SLIDE 1
Parsons Brinckerhoff
- Parson Brinckerhoff
– >14,000 people – 150 offices – six continents
- PB Power Asia
– 500 people – Most major Asian cities
- PB Power Wind
Energy Prediction Uncertainty Assessment for Wind Power Project - - PowerPoint PPT Presentation
Energy Prediction Uncertainty Assessment for Wind Power Project - - PowerPoint PPT Presentation
Energy Prediction Uncertainty Assessment for Wind Power Project Using Monte Carlo Analysis Qi Jianan, Manoj Kumar Singh Parsons Brinckerhoff Parson Brinckerhoff >14,000 people 150 offices six continents PB Power Asia
SLIDE 2
SLIDE 3
An example project in India
25 units of 2MW turbine Hub height: 80 m 1 wind masts at 65 m height 12 months’ site measurement Reference wind data –NCAR/MERRA Net Energy output: 120 GWh per year
50% confidence level (P50)
Investors and banks need P75, P90, P95 or even
- P99. This can be done through uncertainty analysis
SLIDE 4
Uncertainty sources
Sources of uncertainty in predicted wind farm performance
Uncertainty Wind measurement Cross-correlation Extrapolation to hub height Wind flow/wake model Loss factor assumptions Variability Inter-annual wind speed variation Turbine performance, etc
SLIDE 5
Uncertainty- measurement
Sensor type, class Calibrations Mounting arrangement on the mast IEC 61400-12-1 as a guideline The measurement made at the example project is satisfactory
2% uncertainty
SLIDE 6
Uncertainty – cross correlation
- High quality reference data is important to
reduce uncertainty
- The quality of the correlation depends on
Distance from the site to the reference Measurement height of site & reference Complexity of the site terrain
- Correlation quality can be judged from R2
5 10 15 20 25 5 10 15 20 Traces Linear Fit Zero Fit
Overall Correlation
Reference Site (m/s) Measured Site (m/s) mean_Ref_match 5 10 15 20 5 10 15 20 Traces Linear Fit Zero Fit
Overall Correlation
Reference Site (m/s) Measured Site (m/s) mean_Ref_match
The cross-correlation for the example project is acceptable
3% uncertainty
SLIDE 7
Uncertainty- vertical extrapolation (shear)
6 6.5 7 7.5 8 8.5 10 20 30 40
Corrected wind speeds Wind shear profile Uncorrected wind speeds Wind Shear Profile
Wind speed (m/s) Height (mAGL)
If no hub height measurement present, extrapolation is needed Wind shear varies at different wind speed, direction and time Parsons Brinckerhoff uses shear matrix to do the vertical extrapolation Vertical extrapolation from 65 m to 80 m for the example project
2% uncertainty
SLIDE 8
Variability – inter-annual wind speed
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 4 4.25 4.5 4.75 5 5.25 5.5 5.75 6
Annual Averages
Years Average Wind Speed (m/s)
Annual mean wind speed exhibit yearly variation Vary from site to site and can be estimated from long-term data Proportional to length and quality of reference data For the example project, the inter-annual variation is 7% uncertainty!!
SLIDE 9
Uncertainty – wind flow model
Resolution of map Site complexity Ground cover Distance from the mast to the turbines Size of the project (No. of turbines) For the example project, the wind flow/wake model is estimated
8% uncertainty
SLIDE 10
Monte Carlo uncertainty analysis
Source: http://www.vertex42.com/ExcelArticles/
Stochastic/random method Applicable to most probability distribution Sufficient sample to achieve convergence Final uncertainty obtained from actual energy distributions from random simulations Allows for non-linear model
Parsons Brinckerhoff uses Monte Carlo !
SLIDE 11
Results
Uncertainty source Uncertainty Correlation uncertainty 3% Anemometer calibration 2% Shear 2% Wind speed annual variation 7% Wind flow model 8%
Example project in India 5000 iteration Results-normal distribution Estimate P99, P90, P75 P90- 105 GWh per year for 20 years period
SLIDE 12
What happens if…
Uncertainty source Uncertainty Correlation uncertainty 3% Anemometer calibration 3% Shear 2% Wind speed annual variation 7% Wind flow model 12%
Example project in India P90 Change from 105 GWh to 101 GWh !! 4% revenue drop if sell the project using P90 Less debt size and lower IRR for debt financing 4% drop
SLIDE 13