Better Wind Resource Estimation through Detailed Forest Characterization

Better Wind Resource Estimation through Detailed Forest Characterization

Jens Madsen & Adrien CorreVattenfall R&D

EWEA 2011 (Session: Siting Challenges)

Bruxelles, March 14.17, 2011

2 | EWEA 2011 | Jens Madsen | March 2011

Locating Wind Farms in Forested Areas …

What is the problem?• Forests induce high wind shear and turbulence• Reasonable wind speeds only at higher hub heights

So, build somewhere else then …• Sweden has 60-65% forest coverage• Onshore projects in/near forested areas is the rule

rather than the exception

>50% Forest

Dairy farming

Meadows / pastures

Tundra

Intensive farming

3 | EWEA 2011 | Jens Madsen | March 2011

CFD and Forest Flows

• CFD as a Wind Resource Assessment Tool

- Overcomes shortcomings of linearized flow models … including forested areas

- Technical Risk Mitigation: Mapping of severe conditions (turbulence, shear, inflow angle)

- Economical Risk Mitigation: Potential to significantly reduce wind resource uncertainty

• CFD approach to Forest Canopy Modeling

- Canopy represented by porous zone: drag resistance & turbulence modulation

- Applies first principles

4 | EWEA 2011 | Jens Madsen | March 2011

CFD Forest Canopy Model (by Katul et al.)

• Inside a forest canopy, (z<H) momentum sinks are applied

- This drag resistance depends on Leaf Area Density (LAD or α) [m2/m3]

- Leaf Area Index (LAI) is the corresponding integral forest density [-]

G.Katul et al. : ”One- and Two-Equation Models for Canopy Turbulence”

Boundary-Layer Meteorology, Vol.113, pp.81-109, 2004

H

z

dzLAI0

5 | EWEA 2011 | Jens Madsen | March 2011

Model Parameter Sensitivity – Uniform Forest, Flat Terrain

• Preliminary sanity check

using 2D model

• LAD profiles mostly impact

wind speeds within canopy

• Correct tree height H is much

more important than forest

density

• In particular true at typical

hub heights

• Highest sensitivity to model

parameters occur near step

changes in height and density

(forest edges, clearings)

6 | EWEA 2011 | Jens Madsen | March 2011

MM-1MM-2

Model Validation – Risø/DTU Forest Edge Experiments

7 | EWEA 2011 | Jens Madsen | March 2011

Risø/DTU Forest Edge Experiments

LAD data backed out from CFD (courtesy of Andrey Sogachev)Seasonal variation in forest density

8 | EWEA 2011 | Jens Madsen | March 2011

Model Validation – MM-2 (inside forest)

• Canopy models are sufficiently good …- … considering the poor parameters we feed into them (GIGO principle applies)

• Implement advanced forest characterization techniques- The idealized, homogeneous forest does not exist

- Spatial distribution of forest height (and density)

- What is the impact of a considering a more realistic, heterogeneous forest layout?

9 | EWEA 2011 | Jens Madsen | March 2011

Forest Characterization – Classical approaches

Spatial Layout• Classification from satellite/aerial images• Forest perimeters identified • Digitized vegetation map

Tree Height• Assessed through site inspection• Extensive lumping is necessary

10 | EWEA 2011 | Jens Madsen | March 2011

• Direct methods:

- Foliage samples, destructive testing

• Indirect methods:

- Measure Fraction of transmitted radiance

- Hemispherical image analysis (LAI)

Forest Characterization – Canopy density (LAD/LAI)

Z

LAD

hemispherical photography

• Tedious methods!• Limited point-wise sampling• Calibration of density level

11 | EWEA 2011 | Jens Madsen | March 2011

LIDAR Airborne Forest Imaging (cont’d)

• Technology used in Forest Inventory Management

- Laser beam is reflected either by vegetation or ground

- Scans 500-800 meter wide section per flight leg

- <10 cm accuracy (depending on flight height)

• Data acquired (resolutions up tp 1x1 m2)

- Digital Terrain Model (DTM)

- Digital forest model

• Spatial variation of tree heights and density profiles inferred from

point cloud percentile values, e.g. zP-90, zP-75, zP-50, and zP-25

12 | EWEA 2011 | Jens Madsen | March 2011

Int.

Time

LIDAR Airborne Forest Imaging

13 | EWEA 2011 | Jens Madsen | March 2011

LIDAR Point Sky Rendering (colors by veg.height)

14 | EWEA 2011 | Jens Madsen | March 2011

Detailed forest layout: wind speed distribution

Forest Layout Wind Speed @ 10m agl.

Lake with associated speed-up

15 | EWEA 2011 | Jens Madsen | March 2011

Detailed forest layout: CFD vs. met mast data

• Use of detailed forest layouts yield good agreement between CFD and on-site measurements for this Vattenfall site in Southern Sweden

- Instrumented telecom mast (93m) and LIDAR campaign (Vestas)

0

20

40

60

80

100

120

140

160

0 0,5 1 1,5 2

Met Mast

LIDAR

CFD

0

20

40

60

80

100

120

140

160

0 0,5 1 1,5 2 2,5

Met Mast

LIDAR

CFD

16 | EWEA 2011 | Jens Madsen | March 2011

Conclusions

• General on wind power in forest- Forested areas are problematic but inevitable sites (in some geographies)

- Extensive measurement campaigns (tall masts, SODAR and/or LIDAR)

- Use taller hub heights than you would normally do

• Wind resource assessment in forested areas- The CFD canopy models perform well

- Detailed forest characterization provides more accurate results than modeling based

on idealized, homogeneous forest

- Correct tree height distribution is the key information (canopy density, less so)

- Beware of information overkill.

• Investigate trade-off between forest data resolution and accuracy

17 | EWEA 2011 | Jens Madsen | March 2011

Thanks for your attention

Made possible throguh the collaboration with:

Risø/DTU, Vestas Technology R&D, and ETS Montreal