• No results found

Comparison of optimization for non linear and linear wind resource grids

N/A
N/A
Protected

Academic year: 2022

Share "Comparison of optimization for non linear and linear wind resource grids"

Copied!
82
0
0

Loading.... (view fulltext now)

Full text

(1)

COMPARISON OF OPTIMIZATION FOR NON LINEAR AND LINEAR

WIND RESOURCE GRIDS

Ion Dragoi

Submitted to the Office of Graduate Studies of Uppsala University campus Gotland

in partial fulfillment of the requirements for the degree of MSc Wind Power Project Management, Master Thesis 15 ECTS

Supervisor: Dr Bahri Uzunoglu Examiner: Prof. Jens N. Sørensen

Master of Science Programme in Wind Power Project Management, Department of Wind Energy,

Uppsala University campus Gotland Cramérgatan 3

621 57 Visby, Sweden

(2)

COMPARISON OF OPTIMIZATION FOR NON LINEAR AND LINEAR WIND

RESOURCE GRIDS

A Thesis By Ion Dragoi

Submitted to the Office of Graduate Studies of Uppsala University campus Gotland

in partial fulfillment of the requirements for the degree of

MSC WIND POWER PROJECT MANAGEMENT June 2013

Major Subject: "Energy Technology"

Master of Science in Wind Power Project Management

2013, Ion Dragoi

(3)

COMPARISON OF OPTIMIZATION FOR NON LINEAR AND LINEAR WIND

RESOURCE GRIDS

A Thesis By Ion Dragoi

Submitted to the Office of Graduate Studies of Uppsala University campus Gotland

in partial fulfillment of the requirements for the degree of

MSC WIND POWER PROJECT MANAGEMENT

Approved by:

Supervisor, Dr Bahri Uzunoglu

Examiner, Prof. Jens N. Sørensen

June 2013

Major Subject: "Energy Technology"

(4)

ABSTRACT

The aim of this thesis is to assess how the configuration of linear and non-linear wind resource grids impacts the optimization.

Three different software tools are used for this study: WAsP (linear model) included in WindPRO, and WindSim (a non-linear model) - a CFD tool, and WindPRO for the optimization. With the same configuration for wind resources, WAsP and WindSim will run to calculate the wind resource grids, .rsf or .wrg format, which will be compared in the post processing tab of WindPRO (from CFD interface).

Using different optimization algorithms, the results from two software will be compared. The test site is flat terrain in the sea with no complexity (0,0002 roughness and no orography or obstacle), and the chosen turbine here is Enercon 40.3 (55m hub height, with the rated power at 14 m/s), and the wind is coming from one direction, in our case North, which means sector 0.

After comparison of the resource files from linear and non-linear wind resource grids, the optimization and comparison is ran for the two wind resource grids (linear and non-linear). The results of the optimization are also compared with optimization results of Eftun Yilmaz’s thesis (Eftun Yilmaz, 2013). We can see from the results that WindSim gives almost 40% bigger values for the production. The results are comparable with findings of Eftun Yilmaz thesis.

Keywords: Wind resource grid, Flat terrain, Boundary layer, Wind, WAsP, WindSim, WindPRO, Optimization, Linear and Nonlinear wind resources grid.

(5)
(6)

ACKNOWLEDGEMENTS

First, I would like to thank to my supervisor Associate professor Dr. Bahri Uzunoglu for his supervision and help, and to the teachers and staff at the energy section from Gotland University (now Uppsala University campus Gotland), for their support and guidance throughout the year.

I want to thank to my parents Gheorghe and Maria, from all my heart, for the overall support during this time and family members Mihaela, Aurelian, Mihai, for my friends who helped me PS Macarie, Horatiu, Avram, Anisoara, Catalin, Dan, Grigore, David and Erik.

Last but not least I would like to thank to all my classmates for this beautiful time spent together in this beautiful city and island.

Thank you very much to everyone.

Ion Dragoi Visby, June 2013

(7)

TABLE OF CONTENTS

ABSTRACT... 4

ACKNOWLEDGMENTS... 6

TABLE OF CONTENTS... 7

LIST OF FIGURES... 10

LIST OF TABLES... 13

CHAPTER I. INTRODUCTION AND LITERATURE REVIEW ... 14

Scientific methodology... 15

II. ATMOSPHERIC BOUNDARY LAYER. MARINE ATMOSPHERIC BOUNDARY LAYER... 17

The importance of ABL. Introduction... 17

Vertical structure of ABL………. 17

Diurnal cycle of ABL……….. 18

Friction………... 21

Logarithmic wind profile. Monin-Obukhov similarity theory. Temperature………. 22

Offshore winds... 26

Land sea winds……….. 27

Low level jets………. 27

III. DIFFERENCES BETWEEN WASP AND WINDSIM, WIND RESOURCE GRID METHODOLOGY... 28

(8)

Software tools and optimization module... 28

WAsP utility software……… 29

WindSim utility software………. 29

WindPRO utility software……….. 29

Software tools, differences between WAsP and WindSim and Wind resource grid……… 30

Problem set-up Wind Resource Map………. 31

Background data for the Wind Resource Map………... 31

Optimization and Object of optimization……… 31

Overview to Objective function………. 32

IV. OPTIMIZATION RESULTS... 34

Test case set-up………... 34

Location on the earth………... 35

Terrain conditions... 37

Wind conditions………..……….………... 38

Turbine type……… 43

Wind Resource Grid generation……… 44

Wind Resource Grids comparison……….. 55

Optimization set-up and results based on optimization of linear and nonlinear wind resource grids………. 60

Comparison Optimization results with Eftun Yilmaz’s optimization results……… 78

V. CONCLUSION... 82

REFERENCES... 83

(9)

LIST OF FIGURES

Page Figure 1 - Atmospheric boundary layer (Lyndon State College Atmospheric Science,

2001)………. 18

Figure 2 - Vertical structure of ABL (Paul R. Houser, 2012)……… 18

Figure 3 - Atmospheric boundary layer structure Vertical temperature gradient (NikNaks from Stull, 1988)………. 19

Figure 4 - Left – ABL profile in fair weather, in mid-latitudes, Right – Eddy structure and ABL circulation (University of Reading, 2013)…… 20

Figure 5 - Friction within ABL (Symscape, 2013)……….……… 21

Figure 6 – Temperature conditions height (University of reading, 2013)……….……… 25

Figure 7 - MABL vertical structures over a wavy sea surface (S.Emeis, 2013) 26

Figure 8 - Park Map... 35

Figure 9 - Basis – Project data overview……… 36

Figure 10 - STATGEN – Terrain files... 37

Figure 11 - Wind statistic info... 38

Figure 12 - STATGEN – Mean wind speed, Wind energy and WTG energy 39

Figure 13 - Wind data analysis... 40

Figure 14 - WAsP – Wind profile detailed... 41

Figure 15 - WAsP – Wind profile detailed 2... 42

Figure 16 - WAsP – Main results... 43

Figure 17 - Resource – Main results. Generate .rsf file……… 44

Figure 18 - Terrain elevation………. 45

Figure 19 – Digital terrain conversion with the properties……… 46

Figure 20 - Wind fields sector 0 spot values………... 46

Figure 21 – Wind fields sector 180 spot values………. 47

(10)

Figure 22 - Wind fields report……… 47

Figure 23 - Objects. Climatology conversions……… 47

Figure 24 - Park layout……….. 48

Figure 25 – Velocity report sector 0…...………. 49

Figure 26 – Velocity report sector 180……… 50

Figure 27 – Climatology report 1 (Frequency in %)………. 51

Figure 28 – Climatology report 2 (Wind speed in m/s)……… 51

Figure 29 – Wind Resources. Variable: Mean wind speed 2D (m/s) 52 Figure 30 – Wind Resources. Weibull shape parameter………. 52

Figure 31 – Wind Resources. Variable: Weibull scale parameter, A 54 Figure 32 – WindSim wind resource grid .rsf and .wrg files are generated 54 Figure 33 – WRG file………. 55

Figure 34 – CFD wind resource grid comparison model form WindPRO 56

Figure 35 – Wind resource grids comparison 1 (ratio between mean wind speed values WAsP/WindSim)……….. 57

Figure 36 – Wind resource grids comparison 2 (ratio between mean A weibull parameter values WAsP/WindSim)……….. 58

Figure 37 – Wind resource grids comparison 3 (ratio between mean k weibull parameter values WAsP/WindSim)………. 59

Figure 38 - Park calculation – Main results, page 1………. 62

Figure 39 - Park calculation – Main results, page 2………. 63

Figure 40 - Park – Production analysis………. 64

Figure 41 - Park – Power curve analysis………. 65

Figure 42 - Park – Wind data analysis……….. 66

Figure 43 - Optimize wind farm layout 1……….. 68

Figure 44 - Optimize wind farm layout 2………... 69

Figure 45 - Optimizer controller……… 69

Figure 46 - Optimization WAsP random pattern results……….. 70

(11)

Figure 47 - Optimization WAsP Random pattern map………. 71 Figure 48 – Optimization WindSim random pattern results 1…………. 72 Figure 49 - Optimization WindSim random pattern results 2………….. 73 Figure 50 – Optimization WindSim Regular pattern results 1…………. 74 Figure 51 – Optimization WindSim Regular pattern results 2…………. 75 Figure 52 – Optimization WAsP Regular pattern results 1……… 76 Figure 53 – Optimization WAsP Regular pattern results 2……… 77 Figure 54 – Optimization Regular pattern map………. 78

(12)

LIST OF TABLES

Page

Table 1 - Vertical profile climatology for sector 0 (N)……… 55 Table 2 – Regular pattern optimization results comparison

(Eftun Yilmaz, 2013)……… 79 Table 3 – Random pattern optimization results comparison

(Eftun Yilmaz, 2013)……… 80 Table 5 – Random pattern optimization final results comparison

(Eftun Yilmaz, 2013)……… 81

(13)

CHAPTER I

INTRODUCTION AND LITERATURE REVIEW

The aim of this Master thesis is the comparison of optimization for two wind resource grids, linear and non-linear. The configuration of the wind resource grid is playing an important role in the optimization calculation.

Nowadays cost effective wind parks need to be developed and to reach this goal, there are some important steps to be followed and factors to be taken into account, like the placement of the turbines in the park area, the number and type of turbines to be used, investment costs, park efficiency, terrain and location and not the least, wind conditions (Mustakerov et al, 2010).

Wind farm design industrial software tools such as WindSim and WindPRO have implemented algorithms for the energy optimization and the impact of the wind resource grid on optimization in these two software will be investigated in this study.

The energy capture has to be maximized for a wind park layout and the new wind farms needs to be improved for the better utilization of the land. Maintenance and operation costs are neglected in this study.

By this thesis, we address the study of two wind resource grids with two software, using the same inputs for the wind resource data for linear and non-linear grids and we want to observe what is the impact on optimization of these wind resource grids’

configuration.

Question Formulation

How the configuration of the wind resource grid impacts the optimization?

(14)

Scientific Methodology

The scientific methodology followed the next steps:

Theoretical Part covers theory about atmospheric boundary layer and marine atmospheric boundary layer, boundary conditions, wind resource grid configurations differences between WAsP and WindSim and optimization algorithms based on literature study.

Simulation Part includes the work on wind resource grid data created in two file formats .rsf and .wrg for different (linear, WAsP and non linear, WindSim) wind farm software tools.

In the Experimental and Analysis Part the data and simulation results are presented. A comparison has been conducted between optimization results with both wind resource grid models.

Thesis addresses the following topics:

In Chapter 2, the basic theory behind atmospheric boundary layer is given.

There will be also briefly stated the importance, conditions and structure of ABL, logarithmic wind profile, Monin-Obukhov similarity theory, temperature and offshore winds.

In Chapter 3, software tools, wind resource grid and optimization module is presented, regarding to their working principles. WAsP 10 is used for the preparation part; EMD WindPRO 2.8 and WindSim 5.1.0 are used in the simulation part. The simulation process for optimization in terms of wind resource grid for linear and non linear grid is outlined and described. Separate results are presented for WindSim and WindPRO for same configuration. Afterwards, the results are compared.

(15)

In Chapter 4 and Chapter 5, the optimization results from the simulated cases will be presented, analyzed and reviewed in context theory in discussion.

In the conclusion section, we can see that the coding system and the algorithms used by the software are different and the results may differ, even for a simple case with the same input data.

To realize if software is overestimating or underestimating the results, or to see which software is more suitable for these conditions and giving the right results, the predicted data should be compared with real data for a real case (wind farm).

(16)

CHAPTER II

ATMOSPHERIC BOUNDARY LAYER AND MARINE ATMOSPHERIC BOUNDARY LAYER

This chapter presents the theory regarding Atmospheric Boundary Layer, Marine Atmospheric Boundary Layer, Logarithmic law, Monin-Obukhov similarity theory and Temperature.

The importance of ABL – Introduction

The clouds from ABL are very important for climate, which means that ABL is important also for local forecasting, in this area takes place the pollution dispersion and the fluxes are mediated, ABL has an effect on the rest of the atmosphere (because of the turbulences and stronger friction) and the most important thing is that humans are living in the boundary layer. ABL is not affected by the geostrophic winds (University of Reading, 2013).

Ludwig Prandtl described first in 1904 the boundary layer concept as a flow of a moving fluid who can be split in two regions when meets a solid boundary and the atmospheric boundary layer has many properties who can define it, such as: around 1 km deep, but in the mid-latitudes can vary between 100 m – to 3 km; in comparison with the free atmosphere above, the temperatures vary diurnally (University of Reading, 2013).

(17)

Figure 1 - Atmospheric boundary layer (Lyndon State College Atmospheric Science, 2001)

Vertical structure of ABL

The atmospheric boundary layer can be defined as that part of the troposphere that is directly influenced by the presence of the earth‘s surface, and responds to surface forcing with a timescale of about an hour or less (Stull 1988).

From the surface to the top, the boundary layer can be divided in the following parts:

• Interfacial layer (0-1 cm): molecular transport, no turbulence

• Surface layer (0-100 m): strong gradient, very vigorous turbulence

• Mixed layer (100 m - 1 km): well-mixed, vigorous turbulence

• Entrainment layer: inversion, intermittent turbulence (Lindon College).

(18)

Figure 2 - Vertical structure of ABL (Lindon College)

Diurnal cycle of ABL

The structure of ABL is well defined and it evolves with the diurnal cycle:

(19)

Figure 3 - Atmospheric boundary layer structure Vertical temperature gradient (NikNaks from Stull, 1988)

The surface energy balance equation is:

(The small thickness of the active layer makes the storage of energy in the layer to be neglected).

0

H E G

Rn ,

E H G

Rn    ,

R = Net irradiance into the surface, n

G = Ground flux heat density, H = Sensible heat flux density,

E = Latent heat flux density (University of Reading, 2013).

(20)

Figure 4 - Left – ABL profile in fair weather, in mid-latitudes

Right – Eddy structure and ABL circulation (University of Reading, 2013)

In figure 4 can be seen how the ABL is structured during a diurnal cycle and how the ABL can be divided in sub-layers during a clear day. First is the roughness sub-layer which is close to the earth’s surface and it is influenced by the roughness (the air flows around different obstacles); the second sub-layer is the surface layer also called constant flux layer (varies from 100 to 200 m height) where the parameters (humidity, wind, temperature) vary with altitude and is affected by the surface (turbulence); the wind direction change with height; the fourth sub-layer is the capping inversion the third sub-layer is the well-mixed layer, which is influenced by the earth’s rotation and, where the temperature inversion is capping the convective boundary layer. This happens during the day, when the sun is heating the ground and the air at the surface. During night, when the air is cooling from the surface, a new stable nocturnal boundary layer grows and the daytime mixed layer remains as residual layer and the capping inversion is eroded as we can see also in figure 3 (University of Reading, 2013).

(21)

The ABL Diurnal cycle is important to know because people is living inside the ABL, also for the understanding of daily weather forecasts, human activities, air mass generation, atmospheric kinetic energy and for specially for wind turbines extract energy from ABL winds (Lindon State College, 2001).

Friction

The friction caused by surfaces creates heat fluxes at the ground and influences the atmospheric boundary layer; the wind shear creates turbulence (temperature gradient influence turbulence) which characterize the ABL (University of Reading, 2013).

Because of the Earth’s surface, which is not smooth or frictionless and influences directly the movement of the air masses from the lower part of the troposphere (ABL) and responds to this influences created by obstacles and Earth’s surface with a time scale of about an hour (Lyndon State College Atmospheric Science, 2001).

Figure 5 – Friction within ABL (Symscape, 2013)

(22)

Logarithmic wind profile. Monin-Obukhov similarity theory. Temperature

Wind energy generation is influenced by the wind speed, which increases with height. The logarithmic wind profile has different descriptions and the classical logarithmic wind profile will be described here (S.Emeis, 2013).

The stratifications conditions of an atmospheric boundary layer are important and there are three conditions stratifications such as: Neutral, Unstable and Stable.

In Neutral stratification layer, starting from the vertical wind speed gradient (wind shear) equation:

kz u l u z

u * *

 

, can be found the logarithmic wind profile for this layer:

0

*ln )

( z

d z k z u

u

, where z = roughness length, d = displacement height, k = van 0 Karman constant = 0,4; z = height, u = wind speed, u = friction velocity (S.Emeis, 2013).

Depending on the surface roughness we can identify the turbulence intensity in the surface layer neutrally stratified:

) / ln(

1 )

) ( (

z0

z z

z u

Iuu

,

A thermal stratification surface layer can be rarely found in the unstable stratification layer, and a heat flux and friction velocity (Monin-Obukhov length) can be formed.

' '

3

*

* w

u L kg



, = virtual potential temperature, L = length scale, * u = friction velocity, *'w'

=

(23)

virtual potential heat flux, g = gravity, w = vertical wind component, k= van Karman constant = 0,4.

If it’s directed from the atmosphere towards the ground the heat flux is considered positive and it is cooling the ground, and vice versa it is considered to be negative, it is heating the atmosphere (S.Emeis, 2013).

A negative Monin-Obukhov length is characterizing an unstable surface layer.

The Bowen ratio, B, represents the ratio between the turbulent sensitive heat flux and humidity flux:

' '

' '

w q L

w B cp

 

, cp

= specific heat, q = specific humidity, L

= latent heat of vaporization,

' 'w

= virtual potential heat flux (S.Emeis, 2013).

The buoyancy ratio is the ratio between vertical heat and humidity gradients and is inversely proportional to the Bowen ratio, B:

B L

c w

w

BR 0,61 q 0,61 p 1

' '

' '

 

 

The surfaces heated by the sun and over waters, which are warmer than higher up, (because of the big heat storage capacity of the sea) can create an unstable surface layer during daytime. As a stability parameter is used the ratio between Monin- Obukhov length L and height z. The warm air from the surface rise to the top and create an unstable layer, and according to the adiabatic lapse rate, the temperature decreases with height and the lapse rate near the surface is even stronger. The height of an unstable layer is determined by z (boundary layer height). (S.Emeis, 2013) i

In an unstable surface layer, the Monin-Obukhov length L is an additional length scale. If the stability parameter z/ L is negative, then there is an unstable

(24)

stratification while the parameter is positive, then there is a stable stratification and when it is zero, then there is a neutral stratification (S.Emeis, 2013).

For the vertical wind profile in the surface layer and for small z/L negative values a correction function it is introduced:

) 2 ( 2 2

ln 1 2

ln 1 2

2





  



 

  

x x arctg x

m

,

Where 4

1

*) / 1

( bz L

x  , and b=16. m

= integral stability correction function.

The vertical wind profile equation will be:

)) / ( ) / (ln(

/ )

(z u* k z z0 z L*

u  m

If in the convective boundary layer the vertical motion, thermally induced through the layer, depends on the height, then there will be another stability parameter zi/L (S.Emeis, 2013).

The horizontal and vertical wind components have the standard deviation independent (the horizontal) and increases with height (vertical) in an unstable surface layer (10 Hz fluctuations of wind components in unstably stratified Ekman layer above the Prandtl layer) (S.Emeis, 2013)

* ,

,vw 0 w,6

u

,

w v u ,,

= standard deviation of wind components, w = convective velocity scale. *

In a Stable stratification, the heat flux L >0 and it is directed to the surface. A stable condition of the air stratification can be found usually during night time, over snow-covered, ice or water surfaces that are colder than the air above.

z/L factor should be positive and there should be meet some conditions, such as: *

(25)









7 / 5 , 0 , / ) / exp(

) / /

( /

5 , 0 / 0 , / )

/ (

*

*

*

*

*

*

*

L z for D BC L

Dz

D C L z B L Az

L z for L az L

m z

, where a=5, A=1, B=2/3, C=5, D=0,35.

m

= integral stability correction function, z = vertical coordinate, L = length scale, a = reduction factor, constant in stability correction function, power-law exponent; A

= scale factor in Weibull distribution, B = Bowen ratio, C = drag factor, D = distance (S.Emeis, 2013).

In the stable boundary layer there will be following conditions:

- the potential temperature will increase with height;

- the air temperature vertically decreases less than the adiabatic lapse rate (S.Emeis, 2013).

Figure 6 - Temperature conditions height (University of Reading, 2013).

The empirical power law can be used to describe the vertical wind profile:

a

r

r z

z z u z

u 



 ( ) )

(

,

Where z = reference height, a = power law exponent (Hellmann exponent), depends r on the surface roughness and the thermal stability of the Prandtl layer (S.Emeis, 2013).

(26)

Between logarithmic and power law the choice is made by practical arguments and the power law is mathematically simpler. The thermal stability is described differently in these two laws (S.Emeis, 2013).

Offshore winds

Offshore wind parks number is increasing all over the world and the profile of winds and turbulence over the sea is very important to be known. The sea surface may be flat, but there are some differences between the sea flat surface and the homogenous land surface and the sea surface is smoother than the land surface. It results that the wind speed at a certain height is higher on the sea than on the land and with smaller turbulence intensities and the surface layer depths are shallower and the turbines over the rotor area meet less wind shear (S.Emeis, 2013).

The waves are playing an important role in the wind speed over the sea. The water has the capacity to store a large amount of heat and the diurnal cycles of temperature are almost absent. The marine atmospheric boundary layer (MABL) has unstable layer due to the high amount of humidity that can be found. (S.Emeis, 2013)

Figure 7 - MABL vertical structures over a wavy sea surface (S.Emeis, 2013)

(27)

Due to the wind speed (which is generating the waves through the frictional forces created when it touches the sea surface), the roughness of the sea surface is not constant and the surface roughness length increases with the wind speed. In general, depending on the state of the sea (wave steepness or slope, or wave age) could be observed differences in the measurements of the drag coefficients (S.Emeis, 2013).

The wind conditions and the wave age (wind speed and the phase speed of the wave’s ratio) depend on atmospheric conditions and the properties of the wave field.

The surface friction is low and the wind shear over the seas is low also, due to thermal winds (which makes the wind speed to increase with height) (S.Emeis, 2013).

Land sea winds

In the day time the land surface heats from the sun radiation which makes the air on land to be warmer than the air at the sea surface and moves the cold air from sea to land while the warm air on land rises above the land. During the night time the process inverts and the air from land cools down while the temperature on the sea keeps warm and the cold air from the land flows from the warm air masses from the land flows to the sea. On every coast these land-sea winds are well-known

(Meteoblue, 2013).

Low level jets

The low-level jet is a fast moving mass of air which is transporting moisture and warm temperatures to the North, in the low level of the atmosphere. They are divided in 2 parts: nocturnal low level jet and the mid-latitude cyclone induced low level (Jeff Haby).

(28)

CHAPTER III

DIFFERENCES BETWEEN WASP AND WINDSIM, WIND RESOURCE GRID METHODOLOGY

We willl do a control experiment to compare 2 software tools, so we will do a control experiment exercise to demonstrate the differences between 2 WRG. The aim of this thesis is to see how the configuration of linear and non-linear wind resource grids impacts optimization.

Three different software tools are used for this study: WAsP (linear model) included in WindPRO, and WindSim (a non-linear model) - a CFD tool, and WindPRO for the optimization report. With the same inputs, WAsP and WindSim will run to calculate the wind resource grids, .rsf or .wrg format, which will be compared in the post processing tab of WindPRO (from CFD interface). Using different algorithms, the two software results are compared.

After comparison of the files, the optimization and comparison is run and the results of the optimization is compared with optimization results of Eftun Yilmaz (2013).

The simulations were conducted for same wind resource grid, different software.

Software Tools and Optimization Module

In this study, two wind resource grids with two algorithms are studied, using the same inputs for the wind resource data for linear and non linear grids. It has to be observed the impact of the configuration of these wind resource grids on

optimization.

(29)

WAsP Utility Software

In the field of Wind Resource Assessment, WAsP (Wind Atlas Analysis and

Application Program) is the standard software tool used in industry for predicting the power productions developed by wind farms, using data for wind measured at

stations that are located close to the studied field. The things that have to be taken into account when it is used WAsP are the roughness, obstacles and complex terrain flow for which WAsP engineers have developed a model to predict the wind

resources, climates and power productions for wind turbines used in wind farms (DTU Wind Energy, WAsP, 2013).

It is software based on the linear flow model used to estimate the efficiency of wind farms and their production, to map the wind resources, for digital maps and gives information about topographic and orographic effects (DTU Wind Energy, WAsP, 2013).

WindSim Utility Software

WindSim is leading software, very powerful, which combines advanced numeric processing with compelling 3D visualization, based on computational fluid dynamics (CFD) which has a user-friendly interface (WindSim, 2013).

It is a modern wind farm design tool based on a non-linear flow model used in wind resource and assessment to design more profitable wind farms (WindSim, 2013).

WindPRO Utility Software optimization

WindPRO is a software tool developed by EMD International A/S in Aalborg, Denmark, and it is designed for wind energy calculations (annual energy production

(30)

and efficiency estimations of wind farms), to digitalize maps, noise and shadow calculations, optimization, photomontages and simple energy estimations for single wind turbines and wind farm using different modules for developing, assessing, designing, planning, sale and approval of wind energy projects (EMD, WindPRO, 2013).

Software Tools, Differences between WAsP and WindSim and Wind resource grid

WAsP is linear model and fast computational time software in comparison with WindSim which is computational time demanding and need large memory space to save the working files. WAsP is based on the linear flow model and it is estimating, as it was said before, the efficiency of wind farms and their production, to map the wind resources, for digital maps and gives information about topographic and orographic effects (DTU Wind Energy, WAsP, 2013).

WindSim is non linear model and it is based on solving the Reynolds Averaged Navier-Stokes flow equations for the non linear transport equation of momentum, mass and energy. WindSim uses a digital terrain model (used as an input basis is on a proper length scale), a BFC (body fitted coordinates - with refinement towards the ground). It starts to run with a logarithmic wind profile and can perform nested runs with very good resolution (Pep Moreno, Arne R. Gravdahl, & Manel Romero, Wind Flow over Complex Terrain: Application of Linear and CFD Models).

The WindSim package contains several turbulence models and from their experience, the model which works best under most conditions is the standard k- epsilon model and it produces the most reliable results. The modified k-epsilon model is better modeling wind flows in a neutral atmospheric boundary layer. The 3- d turbulence model included is the RNG k-epsilon model. (WindSim 2012)

(31)

As it will be seen, the simulated energy production for low complex terrain is not identical for WAsP and WindSim calculation.

Problem setup

Wind Resource Map

Wind resource map is used as a tool for evaluating the wind energy potential in an area, to plan wind energy projects and search for good sites, to optimize the wind farm layout.

WindPRO can process very large areas and unlimited number of .map files and work with several wind statistics using WAsP as a calculation engine (WindPRO, 2013).

Background Data for the Wind Resource Map

The following data is needed to generate a Wind Resource Map:

· A roughness maps for the calculation area

· Height contour line maps for the calculation area

· Wind Statistics valid for the calculation area.

· Optional: Obstacles in the area.

These will be included in the Site Data Object. (WindPRO, 2013)

Optimization and Object of optimization

The Optimization calculation module used in WindPRO to run for both WRG is with the same parameters. The results are compared as it will be seen in the next chapter.

The optimization is used for the 2 WRG files to calculate the optimal turbine configuration of layouts. Energy optimization objectives like the total cost/year for the entire wind park, including maintenance and operation costs (which are neglected

(32)

in this study) and turbine investment costs which were implemented in algorithms of WindPRO software tool (Eftun Yilmaz, 2013).

For the offshore turbine micrositing there have been used the most the following optimization methods classified as: greedy heuristic methodology, genetic algorithms, and gradient search, swarm optimization, simulated annealing and evolutionary algorithms (for the boundaries) such as maximization of energy production regarding the wake model and minimization of the cost functions (Eftun Yilmaz, 2013).

Overview to Objective Function

For the optimization part it is very important to choose a good objective function and this can be done by maximizing or minimizing a developing objective function choose for the optimization problem and the best way to do it is to minimize the cost of energy or to maximize the energy production of a wind farm in three possible ways (Eftun Yilmaz, 2013).

First is to minimize the cost per energy unit, which means balancing the minimum cost and maximum energy (to have maximum energy production at a low cost), second is to maximize the production of energy which is available and used by some software optimization packages and it means to doesn’t take into account the cost of the produced energy and to produce the maximum possible energy ( it is focusing only on the production of energy) and third is to maximize the profit, which is an economical approach (needs to know tax incentives packages, credits used in

renewable energy production, the price of the electricity in the grid), but it is the best for investors and the owners of wind farms (Wagner et al, 2012).

(33)

The minimum cost per unit energy is the most used and widely accepted criterion for wind park design and the optimization proceed based on following objective

function, Cost/P, where P (Park Yield) is the total power output from all the N turbines in the wind farm (Mustakerov et al, 2010).

Maintenance and operation costs where neglected in this study.

An important factor when is calculated the total cost of the wind turbines is only to take into account the number wind turbines (Mosetti et al, 1998). They also consider that if a maximum number of wind turbines are installed, the best way for an

objective function like minimizing the cost of a wind turbine with a reduction of 1/3 in cost/year for each turbine added and the total cost/year for the wind farm can be calculated with the following equation (Elkinton CN et al):

Costs= 2/3+ 1/3exp (-0.00174N^2)]

In this thesis the parameters used are constant wind from one direction (North) and the same wind turbine type used.

As it was said before, the maintenance and operation costs are neglected in this study, which is not so good for the total investment cost (Eftun Yilmaz, 2013).

(34)

CHAPTER IV

OPTIMIZATION RESULTS

The objective of this study is to see how the configuration of wind resource grid impacts the optimization. The set-up of the test case is as follows.

Test Case Set-up

A regular grid for with uniform height contour and roughness is chosen. A wind climate data with a single wind distribution for a single wind speed and direction is employed for the computational domain.

In this study, the following set-up has been used for optimization,

Location on the earth

The geographical area was chosen regardless of climatic impact of wind conditions.

Plus, coordinate system of UTM with WGS84 Datum in zone 24 was selected without correlation to actual earth coordinates.

In the figure number 8 we have the configuration of the wind park with 10 Enercon turbines and the figure number 9 includes the basis of the project data overview, with the name of the country, site center, the coordinates system, the localization of each of the 10 Enercon turbines, meteorological data, line objects and WTG area objects.

(35)

Figure 8 - Park Map

(36)

Figure 9 - Basis – Project data overview

(37)

Terrain conditions

Domain boundaries are the limits of the map spanned by the design variables of height contours of 0 m and roughness length of 0.0002 m were taken. Figure number 10 includes terrain data: Roughness and Orography data with the properties of the roughness and orography map files.

Figure 10 - STATGEN – Terrain files

(38)

Wind conditions

Wind conditions: Basic wind conditions with constant wind speed magnitude and single direction were taken. The wind speed is 18m/s coming from the north direction. The next figure (11) includes wind statistics info with main data and additional info for wind statistic (coordinates of the met mast, number of sectors, source data and the measurement length).

Figure 11 - Wind statistics info

(39)

Figure number 12 includes the mean wind speed, wind energy and WTG energy for different heights and key numbers for wind energy and WTG energy for the site roughness class and hub height.

Figure 12 - STATGEN – Mean wind speed, Wind energy and WTG energy

(40)

Figure number 13 presents a complex wind data analysis with the weibull data for all the sectors and the graphics for weibull distribution, energy rose, mean wind speed and frequency.

Figure 13 - Wind data analysis

(41)

In the next figures (14 and 15), we have a detailed wind profile calculation included WTG wind test for different heights (from 5 to 5 meters till 200 m height) for all the sectors.

Figure 14 - WAsP – Wind profile detailed

(42)

Figure 15 - WAsP – Wind profile detailed 2

(43)

Turbine type:

A 500 kW turbine with a rotor diameter of 40,3 m, hub height of 55 m, rated speed 14m/s was chosen. Figure number 16 shows the main results for WAsP calculation with the WTG energy calculated for one year.

Figure 16 - WAsP – Main results

(44)

Wind resource grid generation

In the next part we will generate the 2 WRG that will be compared and optimized.

Figure number 17 shows how was calculated and generated the WRG with the map files used for this calculation and the setup.

Figure 17 - Resource – Main results. Generate .rsf file

(45)

In the figures above it was shown how it was generated the wind resource grid file .rsf for WAsP using WindPRO software tool. This WRG file will be compared with the WRG file generated from WindSim and when we got the 2 WRG files, they will be compared and after that, it will be run the optimization and we got the results, based on those 2 WRG files optimization.

The following steps were performed in WindSim to get the wind resource file generated.

First, we introduce the data and begin running simulations starting with terrain elevation map (figure number 18) and the digital terrain conversion with properties (figure number 19).

Figure 18 - Terrain elevation

(46)

Figure 19 – Digital terrain conversion with the properties

The next step is to generate the wind fields, based on the data get from the digital terrain conversion. In figure number 20 we have the wind fields’ spot values for sector 0 and in figure number 21 for sector 180. Figure number 22 shows the generated wind fields report.

Figure 20 - Wind fields sector 0 spot values

(47)

Figure 21 - Wind fields sector 180 spot values

Figure 22 - Wind fields report

In figure number 23 we can see how we placed the climatology within the 3D terrain model and the climatology file conversion.

(48)

Figure 23 – Objects - Climatology conversions

In figure number 24 we can see the park layout.

Figure 24 - Park layout

(49)

After placing the climatology and the climatology file conversion we explore the wind data base and we got results report for wind velocity for sector 0 (figure number 25) and sector number 180 (figure number 26).

Figure 25 – Velocity report sector 0

(50)

Figure 26 – Velocity report sector 180

Before generating the WRG file, it will be accumulated the results of the climatology for all the sectors, like we will see in the figure number 27, where we have the climatology report for frequency, and in figure number 28, the weibull distribution (wind speed and frequency).

(51)

Figure 27 - Climatology report 1 (Frequency in %)

Figure 28 - Climatology report 2 (Wind speed in m/s)

After that we got the wind resource map: in figure number 29 we can see the mean wind speed map, in figure number 30 the weibull shape parameter (k), and in figure number 31 the weibull scale parameter (A).

(52)

Figure 29 – Wind Resources. Variable: Mean wind speed 2D (m/s)

Figure 30 – Wind Resources. Variable: Weibull shape parameter, k

(53)

Figure 31 – Wind Resources. Variable: Weibull scale parameter, A The wind resource grid file is finally generated (figure number 32) in two different file extensions .rsf and .wrg.

Figure 32 – WindSim wind resource grid .rsf and .wrg files are generated A WRG is used to get an estimation of the wind energy potential in the assessment process and all wind farm design and optimization packages work with WRG (AWS Truepower, 2013).

A Basic Wind Resource Grid contains data that it is set for fixed 25 x 25 km area, at 200 m resolution and the hub height at 50 m with the following data:

Elevation

(54)

Roughness

Wind Speed Distribution (Weibull A, k and sector-wise probability)

Power Density (AWS Truepower, 2013).

An example of how it looks like a WRG we can see in figure number 33.

Figure 33 – WRG file

Table 1 - Vertical profile climatology for sector 0 (N)

The vertical profile climatology exported from WindSim simulations for 100.000, 250.000, 500.000, 1.000.000, 2.000.000, 2.500.000 and 3.000.000 are similar with no difference as we can see in Table number 1.

Wind resource grids comparison

Now that both wind resource grids are generated, their comparison is done with the CFD tool calculation model, from WindPRO in Post Processing tab (figure 34).

(55)

After calculation, graphs with different values are analyzed to see if there is any difference between both wind resource grids.

Figure 34 – CFD wind resource grid comparison model form WindPRO

First two tables are showing the same data (first table is in 3D, figure 35) and are giving the ratio between the mean wind speed values, with standard deviation. As we can see the mean value for the ration between wind speed from WAsP wind resource grid and WindSim wind resource grid are 1,05 and standard deviation value 0,0022, which means that the WAsP value is a bit bigger than WindSim wind speed mean value, but there difference is almost insignificant and almost equal.

(56)

Figure 35 – Wind resource grids comparison 1 (ratio between mean wind speed values WAsP/WindSim)

7 33 .9 0 0 7 3 3 .8 00 7 33 .7 0 0 7 3 3 .6 0 0 7 3 3 .5 00 1 ,0 5 2

1 ,0 5 1

1 ,0 5 1

1 ,0 5 1 ,0 5

1,0 4 9 1 ,0 4 9 1 ,0 4 8 1,0 4 8 1 ,0 47 1 ,0 4 7

9 1 7 .55 0 9 1 7 .5 0 0 9 1 7 .4 5 0 9 1 7 .4 0 0 9 1 7 .3 5 0 9 1 7 .3 00 9 1 7.2 5 0 9 1 7 .2 0 0 9 1 7 .1 50

1,05 1,05 1,05 1,05 1,05 1,05

(57)

Figure 36 – Wind resource grids comparison 2 (ratio between mean A weibull parameter values WAsP/WindSim)

7 3 3 . 9 0 0 7 3 3 . 8 0 0 7 3 3 . 7 0 0 7 3 3 . 6 0 0 7 3 3 . 5 0 0 1 , 1 0 3 1 , 1 0 3 1 , 1 0 2 1 ,1 0 2 1 , 1 0 1 1 , 1 0 1 1 , 1 1 , 1 1 , 0 9 9 1 , 0 9 9 1 , 0 9 8

1 , 0 9 8 9 1 7 . 5 5 0

9 1 7 .5 0 0 9 1 7 . 4 5 0 9 1 7 . 4 0 0 9 1 7 . 3 5 0 9 1 7 . 3 0 0 9 1 7 . 2 5 0 9 1 7 .2 0 0 9 1 7 . 1 5 0 9 1 7 . 1 0 0 9 1 7 . 0 5 0 1 ,1 0

1 ,1 0 1 ,1 0 1 ,1 0 1 ,1 0 1 ,1 0

(58)

Figure 37 – Wind resource grids comparison 3 (ratio between mean k weibull parameter values WAsP/WindSim)

7 3 4 .0 0 0 7 3 3 .9 0 0 7 3 3 .8 0 0 7 3 3 .7 0 0 7 3 3 .6 0 0 7 3 3 .5 0 0 0 ,3 2 3 0 ,3 2 3 0 ,3 2 3 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2 0 ,3 2 2

9 1 7 .5 5 0 9 1 7 .5 0 0 9 1 7 .4 5 0 9 1 7 .4 0 0 9 1 7 .3 5 0 9 1 7 .3 0 0 9 1 7 .2 5 0 9 1 7 .2 0 0 9 1 7 .1 5 0 9 1 7 .1 0 0 9 1 7 .0 5 0 9 1 7 .0 0 0 0,32

0,32 0,32 0,32 0,32 0,32

(59)

In figure 36, the two graphs are showing also the same data and are giving the ratio between the mean A weibull parameter values, with standard deviation. As we can see the mean value for the ratio between A parameter from WAsP wind resource grid and WindSim wind resource grid is 1,1 and standard deviation value 0,0024, which means that the WAsP A parameter value is also a bit bigger than WindSim A parameter value, but their difference is almost insignificant and are almost equal on the site area.

In figure 37, the two graphs are showing also the same data and are giving the ratio between the mean k weibull parameter values, with standard deviation. Here the mean value for the ratio between k parameter from WAsP wind resource grid and WindSim wind resource grid is 0,322 and standard deviation value 0,0003, which means that the WAsP k parameter value is almost 3 times smaller than WindSim k parameter value, and the difference is almost constant on the site area (standard deviation is very small).

As a conclusion, the mean values are almost the same, only k weibull parameter is 3 times bigger for WindSim than WAsP for the same input data.

Optimization set-up and results based on optimization of linear and nonlinear wind resource grids

For optimization calculation, Park calculation is used and this includes resource file, calculated annual energy for wind farm, calculated annual energy for each of 10 new WTG with total 5,0 MW rated power and WTG sitting. Below are presented detailed results (figure 38 and 39).

Wake model and wake decay constant is defined by the original park model that is Jensen (Samorani, M., 2007) model is chosen with a wake decay constant value of

(60)

0.04 which is used for offshore applications. Number of hours in complete year was chosen not to be a leap year: the software tool used 8760 hr/year for wind park power output calculation per year. Number of iterations in optimization was based on maximum number of iterations before stop of full optimization convergence which was defined by the software default values.

Parameters used in the optimization consists of,

I. Number of turbines: Optimal number of turbines packed into the predefined area.

II. Resolution: The required grid cell size for turbine placement which was taken is 40 m.

III. Geometry of project site area: Quadratic area was selected.

(61)

Figure 38 - Park calculation – Main results, page 1

(62)

Figure 39 - Park calculation – Main results, page 2

(63)

Production analysis (figure 40) and power curve analysis (figure 41) are shown below:

Figure 40 - Park – Production analysis

(64)

Figure 41 - Park – Power curve analysis

(65)

Wind data analysis is shown in the figure 42 and includes the weibull distribution, energy rose and mean wind speed.

Figure 42 - Park – Wind data analysis

(66)

After the Park calculation and data analysis there was performed the optimization module for the WAsP and WindSim .rsf grid resource files to see if there is any difference between them.

For the optimization it is used a quadratic area 400m X 400m, with constant wind and predominant wind direction (in this case from North), Random and Regular pattern calculation modules (automatic) with 40 m resolution.

First is calculated WindSim .rsf wind resource grid file figure 43) and after that the WAsP .rsf wind resource file (figure 44) with the following inputs:

Figure 43 - Optimize wind farm layout 1

(67)

Figure 44 - Optimize wind farm layout 2

The optimize controller with the setup and calculation for optimization tools are shown in figure number 45:

Figure 45 - Optimizer controller

(68)

The optimization results for WAsP WRG file with random pattern configuration is shown in the figure number 46 and in the figure 47 we got the map of the

optimization.

Figure 46 - Optimization WAsP random pattern results

(69)

Figure 47 - Optimization WAsP Random pattern map

(70)

The optimization results for WindSim WRG file with Random pattern configuration are shown in the figures number 48 and 49.

Figure 48 – Optimization WindSim random pattern results 1

(71)

Figure 49 - Optimization WindSim random pattern results 2

The optimization results for WindSim WRG file with Regular pattern configuration are shown in the figures number 50 and 51.

(72)

Figure 50 – Optimization WindSim Regular pattern results 1

(73)

Figure 51 – Optimization WindSim Regular pattern results 2

(74)

The optimization results for WAsP WRG file with regular pattern configuration are shown in the figures number 52 and 53; in the figure 54 we got the map of the optimization.

Figure 52 – Optimization WAsP Regular pattern results 1

(75)

Figure 53 – Optimization WAsP Regular pattern results 2

(76)

Figure 54 – Optimization Regular pattern map

Comparison Optimization results with Eftun Yilmaz’s optimization results

In table number 2 and 3 we have the optimization results comparison.

400x400 WTG Park yield Efficiency Type of calculation

Eftun 14 61320 100 Regular pattern OpenWind Eftun 20 47755 90,9 Regular pattern WAsP

Ion 20 51637 98,2 Regular pattern WAsP

Ion 20 85504 100 Regular pattern WindSim

Table 2 – Regular pattern optimization results comparison (Eftun Yilmaz, 2013)

(77)

400x400 WTG Park yield Efficiency Type of calculation

Eftun 17 42584 95,3 Random pattern WAsP

Eftun 15 65700 100 Random pattern OpenWind

Ion 4 10720 100 Random pattern WAsP

Ion 24 100497 97,9 Random pattern WindSim

Table 3 – Random pattern optimization results comparison (Eftun Yilmaz, 2013)

In the table 2 and 3 is the optimization data from Eftun Yilmaz and the data of this thesis (Eftun Yilmaz, 2013). There are 3 different software results to compare from WAsP used in WindPRO, from OpenWind and WindSim.

As it can be seen, the numbers of turbines differ for some optimization modules, because of the automatic set up of the optimization module. To be able to compare more accurate, the production will be calculated for an equal number of turbines.

Then the results will be:

400x400 WTG Park yield Efficiency Type of calculation

Eftun 20 87600 100 Regular pattern OpenWind Eftun 20 47755 90,9 Regular pattern WAsP

Ion 20 51637 98,2 Regular pattern WAsP

Ion 20 85504 100 Regular pattern WindSim

Table 4 – Regular pattern optimization final results comparison (Eftun Yilmaz, 2013)

400x400 WTG Park yield Efficiency Type of calculation

Eftun 20 87600 100 Random pattern OpenWind

Eftun 20 50100 95,3 Random pattern WAsP

Ion 20 53600 100 Random pattern WAsP

Ion 20 83740 97,9 Random pattern WindSim

Table 5 – Random pattern optimization final results comparison (Eftun Yilmaz, 2013)

(78)

From the table 4 and 5 it can be seen a very interesting thing like, the values for WAsP for both Regular and Random pattern calculations are almost similar, as well as for WindSim and OpenWind, which are both higher than WAsP results in both cases and for both type of calculations (40% higher for Regular pattern and 35%

higher for Random pattern calculation).

(79)

CHAPTER V CONCLUSIONS

The objective of this thesis is to see how the linear and nonlinear wind resource grids impact the layout of the optimization. This is an important factor in wind industry and for wind farm developers. Employing wind resource files generated by linear approach based WAsP and nonlinear approach based WindSim wind resource assessment software tools, comparison is made for the wind resource grids and optimization using these data files. In this way it can be seen if there is any difference between both of the software, using the same input data, for simple case, under simple conditions (flat terrain – ocean, predominant winds, and constant wind direction).

First the wind resource grids generated by the two software were analyzed. The mean wind speed values and A weibull parameter were observed to be almost the same, the only noticeable difference is for the k weibull parameter that is for WindSim wind resource grid 3 times higher than for WAsP.

The optimization study showed approximate 40% between linear and nonlinear wind resource grid for random optimization and 35% between linear and nonlinear wind resource grid for regular optimization for a simple case with the same input data..

There was at the same order of error observed in the results of E. Yilmaz 2012 for a simple case with the same input data however with different optimizer in contrast to this study and with different weibull distribution, but with the same mean wind speed of the wind.

To realize if software is overestimating or underestimating the results, or to see which software is more suitable for these conditions and giving the right results, the

(80)

predicted data should be compared with real data for a real case (wind farm). This was not the objective of this thesis.

(81)

REFERENCES

1. University of Reading, Meteorology, MT36E Boundary Layer Meteorology, retrieved on 10.05.2013 from:

http://www.met.rdg.ac.uk/~swrhgnrj/teaching/MT36E/MT36E_BL_lecture_n otes.pdf.

2. Stefan Emeis, Wind Energy Meteorology, Green Energy and Technology, Springer-Verlag, Berlin Heidelberg 2013.

3. Figure 3 - NikNaks from Stull, 1988, retrieved from:

http://commons.wikimedia.org/wiki/File:Atmospheric_boundary_layer.svg, on 12.05.2013.

4. Stull, R.B., An introduction to boundary layer meteorology (Kluwer Academic Publishers, Dordrecht, 1998), retrieved from:

http://lidar.ssec.wisc.edu/papers/akp_thes/node31.htm#stull88, http://lidar.ssec.wisc.edu/papers/akp_thes/node6.htm, on 14.05.2013 5. Lyndon State College Atmospheric Science, 20 June 2001, retrieved from:

http://apollo.lsc.vsc.edu/classes/met455/notes/section1/1.html, on 14.05.2013 http://apollo.lsc.vsc.edu/classes/met455/notes/section1/5.html, on 31.08.2013 6. Lyndon State College Atmospheric Science, retrieved from:

http://apollo.lsc.vsc.edu/classes/met455/notes/section2/6.html, on 31.08.2013 7. Symscape, 2013, retrieved from:

http://www.symscape.com/, on 31.08.2013

8. DTU Wind Energy, WAsP, 2013, retrieved from: http://www.wasp.dk/, on 20.05.2013

9. EMD, WindPRO, 2013, Retrieved from:

http://help.emd.dk/knowledgebase/, on 21.05.2013

10. Samorani, M. “Wind farm optimization problem.”, University of Colorado, Boulder, 2007, retrieved on 29.05.2013

11. Eftun Yilmaz, 2013, Benchmarking of Optimization Modules for Two Wind Farm Design Software Tools

(82)

12. AWS Truepower, 2103, retrieved from:

https://dashboards.awstruepower.com/index.php/cms/pages/wind-site- assessment/wind-resource-grids, on 17.09.2013

13. Pep Moreno, Arne R. Gravdahl, & Manel Romero, Wind Flow over Complex Terrain: Application of Linear and CFD Models, retrieved on 08.10.2013 14. WindSim, 2012, retrieved from:

http://user.windsim.com/index.php?action=artikel&cat=14&id=16&artlang=e n, on 08.10.2013

15. Mustakerov I. and Borissova D., “Wind park layout design using

combinatorial optimization”, Institute of Information and Communication Technologies of Bulgarian Academy of Sciences, 2010, Bulgaria

16. Wagner M., Day J., and Neumann F., “A fast and effective local search algorithm for optimizing the placement of wind turbines”, School of Computer Science, University of Adelaide, 20 April 2012, Adelaide, Australia.

17. Elkinton CN, Manwell JF, McGowan JG., “Algorithms for offshore wind farm layout optimization”, Renewable Energy Research Laboratory Dept. of Mechanical and Industrial Engineering University of Massachusetts, Wind Engineering Volume 32, No. 1, PP 67–83, 2008,Massachusetts.

18. Mosetti G, Poloni C, Diviacco B., “Optimization of wind turbine positioning in large wind farms by means of a genetic algorithm”, Journal of Wind Engineering and Industrial Aerodynamics, Elsevier, PP 105-116, 1998.

19. Meteoblue, Land-sea wind, 2013, retrieved from:

http://www.meteoblue.com/en_GB/content/703, on 16.10.2013 20. Jeff Haby, Low-level jet, retrieved from:

http://www.theweatherprediction.com/severe/llj/, on 16.10.2013 21. WindSim, 2013, retrieved from:

http://www.windsim.com/products.aspx, on 29.10.2013

References

Related documents

Chart 6.2.1.1: Cassandra pods total memory usage and the output rates of the Producer and the Logstasher under the Cassandra default configuration benchmark. (Logstasher

Figure 37: The pressure difference between the front and the back of the cylin- drical object as a function of time for the streamline upwind Petrov Galerkin method for

The result of the current study showed that NEWA mesoscale data represents wind climate very well for the onshore site with simple terrain.. On the other hand, NEWA

Similar results were found by Honnery and Moriarty (2011) who used three different exponential growth rates reaching two different installed capacities of wind power

The economic analysis of the 10 MW and the 11.5 MW model offshore wind farms using Vestas and Siemens wind turbines in Bligh Waters Fiji are presented in Tables 8 and 9 in terms

Carl Edström tried a non-linear method when he was sound supervisor for the TV- series Ängelby 17 .He reports that the idea came out of several collaborations with post

Figure 11 displays the factor obtained, when block triangularization and the constrained approximate minimum degree algorithm are applied to the matrix in Figure 2, followed by

In the Linux kernel project, the development is divided into different branches to allow a large number of people to contribute to the source code repository at the same