American Institute of Aeronautics and Astronautics 1

A new technique to improve expected aep estimation in very

complex terrain**

F.Castellani* and G. Franceschini†

University of Perugia, Department of Industrial Engineering, Perugia, Italy, 06125

Annual Energy Production (AEP) estimation is a key issue in the development and financing of wind projects; it is fundamental not only for investment evaluation but also in defining the plant lay-out and dimensions. In order to obtain reliable results, detailed analysis of the wind conditions is needed; for wind sites in complex terrains each stage of this analysis is more difficult than in flat land because of the particular wind field conditions that are characterized by high turbulences and different shapes of the wind profile. It is not well known how these conditions affect wind turbine operation; as a result wind energy production can be different than expected. This is mainly due to the misunderstanding of the wind flow characteristics and the turbine efficiency. The present work is focused on a new method which takes into account the effect of orography in terms of available energy rather than the efficiency for the conversion. A technique to customize power curve is proposed and applied with success in the test case of Fossato di Vico (ITALY) wind farm bringing error in estimating AEP below 10%. Present work has been carried out with Anemon s.p.a. (ITALY) which provided turbine production data and participated in the measuring campaign.

Nomenclature AEP = Annual Energy Production vu = wind speed at hub height level vm = measured wind speed hu = turbine's hub height hm = speed measurement height

= wind profile exponent z0 = roughness parameter P = power available for the turbine’s rotor v = wind speed z = height from ground level Dr = rotor diameter Hr = turbine’s hub height

= air density

R = correction parameter Pcor = corrected power curve Pcer = certified power curve (adjusted according the air density) m = number of sector for wind direction

I. Introduction N Europe the wind energy industry was first developed in the nineties, mainly by northern countries. Since the late nineties interest in wind energy conversion systems (WECS) has been growing in other European countries,

such as Italy, and knowledge of wind technology was transferred from Denmark, Germany and other such regions where WECS industry was already fully experimented. Designing techniques and methods to estimate wind energy potential that were developed for the typical northern European environment were applied, in some cases, to regions of a very different orographic nature.

* Assistant Professor, Department of Industrial Engineering, Via G. Duranti 67, Perugia, Italy 06125. † Associate Professor, Department of Industrial Engineering, Via G. Duranti 67, Perugia, Italy 06125.

I

43rd AIAA Aerospace Sciences Meeting and Exhibit10 - 13 January 2005, Reno, Nevada

AIAA 2005-1331

Copyright © 2005 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

American Institute of Aeronautics and Astronautics 2

The growing interest in wind energy and the verification that knowledge of wind technology could not always be

transferred in its original formulation have increased the need to set up reliable methods of wind resource assessment for very complex terrain areas.

Early experiment showed that wind resource could be exploited even in such areas that were classified unsuitable by first estimations1.

Installation of wind turbines in very complex terrain environment gives interesting data which helps us to understand their operation and to formulate new methods for lay-out optimization and AEP estimation.

In the Fossato di Vico Wind Farm (the first installation in Umbria Region) two Neg-Micon NM740/44 aerogenerators have been operating for five years; data on energy production and anemometer conditions collected by Anemon s.p.a.‡ can be exploited to verify the accuracy of different methods for AEP estimation.

Such data, in conjunction with SODAR measurements and wind field fluid dynamics simulation performed by the Department of Industrial Engineering of University of Perugia, has been used to formulate a new technique to improve expected AEP estimation in very complex terrain.

II. Wind Energy Production Estimation Production data from the operating turbines can be used to test different methods for AEP estimation feeding as

input wind data collected by the anemometer stations placed nearby (the nearest one is placed at a distance of 550 m but also transferred data from another station 8 km away from the plant were used).

The high degree of complexity of the orography makes the analysis of wind resources very difficult. Figure 1 shows the plant lay-out with the existing turbines and two additional aerogenerators that are still in the

design stage§. In complex terrain environment AEP estimation is generally carried out through, among other assessment

techniques, field measurements of wind speed and direction at a level above the ground as close as possible to the turbine's hub height. When it is necessary to transfer climatology data up to hub height level the power law formula can be applied:

oz

m

umu h

hvv

(1)

where hm is the anemometer height, vm is the measured wind speed, hu is the hub height, vu is the wind speed at

hub height and oz is the exponent:

o

o

z

z10

ln

1

(2)

where zo is the roughness parameter. The exponential profile can be used in flat terrains but in complex orography it is not applicable.

For this reason generally a micro-site analysis in complex terrain is generally carried out by measuring wind speed at different levels above the ground to understand wind profile behavior.

Field measurements are available only on a limited number of positions and investigations on wider areas can be performed with numerical models.

Expected energy production can be calculated by using turbine power curve data

‡ The Italian industry that has built and keep in operation the Wind Plant. § This increases interest towards a deepen analysis of wind field and plant operation.

Figure 1. Plant lay-out of Fossato di Vico Wind Farm.

American Institute of Aeronautics and Astronautics 3

provided (fig.2) by the turbine manufacturer (this curve has to be adjusted according to the site’s air density).

Energy production data was used to test the accuracy of two different wind simulation models: the linear model WAsP (Wind Atlas Analysis and Application Program), and the fluid dynamics model Windsim.

WAsP calculates wind field using a linear version of the boundary layer equation so that it cannot include recirculation phenomena. Obviously WAsP calculus has no problem of convergence in any kind of domain: wind resources are derived from climatology data according to orography and roughness conditions.

Windsim calculates wind field using a CFD (Computational Fluid Dynamics) model that is used in each direction sector to understand the influences of orography and roughness on a meshed domain. It includes recirculation phenomena and in a very refined mesh it is able to model vortexes solving the Navier-Stokes equation on a three-dimensional domain.

Windsim needs a larger amount of calculation resources (in terms of hardware performances and calculation time) and in some cases can have problems of convergence. Climatology data are used to scale wind resources and energy production estimation according to the results from wind fields.

Windsim can give information on wind profiles in any point of the domain and uses exponential profiles only for wind field initialization at the boundaries.

In the same period energy production data was collected from turbine nr. 1 and anemometer data from measurement station; the database has been purified by possible malfunction periods.

Giving as input wind data AEP for actual and future turbine positions was estimated with numerical models using the manufacturer’s power curve adjusted for the site’s air density (1.112 kg/m3). Calculation domains were prepared using a digital terrain model (dtm) with a resolution of 25 m and the vegetation map for roughness definition.

Windsim simulation was carried out using the nesting technique: refined calculation was performed on a small domain (including turbines) for which wind field was initialized with results from the meso-model.

For AEP estimation anemometer data from meteorological stations and the production data from turbine nr. 1 that was used for validation were collected in the

same acquisition period. The acquisition period covers approximately 6 months mainly during spring and summer to avoid icing and represents quite well the mean average conditions for the site; the data set was purged from turbine’s unavailability periods (the turbine’s availability in the acquistion period was about 92%).

The complexity of local orography and the low measurement level (10 m) make the wind data collected from the nearest anemometer more difficult to be managed; for this reason in some cases AEP calculations are less reliable

Figure 2. Neg-Micon NM750/44 power curve provided by the manufacturer.

AEP (GWh)

turbine nr. WAsP Windsim 1 3.489 2.289 2 3.552 2.241 3 3.544 2.2 4 3.525 2.284

Table I. AEP calculation results using manufacturer’s power curve and climatology from the transferred anemometer station data (20 m a.g.l.).

Figure 3. Climatology resulting from the wind databases collected from the two stations used for the investigation.

American Institute of Aeronautics and Astronautics 4

using this data set. The use of the advanced option called “blocking” in the CFD model can be useful to avoid this problem performing the mesh refinement near the station; unfortunately the application of this tool is very hard-working and was not yet implemented in the present work. Estimations carried out with climatology data transferred from the farthest station (which collects wind speed data 20 m above ground and is placed in a less complex environment) are more reliable because climatology is less affected by local orography and can be considered more representative of the wind conditions of the overall region.

Figure 3 shows the wind rose and weibull distribution from the two stations: for the station with a tower 10 m high it is clear that wind rose is strongly affected by local conditions.

Results for AEP calculation using manufacturer’s power curve and climatology from the transferred anemometer station data are summarized in the table I.

The energy production for aerogenerator nr.1 in the acquisition period (3977 hours) was 803319 kWh; the AEP in this case can be estimated with a simple proportion equation:

GWh 769443.18760*3977

0.803319AEP

(3)

From this result it is clear that, for this test case, complex terrain leads numerical models to overestimate AEP: WAsP gives an error of about 80% which is considered unacceptable while for Windsim error is about 30% that cannot be considered a good result.

In any case Windsim in complex terrain works better than WAsP2 because wind field is calculated by solving Navier-Stokes equation in three dimensions so that turbulence induced by mountains can be taken into account.

III. The Aerogenerator Power Curve In the estimation of AEP a fundamental parameter is the aerogenerator power curve which describes the

variation of the machine power output in relation to the variation of windiness. Power curves are generally obtained after a testing period of a prototype aerogenerator in a test field. Measurements are carried out following international standard rules when certification of the power curve is

needed. IEC standards3 recommend that tests are to be performed in flat terrain environment (fig. 4) and, if this is not possible, site calibration is needed.

Test fields are generally located in flat terrain to avoid the site calibration. For a significant time period wind speed at hub height is measured in conjunction with net energy output to obtain the power curve.

In flat land wind profile is generally exponential and speeds experienced at different levels by the turbine’s rotors can be defined by equation (1).

This is not true in complex terrain environment where hills and ridges can induce different shapes for wind profiles4. For this reason first measurements with Sodar (Sonic detection and ranging) were performed in the Fossato di Vico Wind Farm in the location selected for turbine nr.4 (fig. 5). Sodar is able to measure wind speed at different heights above the ground detecting Doppler shift of sound signals.

Sodar measurements were performed in some spot (one-day) periods and were used to investigate the tuning of the numerical model: comparing numerical and experimental wind profiles the optimal settings for roughness and for the height of the boundary layer (which represents the mean stratification condition) can be evaluated. The tuning of the CFD model can obviously be refined using a wider set of sodar data.

Sodar data were collected with sodar model VT-1 from Atmospheric & Research Technology LLC provided by the Department of Industrial Engineering of the University of Perugia. For this equipment the vertical resolution can

Figure 4. Topography requirements recommendations for wind turbines test fields3.

American Institute of Aeronautics and Astronautics 5

be from 10 to 40 m depending on the maximum measurement height; in present work the height was set according

with a vertical resolution of 20 m that is useful to investigate wind profiles.

The sodar manufacturer has performed a measurements campaign to compare wind speed and direction collected by sodar model VT-1 and by a tower equipped with anemometers: for wind speed and wind directions correlation coefficients of about 0.99 between data collected from the two different instrumentation systems were discovered.

Measurements were compared with wind profiles calculated using Windsim setting for the meso-model different boundary layer heights related to the stratification condition hypothesis (stable, neutral and unstable). The values for boundary layer heights were derived applying the physics of atmosphere formulas5,6; such equations were useful for the initialization of the meso-model, while in the micro model boundary layer conditions were found out to be

strictly related to orography particularly near the ground.

In figure 6 the numerical profile is compared with the measured one: the closest results were found with the profile deriving from stable conditions in the meso-model, this is due to the meteorological conditions of the acquisition period.

From this it is clear that: 1. the wind profile in complex

terrain is different from the exponential profile of flat land;

2. windsim can be tuned up to reproduce wind profiles in a reliable way;

3. differences in wind profiles can induce different operational conditions from that of test fields for wind turbine. For this reason in the acquisition period for turbine nr. 1 the measured power curve was found to be quite different from the certified one

(fig. 7)7. Figure 7 shows the power curve provided by the turbine’s manufacturer in the certification documentation

(adjusted according to the site specific air density)7 compared with the curve collected for the turbine nr. 1 during the acquisition period; even if in the measured power curve the reference wind speed is collected through the

Figure 5. Sodar measurements lay-out.

Figure 6. Analysis of the wind profile resulting from measurements and calculation.

Power curves for turbine nr. 1

0

100

200

300

400

500

600

700

800

0 2 4 6 8 10 12 14 16 18 20

wind speed (m/s)

P (

kW)

certified power curve (air density 1.112 kg/m^3)

measured power curve

Figure 7. Comparison of measured and certified power curve for turbine nr. 1.

American Institute of Aeronautics and Astronautics 6

turbine’s nacelle anemometer (which is a part of the machine control system) and can be different from the undisturbed speed, the differences in production levels are too large.

IV. The power curve correction Differences in wind profile can be read as differences in energy content in the turbine’s swept range levels. As a

consequence the energy available for a wind turbine operating in flat land can be different from the energy available for a wind turbine operating in complex terrain even with the same wind speed at hub height.

In complex terrain wind profile can have different shapes depending on wind direction (fig. 8). This means that theoretical energy availability depends on wind direction even if the wind speed at the hub level is the same. From these observations came the idea to correct the wind turbine power curve using the wind profiles calculated by Windsim; in this way it is possible to build a method to customize power curve in relation to the turbine’s location orography. This can be done by comparing the energy content in the wind profile calculated for the turbine’s location with the exponential profile characterized by the same wind speed at hub height (representative of certifying conditions). The calculus of energy content can be performed by integration of Betz’s formula within the rotor swept area:

dzzvzHD

zvP rr

DH

DH

rr

rr

322

2

2

)(2

227

16

2

1)(

(4)

where P is the available power, v(z) is the function describing wind speed variation with height from ground z, Dr is the rotor diameter, Hr is the hub height and ? is the air density (fig. 9).

The energy content can be calculated for the exponential profiles and for the numerical profiles (in this case v(z) is the function interpolating numerical results from the fluid dynamics model) experiencing the same wind speed at hub level. By repeating this calculus for different wind speed levels for each direction sector, power curve correction can be obtained defining the correction factor as:

profile) ialP(exponentprofile) lP(numerica

R

(5)

R calculated for different wind speeds at hub height for each sector can be used as multiplier in the customization process for the power curve:

vP*mv,Rmv,P cercor

(6)

Figure 9. Energy content calculation for turbine’s swept area.

Figure 8. Calculated profiles for different direction sectors.

American Institute of Aeronautics and Astronautics 7

where Pcor is the corrected power curve, v is the wind speed at hub height, m is the index indicating the sector

number, Pcer is the certified power curve (adjusted according to air density) and R is the multiplier for correction.

In equation (6) m depends on the number of sectors for which CFD simulations were performed; the maximum number of sectors for Windsim is 24 but in present work only 12 sectors were used because in this way an accurate description of wind field directionality can still be obtained.

Even small differences in the speed profile can induce notable value for R because in equation (4) speed is characterized by the third exponent (fig. 10).

It is possible to customize the power curve for each turbine’s location for each sector; numerical models can work with different power curves for different locations but only need one power curve as input for each turbine.

A unique power curve for each turbine location can be derived according to climatology observations weighting sector power curve using sector frequencies as weight parameters.

Simulations for the calculus of the customized power curve were performed initializing the meso-model with a height of boundary layer which represents an intermediate level between neutral and stable stratification conditions to reproduce the mean climatology.

Power curve correction analysis for turbine nr. 1 revealed many situations such as that in figure 10 where available energy in complex terrain is less than that

available in flat land; as a consequence the customized power curve is below the certified one (fig. 12). The similarity of the customized power curve of figure 12 to the measured power curve of figure 7 is a

confirmation of the accuracy of the proposed method. The customized power curve is the most important result of this work; it depends only on the following

parameter: a) the wind profiles resulting from wind field simulation performed in each direction sector;

Figure 10. Areas of v and v3

for the exponential profile a), the numerical profile b) and interpolated experimental profile c) within the swept interval.

Figure 11. Variation of R with wind speed calculated for turbine nr. 1 for all the 12 direction sectors.

Figure 12. Comparison of the customized power curve with the certified one. .

a)

b)

c)

American Institute of Aeronautics and Astronautics 8

b) the frequencies for each direction sector derived from climatology data that are used to weigh the power

curves and obtain a unique customized power curve. Even if climatology data are not available for the turbine’s site the customization of the power curve for each

direction sector is still possible running CFD simulations.

V. Results Giving as input the customized power curve rather than the certified curve it is possible to perform a more

reliable AEP estimation. Results from table II demonstrate that a strong improvement in the assessment of AEP can be reached using the proposed method for the correction of power curve.

Using this method in conjunction with the wind flow simulation performed by Windsim the error for the overall process drops below 10%. The error in the AEP estimation for linear models (such as WasP) derives both from errors in the wind field calculation and from the misunderstanding of the turbine operation in complex terrain environment: in the results from table II the estimated mean wind speed at hub height has a large error in the case of WasP while Windsim gives a result closer to the real value (8.4 m/s).

In future a deepen experimental analysis of wind profiles is going to be performed with sodar and will be useful to tune up numerical simulation to obtain better results.

Further information on stratification will be derived collecting meteorological data from weather stations installed nearby the Wind Farm. Power curve correction method will also be validated with data collected from wind farms operating in similar terrains.

References 1Troen, I., Petersen, E. L., 1989, European Wind Atlas, RisØ National Laboratory. 2Moreno, P., Gravdhal, A. R., Romero, M., 2003, “Wind flow over complex terrain: application of linear and CFD model,”

European Wind Energy Conference and Exhibition, 16-19 June 2003, Madrid (Spain). 3IEC, 1998, “Wind turbine generator systems – Part 12: Wind turbine power performance testing,” – IEC International

Standard, first edition 1998-02. 4Castellani, F., and Franceschini G., 2002, “Advanced aerodynamics methods for wind site selection,” 21st AIAA/ASME

Wind energy symposium (AIAA-2002-0060) – 14-17 January 2002 Reno, Nevada. 5Zilitinkevich, S., Baklanov, A., 2002, “Calculation of the height of the stable boundary layer in practical applications”,

Boundary layer Meteorology, 105: 389–409, Kluwer, 2002. 6Georgieva, E., 1999, “WINDS – User’s Guide”, Release 3.2, Genova (ITALY), Fiat Lux Publications. 7Det Norske Veritas, 2003, “Type approval of: NM44/750 alternatively NM750-175/44 or NM44,” Det Norske Veritas,

Danmark, A-641063-5, 25-4-2003.

AEP (GWh) Wind speed (m/s)

WAsP 3.204 15 Windsim 1.842 8.0

experimental 1.769443 8.4 Table II AEP calculation and hub height wind speed results for turbine nr. 1 using the customized power curve and climatology from the transferred anemometer station data (20 m a.g.l.).