Simulated Collection Parameters

Several techniques have been developed over the years to estimate the spatio-temporal pattern of the wind resource (Veronesi & Grassi, 2016). Brower (2012) opined that majority of these techniques are part of a set referred to as numerical wind flow models, which estimate the wind resource solving the physical equations that govern the motion of air in the atmosphere. These methods have varying level of complexity, derived by the type and amount of equations they include. The simplest ones are the mass-consistent models (Philips, 1979), first developed in the 1970s, which only solve the equation of conservation of mass.

On the other end of the complexity spectrum are numerical weather prediction models (Brower, 2012), which solve all the computational fluid dynamics equations plus others that govern the energy exchanges between soil and atmosphere. These methods are able to

estimate the long term wind resource and its time variability, even though they tend to be time-consuming and computationally expensive (Veronesi, Grassi, & Raubal, 2016).

Interestingly, another branch of research has been dedicated to the development of techniques for wind resource assessment based purely on statistical algorithms (Veronesi &

Grassi, 2016). Statistical models correlate wind speed data from weather stations, with remotely sensed physical parameters, to infer the wind spatio-temporal pattern (Veronesi &

Grassi, 2016). Veronesi et al. (2016) stated that statistical methods are accurate, computationally efficient, and less time-consuming than physical models. These methods have been tested in the literature for estimating both the long term pattern of the wind resource ((Veronesi et al., 2016; Aksoy, Toprak, Aytek, & Unal, 2004; Luo, Taylor, &

Parker, 2008; Foresti, Tuia, Kanevski, & Pozdnoukhov, 2011; Cellura, Cirrincione, Marvuglia, & Miraoui, 2008) and for time-series estimations with models such as (Auto Regressive Moving-Average (ARMA) (Castellanos & Ramesar, 2006; Philippopoulos &

Deligiorgi, 2009), Markov chain (Shamshad, Bawadi, Hussin, Majid, & Sanusi, 2005) and autoregressive models (Poggi, Muselli, Notton, Cristofari, & Louche, 2003). However, in 2012, Ohashi and Torgo stated that spatio-temporal prediction, that is, the estimation of the hourly wind speed pattern in areas where no direct observations are available, of wind speed time-series using machine learning techniques, is a recent research topic. The major problem in wind resource assessment is the large amount of uncertainty involved, which ranges from malfunctions of the weather stations to the extrapolation of the wind speed profile in complex terrains. Assessing this uncertainty is difficult with numerical wind flow models, but straightforward with statistical wind resource assessment, which can precisely account for all these sources of uncertainties (Cellura et. al., 2008).

In their study, Veronesi and Gassi (2016) presented a new generalized statistical methodology to generate the spatial distribution of wind speed time-series, using Switzerland

51 as a case study. The research was based upon a machine learning model and demonstrated that statistical wind resource assessment can successfully be used for estimating wind speed time-series. In fact, the method was able to obtain reliable wind speed estimates and propagated all the sources of uncertainty (from the measurements to the mapping process) in an efficient way, that is minimizing computational time and load. This allows not only an accurate estimation, but the creation of precise confidence intervals to map the stochasticity of the wind resource for a particular site. The validation shows that machine learning can minimize the bias of the wind speed hourly estimates. Moreover, for each mapped location, this method delivers not only the mean wind speed, but also its confidence interval, which are crucial data for planners.

ThiThi, Boopathi, Bastin, Rangaraj, and Gomathinayagam (2017) technically studied the wind power potential at 100 m AGL (above ground level) in Myanmar by utilizing MERRA_2 (Modern-Era Retrospective analysis for Research and Applications) reanalysis datasets, Geo-informatics data sets and maps. The results show that promising wind potential areas are in Ayeyarwaddy, Yagon, Tanintharyi, Mandalay, Magway, Sagaing Regions and Rakhine States, the highest wind power density is 261 W/m² and installable technical wind power potential is 153 GW approximately. Based on the analysis by using industry-standard software, Annual Energy Production of 4454.88 MWh/year may be obtained with capacity factor of 25 percent. Subsequently, they determined the wind power potential in different hub heights at 50 m, 80 m and 120 m. The results can fulfill to facilitate the development of wind energy not only for utility-scale generation but also village power and other off-grid applications in Myanmar. Therefore, their study provides technical wind power potential estimation to perform future wind feasibility investigations in Myanmar.

The traditional utilization of reanalysis data is as a historical record of wind speed patterns which can be employed to correlate with actual short‐term wind speed measurements from

meteorological masts (ThiThi et. al., 2017). The reanalysis data can also reduce the costs and risks of wind farm development by providing a source of long-term meteorological data that is difficult or expensive to acquire through normal meteorological measurement campaigns.

Moreover, the reanalysis data are gridded datasets that combine data obtained from global circulation models (GCM‟s) with measured data. In mapping the average wind resource over large areas, NCEP-R2 (National Centers for Environmental Prediction), ERA-Interim (European Center for Medium‐Range Weather Forecasts reanalysis series), NCEP-CFSR (Climate Forecast System Reanalysis), NASA-MERRA (Modern-Era Retrospective analysis for Research and Applications) are currently freely and publicly available (Carvalho, Rocha, Gomez-Gesteira, & Santos, 2014). Among them, ThiThi et al. (2017) chose MERRA dataset for the estimation of wind power potential in their study.

ThiThi et al. (2017) stated that MERRA is a NASA (National Aeronautics and Space Administration) reanalysis product with coupled numerical modeling with large quantities of empirical data such as surface measurements and earth observation satellite data to generate a long term continuous dataset. The main advantage of MERRA data is the availability of long-term wind speed data on a global grid and the original MERRA wind is in the public domain to construct the wind power density dataset. Therefore, the MERRA dataset had been used in several studies to estimate the potential wind resource such as in UK (The Crown Estate, 2014) and also other countries (Olauson & Bergkvist, 2015; Cosseron, Schlosser, & Gunturu, 2014; Zhang et. al., 2015; Gunturu & Schlosser, 2015; Ritter, Shena, Cabrera, Odening, &

Deckert, 2015). Olauson and Bergkvist (2015) investigated a model for the Swedish wind power production based on MERRA reanalysis data and noted that MERRA dataset has a relatively high temporal and spatial resolution. Cosseron, et al. (2014) also used reanalysis data from the MERRA data product and computed wind power density for the assessment of the wind power resource over Europe. In 2015, Zhang et al. compared three NWP-based

53 wind resource assessment methods (MERRA, AnEn based on MERRA, and WIND Toolkit) across the United States. Gunturu and Schlosser (2015) used MERRA boundary layer flux data to construct wind profile at 50 m, 80 m, 100 m and 120 m turbine hub heights and estimated wind power density of each level by comparing with NREL (National Renewable Energy Laboratory) wind map. One thing was observed that MERRA could provide a more accurate dataset using the comprehensive suite of satellite based information for climate and atmospheric research”. Ritter et al. (2015) applied MERRA data to obtain wind speed data at an unobserved location in Germany. Hallgren, Gunturu and Schlosser (2014) found that MERRA data provided a more robust assessment of the temporal characteristics (i.e. mean, median, availability, intermittency, etc.) of wind power than that used in other studies. In this study, MERRA_2 reanalysis dataset is used for technical wind energy potential assessment in Myanmar. It hopes to perform the starting point for future wind feasibility investigations in Myanmar.

Recently, several authors have explored the wind potential assessment all over the world (Ohunakin, 2015; Mentis, 2013; European Environment Agency Technical Report, 2009;

Mentis, Siyal, Korkovelos, & Howells, 2016). ThiThi et al. (2017) technically estimated wind energy potential by using input data such as gridded reanalysis weather data, terrain elevations, land cover and socioeconomic data. The results were presented in color-coded maps by using GIS (Geographic Information Systems) analysis tools. The study prompted to support whether government needs to figure out certain policies and regulatory frameworks that are required when there is enough potential to utilize wind energy through technical wind potential in Myanmar.