1. introduction & objectives - istos-lab.gr run a conventional wasp model to generate comparable...

10
WINDSIM CFD MODEL VALIDATION IN A MIXED COASTAL & MOUNTAINOUS REGION WITH COMPLEX TERRAIN Petros Theodoropoulos 1 Dipl. Mechanical Engineer MSc, Nick Deligiorgis, Dipl. Mechanical Engineer, MBA 1: Laboratory of Istos Renewables Ltd Summary The study examines the resource modelling capabilities of modern CFD software on a difficult and extensive region with rough terrain in-between sea and high mountains. A network of 7 masts is used to generate cross-predictions with a WindSim model (RANS equations). Predictions are checked against actual measurements as well as a conventional WAsP model. The study concludes that WindSim can be no replacement for lack of measurements in bigger heights, for knowledge of local scale climatology induced thermal effects or imperfect terrain data. In such conditions, the CFD model cannot predict accurately and fails to outperform a conventional WAsP estimation. 1. Introduction & Objectives Wind energy projects are mostly at the planning phase in eastern Mediterranean and Balkans region. In Greece particularly, less than 1.000 ΜW is operating while tenths of GW of projects are currently under development. One of the most important factors during the development stage of this “bulk” of projects is without doubt wind resource modeling. This new and unexploited geographic territory for wind energy presents a sizeable challenge for modern wind flow modeling software. Why? Because most of the sites are really complex in terrain and climatology, making it hard for wind experts to predict the wind resource using the “best practice” methods originating from namely northwestern Europe. Another alarming issue is that wind energy is a new thing in those countries and local people involved in this business sector are not much acquainted with the capabilities and limitations of wind resource prediction. The knowledge is limited to a small group of experts who use their expertise dangerously selectively. The questions thus are still pending: How can modern commercial CFD software cope with a typical “difficult” eastern Mediterranean region? How many masts should a project developer install in order to get a good picture of local wind flow? Is one mast enough for lets say a 20 or 50 MW project that could span for kms? Is a met mast 5 or 10 kms away acceptable by any chance as a reference for wind park energy predictions? The objective thus of this study is to give an initial but otherwise clear answer to these questions. It differs from other past studies because it is based on a large number (seven) of met masts installed in an exemplary site, with varying altitude, neighboring with both the sea and tall (often snow covered) mountains. The results can thus be used by the local wind energy community and authorities as a rough guide on general wind prediction validity with the scientifically most promising tools, namely CFD. A short interpretation of the results, as well as some recommendations are laid out. 2. Methodology The procedure and methodology that was chosen for the study can be summarized as follows: - Select a geographical region that complies with all the prerequisites to be characterized as “difficult” for wind flow modeling. Meaning that it should have a generally rugged terrain (but not technically inadequate for a wind park), some common climatology affecting factors (sea or tall mountains around) and varying altitude. - Establish a dense met mast network, able to collect reliable measurements, meeting IEC 61400 standards. - Gather concurrent data from a large enough period of time, theoretically enabling correlations and cross-predictions. Select a not much thermally active period, leaving out summer months which could possibly over-distort the climate model of the region.

Upload: phamkhanh

Post on 14-May-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

WINDSIM CFD MODEL VALIDATION IN A MIXED COASTAL & MOUNTAINOUS REGION WITH COMPLEX TERRAIN

Petros Theodoropoulos

1 Dipl. Mechanical Engineer MSc, Nick Deligiorgis, Dipl. Mechanical Engineer, MBA

1: Laboratory of Istos Renewables Ltd

Summary

The study examines the resource modelling capabilities of modern CFD software on a difficult and extensive region with rough terrain in-between sea and high mountains. A network of 7 masts is used to generate cross-predictions with a WindSim model (RANS equations). Predictions are checked against actual measurements as well as a conventional WAsP model. The study concludes that WindSim can be no replacement for lack of measurements in bigger heights, for knowledge of local scale climatology induced thermal effects or imperfect terrain data. In such conditions, the CFD model cannot predict accurately and fails to outperform a conventional WAsP estimation.

1. Introduction & Objectives

Wind energy projects are mostly at the planning phase in eastern Mediterranean and Balkans region. In Greece particularly, less than 1.000 ΜW is operating while tenths of GW of projects are currently under development. One of the most important factors during the development stage of this “bulk” of projects is without doubt wind resource modeling. This new and unexploited geographic territory for wind energy presents a sizeable challenge for modern wind flow modeling software. Why? Because most of the sites are really complex in terrain and climatology, making it hard for wind experts to predict the wind resource using the “best practice” methods originating from namely northwestern Europe. Another alarming issue is that wind energy is a new thing in those countries and local people involved in this business sector are not much acquainted with the capabilities and limitations of wind resource prediction. The knowledge is limited to a small group of experts who use their expertise dangerously selectively. The questions thus are still pending: How can modern commercial CFD software cope with a typical “difficult” eastern Mediterranean region? How many masts should a project developer install in order to get a good picture of local wind flow? Is one mast enough for lets say a 20 or 50 MW project that could span for kms? Is a met mast 5 or 10 kms away acceptable by any chance as a reference for wind park energy predictions? The objective thus of this study is to give an initial but otherwise clear answer to these questions. It differs from other past studies because it is based on a large number (seven) of met masts installed in an exemplary site, with varying altitude, neighboring with both the sea and tall (often snow covered) mountains. The results can thus be used by the local wind energy community and authorities as a rough guide on general wind prediction validity with the scientifically most promising tools, namely CFD. A short interpretation of the results, as well as some recommendations are laid out. 2. Methodology

The procedure and methodology that was chosen for the study can be summarized as follows:

- Select a geographical region that complies with all the prerequisites to be characterized as “difficult” for wind flow modeling. Meaning that it should have a generally rugged terrain (but not technically inadequate for a wind park), some common climatology affecting factors (sea or tall mountains around) and varying altitude.

- Establish a dense met mast network, able to collect reliable measurements, meeting IEC 61400 standards.

- Gather concurrent data from a large enough period of time, theoretically enabling correlations and cross-predictions. Select a not much thermally active period, leaving out summer months which could possibly over-distort the climate model of the region.

Page 2: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

- Setup a terrain model from best publicly available sources, both for height and roughness.

- Run WindSim CFD (based on RANS1 equations) to generate cross-predictions between

the 7 met masts. The use of nesting technique is necessary due to the large extent of the modeled region.

- Run a conventional WAsP model to generate comparable predictions.

- Compare cross-predictions of WindSim with actual measurements and respective WAsP products.

- Critically examine results.

It is expected that the incorporation of the dense met mast network will eliminate possible “random” prediction failures.

The procedure reflects in reality the transformation of wind measurements from original positions to other positions within an imaginary Wind Farm. It is in other words the crucial task that wind project developers and evaluation authorities are faced with. Taking into account that it is neither practical nor economical to move masts around every possible turbine location within a large region, it would be in the project developer’s interest to use the minimum number of masts to capture the local wind resource.

3. Description of Study Area

The validation study took place in Western Greece. The area is surrounded by sea in the North, East and South and margined by high mountains in the West. The area extent is 27Χ24km

2. The western part of the area is characterized by hills not over 500m and is

surrounded by the sea. The eastern part of the study area is mountainous with steep ridges and valleys. The mountains reach altitudes to over 1.400m.

The climate of the area is particularly wet for a typical Mediterranean region. But this is usual for western Greece as it is divided by the main mountain chain of Pindus that holds water vapors from westerly coming meteorological systems. In this part of Greece northern is not the principal blow direction contrast to the eastern Greece, where the northern Meltemi winds dominate each summer period. In contrary most common directions in the area of study are southern. In the warmer days local thermal induced wind is a very common phenomenon.

In general the wind potential of the area is not so strong; a fact that strengthens the local thermal effects and makes the validation project more difficult.

Picture 1: The area of validation study.

4. Measurement Stations

Seven anemological masts of 20m height were installed during November of 2007 for the purpose of this study. The met mast network was spread within a total area of 340km

2

(17X20km). All measurements were concurrent. The first mast began measurements on 23rd

of November and the last on 9

th of December. The measurement campaign was chosen to be

1 RANS: Raynolds Averaged Navier-Stoke equations

27km

1

2 3

4

5 6 7

24km

N

Page 3: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

from 23th of November until 21

st of April in order to avoid thermal effects during late spring,

summer and early autumn. All seven measurement stations were installed in accordance with IEC-61400-12-1 by the accredited laboratory of Istos Renewables. All anemometers were calibrated by a Measnet member lab and in addition control anemometers were installed for in situ comparison at every mast.

4.1. Correlation of Measurements

All measurements are well correlated except two couples of masts (see purple numbers in Table I). All calculations were executed by the famous WindRose

[1] application from CRES.

The correlation is calculated using Pearson formula and a linear regression analysis is carried out for every bin of velocity for the reference station and the target station.

In the next table the goodness of fitness for the linear regression between the reference average speed of each bin (X) and the corresponding average speed (Y) of the concurrent target measurements is depicted for every combination of masts.

MAST 01 MAST 02 MAST 03 MAST 04 MAST 05 MAST 06 MAST 07

MAST 01

MAST 02 0.9984

MAST 03 0.9983 0.9993

MAST 04 0.9321 0.8918 0.8744

MAST 05 0.9449 0.9807 0.9328 0.9958

MAST 06 0.8994 0.8929 0.5448 0.9844 0.9628

MAST 07 0.8927 0.965 0.8081 0.9968 0.9700 0.9667

R2

Table I: Correlation of simultaneous measurements

The very low correlation between Mast 3 and Mast 6 can be explained as following. Mast 3 is exposed to intense thermal effects that strengthen the dominant south winds. On the other hand mast 6 is hindered in the south from high rocky ridges, which weaken the southerly wind, when thermal phenomena dominate.

4.2. Filtering, Analysis & Prediction of Missing Data

The measurement data has been automatically & manually inspected and filtered, identifying inconsistencies and missing data. The missing data for each station is only a very little percentage of the whole volume of data (2-10%) caused mainly due to the different dates of installation. Therefore the missing data are predicted using the parameters from the correlation analysis. In that way a frequency table of measurements that refers to exactly the same time period is used.

5. Wind Field Simulation

Two types of Wind Field Models dominate in wind energy study area: linearised (ie WaSP) and CFD (ie WindSim) models. Linearised are very famous because they require very limited computer power, they are fast and easy to use. But what are their limitations? In the following paragraph the main characteristics of both types of models are presented:

5.1. Presentation of Wind Field Models

a. Linearised Models

Linearised models are based on linearised solutions of the dynamic equations for boundary layer flow perturbed by terrain (simplified steady-state solutions of the Navier-Stokes equations - e.g. linearisation techniques). The theory of Jackson and Hunt (1975) provided a basis for numerical modelling of 2D steady-state turbulent flow over a low hill of gentle slope. The governing momentum equations are linearised using scale analysis and assuming uniform rough surface and small slope. The flow is broadly divided into an inner layer, where turbulence is important, and an outer layer, where the flow can be considered inviscid. The atmospheric stability is considered to be nearly neutral and the oncoming vertical profile of wind speed is therefore supposed logarithmic.

Page 4: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

The terrain shape is analysed in terms of Fourier components [2]

; the equations are thus solved in Fourier space; the Fourier transform is numerically inverted to give the solution in real space. For that purpose terrain model is developed around the anemometry mast using polar coordinate system. Simple closure assumptions are normally used to model the Reynolds stresses. Moreover the linearised theory can be extended in order to model the effect on the wind field of roughness length variations. Application of the theory to the wind energy sector led to the development of two most popular microscale modelling products: WAsP (Troen and Petersen, 1989) and MsMicro (Walmsley et al. 1990). According to bibliography

[2] linearised models fail to predict and describe correctly the lee

region where turbulent wake and separation develop. Further tests and applications proved the linearised theory to give reasonably good results on the upstream side and on the top of hills with H/L≤0.3-0.4, which corresponds to maximum inclinations of 22

ο.

Picture 2: Representation of the 2D hill simple modeling.

b. CFD non linearised models (especially WindSim)

As already known, the most general equations that govern the fluid flow in the Atmospheric Boundary Layer are the Navier-Stokes equations. These equations express the physical principles of conservation of mass, momentum and energy. WindSim CFD software solves the non-linear transport equations for mass, momentum and energy by the means of finite volume method. The finite volume method discretises the governing analytical equations over each finite volume. In particular, WindSim solves the RANS equations. RANS (Raynolds Averaged Navier Stokes) equations are deduced from the Navier-Stokes equation, using a time averaging procedure. In this way all the turbulence is modelled and only the time averaged variables are found with the simulations. These equations aren’t linear and thus the only way to solve them is to use an iterative method. WindSim uses a core constituted of the solver Phoenics.

5.2. Orography Data

The digital terrain data in this project derived from publicly available sources (SRTM & D4W). Grid resolution from those sources cannot be considered fine. The best grid dimensions obtained were 30x90m

[3], which alone stands for a basic inaccuracy source. The grid data

were converted to contour altitude isolines for use in WAsP software. The positioning of the masts on the Digital Terrain Model was carefully accomplished by GPS data and cross-checked with photographs and on-site observations.

5.3. Roughness

The whole region of this study is very complicated concerning the roughness. As described in paragraph 3, the sea areas around the west part correspond to roughness length of 0.001m. In the inland the roughness length which prevails is 0.03m (corresponds to open grass-land with few bushes). But there are many areas with steep rocks or forests that have a roughness length from 0.3m up to 0.8m and even more.

5.4. WindSim Modelling

Due to the large domain size, the nesting technique was used. Flow fields were solved using 16 direction sectors in both the meso-scale, and the two micro-scale models. The meso-scale

Page 5: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

model has simple logarithmic wind profiles imposed at each boundary, with the nested models using the solved meso-scale flow field as boundary conditions. The nesting of the micro scale models within the meso scale model attempts to impose realistic boundary conditions in the principal areas of interest, the boundaries of the micro-scale models.

The WindSim version used was 4.8.1. At first the wind conditions for the meso model were determined with the following parameters:

Domain size: 27x24 km Model height: 6.087 m Grid spacing x-y: 168 m

Vertical nodes: 20, height distribution factor: 0.05 Number of nodes: 454 400 Iterations: 100

Picture 3: The meso-scale model and the two micro-scale models.

To get more accurate results the results of the meso-model are used as an input (boundary conditions) to both micro-models engaged:

1) Eastern micro-model with the following characteristics:

Area size: 11x19 km Model height: 14417m Grid spacing x-y: 50x60 m

Vertical nodes: 30, height distribution factor: 0.025 Number of nodes: 913 710 Iterations: 100

2) Western micro model has the following characteristics:

Area size: 10x13 km Model height: 3081m Grid spacing x-y: 45x43 m

Vertical nodes: 20, height distribution factor: 0.5 Number of nodes: 839040 Iterations: 100

Measured wind data was imported to each mast in the form of frequency-direction tables. Both micro models had a refinement grid area close to the masts. The predictor climatologies were placed at the relevant mast location and height and then transferred by vertical translation to 150m above ground level, using the micro model for the relevant mast. This climatology is applied to the meso-scale model and a horizontal translation is performed to generate climatology at each of the remaining mast locations based on the meso-scale flow field. A further vertical translation is performed on this transferred climatology within the relevant micro scale model to generate climatology at predicted mast height and location.

The process was executed for the round robin test generating 42 predicted climatologies, 6 at each mast location. These predicted climatologies were compared with the observed wind climate at the given mast.

5.5. WAsP Modelling

In order to provide a reference and comparison measure for CFD, the study includes a similar “conventional” WAsP based analysis. The model is based on WAsP 8.0 engine, set up through the WindPro 2.5 frontend. The procedure followed to establish the WAsP model was a “by the book” approach. The same Data4Wind XYZ grid data file used in WindSim is

Page 6: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

imported and converted to contours with the built in map editor, and a separation of 5m height. As far as roughness is concerned, the Data4Wind XYZ grid data used in WindSim was too large and with too many open lines to be handled by the software. It was thus converted to a graphic file (.jpg) with coordinate references and imported in Autocad. Closed contours were drawn by hand to match the most important roughness changes depicted. The resulting .dxf file was imported. Measured wind data was imported to each mast in the form of frequency-direction tables. Seven statistics were generated, one from each mast. 7 x 6 WAsP calculations were executed from each wind statistic to the location of the remaining 6 masts.

Apart from the recurring procedures, calculations were quite speedy, due to the fact that only mast positions were subject to predictions, one of the advantages of the WAsP modeling.

6. Results

With strong variations in measured resource, the results of cross-prediction were calculated. 42 cross-predictions were carried out. The procedure was to calculate for each anemometry position the differences between:

• Measured Average Velocity and predicted Average Velocity.

• Exploitable Energy Content of wind from real measurements and from predicted climatologies. The exploitable energy content was estimated using a typical Power Curve @ 20m hub height and calculating the corresponding Energy Production.

Finally, the accuracy of transferred wind-rose prediction was examined by quantifying the differences per sector between the original windroses and the predicted ones. This corresponds to 6X7X16 = 672 comparisons.

The aforementioned comparisons were carried out also for the WaSP model in order to generate comparable predictions.

6.1. Mean Wind Speed

In diagram 1 the prediction error in average speed for each mast against every predictor mast is depicted. The overall average error in Umean prediction was for WindSim 1.4%, while for WAsP was 2.3%. This slightly better performance of WindSim was counterbalanced by the better performance in average standard deviation of WaSP. In other words average Std Deviation of errors in Umean prediction was for WindSim 11% and for WAsP 8%

Diagram 1: Errors of Umean estimation from WindSim prediction.

Page 7: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

Diagram 2: Errors of Umean estimation from WaSP prediction.

6.2. Exploitable Energy Content

The same tendency appeared in the prediction of the exploitable energy content: The Overall Average error for WindSim was -1.7% while for WAsP was 3.1%, but the average Std Deviation for WindSim was 19% while for WAsP was 15%.

Diagram 3: Errors of Exploitable Energy Content estimation from WindSim (L) & WaSP (R).

6.3. Quantitative Analysis of Windroses fit

In table II cross-predictions of windroses from WindSim are depicted. Under each windrose a percentage value is written. This percentage P expresses the relative weighted difference of the predicted windrose from the original one.

The quantisation has been carried out using the following norm:

, whereas Ei is the measured energy content per sector i and EPi is the predicted energy content per sector.

=

=

=16

1

16

1

i

i

i

iPi

E

EE

P

Page 8: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

PREDICTORS (SOURCE)

1 2 3 4 5 6 7

1

77%

98%

107%

83%

124%

80%

2

174%

56%

80%

116%

157%

118%

3

178%

56%

76%

90%

196%

124%

4

211%

81%

75%

59%

241%

139%

5

127%

58%

67%

64%

180%

86%

6

124%

91%

94% 90%

95%

85%

PR

ED

ICT

ION

S (

TA

RG

ET

)

7

72%

83%

94%

79%

60%

76%

Table II: Cross-predictions of windroses by WindSim.

7. Conclusion - Possible Error Inducing Factors & Interpretation

Looking at the results in the previous chapter, it is obvious that notable errors are widespread. The depicted deviations between actual measurements and predictions could be put down to a number of factors. The authors have tried to judge their severity and describe their possible effects based on the arithmetic results of the study.

- Climatic variations within the region: One obvious disadvantage of both CFD and WAsP models is that climatic meso-scale

[4] wind variations are not taken into account in the

modelling. The effects of thermal vertical stability and horizontal microclimatic differences on both the meso-scale and the micro-scale flow are judged to be very important. A clear sign of this issue is that although masts are nicely correlated they fail to predict each other, even at close distances. Particular masts are constantly over or under predicted (ex. mast 1), giving further proof of thermally induced effects that affect some parts of the region more than others.

- Imperfect digital terrain data: Although erroneous predictions exist between all masts and therefore no particular glitch can be attributed to specific spot wrong data, the Data4Wind is surely not enough to guarantee trouble free flow predictions. The variable grid

Page 9: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

resolution (about 30x90m) of the original source (SRTM) often exceeds the practically accepted 30x30m limit resolution

[5] [6] [7]. Roughness data imported through Data4Wind

seems to enclose correctly the terrain roughness changes, but absolute values are highly debatable.

- Low measuring height is another error factor. In a low grid resolution local idiomorphies of the terrain are of the same magnitude scale as the mast height,thus they cannot be defined in the DTM. In addition thermal induced effects are stronger closer to the surface.

- The distance between cross-predicted masts varies from 2 to 20 kms and yet no mathematical link with prediction errors can be established. This is actually one of the most worrying products of the study.

- DeltaRIX[8] [9] [10]

seems slightly influencing for WAsP (R2=0.24). However and to our initial

surprise the correlation coefficient between prediction error and DeltaRIX is more than doubled for WindSim (R

2=0.51), indicating a not dominant but notable connection. This

relation is surely not based on a theoretical shortfall of CFD but clearly on a technical occurrence. What happens in reality is that this correlation conceals the effect of nesting technique. As it is obtained masts that lie on the western micro scale model have smaller RIX factor because of the smoothest topography and vicinity with sea. On the other hand masts that lie on the eastern micro scale model have greater RIX factor because of the more rugged topography and highest mountains. So, when DeltaRIX is close to zero that means that both masts belong to the same micro model and thus the additional errors due to nesting procedure are avoided. However, when the examined masts lie on different micro models this will conclude to a bigger DeltaRIX (negative or positive) and the errors due to nesting technique will be added.

- Finally, one other very probable error factor is the discretisation procedure in the redistribution of transferred climatologies.

[7]

All previous possible errors when combined together exceed the errors that derive from the use of conventional linearised model instead of more sophisticated CFD WindSim code.

8. Recommendations

Based on the study’s results and prediction errors interpretation, we can also conclude to a list of recommendations towards members of the wind energy community. It is generally obvious that in difficult regions like eastern Mediterranean and Balkans, we desperately need all round improvement of the resource monitoring and modelling procedure to work together with CFD.

8.1. For future research, studies and code developers.

For future validation studies it is emphasised by the authors the need for: - Finer resolution and high accuracy digital terrain data around mast positions (<30m). (To

be implemented by WindSim users) - Bigger measuring heights (over 40m). (To be implemented by users & consultants) - Models with more cells, more flexible grid generation (trihedral cells) and sophisticated

refinement around masts in order to avoid nesting technique. (implemented by WindSim developers)

- Wind profile validation against measurements from tall masts (>40m). There are many validations

[6] [11] [12] already performed but none is taken place in a high rugged terrain like

this in the present paper. - Coupling of temperature equations and practical implementation in coastal areas (To be

implemented by WindSim developers).

8.2. For wind park developers and owners

A piece of advice for Wind park developers is to use human expertise to counterbalance the probable errors of software predictions. A local wind expert would perform a site visit and should be able to rate any region’s difficulties based on the terrain topography, the prevailing wind directions & the local climatology. His experience should be used in conjunction with reliable wind measurements to “generate and rate” a resource estimation study. In addition, as project developers move towards areas with lower wind potential, they have not realised

Page 10: 1. Introduction & Objectives - istos-lab.gr Run a conventional WAsP model to generate comparable predictions. - Compare cross-predictions of WindSim with actual measurements and respective

how crucial it becomes in such sites to have accurate predictions all along. A small error in wind speeds can lead to large deviations of predicted energy

[13], unlike high wind sites.

Moreover, it is more than evident that large wind parks demand the installation of more than one wind mast. It is concluded from the results of this study that a 20m mast can be held accountable for as less as a circle with 1.5km radius around it and no more.

8.3. For approving/regulating authorities

Authorities should be overly careful when reviewing 3rd

party wind park energy production predictions. If the authorities get it wrong, we might end up with a batch of hastily planned wind farms that would destroy the image and usefulness of wind power beyond repair. In “difficult” regions, (a large proportion of project venues in eastern Mediterranean & Balkans), modern wind prediction software needs some mandatory prerequisites. Therefore based on the conclusions of this study, regulating authorities should request from project developers:

- To present measurements from a mast inside the planned wind park.

- To install more and taller masts for larger wind parks. It must be clear that in the subject of resource modelling, eastern Mediterranean authorities cannot adopt the “best practice” from Germany or Denmark simply because it is not right for these areas’ particularities. This means that general “Wind Maps” or “Wind Atlases” are totally inappropriate in complex terrain countries, even for the initial planning stages of a new project. 9. Credits

Credits to Mrs Andrianna Kakoura and Mr Christoforos Magoulas, scientific members of Istos Renewables Laboratory, for attending the measurement stations, downloading & analysing the data and making the correlation analysis for all masts used in the present paper. References:

1. Windrose User’s Guide, D. Fousekis, CRES, 2004.

2. Wind Flow Models over Complex Terrain for Dispersion Calculations, Sandro Finard, Maria Grazia Morselli, Cost Action 710, 1997

3. An Assesment of the SRTM Topographic Products / E. Rodriguez et al. / Jet Propulsion Laboratory, NASA / http://www2.jpl.nasa.gov/srtm/srtmBibliography.html, 2008

4. Wind in complex terrain - A comparison of WAsP and two CFD-models / Erik Berge et al., Kjeller Vindteknikk AS, Arne R. Gravdahl, / EWEC 2006 / Dec 2005

5. Influence of topographical input data on the accuracy of wind flow modelling in complex terrain / Niels G. Mortensen & Erik L. Petersen, Wind Energy and Atmoshperic Physics Dept, Risoe National Laboratory, Roskilde, Denmark / EWEC 1997 / Oct 1997

6. Wind Energy Resource Evaluation In A Site Of Central Italy By Cfd Simulations, Daniele Fallo, Diploma Thesis, University of Cagliary, Italy

7. Meso Scale Modeling with a Reynolds Averaged Navier-Stokes Solver. Assessment of wind resources along the Norwegian coast, Gravdahl A.R., IEA Annex XI, Risø Denmark, 1998.

8. WAsP Manual Vol 2 Users Guide / Risø National Laboratory, Roskilde, Denmark, 1993.

9. Comparison of wind flow models in complex terrain / Graeme Watson et al. / Natural Power Consultants Ltd, Dumfries and Galloway, Scotland, 2001

10. Improving Wasp Predictions in (too) Complex Terrain / Niels G. Mortensen et al. / Wind Energy Department, Risø National Laboratory, Roskilde, Denmark / EWEC 2006

11. Three-dimensional Wind Field Calculation Above Orographic Complex Terrain In Southern Europe, Dipl. Inf. Carsten Albrecht, EWEC 2006

12. WindSim Validation Study CFD validation in Complex terrain, Tristan Wallbank, 2008

13. Wind Energy Handbook, Tony Burton, David Sharpe, Nick Jeckins, Ervin Bossanyi, Wiley editions, 2001