You are on page 1of 23

Assessment of Wind Energy Resources Using Reliable Multi-modal Wind

Distribution Model
Maja Celeska*, Krste Najdenkoski, Vladimir Dimchev, Vlatko Stoilkov
Faculty of Electrical Engineering and Information Technologies
Ss Cyril and Methodius University of Skopje, R. Macedonia
email: celeska@feit.ukim.edu.mk

ABSTRACT
In the past few years, wind energy has been established as a fastest-growing source of
electricity in the world from all renewable and clean energy sources. Precise and reliable wind
modelling is of major significance in the process of wind field development. Previous well-
known practice for representing wind data by sector-wise Weibull distribution, lately is
replaced with different multivariate and multimodal wind distribution models which are far
more precise. The measured wind data used in this paper are from five locations investigated
in the second wind measuring campaign in R. Macedonia. For the five sites- characterized
with completely different wind regimes, two different joint distributions are obtained and
their good of fitness is proven by coefficient of determination R2.
The selection and assignment of appropriate, site-specific turbine is a complex problem.
Therefore, in the paper a suitable modal procedure is carried out in order to achieve the best
compromise between the decision criteria. The importance of identifying a class of wind
turbine (WT) is due to the complex environmental conditions that produce turbulent air. Тhe
turbulence intensity of the winds in the studied areas is calculated according to IEC-61400
standard. Afterwards, wind energy potential is being estimated and annual energy production
and capacity factor (CF) for six different WT types is calculated. From the analysis based on
the computed CF, only two of the tested locations perform a promising potential for future
wind field development, while at the other sites WTs from a smaller scale are applicable.

KEYWORDS
Wind modelling, multi-modal distribution, wind energy; capacity factor

INTRODUCTION
From 2011 to 2015, the European Wind Energy Association (EWEA) arranged four
Comparison of Resource and Energy Yield Assessment Procedures (CREYAP) exercises, in
which parties from the wind energy industry and academia were invited to carry out and
submit wind climate and energy yield predictions for case study wind farms, [ CITATION
NGM15 \l 1033 ]. It was clearly stated that in specific areas current engineering practices can
be improved and more research and development are needed.
The process of estimating and analysing the wind energy yield of a certain location,
seemingly is an easy and defined procedure when using commercial softwares for that
purpose. On the one side, that implies big expenditures, but on the other- every individual
without extensive and detailed knowledge in wind engineering area, can work with those
software packages. The aim of this paper is to present a procedure for assessment of wind
energy resources on some certain locations, using multi-staged and modal estimations, by
using easy accessible and free softwares.

**
Corresponding author

1
Following the example of the most countries worldwide, Republic of Macedonia also
recognized the significance of renewable energy as a complement to the conventional fuel that
serves as a source for generating electricity, [ CITATION AAl17 \l 1033 ]. In 2010, Macedonian
Academy of Sciences and Arts, published a strategy for utilization of renewable sources of
energy. In the study, it was predicted that by 2030, the total amount of installed capacity of
wind energy sources will be around 360 MW, [ CITATION Art10 \l 1033 ]. By now, only one
stage of the first wind field (WF) in Macedonia is accomplished- WF Bogdanci, with installed
capacity of 36.8 MW. Because of these reasons, an intensification in the process of
assessment of wind energy resources is indispensable. The data used in this paper are gained
by gathering measurement data from the second measuring wind campaign in Macedonia in
the period July 2012-March 2016, [ CITATION VDi14 \l 1033 ] . It has been covered a continuous
process of data acquisition, except for some time intervals when certain malfunction in the
equipment took place, [CITATION KDe16 \l 1033 ]. Hence appears the discrepancy in the number
of data rows collected at the measurement stations, Table 1. Five locations were chosen
according to several relevant factors, [ CITATION NGM15 \l 1033 \m DKK14] such as: wind atlas
of Macedonia; terrain configuration; local infrastructure - proximity of electrical network and
roads, etc. The sites and the general data for each location are given in Table 1, [ CITATION
MCe17 \l 1033 ]. All measuring stations are equipped with standardized equipment for
measuring wind characteristics, ensuring high accuracy and reliability of all measuring
sensors. Additionally, a 29.25 km2 (6.5x4.5 km2) elevation maps are presented from the stance
of the different colors, as illustrated in Figure 1.

Table 1. General data for every selected site and measurement equipment, [ CITATION MCe17 \l
1033 ]

Data parameters, Measurement


Geographical Elevation
Site heights above site elevation and start and number
coordinates [m]
accuracies of data rows
Berovo
N 041° 44.042'
E 022° 47.059'
608
Wind speed, 30/40/50 m, [±0.4 m/s]
Wind direction, 38/48 m, [±1o

2
]

Temperature, 5 m, [±0.5

3
C]

Pressure, 4 m, [±0.5 mbar] 21.09.2012


171155 data rows
Mogila
N 041° 11.303'
E 021° 21.546'
702
Wind speed, 30/40/50 m, [±0.4 m/s]
Wind direction, 38/48 m, [±1o

4
]

Temperature, 5 m, [±0.5

5
C]

Pressure, 4 m, [±0.5 mbar] 20.07.2012


143215 data rows
Sopishte
N 041° 11.299'
E 021° 21.562'
730
Wind speed, 30/40/50 m, [±0.4 m/s]
Wind direction, 38/48 m, [±1o

6
]

Temperature, 5 m, [±0.5

7
C]

Pressure, 4 m, [±0.5 mbar] 22.09.2012


166100 data rows
St. Nagorichane
N 042° 18.070'
E 021° 58.504'
1179
Wind speed, 30/40/50 m, [±0.4 m/s]
Wind direction, 38/48 m, [±1o

8
]

Temperature, 5 m, [±0.5

9
C]

Pressure, 4 m, [±0.5 mbar] 26.08.2012


138083 data rows
Sv. Nikole
N 041° 55.093'
E 021° 56.792'
431
Wind speed, 30/40/50 m, [±0.4 m/s]
Wind direction, 38/48 m, [±1o

10
]

Temperature, 5 m, [±0.5

11
C]

Pressure, 4 m, [±0.5 mbar] 01.07.2012


180459 data rows

WIND MODELING
In order to illustrate the distribution of wind speed, the converted wind data at hub height, first
are processed by most widely used distribution for the characterization of wind speed- the two
parameter Weibull distribution, [ CITATION TBu01 \l 1033 ]:
k v k−1 −v k
( )( )
f ( v , k ,c )=
c c
exp( )
c
(1)

a) Berovo b) Mogila

c) Sopishte d) St. Nagorichane

e) Sv. Nikole
Figure 1. The 29.25 km2 elevation maps for every selection site

12
Amongst all know methods for calculating Weibull parameters (scale- k and shape-c), here it is
used the Maximum Likelihood Method (MLM) as a method whitch is proven to has the highest
accuracy for calculations at sites similar like the mentioned above, [ CITATION MCe15 \l 1033 ]:
n n −1

[ ]
k
∑v i ln ⁡(v i ) ∑ ln ⁡( v i)
i=1
k= n
− i=1 (2)
k
n
∑v i
i=1
n 1/k
1
c= ∑ v ki
n i=1 ( ) (3)

In order to demonstrate the advantage of the multivariable model for displaying wind regimes
at a certain cite, Figure 2 depicts probability density function (PDF), together with the
frequency distribution for each location. Additionally, Weibull PDF’s for 8 sectors (N, NE, E,
SE, S, SW, W, NW) are displayed individually, on each graph.

a) Berovo b) Mogila

c) Sopishte d) St. Nagorichane

e) Sv. Nikole
Figure 2. Frequency of occurrence, overall and 8 sectors Weibull PDFs for every selected site

13
Wind Power Density (WPD) provides information for the wind energy resource available at a
potential site, in other words indicates how much energy is available at the site. WPD [W/m2]
is a nonlinear function of the PDF of wind velocity and air density, which is expressed as,
[ CITATION JZh111 \l 1033 \m JZh13]:
o
359 vmax ρ max
WPD= ∫
o
∫ ∫ 0.5 ρ v 3 f ( v , θ , ρ ) dρdvdθ (4)
0 0 ρmin
where v and θ represent the wind speed and wind direction, respectively; vmax is the maximum
possible wind speed at that location; ρ represents the air density; ρmin and ρmax are the maximum
and minimum air density in that location, respectively; and f(v, θ, ρ) is the PDF of the wind
condition (speed, direction and air density). Air density is determined using the ideal gas law,
from measured air temperature and air pressure according to the equation:
p10 min
ρ10 min = (5)
R 0 ∙ T 10 min
where ρ10min is the averaged air density, T10min is the measured absolute air temperature averaged
over 10 min, p10min is the measured air pressure averaged over 10 min and ρmin is the gas constant
of dry air (287.05 [J/kg∙K]).
Evaluated values for annual average temperature, pressure, air density, Weibull scale, Weibull
shape parameter and WPD for all five locations are presented in Table 2. From this statistical
data it can be concluded that Sv. Nikole site has the smallest elevation (431 m) and that is why
the air density has the highest value there, also taking into account the air temperature. When
analyzing the mean wind speed, Weibull parameters (k=1.30, c=4.76m/s) and WPD (56.24
W/m2), which are parameters that provide general information for the energy yield at one site,
Sv. Nikole is ranked on the second position. Most promising location for further investigation is
St. Nagorcihane, with highest values for elevation (1179 m), mean wind speed (6.27 m/s), WPD
(133.72 W/m2) and lowest values for pressure, air temperature and air density (1.0845 kg/m 3).
Location with smallest value for WPD (22.92 W/m2) is Mogila, next is Berovo (26.47 W/m2)
and Sopishte (26.47 W/m2).

Table 2. Statistical parameters

Site t [oC] p [mbar] ρ [kg/m3] k c [m/s] v [m/s] WPD [W/m2] R2


Berovo 8.76 908.44 1.1159 1.51 4.01 3.62 26.47 0.9826
Mogila 11.62 925.99 1.1338 1.24 3.71 3.47 22.92 0.9439
Sopishte 9.07 880.78 1.0879 1.35 4.15 3.80 29.86 0.9893
St. Nagorichane 7.67 873.59 1.0845 1.73 7.04 6.27 133.72 0.9977
Sv. Nikole 13.64 955.37 1.3289 1.30 4.76 4.39 56.24 0.9830

CONSTRUCTION OF MULTIVARIATE WIND DISTRIBUTION MODEL


Univariate and unimodal distribution models make limiting assumptions regarding the
correlativity and the modality of the distribution of the wind, [ CITATION JZh13 \l 1033 ].
Through this type of calculations, a lot of assumptions are made that lead to approximations
which deviate significantly from the real scenario. That is why a more complex analysis have
been proposed lately.
Multivariate Kernel Density Estimation (MKDE) method (also known as Parzen–Rosenblatt
window method) is a non-parametric approach to estimate the PDF of a random variable. For
an independent and identically distributed sample, x1,x2,…,xn, drawn from some distribution
with an unknown density f, the KDE is defined to be:

14
n n
x−x i
^p ( x , h ) = 1 ∑ K h ( x−x i )= 1 ∑ K h ( ) (6)
n i=1 nh i=1 h
where K(∙)=(1/h)K(∙/h) for a kernel function K (often taken to be a symmetric probability
density) and a bandwidth h (the smoothing parameter), [ CITATION EPa621 \l 1033 ].
Before representing the multivariable wind distribution, a piecewise Bivariate PDF is
constructed for each site (Figure 3- i). Note that when modelling wind with sector-wise
Weibull distributions as described above, we are assuming that the wind speed satisfies the
same probability distribution inside a direction sector, [ CITATION JFe151 \l 1033 ]. Following, a
Bivariate PDF using piecewise joint distribution is done (Figure 3- ii). Namely, this
distribution contains all input parameters for calculating the bivariant Kernel distribution. For
comparison of these two distributions, a coefficient of determination- R2 is used. R2, is the
squared of correlation coefficient between the observed and modeled (predicted) data values,
which is expressed as:
n n n

{ √[ }
2 n ∑ U i U^ i−∑ U i ∑ U^ i
cov (U , U^ )
2
R=
[√ var (U )var( U^ ) ] =
n
i=1

n 2
i=1

n
i=1

n 2
n∑ U
i=1
2
i− (∑ ) ][
i=1
Ui n ∑ U^ 2i −
i =1
( )]
∑ U^ i
i=1
(7)

where U and U ^ are the observed and fitted quantiles, respectively; cov and var mean
covariance and variance, respectively. The closer the value of R2 is to one, the more the fitted
distribution agrees with the observed data, [ CITATION JZh13 \l 1033 ]. Table 2 contains values
of the comparison of R2.
From Kernel's probability distribution function (Figure 3-iii), all parameters of the wind are
further treated as continuous variables, in function from the angle, which facilitates and
refines all further steps for optimizing the distribution of WTs in one wind field.

i
) ii) iii)
a) Berovo

15
i) ii) iii)
b) Mogila

i) ii) iii)
c) Sopishte

i) ii) iii)
d) St. Nagorichane

i) ii) iii)
e) Sv. Nikole

16
Figure 3. Multivariable multi-modal wind distribution estimations of wind speed and wind
direction

WIND TURBINE CLASSES AND TURBULENCE INTENSITY MEASUREMENT


Before setting a turbine in a wind farms allocated for power generation, it must be known the
appropriate turbine class for that site depending on the turbulence intensity of the winds in the
studied area and the IEC-61400 standard. The importance of identifying a class of WT is due to
the complex environmental conditions that produce turbulent air which, in turn, may cause
damage to the turbine blades and weakness in the performance. At lower wind speeds, the
calculated turbulence intensity is higher. However, the higher turbulence at low wind speeds is
not a concern because of the low power available at those low wind speeds. Turbulence at higher
winds speeds is of greater interest and concern to WT manufacturers, [ CITATION Moh14 \l 1033 ].
Turbulence Intensity (TI) is defined in [CITATION Int05 \l 1033 ] as “the ratio of wind speed
standard deviation to the mean wind speed, determined from the same set of measured data
samples of wind speed, and taken over a specified time” and should actually be considered as the
standard deviation of the wind speed-u normalized with the mean wind speed (Eq. 8), [ CITATION
KSu15 \l 1033 ]:
σu
T I u= (8)

TI is a measure of the gustiness of the wind and it is dimensionless quantity. The representative
TI, for a set of 10-minute time steps, is equal to the 90th percentile of the TI values. Assuming a
normal distribution of these values, it represents the mean value plus 1.28 times standard
deviations, [ CITATION Moh14 \l 1033 ]:
T I 90=T I ave + 1.28 σ TI (9)
IEC Classification of WTs, according to turbulence intensity, is presented in Table 3. The
indexes in the table refer to: vave-annual mean wind speed at hub height; vref -50-year extreme
wind speed over 10 minutes; v50 gust-50-year extreme gust over 3 seconds; Iref is the mean
turbulence intensity at 15 m/s. A, B and C are the categories of higher, medium and lower
turbulence intensity characteristics, respectively [ CITATION Int05 \l 1033 ].
Measurement data regressions are compared to Normal Turbulence Model (NTM) as it is
defined for the standard small to large WTs, in IEC 61400-1 (2005), [CITATION Int05 \l 1033 ].
The locations selected, primarily are investigated for designing wind fields and therefore the
comparison with big WTs NTM 90th percentile is done.
Table 4. Wind turbine classes, [ CITATION Int05 \l 1033 \m Mat15]

High wind Medium wind Low wind


Wind turbine class S
I II III
vave [m/s] 10 8.5 7.5
vref [m/s] 50 42.5 37.5
v50 gust [m/s] 70 59.5 52.5 Defined by
A 0.16 manufacturer
Iref B 0.14
C 0.12

Table 5. Calculated representative turbulence intensity at every selected site

Site Berovo Mogila Sopishte St. Nagorichane Sv. Nikole


TIu 0.2379 0.2549 0.2379 0.1536 0.2204

17
At Figure 4, the y-axes of the all preceding graphs are cut at 0.6 for convenience, values greater
than this are only found for low wind speeds which corresponds to minimal loadings on the WTs
and are thus irrelevant for the results, [ CITATION Moh14 \l 1033 ].

a) Berovo b) Mogila c) Sopishte

d) St. Nagorichane e) Sv. Nikole Legend


Figure 4. Standard deviation of wind speed as a function of mean wind speed at five
measuring locations, for big wind turbine.

a) Berovo b) Mogila c) Sopishte

d) St. Nagorichane e) Sv. Nikole Legend


Figure 5. Turbulence intensity as a function of mean wind speed at five measuring locations,
for big wind turbine.

At all sites, big turbulence intensity is evaluated. On the other side, the annual mean wind speed
at 50 m is between 3.47-6.27 m/s, which characterize all five sites as locations suitable for
installing WTs from type IV. Available data were found only for one turbine form class III (A, B
and C) and none for class IV(S). This class is a novelty and so WT manufacturers still struggle
with a design that will belong to IV Class.

18
Analyzing Figure 4 and 5, closest to the function of interpolation of measurement data, without
exception is the curve from NMT IEC-90% class C. This graphical representation leads us to the
conclusion that WTs designed for very low wind speeds and low turbulence intensity are the
most adequate for the selected sites.
The slow growth of with increasing wind speed (Figure 4), resulting in low turbulence
intensities, again proves the principle that turbulence intensity at high wind speeds is more
provident and sometimes even has smaller value compared to turbulences identified at low wind
speeds. The question is how many of WTs operational hours annually are under high wind
speeds on locations with predominant low wind speeds?

WIND TURBINE POWER GENERATION ESTIMATION


Wind turbines need to be designed for optimal performance and reliability in whatever weather
conditions they may face throughout their period of exploitation. In the paper the subject of
interest is only identification of the most suitable class and type of WT for areas with similar
characteristics as those of the measurements locations presented. The selection of the presented
six types of WTs is done based on the calculations of the turbulence intensity at each location
(mean wind speed, roughness class, WPD and TI). When it comes to give a relevant estimation
on the suitability of a WT for a certain site, a couple of parameters need to be calculate: Annual
Energy Production (AEP), Capacity Factor (CF) and the Levelized Cost of Energy (LCOE).
An easy and commonly used indicator to describe the performance of a WT is the capacity factor
(CF), which is the ratio of the turbines actual power output over a period of time to its theoretical
(rated) power output, [ CITATION IEC13 \l 1033 ]. For the calculation of the capacity factor the
average power output is divided by the rated power of the WT, as described in Eq. 10. The
average power output has to be calculated including all operational states. Due to physical
principles, the capacity factor heavily depends on the available wind condition, [ CITATION
SPf17 \l 1033 ]:

CF= (10)
Prated
where CF is the capacity factor, Ṕ is average power output of WT and Prated is rated power of
WT. In this phase, we assume a power output from each type of WT, according to the known
correlation between wind speed and power output, i.e. power curve.
For any given selected site, the the wind farm designer could choose turbines that would give a
capacity factor of 1%, or a capacity factor of 80%, or anything in between. It’s an economic
decision. Yes, over a minute or a month, the weather is the primary driver of the capacity factor;
but over the 20-30-year life of a windfarm, almost all those weather variations will average out,
leaving a capacity factor that’s almost entirely driven by the windfarm developer’s trade-offs
between the costs of blades and structure; the cost of the mechanics and electronics in the
nacelle; and the cost of the grid connection. The longer the blades, the more expensive they are,
but they can harness more wind to generate power. The lower the maximum power rating of the
electrics and electronics, the cheaper they are, and the cheaper the connection to the grid will be.
So, a developer could choose a turbine with very long blades, and very low rated electronics, so
it would generate its peak rated power pretty much any time when there was more than a breeze,
and that would result in a capacity factor at a typical onshore wind site of about 80%, but it
would deliver much less electricity almost all the time. On the other hand, at the very same site,
the developer could instead choose to have shorter blades, and electronics and grid connection
rated at a very high power. In this case, the turbine will never generate its peak power, and the
average power generated will only be 1% of the rated power: and the electricity it generated
would be at a much higher unit cost, [CITATION And \l 1033 ].

19
Since the WT should work continuously throughout the year (365×24), the annual production
of electricity (AEP) [kWh / year] from one WT, can be represented by the following
expression:
o
359 v max
AEP=(365 ×24 ) ∫ ∫ Pfarm (v ,θ) p( v , θ) dvθ (11)
o
0 0
where: vmax is the maximum possible wind speed at that location and p is the PDF of the wind
condition (speed, direction).
The LCOE is defined as ratio between total generated electricity and the total costs (including
the costs during the exploitation and those at the end of the exploitation period), [CITATION
GCo09 \m USE17 \l 1033 ]:
n
C
∑ ( 1+rt )t
t=1
LCOE=
n E farm
∑ ( 1+r )t t
t=1 (12)
th
where n=20 years is a lifetime period of the WT, Ct -total cost in t year, r- discount rate
predicted for the whole project (we assume r=0.08). LCOE is calculated using standardized
percentage ratios for capital expenditures-investment (CAPEX), operation and maintenance
(OPEX) and project decommissioning expenditures (DEPEX). Important to mention is the
feed in tariff for electrical energy form wind energy in Macedonia, which is 0.089 EUR/kWh
for energy produced from wind fields with maximal installed capacity of 65 MW. In the
calculations for LCOE, an unexpected loss in the production process is estimated with 10 %
of the annual production for each WT.
A Wind Class III turbine is designed for an easy life with average wind speeds up to 7.5 m/s, and
these turbines typically have extra-large rotors to allow them to capture as much energy as
possible from the lower wind speeds they are subjected to, [ CITATION htt182 \l 1033 ] . It is
assumed that this wind class is the most suitable for all selected sites. St. Nagorichane site,
because of its wind energy prolific characteristics, could be suitable for turbines from class II
too.

Table 6. Wind turbines general specifications

cut in Nom. cut out


Nominal Rotor
Hub wind power at wind IEC
WT type power diameter
height [m] speed wind speed speed class
[kW] [m]
[m/s] [m/s] [m/s]
SWT-2.3-108 2300 108 78.3-115 3.5 11-12 25 II B
NORDEX-S77 1500 77 85 3 13 25 III A
VESTAS V52 850 65 36.5-74 4 17 25 III B
WinWinD-1 D65 1020 56 45-75 4 14 25 III C
ENERCON E-53 800 52.9 50-73 2 13 25 III C
VENSYS87 1.5 1500 86.6 75-100 3 12 22 III B

Table 7. Estimated performance data for six WT at five selected sites, [CITATION SIE151 \l
1033 \m Nor07 \m ENE15 \m Ves05 \m Win13 \m VEN]

20
WT 1. SWT-2.3-108 2. NORDEX-S77 3. VESTAS V52
LCOE LCOE LCOE
Site AEP[MWh] CF[%]
[EUR/kWh]
AEP[MWh] CF[%]
[EUR/kWh]
AEP[MWh] CF[%]
[EUR/kWh]
Berovo 4751.96 23.59 0.0803 2931.03 22.31 0.1118 1501.7 20.17 0.1824
Mogila 2636.64 13.09 0.1448 2868.6 21.83 0.1143 1471.8 19.77 0.1861
Sopishte 5121.1 25.42 0.0745 3788.02 28.83 0.0868 1943.7 26.10 0.1409
St. Nagor. 8813.887 43.75 0.0433 9162.4 69.73 0.0358 4323.8 63.44 0.0579
Sv.
6402.7605 31.78 0.0596 6646.2 32.99 0.0493 3410.6 16.93 0.0803
Nikole
WT 4. WinWinD-1 D65 5. ENERCON E-53 6. VENSYS87 1.5
LCOE LCOE LCOE
Site AEP[MWh] CF[%]
[EUR/kWh]
AEP[MWh] CF[%]
[EUR/kWh]
AEP[MWh] CF[%]
[EUR/kWh]
Berovo 985.303 11.04 0.3327 1787.7 25.11 0.1537 4193.88 31.92 0.0782
Mogila 1107.1 12.40 0.2962 1701.6 23.98 0.1610 3868.15 29.44 0.0847
Sopishte 1456.3 16.31 0.2252 2245.8 31.65 0.1220 5131.3 39.05 0.0639
St. Nagor. 4056.8 45.45 0.0808 5190.2 0.0528 73.15 11430.4 86.99 0.0287
Sv.
2778.2 13.79 0.1181 3818.7 18.95 0.0717 8540.6 42.39 0.0389
Nikole

GREENHOUSE GASES SAVINGS


In order to emphasize the importance and the opportunity to demonstrate the positive impact
over environment that one wind field could have, we impose an equation for calculating the
saving of greenhouse gases (GHG) emission. After determining the value of AEP, the annual
reduction in GHG emission could be computed, [ CITATION AAl17 \l 1033 ] . Currently, that
combined margin (CM) emission factor for Macedonian electricity grid is estimated at 0.915 t
CO2/MWh, [ CITATION The111 \l 1033 ] . Thus, the following equation had been employed to
compute the reduction in GHG emission:
MWh C O2
GHG= AEP
ann (
. ∙0.915 (t ∙
MWh ) ) (13)
where GHG is the emission reduction in Tone CO2/Year, and AEP is the annual energy
production in kWh/Year. This type of calculations it is logical to be implied when an AEP from a
wind field is known.

CONCLUSION
In this paper, a comprehensive analysis on methods of processing measurement wind data is
done. A Multivariate and Multimodal Wind Distribution (MMWD) model is used, to
represent the distribution of wind conditions (speed, direction and air density), based on
recorded data. The usage of MMWD reduces uncertainty and increases accuracy in further
algorithmic simulations and calculations.
Furthermore, comparisons on measured TI with prescribed values form IEC 61400-1 was
carried out in order to ease the next step- selection of appropriate WT for each location. It is
provided a sample evaluation of the performance indexes for five measurement locations with
six turbine types
Analytical analyzes and predictions presented in the above chapters of the paper (TI) are not
quite in line with the results for AEP, CF and LCOE. It was expected, wind turbine class IIIC
to be the most appropriate for all sites, assuming the similar wind regimes and terrain
conditions, but the analysis in penultimate chapter undoubtedly showed that a turbine from
class IIIB is the most suitable for all five selected sites.

21
It must be emphasized the complexity in the subject, which is again confirmed with this
paper. If we strive to develop precise predictable algorithms for assessment of wind energy
resources, а comprehensive and thorough analysis from multiple aspects and at multiple levels
have to be carried out.

REFERENCES
[1] N. G. Mortensen, M. Nielsen, H. Ejsing Jørgensen, "Comparison of Resource and Energy Yield
Assessment Procedures 2011-2015: What General rightsCopyright and moral rights for the
publications made accessible in the public portal are retained by the authors and/or other
copyright owners and it is a conditio," in Proceedings of the EWEA Annual Event and Exhibition
2015, Paris, France, 2015.
[2] A. Albani, M. Z. Ibrahim, "Wind Energy Potential and Power Law Indexes Assessment for
Selected Near-Coastal Sites in Malaysia," Energies, vol. 10, no. 307, p. 21, 2017.
[3] The Macedonian Academy of Sciences and Arts, "Strategy for the utilization of renewable
sources of energy in the Republic of Macedonia by 2020," The Macedonian Academy of
Sciences and Arts, Skopje, June, 2010.
[4] V. Stoilkov, K. Najdenkoski, Z. Kokolanski, V. Dimcev, "Wind Measurement Campaign in
Republic of Macedonia," in The World Wind Energy Conference and Exhibition, Istanbul,
Turkey, 2014.
[5] K. Demerdziev, Z. Kokolanski, V. DImcev, M. Celeska, K. Najdenkoski, V. Stoilkov, "Wind
Parameters Analysis on Five Locations in Macedonia-Second Measurement Campaign," Journal
of Electrical Engineering and Information Technologies - JEEIT, vol. 1, pp. 1-7, 2016.
[6] D.K. Kaoga, D. Raidandi, N. Djongyang, S.Y. Duka, "Comparison of five numerical methods for
estimating Weibull parameters for wind energy applications in the district of Kousseri,
Cameroon," Asian Journal of Natural & Applied Sciences, vol. 3, pp. 72-87, 2014.
[7] M. Celeska, K. Najdenkoski, V. Dimchev, V. Stoilkov, "Analysis of Appearance, Variations and
Correlations of Monthly Mean Wind Speed Data," in IEEE EUROCON2017, Ohrid, Rep. of
Macedonia, 2017.
[8] T. Burton, D. Sharpe, N. Jenkins, E. Bossanyi, Wind energy handbook, Wiley, 2001.
[9] M. Celeska, K. Najdenkoski, V. Stoilkov, A. Buchkovska, Z. Kokolanski, V. Dimchev,
"Estimation of Weibull Parameters from Wind Measurement Data by Comparison of Statistical
Methods," in IEEE EUROCON2015, Salamanca, Spain, 2015.
[10] J. Zhang, S. Chowdhury, A. Messac, L Castillo, "Multivariate and Multimodal Wind Distribution
Model Based on Kernel Density Estimation," in ASME 2011 5th International Conference on
Energy Sustainability & 9th Fuel Cell Science, Engineering and Technology Conference,
Washington, DC, August 7-10, 2011.
[11] J. Zhang, S. Chowdhury, A. Messac, L Castillo, "A Multivariate and Multimodal Wind
Distribution Model," Renewable Energy, vol. 51, pp. 436-447, 2013.
[12] E. Parzen, " On estimation of a probability density function and mode," The Annals of
Mathematical Statistics, vol. 33, no. 3, pp. 1065-1076, 1962.
[13] J. Feng, W. Zhong Shen, "Modelling Wind for Wind Farm Layout Optimization Using Joint
Distribution of Wind Speed and Wind Direction," Energies, vol. 8, pp. 3075-3092, 2015.
[14] Mohammed A. Saleh, Ayad A. Ani, Firas A. Hadi, "Ambient Turbulence Intensity Calculation
for Al-Nasiriyah Province in Iraq," Iraqui Journal of Science, vol. 55, no. 2A, pp. 561-571, 2014.
[15] International Electrotechnical Commission, Wind Turbines - Part 1: Design requirements,
Switzerland, 2005.
[16] K. Sunderland, T. Woolmington, M. Conlon, J. Blackledge, "Small wind turbines in turbulent
(urban) environments: A consideration of normal and Weibull distributions for power
prediction," Journal of Wind Engineering and Industrial Aerodynamics, vol. 121, pp. 70-81,
2015.

22
[17] M. H. Zhang, Wind Resource Assessment and Micro-siting. Science and Engineering., China:
John Wiley & Sons Singapore Pte Ltd., 2015.
[18] International Electrotechnical Commission, Production Based Availability for Wind Turbines,
Geneva, Switzerland, 2013.
[19] S. Pfaffel, S. Faulstich, K. Rohrig, "Performance and Reliability of Wind Turbines:A Review,"
Energies, vol. 10, no. 1904, pp. 1-27, November 2017.
[20] Andrew, "http://energynumbers.info/," 29 September 2014. [Online]. Available:
http://energynumbers.info/capacity-factor-of-wind. [Accessed 14 March 2018].
[21] G. Cornelis van Kooten, Govinda R. Timilsina, "Wind Power Development Economics and
Policies”-Report, The World Bank Development Research Group Environment and Energy Team
, Washington, DC, March, 2009.
[22] U.S. Energy Information and Administration, "Levelized Cost and Levelized Avoided Cost of
New Generation Resourcesin the Annual Energy Outlook 2017," Indpendent Statistic and
Analysis, April, 2017.
[23] "http://www.renewablesfirst.co.uk/," Brimscombe Hill, UK, [Online]. Available:
http://www.renewablesfirst.co.uk/windpower/windpower-learning-centre/what-is-the-wind-class-
of-a-wind-turbine. [Accessed 14 March 2018].
[24] SIEMENS, Wind Turbine SWT-2.3-108 Technical specifications, Hamburg, Germany: Siemens
AG 2015, 2011.
[25] NORDEX. AG, S70/1500 KW S77/1500 KW The perfect technology for each location,
Norderstedt, Germany: Nordex AG, 2007.
[26] ENERCON, ENERCON product overview, Aurich, Germany: ENERCON Energy for tthe world,
June, 2015.
[27] VESTAS. A/S, V52-850 kW The turbine that goes anywhere, Randers ,Denmark: Vestas Wind
Systems A/S, December, 2005.
[28] WinWind.-1. D60, "Wind-turbine-models," 23 February 2013. [Online]. Available:
https://www.en.wind-turbine-models.com/turbines/486-winwind-wwd-1-d60. [Accessed 15
March 2018].
[29] VENSYS AG, Technical specification VENSYS 87 1.5 MW, Neunkirchen, Germany: VENSYS
Energy AG.
[30] The European Union's CARDS Programme for the Western Balkans, "Wind Park Development
Project Macedonia - Feasibility Study Bogdanci A," WYG International Ltd, March 2011.
[31] R. Robichaud, J. Fields, R. J. Owen, "Wind Resource Assessment," NREL/TP6A20-52801,
February 2012.

23

You might also like