Transcript No Slide Title

Perspectives from Abroad
Sustainable Energy Ireland, Dublin 13 June
Wind Farms in a Gross Pool Market:
Specific issues for wind farms
Hugh Outhred
School of Electrical Engineering and Telecommunications
The University of New South Wales
Sydney, Australia
Tel: +61 2 9385 4035; Fax: +61 2 9385 5993;
Email: [email protected]
www.ergo.ee.unsw.edu.au
Outline
• Trends in wind energy
• Network-related issues
• Power variability issues
– Forecasting wind farm output
– Spot & derivative markets
• Are wind farms viable in the National
Electricity Market?
2
Wind turbine installations in Australia:
history & forecast
Summary of wind farm projects
at 3/03. Approximate, based on
www.auswea.com.au
Completed
105 MW
Under
Construction
106 MW
Tendering
230 MW
Approved
294 MW
Planning
1400 MW
Total
2135 MW
(AusWEA: IEA Annual Report, 2002)
3
Australian wind farm planning
experience to date
• Limited experience to date:
– Some strong support, some strong opposition
• Mixed federal, state & local government
approvals process lacks coherence:
– Project based - may not manage cumulative
issues & interactions well
• Other industries have a comprehensive
planning framework, eg:
– Strong, state-based planning framework for
the minerals industry
4
Network issues for wind farms #1
• Networks are shared, centrally planned resources:
– Must limit disturbances caused by wind farms
– Must survive disturbances from the network
• Renewable resources are often distributed
differently from fossil fuel resources:
– Weak network conditions likely to be more common in
Australia than Europe or North America
• Network must be built to carry peak flows:
– Want good estimates of aggregation & seasonal effects
• Benefits of staged development of wind resources:
– Network savings; reduced voltage & frequency impacts
5
Network issues for wind farms #2
• Wind turbine starting & stopping transients:
– Severity can be alleviated by soft-start &
high wind-speed power-management
• Some wind turbine designs:
– May cause voltage distortions:
• Harmonics &/or transients
– May have poor power factor, eg:
• Uncompensated induction generator
– May not ride-through system disturbances
• Temporary voltage or frequency excursions
6
Wind turbine type comparison
(Slootweg & Kling, TU Delft, 2003,
http://local.iee.org/ireland/Senior/Wind%20Event.htm)
7
Size of wind turbines used by
Western Power (www.wpc.com.au)
8
Wind turbine starting transients
for Esperance 2 MW wind farm
• 9 x 225 kW turbines with squirrel cage IG
• Magnetisation inrush current may cause a
voltage dip - starts should be spaced out
(Rosser, 1995)
9
Network connection issues &
examples
• Approximate ability of a transmission line to
accept a wind farm:
–
–
–
–
66kV
≤
20MVA
132kV
≤
100MVA
330kV
≤
200MVA
Constraints may be determined by several factors:
• Thermal, voltage, fault clearance, quality of supply
• Thermal ratings depend on line temperature & wind speed
• Relevant wind farm rating is its maximum output,
not the sum of turbine rated powers:
– Coincident output of the connected wind turbines
10
Connection costs to 330kV
(Transgrid, 2002)
Wind farm Total wind Conn. cost Conn.cost
number
MW
$M
$/kW
1
5
12.7
2,500
1
20
12.9
650
2
100
17.7
180
4
200
28.3
150
Important to capture economies of scale of grid connection
11
NEMMCO concerns about
wind energy (NEMMCO, 2003)
• Frequency control in normal operation:
– Frequency regulating service costs ~5 $/MWH
• Security control - largest single contingency
– Will wind farms ride-through disturbances?
• Interconnection flow fluctuations:
– Exceeding flow limit may cause high spot price
• Forecast errors due to wind resource uncertainty:
– Five minute dispatch forecast (spot price)
– Pre-dispatch & longer term (PASA & SOO) forecasts
12
Western Power’s proposed wind
penalty charge (c/kWh) (Western Power, 2002)
13
Demand forecast errors
South Aust,02 Q4 (NECA, 02Q4 Stats, 2003)
14
Spectral analysis of Danish longterm wind data (17 years of data)
Spectral gap between weather
and local turbulence phenomena
(Sorensen, 2001, Fig 2.110, p194)
15
Forecasting the output of
wind farms
• 30 minute horizon (FCAS & spot market):
– Turbulence spectrum - likely to be
uncorrelated for turbines spaced > 20 km:
• Then % power fluctuations ~ N-0.5
– eg for 100 identical wind farms spaced >20 km apart,
%fluctuation in total power ~ 0.1x%fluctuation for 1 farm
• 30 minutes to ~3 hours:
– ARMA model best predictor of future output
• > 3 hours:
– NWP model best predictor
16
One-second power fluctuations at
Esperance 2MW wind farm
• 9 x 225 kW turbines
• Solid line is proportional to N-0.5
– Implies 1-second fluctuations are uncorrelated
(Rosser, 1995)
17
Forecasts for Lake Benton wind farm, USA
138 turbines, 103.5MW, hourly data
(Hirst, 2001)
Two-hour ahead prediction of wind power:
MWPred(T+2) = 2.7 +0.9xMW(T) + [MW(T) - MW(T-1)]
18
Combined output of 2 wind farms
80 km apart (Gardner et al, 2003)
19
Cross-correlations between measured
power outputs of German wind farms
(Giebel (2000) Riso National Lab, Denmark)
20
Cross-correlations between 34 years of 12hourly data for all grid points
(Giebel (2000) Riso National Lab, Denmark)
21
CSIRO WindscapeTM model
(www.clw.csiro.au/products/windenergy)
Windscape derives location-specific wind forecasts
from a Numerical Weather Prediction model
(Steggle et al, CSIRO, March 2002)
22
(Steggle et al, CSIRO, March 2002)
• Windscape predictions of annual mean wind
speed at 65 m, showing nested model results
• More rapid changes in colour
probably imply higher local turbulence
23
SEDA NSW Wind atlas
(www.seda.nsw.gov.au)
24
Issues for NEM spot market
• Wind farms will operate as “price takers”:
– Generate whenever wind is blowing
• NEM spot market prices are volatile with a
“rectangular” price distribution:
– Prices are usually low, sometimes high
– Timing of high prices not easily predicted
• Value of wind energy in the spot market:
– Will depend on how regularly wind farms are
producing when spot prices are high
25
1GW wind contribution to meeting SA Load
(simulation study)
1997-98 Financial Year
SA load duration curve and the reduction from the high wind penetration case
100%
GDC band 1
90%
80%
GDC band 2
GDC band 3
GDC band 4
GDC band 5
% of Peak Load
70%
60%
50%
40%
30%
20%
10%
0%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
55%
60%
65%
70%
75%
80%
85%
90%
95% 100%
% of time
26
Demand
demand - Generation from high wind penetration
Spot price as a function of demand
SA,02 Q4 (NECA, 02Q4 Stats, 2003)
27
Weekly average NEM spot prices
since market inception (NECA, 02Q4 Stats, 2003)
28
Forward prices for wind energy
• Wind farms may
have to accept a
lower price than
“flat contract” due
to uncertainty in
production:
– Daily
– Seasonal,
– Annual
(Giebel (2000) Riso National Lab, Denmark)
29
Flat contract prices, 1999-2006
(NECA, 02Q4 Statistics, 2003)
30
Renewable Energy Certificate
Prices (A$/MWH)
60
55
50
45
40
35
30
25
20
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
31
Wind farms marginal at $70/MWH
(PWC, 2002)
32