-mover the USA. Wakeyama and Ehara (2011) usedmodel-generated wind data at 500-m resolution asavailable from NEDO (New Energy and IndustrialTechnology Development Organization) andassessed the wind power potential over NorthernJapan. Standen et al. (2017) used UKMO (UnitedKingdom Meteorological ODce), uniBed model, to

64 of 19 J. Earth Syst. Sci. (2021) 130:64

generate hourly wind speeds at seven heightsbelow 200 m for the 10 years of 2001–2010, andtheir results indicated the superiority of atmo-spheric model output over the statistical-basedMeasure-Correlate-Predict (MCP) method. Florinet al. (2016) used the assimilation of satellitemeasurements in ECMWF (European Centre forMedium Range Weather Forecasts) and NCEP(National Centres for Environmental Prediction)model-based global reanalysis data to producewind speed data for the Mediterranean Sea region.Their study indicated the usefulness of this winddata over coastal and island locations. Sabiqueet al. (2016) evaluated wind and wave powerpotential over the Red Sea region using AdvancedResearch Weather Research and Forecasting(ARW) model-derived data. As compared to theatmospheric model application, only one studyusing a statistical model was attempted. Jeon andJames (2012) applied a stochastic time seriesmodel to generate 72-hr predictions for one yearover Greece and suggested that the statisticalmodel could be used as an alternate to an atmo-spheric model to derive the wind speed for windpower estimation.Continuous eAorts are being made towards the

assessment of wind energy over the Indian sub-continent, since the 1980s. Mani and Mooley (1983)reported the Brst wind source assessment, in whichregular surface wind data from monitoring towershave been used and were the Brst to identify theGujarat region as the most resourceful for windenergy. Many studies on wind speed and energyover India are followed (Rao 1986; Amin 1999;Bakshi 2002; Prakash and Srinivasa 2017; Maityet al. 2019; Satyanarayana et al. 2019). Althoughthe objective of these studies was the same toprovide wind resource data, they differed only indata type and methodology. The wind powerpotential over India had been assessed by theCentre for Wind Energy Technology (CWET2010), presently National Institute of Wind Energy(NIWE), and compiled ‘Indian Wind Atlas’ pro-viding winds at 50-m from observations and prob-able estimates at 80-m height. CWET, incollaboration with the Riso DTU National Labo-ratory for Sustainable Energy, Denmark computedthe wind Cow at 500 m spatial resolution using ablend of the mesoscale model derived winds and 97micro-scale measurements. So far this is theauthenticated estimation of wind power over India,released from an oDcial source, and is being usedby the industry as well as the government agencies

related to renewable energy. Although this is thebest wind power potential estimation over India,the estimation had the limitation of using data ofonly one year.The standard height for measuring wind speed

as well as model derived product is 10-m. However,for wind resource assessment, it may become nec-essary to estimate the wind speed at higher alti-tudes to choose the heights of wind turbines forpower production (WMO 1981). For this purpose,‘Power and logarithmic laws’ are used to obtainthe change in wind speed with height (WMO1981). The Power law is simpler, empirical, andadequate for wind energy calculations as comparedto the logarithmic law (Farrugia 2002). In recenttimes, wind resource characterization is made atthe turbine heights of 50/80/120 m and windenergy studies use the reconstructed winds fromthe 10-m level. Chauhan et al. (2010) providedestimated wind power potential over differentstates of India based on the ‘Indian Wind Atlas’from CWET. Phadke et al. (2011) have used windspeed data produced using WRF model at 3.6 kmresolution and three heights of 80, 100, and 120 mfor the Indian region and reported that the windenergy potential estimates are higher and slightlybetter than those provided in the ‘Indian WindAtlas’ produced by the CWET (2010). Draxl et al.(2014) reported the use of the Weather ResearchForecasting (WRF) model to generate wind at 1.1km spatial resolution over Gujarat (a state inIndia) for one complete year and demonstrated theadvantages of the model-derived wind data forwind power assessment. Reddy et al. (2015) stud-ied the wind speed observations at 4.5 m level andestimates at 50-m height for the 6 years of2007–2012 and reported that Weibull distributionprovides better estimates than the Rayleigh dis-tribution. Yip et al. (2016) characterized the windresources over the Arabian peninsula using thewind estimates based on 10-m level data from theMERRA dataset and classiBed the wind resource-ful areas. This study concluded that a precise andcomprehensive wind power potential estimationneeds analysis of high-resolution wind data thatcould represent the spatial features and local cir-culations. Satyanarayana et al. (2019) estimatedwind speeds at different levels below 120 m usingthe wind speed observations at the 10-m level forthe period from 1979 to 2015 over the AndhraPradesh and Telangana regions.It could be inferred from the aforesaid reviews,

not withstanding that the use of model-generated

J. Earth Syst. Sci. (2021) 130:64 of 19 64

data is more scientifically and technically better,their generation had been constrained of compu-tational resources to run atmospheric models atsufBciently high horizontal resolution. Due torapid advancements in computer technology dur-ing the last 2–3 decades, it is now possible to runatmospheric models at high resolutions of few tensof kilometres, and atmospheric models are nowbeing used for data assimilation that can includeall synoptic and asynchronous observations fromdifferent sources to generate high-resolutionanalysis Belds over a domain of interest. Theatmospheric data, thus generated would conformto the atmospheric dynamics and physics, at thedesired spatio-temporal resolutions (Bhaskar Rao2019). Hence, at present, this methodology isconsidered superior to the application of statisti-cal methods and a few studies have used thedynamically derived global analysis data for windenergy studies (Wakeyama and Ehara 2011 forNorthern Japan, Willow et al. 2014 for Australia,Florin et al. 2016 for Mediterranean region,Sabique et al. 2016 over the Red Sea, Xi et al.2009 over the USA).The established relationship between wind speed

and wind power production shows that the cut-inspeed (the threshold wind speed needed to begingenerating electricity) is 3.5 m/s, congenial fornormal wind power production from 4.5 m/s, whichincreases rapidly till the wind speed attains opti-mum at 14 m/s and the turbines would stop at thewind speeds [25 m/s (Wolfson 2012; NYSERDA2020). It was also reported that the wind speedthresholds of 4.5–5.4, 5.4–6.7, and [6.7 m/s cor-respond with marginal, good, and exceptional windpower productivity (Ramachandra and Shruthi2003). While wind power generation is consideredas the best renewable energy source, the high costinvolved in the installation of wind turbines forlarge scale power generation is one of the factors ofthe hindrance. Research advancements have led tothe development of low wind turbines that couldproduce micro-generation of power ranging from1–100 kilowatts with the cut-in speed of 3.5 m/sthat are typically usable at home or a farm or beembedded as a part of grid generation (Tummalaet al. 2016; Oluseyi et al. 2019). Low wind turbinesare also being favoured as large turbines aresupposed to contribute to the increase in tempera-tures and precipitation ultimately aAecting theirsustainability (Wang and Prinn 2010).The current status of wind energy assessment

over India shows the limitations of data sources

and applications. The present study is motivatedto provide an assessment of wind speed distribu-tion at different ranges for the entire land region ofthe Indian subcontinent, identify potential regionsof wind energy using high-resolution wind data,which would be useful for wind energy applica-tions. This study is an attempt to provide windpower potential at different heights below the100-m level over the Indian subcontinent. Thenovelty of the present study is the use of hightemporal (1-hourly) and spatial (12.5-km) resolu-tion data for a continuous 40-year period to esti-mate the wind energy potential at these resolutionsover the entire Indian subcontinent.

2. Data and methodology

2.1 Data

1. Wind speed data at 10-m level, at the 30-kmhorizontal resolution, at 1-hr interval (from 0000 to2300 UTC) for the period 1979–2018, availablefrom the ERA5 [ECMWF (European Centre forMedium-Range Weather Forecasts) Re-Analysis5th generation] (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5). ERA5is the Bfth generation of ECMWF atmosphericreanalyses of the global climate, and the Brstreanalysis produced as an operational service.2. Wind speed data at 10-m level, at *12.5-km

resolution, at the 3-hourly interval (0000, 0300,0600, 0900, 1200, 1500, 1800, and 2100 UTC) for theperiod 1979–2018 from the ERA (ECMWF Re-Analysis) Interim datasets (http://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/).ERA-Interim represents a third-generation reanal-ysis, available for the period 1 January 1979–31August 2019.Both the ERA5 and ERA-Interim data are pro-

duced using an atmospheric model and data assimi-lation systemreferred to as an IntegratedForecastingSystem(IFS). IFSuses a comprehensive earth systemmodel developed at ECMWF and 4-dimensionalvariational dataassimilation system(Dee et al.2011).The reanalysis products, currently available from afew sources such as NCEP-USA, ECMWF-UK, JRA(Japanese Meteorological Agency (JMA) reanalysis)Japan, are the most reliable and credible as they aregenerated using the state-of-the-art atmosphericmodels andmethodologies to assimilate observationsfrom different sources such as radar, satellite, buoy,


ship, etc., along with conventional global observa-tions. The global reanalysis products would not haveany bias arising due to physical parameterizationschemes as the model is integrated for very shortperiods of less than 24-hrs (Dee et al. 2011). TheseERA5andERA-Interimatmospheric data (includingwind) were assessed by several research groups allover the world and approved their quality and use-fulness to derive credible scientiBc deductions(Horanyi 2017; Olauson 2018; Urraca et al. 2018;Belmonte Rivas and Stoffelen 2019; Graham et al.2019; Hersbach 2019; HoAmann et al. 2019; Satya-narayana et al. 2019; Hersbach et al. 2020; Lei et al.2020). These two datasets are exclusive as they coverlong temporal periods starting from 1979 at 12.5 and30-km horizontal resolutions, respectively, coveringthe entire globe.3. Wind speed observations, at 10-m level, at

1-hr interval, of the recent Bve year period2011–2018, at 10 Automatic Weather Stations(AWS), located within the high wind speed zonesin the west and south-central parts of India(table 1a) as available from the oDcial Indiansource, i.e., the MOSDAC website (http://mosdac.gov.in/data/jsp/new˙login/main˙login.jsp).4. Daily mean wind speed observations, at the

20-m level at eight wind turbine locations over the

west and southern parts of India (table 1b), asprovided by the National Institute of WindEnergy, India.

2.2 Methodology

1. ERA-Interim wind data at the 10-m level at12.5-km resolution were used to compute dailymean wind speeds, and spatial distributions foreach of the 12 months separately and seasonallyof monthly and seasonal mean wind speeds overthe Indian subcontinent considering the 40-yrdata period 1979–2018.

2. ERA5 data at the 10-m level at 1-hr intervalwere used to compute hourly averages of windspeeds and the average number of hours thatsurpass the wind speed thresholds of 4.5, 5.4,and 6.7 m/s per day for each of the 12 monthsseparately and seasonally considering the 40-yrdata period 1979–2018.The data corresponding to all the days on whichthe Indian subcontinent had the occurrence of aweather disturbance such as a depression or atropical cyclone were discarded and data of allthe other days were considered for analysis.This Bltering was done to avoid the inCuence ofhigh wind speed associated with depressions or

Table 1. Statistical metrics of the evaluation of ERA-Interim wind speed data at different (a) MOSDAC-AWS locations and(b) wind turbine locations.

Sl.

no. Station

Latitude

(�N)

Longitude

(�E)

Length of the data

MAE RMSE CC BIAS IOAStart date End date

(a)

1 Gujarat 23.02 72.51 1 Jan 2011 31 Dec 2018 0.48 0.53 0.88 0.21 0.91

2 Desalpur, Kutch 23.23 69.44 0.52 0.57 0.87 0.34 0.87

3 Maharashtra 17.93 73.67 0.51 0.58 0.78 –0.35 0.84

4 Tamil Nadu 8.70 77.46 0.53 0.56 0.94 0.31 0.93

5 Rajasthan 26.76 77.01 0.53 0.62 0.87 0.34 0.89

6 Bangalore 13.03 77.56 0.57 0.61 0.84 0.31 0.82

7 Ahmedabad 23.03 72.54 0.48 0.54 0.86 0.37 0.86

8 Karanataka 14.86 74.43 0.56 0.56 0.75 0.36 0.78

9 Machilipatnam 16.21 81.17 0.58 0.57 0.93 –0.28 0.94

10 Anantapur 14.46 77.88 0.54 0.6 0.91 0.34 0.91

(b)

1 Muppandal 08.15 77.33 May-1997 Oct-2005 0.62 0.61 0.82 0.45 0.86

2 Ramagiri 14.17 77.30 May-2001 Jun-2005 0.53 0.62 0.87 0.34 0.89

3 Kadavakallu 14.47 77.57 May-2001 Dec-2010 0.63 0.59 0.83 0.58 0.84

4 Chikkodi 16.25 74.34 Aug-1993 Mar-2011 0.66 0.53 0.79 0.41 0.81

5 Bhud 17.20 74.12 Oct-2003 Dec-2006 0.57 0.53 0.88 –0.31 0.81

6 Kamalapur 17.35 76.58 Jun-2004 Mar-2007 0.58 0.55 0.86 0.45 0.83

7 Sivalakha 23.38 70.58 Nov-2000 Mar-2004 0.59 0.54 0.87 0.34 0.88

8 Jaisalmer 26.56 70.53 Jan-1999 Dec-2005 0.57 0.61 0.84 0.31 0.82

MAE: Mean absolute error; RMSE: root mean square error; CC: correlation coefBcient; IOA: index of agreement.


cyclones that would enhance the average value.

3. Statistical metrics of BIAS, MAE (mean abso-lute error), RMSE (root mean square error),IOA (index of agreement) and correlation coef-Bcient (COR) have been computed to evaluatethe ERA-Interim data as compared to dailyobservations. The equations/formulae (Wilks2006) are as follows:

COR ¼Pn

i¼1ðf i � �f Þ oi � �oð ÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPn

i¼1ðf i � �f Þ2q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPn

i¼1ðoi � �oÞ2q : ð1Þ

MAE ¼ 1

n

Xn

i¼1

fi � oij j; ð2Þ

RMSE ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPn

i¼1 fi � oið Þ2

n

s

; ð3Þ

IOA ¼ 1:0�Pn

i¼1 fi � oið Þ2Pn

i¼1 fi � �oj j þ oi � �oj jð Þ2; ð4Þ

where f and o values correspond to ERA-Interimdata and AWS observations, respectively.

4. Wind speeds at four speciBed heights of 20-, 40-,60-, and 80-m were estimated using the Powerlaw (WMO 1981; Petersen et al. 1998)

WS2 ¼ WS1 �H2

H1

� �a

; ð5Þ

where WS1 is the velocity at height Z1, WS2 isthe velocity at height Z2, H1 is the height 1(lower height), H2 is the height 2 (upper height)and a is the wind shear exponent (0.14).In the present computations, the wind speed at10 m level is the base (lower height H1).

5. Wind power density at the four chosen levelswas computed using the formula (Ramachandraand Shruthi 2003)

Pd ¼ 0:5KeDw3e ; ð6Þ

where Ke is the energy pattern factor Ke = [wi3/

Ne]/we3,D is the density of air (1.29 kg/m3),we are

the wind speed in m/s, Pd = power per unit area(watts/m2),wi is wind speed hourly,Ne is numberof wind speed hourly values, Ke energy patternfactor (ERA5 data at the 10-m level at 1-hrinterval were used to calculate the average energypattern factor for each of the 12months separatelyconsidering the 40-year data period 1979–2018).

6. The mean speed and wind power density valueswere calculated using the above formula (6) for

each of the eight boxes (locations and domainsare given in Bgure 4 and table 4), consideringthe box-averaged values. These eight boxeswere chosen to represent the highest wind speedregions and not based on the station height.Box averaging is made to produce area-repre-sentative magnitudes which would be moreconsistent than of a single point.

3. Results and discussions

The study region of the present research is conBnedto the Indian subcontinent. For brevity of under-standing the spatial characteristics, presentation,and discussion of the results are made with refer-ence to 36 subdivisions as per the classiBcation bythe India Meteorological Department (Bgure 1).

3.1 ERA wind data validation

In the present study, ERA-Interim wind data atthe 10-m level for the period 1979–2018 is used tomake an assessment of wind speed and conse-quentially the wind power potential over theIndian subcontinent. Since ERA data is a globalanalysis product, described in section 2, the 10-mlevel wind data is validated by comparison withcorresponding data collected by automatic weatherstations (AWS). For the validation, 1-hourly 10-mwind data from 10 AWS, located within the highwind speed zones over the west and south-centralparts of India, for the period 2011–2018 are used.ERA data at the nearest grid point to the AWSlocation is considered, which implies that themaximum distance between station location andthe grid point is B 8.3 km. The statistical metricsof MAE, RMSE, CC, BIAS, and IOA are computedfor each of the 10 AWS stations and are presentedin table 1(a). The results indicate a very goodagreement and consistency as evidenced by highIOA (*0.78–0.9); high CC (*0.75–0.94); lowestRMSE (0.53– 0.62) and MAE (*0.48–0.58) andlow BIAS (–0.35 to 0.37). To supplement the datavalidation, ERA wind speeds are compared withdata from wind turbine installations over thenorthwest, west, and southwest parts of the Indiansubcontinent. As the wind data at the wind turbineinstallations are made at 20-m level, ERA data atthe 10-m level is used to estimate the wind speedsat 20-m level using equation (5) given in section 2(Data and methodology). The data at the turbine


locations are compared with the nearest ERA gridpoints, which means that the distance between thetwo points is limited to a maximum of 8.3 km. Thestatistical metrics of MAE, RMSE, CC, BIAS, andIOA are computed for each of the turbine locationsseparately and are given in table 1(b). These statis-tical metrics are similar to those of validation withAWS data. The CC is high ranging from 0.79 to 0.88;MAE is lower in the range of 0.53–0.66; RMSE islower in the range of 0.53–0.62; BIAS varying from–0.31 to 0.45, and the IOA is high in the range of0.81–0.89. This validation with data from wind tur-bine locations complements the validation with AWSdata and conBrms the quality of the ERA wind data.These values denote very good correspondence andmuch better consistency of ERA-Interim globalanalysis data than of Draxl et al. (2014) whoreported MAE, RMSE, and BIAS of 1.47–1.57,1.93–2.01, and –0.03 to 0.73 for their model datavalidation over Gujarat. A separate evaluation ofERA5 data is not performed as both the ERA5and ERA-Interim are generated using the samemodel and data assimilation procedure and thedifference being limited to the spatial (30 and 12.5km) and temporal (1- and 3-hourly) resolutions.These results authenticate the quality of the ERAdata and their application to the assessment of thewind power potential of the present study.

3.2 10-m wind analysis

The spatial distribution of mean annual wind speedat 10-m level over the Indian subcontinent wasanalyzed (not shown), which depicts regions on thewest and south-central parts of India to be having

wind speeds higher than 3 m/s, whereas the northand eastern parts of India and the windward side ofthe Western Ghats to have lesser wind speed.Specifically, West Rajasthan, West Gujarat, andSaurashtra regions and the leeward side of theWestern Ghats that include the parts of Maha-rashtra, Karnataka, Andhra Pradesh, and TamilNadu were identiBed as regions with higher windspeed ([3 m/s). Since the variability of the windspeed is important, the spatial distribution of thestandard deviation of the wind speed was com-puted and investigated (not shown). It is observedthat the regions of higher (lower) wind speed areassociated with higher (lower) values of standarddeviation. This implies that higher wind speedregions are to be further analysed for the timedurations of the wind speed.Additionally, the spatial distributions of mean

monthly wind speed distributions were analysed toidentify the months and seasons that would befavourable for wind power generation (Bgure 2). Inthese maps, contours are drawn at 3.5, 4.5, 5.4, and6.7 m/s which are the thresholds considered forwind power generation as described before in sec-tion 1. The monthly distributions showed that thehighest wind speeds occur during June, July, andAugust and marginally higher during May andSeptember as compared to other months, whichwere identiBed over the west and south-centralparts of India. These higher magnitudes favouringwind power generation were observed to occurstarting in the pre-monsoon month of May andextend through the southwest monsoon season(June–September). Indian subcontinent experi-ences higher temperatures during these 5 months,

Figure 1. Meteorological subdivisions over India.


and the increase in wind speed over the northwestand along the west coast is due to the strongwesterly Cow with speeds of 4–6 m/s from theMiddle East and southwesterly wind Cow withspeeds of 6–8 m/s from the Oman coast (DodlaVenkata et al. 2017; Satyanarayana and BhaskarRao 2020). Apart from the general atmosphericCow, the noted higher wind speeds in the planetaryboundary layer during the summer (April andMay) and the monsoon rainy season (June, July,August, and September) are due to higher con-vective instability and unstable conditions ascompared to calm winds and higher atmosphericstability during the winter season (Stull 2015).It is observed that Gujarat, specifically the

regions of Kutch and Saurashtra, had the advan-tage of winds higher than 4.5 m/s in all the 5months, i.e., May through September, which des-ignate this region to be most congenial for windpower generation. Further, the average number ofhours/day with different thresholds of 3.5, 4.5, 5.4,and 6.7 m/s were computed using ERA-Interimdata at 1-hourly temporal resolution consideringthe 40 years. As noted earlier, consideration ofwind speeds between 3.5 and 4.5 m/s are useful forlow turbine wind power generation and thresholdsof[4.5,[5.4, and[6.7 m/s denote marginal, good,

and exceptional wind power production. The spa-tial distributions of the number of hours/day forthe four different thresholds are presented inBgure 3. The pattern of the number of hours issimilar to that of wind speed distribution indicat-ing that the regions with higher wind speed alsohave a longer duration. West and south-centralparts have the highest duration during June, July,and August whereas Gujarat had a longer durationextending to May and September. Considering thethresholds, west and south-central parts have goodwind power productivity with wind speeds higherthan 5.4 m/s and with the duration of *10–16hrs/day, whereas Gujarat and South Karnatakahad an exceptional rate of productivity, where thewind speeds exceeding 6.7 m/s had a duration of*10–12 hrs/day. This analysis clearly brings outthat the wind power potential is highest over thewest and southwest parts of India during thesummer and monsoon seasons. The combination ofthe distributions of wind speed and duration wouldbe helpful to identify the localised regions of thehighest wind power productivity in India.To supplement the monthly analysis, seasonal

distributions of the wind speed and the number ofhours that exceed 4.5 m/s in a day are analysedand presented in Bgure 4. The spatial distributions

Figure 2. Spatial distributions of mean monthly wind speed (m/s) from January through December. Contours selected tocorrespond with wind speed thresholds for wind power production.


of the average wind speed (Bgure 4a) showedhigher values over the western parts of India (ex-cluding extreme north and south). The regions ofRajasthan, Gujarat, Kutch and Saurashtra, EastMaharashtra, Interior Karnataka, South coastalAndhra Pradesh, and Rayalaseema are noted to bemore favourable with wind speeds above 4.5 m/sthan the rest of the Indian subcontinent. Thenumber of hours per day for the wind speedsexceeding 4.5 m/s (Bgure 4b) clearly displayedthat the regions of maxima coincided with theregions of higher wind speeds. Thus, the identiBedregions of Gujarat, Kutch, and Saurashtra, EastMaharashtra, Interior Karnataka, South coastalAndhra Pradesh, and Rayalaseema could be con-strued as the regions of wind power potential. Tofacilitate the application of wind speed distributionin the design and implementation of wind powergeneration over the Indian subcontinent, theaverage number of hours per day, during May–June,for the wind speed thresholds of[3.5,[4.5,[5.4,and[6.7 m/s for different subdivisions are providedin table 2. This data supplements the informationprovided in Bgure 4 and would be useful to easilyassess subdivision-scale wind power potential.

To further understand the wind power potentialover these regions, eight box regions, with each boxcovering 100 9 100 km2 area, representative oflocations within the identiBed higher wind speedregion is selected over which additional analysis ofwind speed has been carried out. The locations ofthe boxes are given in table 3. The time series ofdaily mean wind speed (m/s), computed for theperiod 1979–2018, for each of the eight boxes arepresented in Bgure 5(a). A dashed line corre-sponding to 4.5 m/s is drawn in all the Bgures tofacilitate interpretation concerning the wind powerthreshold. The time series clearly brings out thevariations in the wind power potential in differentareas within the higher wind speed regions. Theregions of southwest Rajasthan, Kutch, andSaurashtra (boxes 3, 2, and 1) had the highestpotential with higher wind speed and longer timedurations of 198, 169, and 149 days. The regions ofInterior Karnataka, Rayalaseema, and Southcoastal Andhra Pradesh (boxes 5, 7, and 6) alsoshow good power potential with durations of 122,102, and 74 days, respectively. Although TamilNadu and Madhya Pradesh regions fall within theregion of wind speeds [3 m/s, they have meagre

Figure 3. Spatial distributions of mean wind speed hours/day for the selected wind speed thresholds of (a–e) [6.7 m/s;(f–j)[5.4 m/s; (k–o)[4.5 m/s; and (p–t)[3.5 m/s for May to September.


Figure 4. Spatial distributions of (a) mean wind speed (m/s); (b) the number of hours/day exceeding[4.5 m/s averaged for the5 months from May to September. Selected box regions located in the high wind speed zones are marked in (a).

Table 2. Mean wind speed hours per day for different wind thresholds for differentmeteorological subdivisions.

Sl.

no. Subdivision

[3.5

m/s

[4.5

m/s

[5.4

m/s

[6.7

m/s

1 Saurashtra & Kutch 22 22 20 11

2 West Rajasthan 20 18 14 8

3 Gujarat 20 16 14 7

4 North interior Karnataka 20 18 14 7

5 South interior Karnataka 20 16 12 5

6 Tamil Nadu 20 16 10 4

7 Rayalaseema 20 16 10 4

8 Madhya Maharashtra 18 16 14 6

9 West Madhya Pradesh 16 12 8 1

10 Telangana 16 12 10 3

11 Coastal Andhra Pradesh 14 10 6 2

12 Gangetic West Bengal 14 6 0 0

13 Jammu & Kashmir 14 6 4 2

14 Marathwada 14 8 6 0

15 Kerala 14 0 0 0

16 Orissa 12 4 0 0

17 Bihar 12 6 3 0

18 East Madhya Pradesh 12 4 1 0

19 Konkan & Goa 12 4 2 0

20 Vidarbha 12 6 2 0

21 Chattisgarh 12 4 0 0

22 Haryana 10 4 1 0

23 Nagaland, Manipur, Mizoram and

Tripura

8 2 0 0

24 East Uttar Pradesh 8 4 2 0

25 West Uttar Pradesh 8 4 1 0

26 Jharkhand 6 2 0 0

27 Uttaranchal 6 2 1 0

28 Punjab 6 2 0 0

29 Assam and Meghalaya 2 2 0 0


power potential. The diurnal variations of themean hourly wind speed computed using hourlydata for the period from 1 May to 30 September forthe 40 years, and averaged for each of the boxregions are presented in Bgure 5(b). This compu-tation is performed to identify the periods in a daythat have wind speeds exceeding 4.5 m/s. The timeseries plots show that the wind speed startsincreasing from 0000 UTC (correspond to 0530

IST, Indian Standard Time), crosses the thresholdof 4.5 m/s between 0100 and 0200 UTC (0630–0730IST) soon after sunrise, increases till 0800 UTC(1330 IST), starts decreasing and dips below 4.5m/s around 1200 UTC (1730 IST). This varia-tion indicates the occurrence of winds exceeding4.5 m/s for*10 hrs continuously during daytime asthe increase and decrease of wind speed are found tobe synchronous with heating from solar radiation.

3.2.1 Low wind farming

Considering the cut-in speed for wind powergeneration to be 3.5 m/s and the advent of lowwind farming in recent times (Wang and Prinn2010; Tummala et al. 2016; Oluseyi et al. 2019),assessment of the spatial extent of the averagewind speed between 3.5 and 4.5 m/s and thenumber of hours/day exceeding 3.5 m/s over theIndian subcontinent was made (Bgure 4). It isinferred from this analysis that most of the

Figure 5. (a) Time series of average daily wind speed (m/s) for the selected box regions and (b) diurnal variation of averagehourly wind speed (m/s).

Table 3. Domains of the 8 selected boxes.

Box

Latitude

(�N)

Longitude

(�E)

1 26–27 70–71

2 23.5–24.5 69–70

3 22–23 69.5–70.5

4 23–24 75–76

5 17–18 75–76

6 15–16 79–80

7 14–15 77–78

8 9–10 78–79


western parts of India, covering the regions ofRajasthan, Gujarat, Kutch and Saurashtra, WestMadhya Pradesh, Madhya Maharashtra, Coastaland interior Karnataka, Telangana, CoastalAndhra Pradesh, and Rayalaseema (Bgure 4a) arefavourable for low wind farming. It is also notedthat these wind speeds occur for at least 10hours/day (Bgure 4b) indicating these regions tobe highly favourable for low wind farming. Thisresult adds to the general analysis with thresholdwind speed[4.5 m/s.

3.3 OAshore wind farming

It would be of interest to consider the generation ofwind power in the open ocean, which is referred toas ‘oAshore wind power’ in contrast to ‘onshore windpower’ overland. The ‘oAshore wind farming’ issupposed to have the advantages of higher windspeeds (due to negligible surface friction), largerfarming area, and non-detrimental climate impacts(Zheng et al. 2016). Since the ERA analysis isavailable at 12.5 km resolution globally, the windspeed data at 10-m level, averaged for the 5 monthsfrom May to September of the 40 years from1979–2018, are used to derive spatial distributionsover the ocean regions of the Arabian Sea and Bayof Bengal along with the Indian subcontinent, andpresented in Bgure 6. The results have shown thatthe wind speeds are between 5.4 and 6.7 m/s nearthe coastline and are higher than 6.7 m/s over theocean part, with a duration of 12–16 hours/day(Bgure 6). This is in sharp contrast to the occur-rence of wind speeds ranging between 4.5 and 5.4m/s with durations of 10–12 hrs/day (Bgure 4b)that are limited to the northwest and south-centralparts of the Indian subcontinent. This analysisbrings forth the higher wind power potential, interms of higher wind speed and longer duration,over the open ocean and substantiate the advan-tages of oAshore wind farming. Although oAshorewind farming has certain advantages of higherpower generation, the involved higher cost ofinstallation and maintenance tend to favour onshorewind farming at the current time (Blanco 2009).

3.4 Wind power potential

The data source provides wind speed at the 10-mlevel as representative of surface wind. It is possibleto estimate the wind speed at different altitudes,within the surface layer (below *100 m) of thePBL (Planetary boundary layer) for which the

mathematical expression, based on theory andobservation, is deBned in equation (5) (WMO 1981;Petersen et al. 1998). The wind speeds at 20, 40, 60,80, and 100-m levels have been computed usingequation (5), taking the 10-m level wind speed asthe base. The spatial distributions of the meanwind speed averaged for May, June, July, August,and September over the Indian subcontinent at 20,40, 60, and 80 m levels are presented in Bgure 7.Contours were drawn considering the thresholds of3.5, 4.5, 5.4, and 6.7 m/s, which provide the windspeed variation with height within the lowest100-m layer. It was observed that the wind speedincreases with height as 19% from 10- to 20-m and20- to 40-m level; 9% from 40- to 60-m level; 8%from 60- to 80-m level and 5% from 80- to 100-mlevel. It was also observed that the area extentincreased with height. The spatial distribution atthe 20-m level shown that West Rajasthan, Kutchand Saurashtra, South Coastal Andhra Pradesh,Rayalaseema, South Telangana, and Interior Kar-nataka are the regions with wind speed exceeding4.5 m/s. The 40-m level distribution shows abroader region extending from Rajasthan throughthe southeast coast, which denotes 50% areaenhancement as compared to 20-m level. The 60-mlevel distribution shows a slight increase in areaextent (*15%) than at 40-m level. At the 80-mlevel, more than 50% of the Indian subcontinentcovering the western parts of India denotes windspeed exceeding 4.5 m/s. The computed windspeeds over Gujarat agree with those of Draxl et al.(2014). These estimations of wind speed at differ-ent levels below 100-m level bring out an importantdeduction that the mast height of windmill shouldbe around 80-m level that would have enhancedand most productive wind power density in termsof not only higher wind speed but also wider spatialextent.The daily wind power density was calculated

using equation (6) for each of the eight boxeschosen to represent different locations within thehigher mean annual wind speed region (above 4.5m/s). The time series (Bgure not shown) showedhigh wind power density in the regions of Kutch,Saurashtra, and Rajasthan and moderate windpower density over Rayalaseema, South AndhraPradesh, and Interior Karnataka. The regions ofTamil Nadu and Madhya Pradesh were not asproductive as the wind power density values were\80 W/m2. The wind power potential at theselected heights of 20-, 40-, 60-, and 80-m wascomputed at each of the grid points over the


Indian subcontinent, and the spatial distributionsof their values averaged for the 5 months fromMay to September at the four heights are pre-sented in Bgure 8. It is to be noted here that thewind power potential distribution would be dif-ferent from the distribution of wind speed as the

computation of wind power potential involves thecube power of the wind speed. These resultsbrought out that the wind power potential regionsare conBned to Gujarat, Kutch and Saurashtra,Interior Karnataka, and Rayalaseema only,although the areas with wind speed[4.5 m/s are

Figure 6. Spatial distributions of (a) mean wind speed (m/s) and (b) number of hours/day exceeding wind speed of 6.7 m/saveraged for May to September.

Figure 7. Spatial distributions of mean wind speed (m/sec) at different heights of (a) 20 m, (b) 40 m, (c) 60 m, and (d) 80 maveraged for May to September.


noted to have much larger area extent. Further,the highest power potential is conBned to Gujarat,Kutch, and Interior Karnataka only, and thespatial extent, as well as the magnitude of thewind power density, marginally increased withheight. The increase in wind power density was*68% from 10- to 20- and 20- to 40-m; *36%from 40- to 60-m; *24% from 60- to 80-m; andwas *18% from 80- to 100-m level. The variationof the wind power density with height (averagedover the eight boxes) showed that the thresholdvalue of 150 W/m2 coincides with the *50-m level,which conBrms the advantages of having the windmillheight to be above 50-m (table 4). These results areconclusive to indicate that windmill heights above50-m are more productive. The results of this studyqualitatively agree with those in ‘Indian Wind Atlas2010’ to the extent of magnitudes and the regions of

high power density ([200 W/m2) are broader in the

area extent and better delineated due to the use of

high-resolution data.

Figure 8. Spatial distributions of mean wind power density (W/m2) at different heights of (a) 20 m, (b) 40 m, (c) 60 m, and(d) 80 m averaged for May to September.

Table 4. Wind speed (m/s) and wind power density (W/m2) atdifferent heights for each of the eight selected box regions.

Box

Wind speed (m/s)

10 m 20 m 40 m 60 m 80 m 100 m

1 5.8 6.8 8.1 9 9.7 10.2

2 6.1 7.2 8.6 9.5 10.2 10.8

3 6.6 7.8 9.3 10.3 11 11.7

4 5.6 6.6 7.9 8.7 9.4 9.9

5 4.3 5.2 6.1 6.8 7.3 7.7

6 5 5.9 7 7.8 8.4 8.8

7 4.1 4.9 5.8 6.4 6.9 7.3

8 4.2 5 6 6.6 7.1 7.6

Wind power density (W/m2)

1 117.2 197.1 331.5 449.4 557.6 659.1

2 136.6 229.7 386.3 523.6 649.7 768.1

3 173.3 291.4 490.1 664.3 824.3 974.5

4 107.1 180.1 302.9 410.5 509.4 602.2

5 50.2 84.4 142 192.4 238.8 282.3

6 75.3 126.6 212.9 288.6 358.1 423.3

7 42.8 72 121.1 164.1 203.6 240.7

8 47 79.1 133 180.3 223.8 264.5


4. Summary and conclusions

The present study is an attempt to provide adistinct and comprehensive assessment of windspeed and wind power potential over the Indiansubcontinent. The novelty of the study is the use ofhigh resolution (*12.5 km) global analysis based10-m wind data spanning a continuous 40-yearperiod from 1979–2018. The use of this data forderiving wind power potential is considered to beadvantageous than the conventional observations,which have the limitations of spatial sparsity thatare hitherto used in earlier studies in India. TheERA global analysis of wind data has been vali-dated by comparison with observations fromselected automatic weather stations before adop-tion for analysis. Since the wind data is to beapplied for the estimation of wind power potential,wind data for the days inCuenced by synopticweather disturbances such as depressions andtropical cyclones have been discarded to avoidenhancement in the computation of averagemonthly and seasonal wind speeds. Spatial distri-butions of the mean monthly wind speeds at the10-m level are analysed to identify the regions andperiods that are conducive for wind power farming.Analyses were performed with different thresholdsof 3.5, 4.5, 5.4, and 6.7 m/s to identify the regionsof low wind, and normal wind farming for marginal,good, and exceptional categories of wind powergeneration. Estimations of wind speed were madeat different heights below the 100-m level todetermine optimized heights for wind farming.Projections on oAshore wind farming have beenattempted for exploring its advantages and via-bility. The conclusions derived from this studywould be appropriate towards the formulation ofpolicies and planning of wind energy productionover the Indian subcontinent.The results are summarized as follows:

1. ERA (ECMWF ReAnalysis) wind data isestablished to be useful for the estimation ofwind power potential. The high resolution(12.5 km) hourly data had shown to yield highcorrelation (0.75–0.94), high IOA (0.78–0.9),low BIAS (0.31–0.45), low MAE (0.48–0.66),and low RMSE (0.53–0.62) when validatedwith corresponding observations from auto-matic weather stations and wind turbinelocations.

2. Spatial distributions of average wind speeddepict regions on the west and south-central

parts of India to have magnitudes higher than3.5 m/s, whereas the north and eastern partsof India and the windward side of the WesternGhats have lesser wind speed.

3. Monthly variations of wind speeds, in terms ofmagnitude as well as the number of hours/day,denote the summer months of May, June, July,August, and September (the later 4 monthsalso represent summer monsoon over India) tohave the required threshold winds ([4.5 m/s)conBrming these periods to be favourable forwind farming.

4. The regions of Gujarat, Kutch and Saurashtra,East Maharashtra, Interior Karnataka, Southcoastal Andhra Pradesh, and Rayalaseema arenoted to have the potential for wind powergeneration, considering the threshold of [4.5m/s. Gujarat, Kutch, and Saurashtra areidentiBed to have the longest duration of 5months (May through September), whereasthe other regions would have the thresholdwinds for 3 months only (June, July, andAugust).

5. The regions of Gujarat and South Karnatakahad the highest power potential where thewind speed exceeding 6.7 m/s occur for*10–12 hrs/day. In contrast, the other regionsover the west and south-central India havegood wind power productivity with windspeeds higher than 5.4 m/s occurring for*10–16 hrs/day.

6. In terms of the average number of dayswith wind speeds exceeding 4.5 m/s peryear, the regions of southwest Rajasthan,Kutch and Saurashtra had longer timeduration of 198, 169, and 149 days, fol-lowed by Interior Karnataka, Rayalaseemaand South coastal AP which had durationsof 122, 102 and 74 days respectively,whereas Tamil Nadu and Madhya Pradeshregions had wind speeds [3 m/s thatindicate meagre power potential.

7. Diurnal variation of wind speeds delineatesthat daytime hours favour higher wind speeds,as the wind speed threshold of 4.5 m/s wasnoted to occur soon after sunrise (*07:30 AMIST) and decrease from sunset (*17:30 IST).

8. It was identiBed that most of the western partsof India, covering the regions of Rajasthan,Gujarat, Kutch and Saurashtra, West MadhyaPradesh, Madhya Maharashtra, Coastal andinterior Karnataka, Telangana, Coastal AndhraPradesh, and Rayalaseema, are favourable for


low wind farming where the wind speeds arehigher than 3.5 m/s.

9. In respect of oAshore wind power potential,this study characterises that the wind speedsare higher over the ocean than land. The windspeeds are between 5.4 and 6.7 m/s near thecoastline and are higher than 6.7 m/s over theopen ocean, with a duration of 12–16 hrs/day,in contrast to the wind speeds of 4.5–5.4 m/sto have durations of 10–12 hrs/day overland.

10. This study brings forth the increase of windspeed with height to be rapid (20%) from thesurface to 40-m level and then gradual (10%)from 40- to 100-m level. This ensures that thespatial extent of the wind speed thresholdsincreases with height below the 100-m level,approximately covering 50% of the Indiansubcontinent to have wind speeds[4.5 m/s at80-m level.

11. Estimation of wind power potential indicatesthat Gujarat, Kutch, and Interior Karnatakahave the highest potential, whereas Saurashtraand Rayalaseema had moderate potential.This result is in sharp contrast to the reportedlarger areas having threshold wind speeds[4.5 m/s.

12. The wind power density is found to haverapidly increased at lower levels, at *68%between 10–20 and 20–40 m layers, graduallyreducing to *18% between 80 and 100 m. Thewind power density is noted to attain thethreshold value of 150 W/m2 at *50-m level,indicating the advantages of positioning thewindmill at a height of *50 m.

The present study conclusively identiBes thepotential areas of wind energy over the Indiansubcontinent, based on high resolution, hourlywind speed data for 40 years from 1979 to 2018.The recently accessible global reanalysis data,generated using model application assimilatingobservations from conventional and remote sensingplatforms, was authenticated through weather andclimate studies by several researchers. The noveltyof the present study is the use of this futuristicreanalysis data, at the high spatial resolution of*12.5 km and temporal resolution of 1–3 hrs fordetermining the regions of threshold ([4.5 m/s)wind speeds and the periods in terms of favourablemonths, seasons, times of the day and the contin-uous number of hours. Further, these data at 10-mlevel was utilised to estimate the wind speeds andto assess the wind power potential at different

heights below 100-m level which helped to inferthe optimized height, favourable periods, and thespatial extents of wind power potential. Theimportant deduction that the congenial times forwind power production are the daytime and thesummer months imply the role of solar heating,whereas the identiBed regions on the west andsouth-central parts of India signify the dominantrole of regional-scale monsoon circulation resultingfrom land–ocean thermal contrast. This study alsoprovided potential regions for low wind farmingand projected the higher oAshore wind powerpotential over the ocean region surrounding theIndian subcontinent, both of which are of interestin the emerging renewable energy scenario. Theauthors have notiBed the wind power potentialinformation at the level of subdivisions for dif-ferent wind speed thresholds, which along withthe other results may provide useful inputs fordifferent applications. The results obtained fromthis study are unique and valued, as they havebeen derived from accredited surface wind datathrough the assimilation of all available observa-tions using modelling strategy, which will beuseful to both the industry and policymakers foreAective planning of wind energy augmentationover India.

Acknowledgements

The authors acknowledge free access to ERA datafrom the European Centre for Medium-RangeWeather Forecasts (ECMWF), UK; AWS datafrom MOSDAC, Government of India; windobservations data at the turbine level fromNational Institute of Wind Energy, India. Thisresearch is supported by the Early Career ResearchAward, Science and Engineering Research Board,Government of India through Bnancial supportunder grant No. ECR/2016/001295.

Author statement

G Ch Satyanarayana: Conceptualization, prob-lem envision, methodology adoption, computa-tion, visualization, data curation, validation,draft preparation, and writing; D V BhaskarRao: Intellectual contribution, original draftpreparation, review and editing, and D Srinivas:Analysis and visualization, draft preparation,review and editing.


References

Ahmed S A and Mahammed H O 2012 A statistical analysis ofwind power density based on the Weibull and Ralyeighmodels of Penjwen Region Sulaimani, Iraq; Jordan J. Mech.Ind. Eng. 6 135–140.

Al-Mohamad A and Karmeh H 2003 Wind energy potential inSyria; Renew. Energy 28 1039–1046.

Al-Nhoud O and Al-Smairan M 2015 Assessment of windenergy potential as a power generation source in the AzraqSouth, Northeast Badia, Jordan; Modern MechanicalEngineering 5 87–96, https://doi.org/10.4236/mme.2015.53008.

Amin A L 1999 Liberalization of the Indian power industry,wind power in Gujarat; Renew. Energy 16 977–980.

Ashtine M, Bello R and Higuch K 2016 Assessment of windenergy potential over Ontario and Great Lakes using theNARR data: 1980–2012; Renew. Sust. Energ. Rev. 56272–282.

Bakshi R 2002 Wind energy in India; IEEE Power Eng. Rev.22 16–18.

Belmonte Rivas M and Stoffelen A 2019 Characterizing ERA-Interim and ERA5 surface wind biases using ASCAT;Ocean Sci. 15 831–852, https://doi.org/10.5194/os-15-831-2019.

Bhaskar Rao D V 2019 Numerical Weather Prediction; BSPublications Private Limited, 266p, ISBN: 9789387593961.

Blanco M I 2009 The economics of wind energy; Renew. Sust.Energ. Rev. 13 1372–1382, https://doi.org/10.1016/j.rser.2008.09.004.

Carrillo C, Cidras J, Diaz-Dorado E and Obando-Montano AF 2014 An approach to determine the Weibull parametersfor wind energy analysis: The case of Galicia (Spain);Energies 7 2676–2700, https://doi.org/10.3390/en7042676.

Central Electricity Authority 2020 Executive Summary onPower Sector; Ministry of Power, Government of India.http://cea.nic.in/reports/monthly/executivesummary/2020/exe˙summary-05.pdf.

Chauhan S, Tcherniak D and Hansen M 2010 Dynamiccharacterization of operational wind turbines using opera-tional modal analysis; In: Proceedings of China WindPower, Beijing, China, pp. 13–15.

CWET 2010 Indian Wind Energy Atlas; Centre for Wind EnergyTechnology, http://www.cwet.tn.nic.in/docu/Indian˙wind˙atlas˙brochure.pdf.

Dahmouni A W, Ben Salah M, Askri F, Kerkeni C and BenNasrallah S 2010 Wind energy in the Gulf of Tunis, Tunisia;Renew. Sust. Energ. Rev. 14 1303–1311.

Dee D P, Uppala S M, Simmons A J, Berrisford P, Poli P,Kobayashi S, Andrae U, Balmaseda M A, Balsamo G,Bauer P, Bechtold P, Beljaars A CM, van de Berg L, BidlotJ, Bormann N, Delsol C, Dragani R, Fuentes M, Geer A J,Haimberger L, Healy S B, Hersbach H, H�olm E V, IsaksenL, Kallberg P, K€ohler M, Matricardi M, McNally A P,Monge-Sanz B M, Morcrette J -J, Park B-K, Peubey C, deRosnay P, Tavolato C, Th�epaut J-N and Vitart F 2011 TheERA-Interim reanalysis: ConBguration and performance ofthe data assimilation system; Quart. J. Roy. Meteorol. Soc.137 553–597, https://doi.org/10.1002/qj.828.

Dikko I and Yahaya D B 2012 Evaluation of wind powerdensity in Gombe, Yola and Maiduguri, NortheasternNigeria; J. Res. Peace, Gender Dev. 2 115–122.

Dodla Venkata B, Satyanarayana G Ch and Srinivas Desam-setti 2017 Analysis and prediction of a catastrophic Indiancoastal heat wave of 2015; Nat. Hazards 87 395–414,https://doi.org/10.1007/s11069-017-2769-7.

Dokur E and Kurban M 2015 Wind speed potential analysisbased on Weibull distribution; Balkan J. Electr. Comput.Eng. 3 231–235, ISSN:2147-284X, https://doi.org/10.17694/bajece.72748.

Draxl C, Purkayastha A and Parker Z 2014 Wind resourceassessment of Gujarat (India); Technical Report, NREL/TP-5000-61741, National Renewable Energy Laboratory,USA, https://doi.org/10.2172/1149654.

Dvorak M J, Archer C L and Jacobson M Z 2010 CaliforniaoAshore wind energy potential;Renew. Energy 35 1244–1254.

Elamouri M and Amar B F 2008 Wind energy potential inTunisia; Renew. Energy 33 758–768.

El-Osta W and Kalifa Y 2003 Prospects of wind power plantsin Libya: A case study; Renew. Energy 28 363–371.

Farrugia R N 2002 The wind shear exponent in a Mediter-ranean Island climate; Renew. Energy 28 647–653.

Florin O, Deleanu L, Rusu L and Georgescu C 2016 Evalu-ation of the wind energy potential along the MediterraneanSea coasts; Energ. Explor. Exploit. 34 766–792.

Fyrippis I, Axaopoulos P J and Panayiotou G 2010 Windenergy potential assessment in Naxos Island, Greece; Appl.Energy 87 577–586.

Ghitas A E, Abulwfa A and Abdel-Hadi Y A 2016 AnAssessment of wind energy potential as a power generationsource in Helwan; J. Clean Energ. Technol. 4 453–456.

Gokcek M, Bayulken A and Bekdemir S 2007 Investigation ofwind characteristics and wind energy potential in Kirk-lareli, Turkey; Renew. Energy 32 1739–1752.

Graham R M, Hudson S R and Maturilli M 2019 Improvedperformance of ERA5 in Arctic Gateway relative to fourglobal atmospheric reanalyses; Geophys. Res. Lett. 466138–6147, https://doi.org/10.1029/2019GL082781.

Hersbach H 2019 ECMWF’s ERA5 reanalysis extends back to1979; ECMWF Newsletter 158 ECMWF, UK.

Hersbach H, Bell B and Berrisford P et al. 2020 The ERA5global reanalysis; Quart. J. Roy. Meteorol. Soc. 1461999–2049, https://doi.org/10.1002/qj.3803.

Himrib Y, Rehmana S, Draouic B and Himri S 2008 Windpower potential assessment for three locations in Algeria;Renew. Sust. Energ. Rev. 12 2495–2504.

HoAmannL,GuntherG,DanLi, SteinOlaf,WuXue,GriessbachS, Heng Yi, Konopka P, M€uller Rolf, Vogel B and WrightJonathon2019FromERA-InterimtoERA5:Theconsiderableimpact of ECMWF’s next-generation reanalysis on Lagran-gian transport simulations; Atmos. Chem. Phys. 193097–3124, https://doi.org/10.5194/acp-19-3097-2019.

Horanyi A 2017 Some aspects on the use and impact ofobservations in the ERA5 Copernicus climate changeservice reanalysis; Idojaras 121 329–344.

Jeon J and James T W 2012 Using conditional kernel densityestimation for wind power density forecasting; J. Am. Stat.Assoc. 107 66–79.

Kim G, Jeong W M, Lee K S, Jun K and Lee M E 2011OAshore and nearshore wave energy assessment around theKorean peninsula; Energy 36 1460–1469.

Kumar T B, Sekhar O C, Ramamoorty M, Rao S K and Rao DV B 2017 Comparative study on wind forecasting modelsfor day ahead power markets; IEEE International


Conference on Signal Processing, Informatics, Communica-tion and Energy Systems (SPICES), Kollam, pp. 1–5,https://doi.org/10.1109/spices.2017.8091273.

Lei Y, Letu H and Shang H et al. 2020 Cloud cover over theTibetan Plateau and eastern China: A comparison of ERA5and ERA-Interim with satellite observations; Clim. Dyn. 542941–2957, https://doi.org/10.1007/s00382-020-05149-x.

Mahyoub A H 2006 Assessment and analysis of wind powerdensity in Taiz, Republic of Yemen; Assiut Univ. CenterEnviron. Stud. 9 13–20.

Maina AW, Kamau J N and Timonah S 2016 Analysis of windenergy potential by different methods based on Weibullstatistics for a site in Juja, Kenya; J. Multidiscip. Eng. Sci.Technol. 3 5123-5129, ISSN: 2458-9403.

Maity S, Das B, Maity R, Maity N P, Guha K and SrinivasaRao K 2019 Improvement of quantum and power conver-sion eDciency through electron transport layer modiBcationof ZnO/perovskite/PEDOT: PSS based organic hetero-junction solar cell; Sol. Energy 185 439–444.

Mani A and Mooley D A 1983 Wind energy data for India;Allied Publishers Private Limited, New Delhi, India, 545p.

NYSERDA 2020 New York Wind Energy Guide for LocalDecision Makers; https://www.nyserda.ny.gov/Researchers%20and%20Policymakers/Power%20Generation/Wind/Large%20Wind/New%20York%20Wind%20Energy%20Guide%20Local%20Decision%20Makers.

Odo F C, ODah S U and Ugwuoke P E 2012 Weibulldistribution-based model for prediction of wind potential inEnugu, Nigeria; Adv. Appl. Sci. Res. 3 1202–1208.

Olauson J 2018 ERA5: The new champion of wind powermodelling?; Renew. Energy 126 322–331.

Oluseyi A O, Ojo O and Vasel A 2019 On the need for thedevelopment of low wind speed turbine generator system;IOP Conf. Ser.: Earth Environ. Sci. 331 012062, https://iopscience.iop.org/article/10.1088/1755-1315/331/1/012062.

Petersen E L, Mortensen N G, Landberg L, H�ujstrup J andFrank H P 1998 Wind power meteorology. Part I: Climateand turbulence; Wind Energy 1 2–22.

Phadke A, Bharvirkar R and Khangura J 2011 Reassessingwind potential estimates for India: Economic and policyimplications; No. LBNL-5077E, Lawrence BerkeleyNational Lab. (LBNL), Berkeley, CA (United States).

Prakash R B R and Srinivasa Varma P 2017 Stability enhance-ment of wind power plant during abnormal conditions withnegative sequence components compensation using statcom;J. Adv. Res. Dyn. Control Syst. 9(18) 85–95.

Ramachandra T V and Shruthi B V 2003 Wind energypotential in Karnataka, India; Wind Energy 27 549–553.

Rao K S 1986 Renewable energy in India; Electr. Conserv. Q6 7–8.

ReddyKKG, Venkatramana R S, SarojammaB and RamkumarT K 2015 Statistical analysis for wind energy estimation overGadanki, India; Res. J. Eng. Technol. 1 30–40.

REN21 2020 Renewables 2020 Global Status Report (Paris:REN21 Secretariat), ISBN 978-3-948393-00-7 (https://www.ren21.net/wp-content/uploads/2019/05/gsr˙2020˙full˙report˙en.pdf).

Rusu L and Soares C G 2012 Wave energy assessments in theAzores islands; Renew. Energy 45 183–196, http://dx.doi.org/10.1016/j.renene.2012.02.027.

Sabique L, Yesubabu V, Hari Prasad D, Omar K and IbrahimHoteit 2016 A high-resolution assessment and wave energypotentials in the Red Sea; Appl. Energy 181 244–255.

Satyanarayana G C and Bhaskar Rao D V 2020 Phenology ofheat waves over India; Atmos. Res. 245 105078, https://doi.org/10.1016/j.atmosres.2020.105078.

Satyanarayana G C, Lucy Supriya R H and Bhaskar Rao D V2019 Wind energy assessment over the Andhra Pradesh andTelangana regions; Meteorol. Appl. 26 14–29, https://doi.org/10.1002/met.1730.

Sempreviva A M, Barthelmie R J and Pryor S C 2008 Reviewof methodologies for oAshore wind resource assessment inEuropean seas; Surv. Geophys. 29 471–497.

Standen J, Wilson C, Vosper S and Clark P 2017 Prediction oflocal wind climatology from Met ODce models: Virtual MetMast techniques; Wind Energy 20 411–430, https://doi.org/10.1002/we.2013.

Stull R 2015 Practical meteorology: An algebra-basedsurvey of atmospheric science; Univ. British Columbia,938p.

Tummala A, Velamati R K, Sinha D K, Indraja V and KrishnaV H 2016 A review on small scale wind turbines; Renew.Sustain. Energy Rev. 56 1351–1371.

Urraca R, Huld T, Gracia-Amillo A, Martinez-de Pison F J,Kaspar F and Sanz-Garcia A 2018 Evaluation of globalhorizontal irradiance estimates from ERA5 and COSMO-REA6 reanalyses using ground and satellite-based data;Sol. Energy 164 339–354.

Wakeyama T and S Ehara 2011 Estimation of renewableenergy potential and use: A case study of Hokkaido;Northern-Tohoku Area and Tokyo Metropolitan; Japan;World Renewable Energy Congress-Sweden; 8–13 May,Link€oping; Sweden. No. 057, Link€oping University Elec-tronic Press, pp. 3090–3097.

Wang C and Prinn R G 2010 Potential climatic impacts andreliability of very large-scale wind farms; Atmos. Chem.Phys. 10 2053–2061.

Wilks D S 2006 Statistical Methods in the AtmosphericSciences; 2nd edn, Academic Press, London.

Willow H, Gunturu U B and Schlosser A 2014 The potentialwind power resource in Australia: A new perspective; PLoSONE 9 e99608, https://doi.org/10.1371/journal.pone.0099608.

WMO (World Meteorological Organization) 1981 Meteoro-logical aspects of the utilization of wind as an energy source,WMO Technical Note 175, World Meteorological Organi-zation, Geneva.

Wolde-Ghiorgis W 1998 Wind energy survey in Ethiopia;Solar Wind Technol. 5 341–351.

Wolfson R 2012 Energy, environment and climate; 2nd edn,WW Norton and Company Inc, New York, USA,pp. 366–370.

Xi L, Michael B, McElroya and Kiviluomac J 2009 Globalpotential for wind-generated electricity; Proc. Natl. Acad.Sci. USA 106 10933–10938.

Yip C M A, Gunturu U B and Stenchikov G L 2016 Windresource characterization in the Arabian Peninsula; Appl.Energy 164 826–836.

Zheng C W and Li C Y 2015 Variation of the wave energyand significant wave height in the China Sea andadjacent waters; Renew. Sustain. Energy Rev. 43381–387.

Zheng C W, Pan J and Li J X 2013 Assessing the China Seawind energy and wave energy resources from 1988 to 2009;Ocean Eng. 65 39–48.


Zheng C W, Li C Y, Pan J, Liu M Y and Xia L L 2016 Anoverview of global ocean wind energy resource evaluations;Sustain. Energy Rev. 53 1240–1251.

Wei Zhou, Hongxing Yang and Zhaohong Fang 2006 Windpower potential and characteristic analysis of the PearlRiver Delta region, China; Renew. Energy 31 739–753.

Corresponding editor: KAVIRAJAN RAJENDRAN


assessment of wind energy potential over india using high

Documents