Offshore Wind Energy Resource Assessment from Satellite Data Observations and WRF in Porto Santo Island

The vast majority of isolated electricity production systems such as Islands depends on fossil fuels. Porto Santo Island, a Portuguese UNESCO Biosphere Reserve candidate from Madeira Archipelago situated in the Atlantic Ocean, aims to become a sustainable territory in order to reduce its carbon footprint. 10 A sustainable pathway goes through the integration of renewable energy in the electricity production system, in particular, the potential of offshore wind energy. The scope of this work has three main purposes: 1) the offshore wind resource assessment in Porto Santo Island, 2) the determination of a zone of interest regarding the combination of different parameters such us the bathymetry, distance to the coastline and integrated in the national situation plan of maritime space 3) the estimation of the annual energy production from 15 the best-fitted Weibull Distribution. In the first place, a methodology for data analysis was defined processing netcdf data regarding a ten year wind hindcast from WRF (Weather Research and Forecasting) atmospheric model at 100 m above mean sea level from Ocean Observatory, annual and monthly mean offshore wind energy resource maps were created and a comparison with about 20 year times series of surface winds derived from remotely satellite scatterometer observations at different locations was made. Results show that 20 the average annual mean wind speeds reach the range of 6,6-7,6 m/s in specific areas, situated in the northern part of Porto Santo Island with a Weibull distribution shape parameter (k) of 2,4-2,9. Based on the results, the wind resource assessment, the estimation of the annual wind energy production and capacity factors were calculated from the best-fitted Weibull distribution for each of the geographical coordinates selected. Comparisons with observational data show that WRF model is a proficient wind generating tool. The technical energy 25 production potential and a priority zoning for offshore wind power development is performed using wind turbine generators of 3.3 MW–8.0 MW capacity, that could generate between 12 and 26 GWh of energy per year, while avoiding CO2 emissions. The results show that an offshore wind farm plan is an eligible choice, with an average annual wind power density reaching about 300 W/m at 100 m height in the north region. 30

development of offshore wind power plants, i.e. government policies, technological advancements, reliability of the equipment 85 and infrastructure, environmental and social responsibility and cost reductions. However, offshore wind resource assessment is the first, and necessary, key step in the development phase of offshore wind power plants. Offshore wind resource assessment is the first key step in order to identify and select a potential zone of interest by integrating specific wind profile, wind direction, variability regarding height and distance from shore with economic, social, political and environmental parameters for a multi criteria decision making solution.
In this study, offshore wind resource maps were created considering two data sources: 1) Weather Research and Forecasting (WRF) atmospheric model regarding 10-year historic data base for the Typical Meteorological year from 2007 until 2016 105 (80x50 km 2 , distance of 10 km, applied under nesting grids on three domains with resolutions of 9km, 3km and 1km, wind speed at 100 m height) source data from Ocean Observatory and 2) Time series of surface winds (10m height) from remotely scatterometer observations (L4) a 20-year historic data from 1994 until 2016, data source from IFREMER.
For this purpose, mean hourly wind speeds data were processed in netcdf format, monthly and annual resource maps were computed.

Prandtl logarithmic law
The Prandtl logarithmic law as described in eq.1 gives a good representation of the variation of the mean wind speed with height. Comparison between measurements and calculations of the log-law model showed that the model can describe the change of mean wind speed extrapolating for different heights all the data at a reference height zR and obtain with reasonable 115 accuracy the mean wind speed turbine at the hub height z which is 100 m.
Where u(z) is the mean wind speed at height z, u(zR) is the mean wind speed measured at the reference height zR and z0 is the roughness length determined by the surface condition of the surface.

Method of bins
This method is characterized by the representation of the wind speed data in a histogram, obtained by splitting the range of data into equally sized bins, called classes. Each class is represented by the middle value of the bin. Therefore, each bin with 1 m/s width has associated a relative frequency which is calculated by the Eq.

Weibull distribution functions
The Weibull Distribution has been used to represent wind speed distributions for application in wind loads studies for a long time. The Weibull distribution function, which is a two-parameter function, in wind energy analysis it is used to represent the 130 wind speed probability density function (pdf), expressed mathematically (Spera, 1995;Persaud et al., 1999) as: Where f (u) is the probability of observing wind speed u, k is the dimensionless Weibull shape parameter (or factor), and c the Weibull scale parameter. The k values range from 1,5 to 3,0 for most wind conditions, depending on the variability the wind.
Smaller k values correspond to more variable (gusty) winds.

135
In this case study the parameters of the Weibull wind speed distribution for wind energy analysis were determined by the least squares method and graphic method. The application of each method is demonstrated using a mean hourly wind speed data set measured in a year.

Analytical procedure -Least Squares Method
It was assumed a dataset that constitute a pair (xi,yi)=(x1,y1),..., (xn,yn). The least squares principle minimizes the vertical 140 distance between the data points and the straight line fitted to the data. Both parameters k and c were calculated in order to minimize the mean squared error (MSE) given in the following equation by changing both k and c to minimize quadratic error: Where

Graphical procedure -Linear Regression
The Weibull cumulative distribution function (cdf) was transformed to allow the computation of a linear regression in order to obtain the shape and scale parameter, mathematically the cdf equation it is given as: and Applying a linear regression it is possible to obtain the correlation between these variables and obtain the linear Eq. (8): = + , where the shape parameter k is the slope (A) and from the interception in the y-axis we can calculate the scale parameter c which is given by the mathematical Eq. (9): = exp (− ).

Power Curve -Sigmoid Approximation
In this procedure was estimated the parameters c1 and c2 of the sigmoid function Eq. (11) in order to minimize the mean squared error by changing both c1 and c2 according to the following eq.
The results are presented as an analytical performance curve after the best fitting of the sigmoid function to the manufacturer 160 power curve currently available that can be used in assessing the power output as subject to the given wind regime, thus providing a practical and straightforward tool for power potential assessment. The sigmoid function applied is the following: https://doi.org/10.5194/wes-2020-7 Preprint. Discussion started: 28 January 2020 c Author(s) 2020. CC BY 4.0 License.

Estimation of annual energy production
The assessment of estimated output energy is a key issue in the design and sizing of wind generation projects. Further, its 165 accuracy and reliability can benefit from the use of actual power curve of the wind turbine generator that furnish the response of the generator to the wind regime under consideration further described in method 1 or by calculating according to method 2 the best sigmoid power curve approximation . Furthermore, regarding the calculation of the cumulative distribution function, F(u) according to Eq. (6) it is estimated the annual energy produced by the wind turbine of the manufacturer.

170
The direct method combines the method of bins which divides the average hourly wind speed data set in classes (bins) with the wind turbine generator power curve in order calculate the annual energy produced.

175
This method allows to estimate the annual energy produced through the analytical computation of the Weibull Distribution approximation for the wind speed and the sigmoid approximation for the wind turbine power curve.
The energy generated by the offshore wind farm is estimated by applying a generic power curve or by calculating the sigmoid approximation for four offshore class wind turbine generators (WTG) available on the market, i.e. 3.3MW, 5MW, 7MW and