Professional Documents
Culture Documents
Khanh T. Tran
AMI Environmental
206 Black Eagle Ave, Henderson, NV 89015
(702)564-9186
http://www.amiace.com
Wind Energy Forecasts
Growing Wind Energy Development
Wind Energy is Intermittent
Reliable Forecasts needed for Load
Scheduling
Forecasts of 48-Hour and Longer
Forecasts Updated Twice Daily
Facility located in Complex Terrain
AMI Forecasting System
AMI System Modules
Mesoscale Model MM5/WRF
> Nested Grids of 36,12,4,1.33 km spacing
> ETA/AVN Forecasts for IC/BC
Diagnostic Wind Model
> 100 m spacing or less
Adaptive Statistical Module
> Bias Removal
> Recent Onsite Data
Internet-based Forecast Delivery (website, email)
System Runs on Linux Multiprocessor PC
System Testing
Testing at FPL Southwest Mesa Plant in
southwest Texas
12-month Testing (April 2002-March 2003)
sponsored by US DOE NREL and EPRI
Forecasts by AMI, TrueWind and Risoe
48-Hour Forecasts twice daily at 00 UTC and
12 UTC (wind speed, wind direction, ambient
temperature and wind energy)
FPL Southwest Mesa Plant
MM5 Modeling Domain
Domain Topography
System Performance Evaluation
Forecasts compared against actual
observations (wind speed and energy)
Forecasts compared against persistence and
climatological forecasts
Statistical measures (mean error ME, mean
absolute error MAE, skill score)
Sample Wind Speed Forecast
Applied Modeling Wind Speed Forecasts
Southwest Mesa Forecasts Beginning at 1800, July 18 2002
Meteorology Fcst Persistence Fcst Climatology Fcst Observed Data
14
12
Wind Speed m/sec
10
0
0 6 12 18 24 30 36 42 48
Forecast Hour Number
Sample Wind Energy Forecast
Applied Modeling Wind Generation Forecasts
Southwest Mesa Forecasts Beginning at 1800, July 18 2002
60,000
50,000
40,000
30,000
20,000
10,000
0
0 6 12 18 24 30 36 42 48
Forecast Hour Number
Forecast Performance Statistics
Wind Speed (m/s) Energy (kw)
Normalized Error
Annual ME -0.2% -5.2%
Annual MAE 30% 48.4%
Skill Score
vs. Persistence 26.1% 28.3%
vs. Climatology 26.8% 28.1%
Normalized MAE for Wind
Speed
Normalized Mean Absolute Errors for
Applied Modeling Wind Speed Forecasts
Southwest Mesa FPL Energy, April 1, 2002 - March 31, 2003
50%
40%
30%
20%
10%
0%
0 6 12 18 24 30 36 42 48
Forecast Hour Number
Normalized MAE for Wind Energy
Normalized Mean Absolute Errors for
Applied Modeling Wind Generation Forecasts
Southwest Mesa FPL Energy, April 1, 2002 - March 31, 2003
80%
60%
40%
20%
0%
0 6 12 18 24 30 36 42 48
Forecast Hour Number
Normalized Monthly MAE for WS
Normalized Monthly Mean Absolute Errors for
AMI Wind Speed Forecasts
Southwest Mesa FPL Energy, April 1, 2002 - March 31, 2003
Climatology Meteorology Persistence
Normalized Mean Abs. Error
70%
60%
50%
40%
30%
20%
10%
0%
y t er
0 2 ay n e l us be
r er er 03 ar
y
rc
h
20 Ju Ju b b b 0
M ug m ct
o m m 2 br
u
M
a
p ril A pte O o ve e ce a ry
F e
A S e N D n u
Ja
Month
Normalized Monthly MAE for WE
120%
100%
80%
60%
40%
20%
0%
y
03
t
e
ry
ay
ch
r
02
er
s
be
be
l
be
n
Ju
gu
ua
20
ob
M
ar
Ju
20
em
em
em
Au
br
M
ct
y
ril
ov
Fe
ar
O
pt
ec
Ap
Se
nu
N
D
Ja Month
Comparison of Forecast Systems
Annual Mean Absolute Errors of Wind Speed
and Energy Forecasts
Southwest Mesa FPL Energy, April 1, 2002 to March 31, 2003
60%
Risoe TrueWind AMI
Normalized Mean Abs. Error
50%
40%
30%
20%
10%
0%
Wind Speed Wind Energy
Conclusions
AMI Forecast System based on advanced models
AMI system tested for 12-month in southwest
Texas
AMI system provides accurate wind speed and
energy forecasts
AMI system is more accurate than other systems
AMI system runs on Linux PC and can be easily
adapted to other sites.