This document discusses mapping icing in Sweden using mesoscale models. It summarizes how mesoscale models work by resolving low pressure systems and fronts down to 1km resolution. It then shows mesoscale wind index maps from 2000-2011 and how power production was lower than expected in some winters due to icing not captured by the index. The document concludes that including icing effects can improve energy production forecasts and that icing variability often dominates annual wind variability in production levels.

1. Mapping of icing in Sweden
The benefits of forecasting icing for
energy production
gy p
Øyvind Byrkjedal
Kjeller Vindteknikk
oyvind.byrkjedal@vindteknikk.no
Wintewind 2012, Skellefteå 07.02.2012

2. Mesoscale model
 The meso-scale refers to the time and spatial scale of the features
resolved by the model. The model resolves low pressure systems and
model low-pressure
fronts.
 WRF – Weather Research and Forecasting
 The
Th model d
d l describes the atmosphere d
ib h h dynamics ( i d temperature
i (wind,
and humidity), and includes physical description of radiation, cloud
formation, precipitation, snow, surface processes, etc.
 The model performes calculations in the time domain, no steady-
state model
 Typical model resolution down to 1km x 1km
2

3. Mesoscale wind
index
2000
3

4. Mesoscale wind
index
2001
4

5. Mesoscale wind
index
2002
5

6. Mesoscale wind
index
2003
6

7. Mesoscale wind
index
2004
7

8. Mesoscale wind
index
2005
8

9. Mesoscale wind
index
2006
9

10. Mesoscale wind
index
2007
10

11. Mesoscale wind
index
2008
11

12. Mesoscale wind
index
2009
12

13. Mesoscale wind
index
2010
13

14. Mesoscale wind
index
2011
14

15. Calculation of in-cloud icing
 We calculate icing on a standard body following
Makkonen.
Makkonen 1m cylinder with diameter 30mm:
dM
  1 2 3  w  A  V α1- collision efficiency, α1=f(V,d,D)
lli i ffi i f(V d D)
dt α2- sticking efficiency, α2 ≈ 1
α3- accretion efficiency, α3=
f(V,d,w,T,e,D,α1)
w – cloud liquid water content
A – collision area, perpendicular to flow
V – Wind speed

16. Power production index
%]
ndex [%
Power in
16

17. Power production index
%]
ndex [%
Power in
The winter 2008/2009:
According to the wind index we would expect more production than normal.
Due to icing we would find less production than normal.
17

18. Power production index
%]
ndex [%
Power in
The winter 2010/2011:
Large ice loads and large production losses
18

19. Production
forecasts
ncy
Relative frequen
The analysis is based on 3 years
of operational data from a wind
farm in Sweden
f
Forecast error [%] 19

20. Production
forecasts
ncy
Relative frequen
The analysis is based on 3 years
of operational data from a wind
farm in Sweden
f
The overall forecast errors
are reduced by incluing
forecasts of ice
Forecast error [%] 20

21. Production
forecasts
ncy
Relative frequen
The analysis is based on 3 years
of operational data from a wind
farm in Sweden
f
With the current model we
particulary reduce the
number of cases when the
production is overpredicted
Forecast error [%] 21

22. Production
forecasts
ncy
Relative frequen
The analysis is based on 3 years
of operational data from a wind
farm in Sweden
f
We get an increase of the
number of cases when we
underpredict production
Forecast error [%] 22

23. Conclusions
 For a large number of sites the variability in icing from
year to year will dominate over the variability in wind
in the production index.
 An improvement in the energy production forecasts
can be achieved by including the effects of icing
23

24. Thank you!
Wind and icing map are now available for Sweden:
www.vindteknikk.no
24