wind power capacity assessment mary johannis, bpa, representing northwest resource adequacy forum...
TRANSCRIPT
Wind Power Capacity Assessment
Mary Johannis, BPA, representing Mary Johannis, BPA, representing Northwest Resource Adequacy ForumNorthwest Resource Adequacy Forum
Northwest Wind Integration ForumNorthwest Wind Integration Forum
Technical Working GroupTechnical Working Group
October 29,2009October 29,2009
October 29, 2009 NW Wind Integration Forum 2
March 2007 NW Wind Integration Action Plan
• ACTION 1: By July 2007, the Northwest Resource Adequacy Forum (NWRA Forum) should reassess its 15 percent pilot sustained wind capacity value using currently available data on wind plant operation during periods of peak load. In 2008, the NWRA Forum should further refine the sustained peaking capacity value of wind power using the improved wind resource data set of Action 3 and other available data.
October 29, 2009 NW Wind Integration Forum 3
Phase I: Reassess15% Wind Capacity Value
• July 25, 2007 Forum Technical Committee Meeting– Wind Capacity Subgroup calculate wind capacity value
based on contribution to meeting load during sustained peak period of cold snap/heat wave events
– Contract with BorisMetrics to translate wind speed data into simulated data, to perform quality control of existing wind generation data and to evaluate the wind capacity value
• January 17 & February 28, 2008 Tech Meetings– BorisMetrics developed a 4th degree polynomial constrained
econometric model to backcast hourly project output as a dependent variable of Pendleton wind speeds (E. & W. Gorge areas)
– Concern that statistical attributes of backcast generation do not match actual wind generation attributes, i.e. many more instances of zero generation in actual records than in backcast simulation
October 29, 2009 NW Wind Integration Forum 4
Phase I: Placeholder Wind Capacity Value of 5% Selected
0.0
0.1
0.2
0.3
0.4
Pro
bab
ilit
y o
f O
ccu
ran
ce
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100Capacity Midpoint (%)
Median = 7.3 %50% of observations are below this value50% of observations are above this value
Distribution of Wind Fleet's Capacity FactorsIn BPA's Control Area Newer Technology Historical DataAveraged over 6 Peak Hours (n = 60) During Winter 3 Day Cold SpellsNov, Dec and Jan from 2002 - 2008
Mean or Average = 17.2 %
• Historical record is insufficient to calculate statistically significant wind capacity factor over 18 hour sustained peak period during cold snaps
• Median capacity factor over 6 peak hours during cold snaps is 7.3%
• Adverse wind capacity factor ≈ 5%
BPA Analysis: 1/
1/ 7/8/08 Forum Tech Committee Meeting
October 29, 2009 NW Wind Integration Forum 5
PHASE II: Long-term Plan to develop Wind Capacity Value
• Need sufficient years of hourly wind generation by wind site for GENESYS to perform Monte Carlo picks
• Options:– Backcast Wind Generation using
historical Anemometer records– Develop Temperature-Correlated
Synthetic Wind Generation Records
October 29, 2009 NW Wind Integration Forum 6
BorisMetrics Contract Identified Issues
• Can wind speed be used to backcast wind generation?– Example: East Gorge Generation Dec 2006– Why is there so little generation when the wind is blowing?– This example points outs problem with using off-site
anemometer to backcast wind generation
October 29, 2009 NW Wind Integration Forum 7
BorisMetrics Contract Identified Issues
• Can a unique function calculating generation based on wind speed be determined?
October 29, 2009 NW Wind Integration Forum 8
BorisMetrics Contract Identified Issues
• Pendleton Anemometer Data not Clean
Pendleton Ave Annual Anemometer DataHeight, reading method, and meter location are noted
5
6
7
8
9
10
11
12
1944
1946
1948
1950
1952
1954
1956
1958
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
Year
Ave
An
nu
al W
ind
Sp
eed
(m
ph
)
37' manualT-107 bldg
53' manterminal bldg
20' manualterminal bldg
20' automaticterminal bldg
30' automaticterminal bldg
October 29, 2009 NW Wind Integration Forum 9
Vansycle Backcast Case Study• Vansycle has an anemometer on-site
– ½ mile from the nearest generator– 6 miles from the furthest generator
• Wind speed data is available in 10 minute intervals for period• Scada data is available in 5 minute intervals for period
• Vansycle Backcast should be doable – Relatively long-term Generation Record – Relatively clean Anemometer Record
• Wind Turbine Power Characteristics:– Cut-in wind speed 4 m/s (8.9 mph)– Nominal wind speed 15 m/s (33.6 mph)– Stop wind speed 25 m/s (55.9 mph)
October 29, 2009 NW Wind Integration Forum 10
Vansycle Study Backcasting not feasible
• Lessons learned:– High R2 of multivariate regression (without zeros) and residual
analysis indicates that Persistence is an important feature in regression
– Other regressions have artificially high R2 by including zeros
– Prediction interval of .3 is not sufficiently tight to backcast
• Backcasting Wind Generation for NW is NOT feasible – Even on-site wind anemometers can be miles from some wind
turbines resulting in the LACK of a unique correlation
– Due to the persistence feature of the regression cannot use other means to reflect randomness in the correlation
– Insufficient on-site anemometer data to backcast the entire NW wind generation fleet
• Conclusion: Develop Temperature-Correlated Synthetic Wind Generation Records
October 29, 2009 NW Wind Integration Forum 11
Synthetic Wind Generation usingKth Nearest Neighbor Method
• What is Kth Nearest Neighbor Method?– For a time series of size N, randomly select a single or two consecutive of
the N observations then select the third based on how “close” the lag(s) for the selected observation are to the randomly selected observation(s)
• For example, select two hours where the capacity is 0.3 & 0.4, respectively, then pull from observations that have capacities that are close to .3 for the observation 2 hours prior and .4 for the hour prior.
• Creating a subset of the K “closest” observations to draw from maintains the structure that is expected in the time series.
• Methodology is undergoing peer review– A cross-correlated time series synthetic study presented to joint
conference of Western North American Region of the Biometric Society and the Institute of Mathematical Statistics
– Paper using method for wind fleet capacity factor data submitted to IEEE
– Kth Nearest Neighbor Method presented to NERC RIS-IVGTF team
October 29, 2009 NW Wind Integration Forum 12
Goal of Historic Temperature Correlated Synthetic Wind Gen• It has been well established that temperatures affect load where
extreme high or low temperatures translate into high loads.• The synthetic wind power generation data recreates certain
statistical characteristics of the original or observed wind power generation data set. The characteristics to focus on are:– Distribution/Density
– Lag Structure or Persistence
– Cross-Correlation
• The long-term temperature-correlated wind generation records will be incorporated into the existing resource adequacy studies using the GENESYS model, which will perform Monte Carlo picks on temperature-years, thus pointing to synthetic wind generation and loads
October 29, 2009 NW Wind Integration Forum 13
Average
Observed
CapacityFactor
Basedon2006through2008observationsof theBPAintegratedWindFleetduringheavyloadhoursasdef i nedbyNERCandmaximumregional averagetemperaturefromSeatac,Portland,andSpokaneairports.
33.1%
20.7%19.54%
October 29, 2009 NW Wind Integration Forum 14
Wind Generation vs. Temperature
October 29, 2009 NW Wind Integration Forum 15
Forum Wind Methods consistent with other Wind Forums
• NERC: Joint Integration of Variable Generation Task Force (IVGTF) – Resources Issues Subcommittee (RIS) Task 1.2 (Capacity Value) and Task 1.4 (Flexible Resources to integrate Variable Generation) Teams– IVGTF Report:
http://www.nerc.com/docs/pc/ivgtf/IVGTF_Report_041609.pdf
• WECC: Variable Generation Subcommittee (VGS) Planning Work Group
• Northwest: PNW Resource Adequacy Forum/NW Wind Integration Forum
October 29, 2009 NW Wind Integration Forum 16
Wind Capacity Value Methods• Effective Load Carrying Capability (ELCC) approach
– Evaluate effective wind capacity contribution based on LOLP studies with and without wind generation & same target (GENESYS Approach)
– Need sufficient wind generation data to simulate full range of generation under various conditions, especially if wind and loads correlated at times
– Need realistic depiction of combined uncertainties• Contribution of variable generation to system capacity
during high-risk hours using historical data– Investigate contribution of wind capacity during heat wave and cold
snap events in PNW because of evidence of statistical relationship between lack of wind generation when it gets very hot or very cold (Forum Wind Capacity Subgroup Approach)
• Correlation between resource contribution and the resource mix by system (e.g. what is appropriate for a hydro based system)– Wind may contribute more in energy-limited system if certain
amount of wind generation can be counted upon during drought
October 29, 2009 NW Wind Integration Forum 17
Counting Wind toward Capacity Adequacy in the NW
BPA Balancing Authority Area Load & Total Wind Generation
Jan. 5-25, 2009
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
1/5/
09
1/6/
09
1/7/
09
1/8/
09
1/9/
09
1/10
/09
1/11
/09
1/12
/09
1/13
/09
1/14
/09
1/15
/09
1/16
/09
1/17
/09
1/18
/09
1/19
/09
1/20
/09
1/21
/09
1/22
/09
1/23
/09
1/24
/09
1/25
/09
Date/Time (5-min increments)
MW
BPA TOTAL WIND GENERATION
BPA BALANCING AUTHORITY AREA LOAD
January 2009 Cold Snap
October 29, 2009 NW Wind Integration Forum 18
Synthetic Wind Generation: Historic Cold Snaps
October 29, 2009 NW Wind Integration Forum 19
Synthetic Wind Generation: Historic Cold Snaps
October 29, 2009 NW Wind Integration Forum 20
Simulated Wind Generation: Historic Heat Waves
October 29, 2009 NW Wind Integration Forum 21
Simulated Wind Generation: Historic Heat Waves
October 29, 2009 NW Wind Integration Forum 22
Observed Wind over all hours:Regional Load Duration
• Treating the wind as negative load changes the duration curve.– Minimum distance
between the two curves is about 1.6% of the nameplate.
– 99.5% of the hours have a 6.8% of the nameplate or greater “contribution” of wind toward reducing the load durations.
0% 20% 40% 60% 80% 100%
15,000
20,000
25,000
30,000
Percent of Hours Greater Than Load
RegionalLoad
(MW)
Duration Curves with and without Wind Adjustment
Minimum Difference = 35.7 MWat Observed Load of 31,962.4 MW
Observed Load Duration
Load Less Wind Duration
Duration curve based on all observed hours from 2006 through 2008, regional load provided by Council,wind fleet adjustment based on observed capacity factors multiplied by current wind fleet nameplate.
October 29, 2009 NW Wind Integration Forum 23
NW Wind Capacity Value using quasi-ELCC Approach
• Difference between the percentiles of the load durations show us:– Between the 10th and 90th
percentiles the contribution of the wind fleet was fairly flat with a slight trend of more energy during the lower loads.
– During the highest observed loads the difference is minimal.
0% 20% 40% 60% 80% 100%
0500
1000
1500
2000
Duration Percentile Corresponding to Difference
Difference
BetweenDurationCurves(M
W)
Difference of Regional Load DurationCompared to Load Less Wind Duration
Minumum Difference at Highest Observed Loads
Maximum Difference at the Lowest Observed Loads
ELCC Focus
October 29, 2009 NW Wind Integration Forum 24
NW Wind Capacity Value over 6 peak hours (high risk hrs)
• Alternately looking at the differences between the six peak hours with and without wind yields:– A minimum difference of
zero.– 97% of the time,
contribution is only .03% of nameplate in aMW.
– 91% of the time, “contribution” of 1% of the nameplate in aMW.
0% 20% 40% 60% 80% 100%
0500
1000
1500
2000
Percentage of Observations Less Than Difference
Difference
(aMW)
Difference Between Six Peak Load HourObserved Load and Observed Load Less Wind
October 29, 2009 NW Wind Integration Forum 25
18-hour wind capacity 2006
First Day18hr Mean Load
(aMW)18hr Mean Load Less
Wind (aMW)Difference
(aMW)% of Integrated Fleet Nameplate
Jan 2006 1/16/2006 26260.09 25471.7 788.39 34.52%
Feb 2006 2/16/2006 28775.7 28097.36 678.34 29.70%
Mar 2006 3/8/2006 26556.67 25366.26 1190.41 52.12%
Apr 2006 4/17/2006 22856.1 22194.6 661.5 28.96%
May 2006 5/16/2006 24012.67 23953.73 58.94 2.58%
Jun 2006 6/26/2006 26325.45 26045.51 279.94 12.26%
Jul 2006 7/23/2006 27300.8 26920.43 380.37 16.65%
Aug 2006 8/7/2006 24653.81 23934.57 719.24 31.49%
Sep 2006 9/5/2006 23707.89 23402.34 305.55 13.38%
Oct 2006 10/30/2006 26282.43 26114.24 168.19 7.36%
Nov 2006 11/27/2006 30132.9 29882.64 250.26 10.96%
Dec 2006 12/18/2006 29122.29 29120.18 2.11 0.09%
October 29, 2009 NW Wind Integration Forum 26
18-hour wind capacity 2007
First Day18hr Mean Load
(aMW)18hr Mean Load Less
Wind (aMW)Difference
(aMW)% of Integrated Fleet Nameplate
Jan 2007 1/15/2007 30014.62 30001.96 12.66 0.55%
Feb 2007 1/31/2007 28100.33 28072.56 27.77 1.22%
Mar 2007 2/27/2007 26696.8 25667.81 1028.99 45.05%
Apr 2007 4/2/2007 23559.33 23089.03 470.3 20.59%
May 2007 5/30/2007 23417.07 23392.29 24.78 1.08%
Jun 2007 6/19/2007 23667.9 23336.97 330.93 14.49%
Jul 2007 7/10/2007 26801.67 26639.63 162.04 7.09%
Aug 2007 8/13/2007 24816.52 24733.04 83.48 3.65%
Sep 2007 9/10/2007 22875.17 22729.48 145.69 6.38%
Oct 2007 10/31/2007 24168.59 23924.87 243.72 10.67%
Nov 2007 11/26/2007 27626.28 27150.03 476.25 20.85%
Dec 2007 12/10/2007 28796.63 28521.3 275.33 12.05%
October 29, 2009 NW Wind Integration Forum 27
18-hour wind capacity 2008
First Day18hr Mean Load
(aMW)18hr Mean Load Less
Wind (aMW)Difference
(aMW)% of Integrated Fleet Nameplate
Jan 2008 1/22/2008 30891.69 30822.25 69.44 3.04%
Feb 2008 2/4/2008 27867.22 26529.09 1338.13 58.59%
Mar 2008 3/26/2008 25613.51 24654.68 958.83 41.98%
Apr 2008 3/31/2008 25289.78 24821.24 468.54 20.51%
May 2008 5/17/2008 23016.08 22142.59 873.49 38.24%
Jun 2008 6/30/2008 26012.58 25591.03 421.55 18.46%
Jul 2008 7/7/2008 25511.94 25088.61 423.33 18.53%
Aug 2008 8/13/2008 26222.16 26016.39 205.77 9.01%
Sep 2008 9/15/2008 22821.76 22705.31 116.45 5.10%
Oct 2008 10/22/2008 22961.12 22564.17 396.95 17.38%
Nov 2008 11/24/2008 24885.25 24809.17 76.08 3.33%
Dec 2008 12/15/2008 32175.08 31638.4 536.68 23.50%
October 29, 2009 NW Wind Integration Forum 28
Status of Wind Discussions
• 10/16/09 Resource Adequacy Forum Technical Committee Meeting– Evidence suggests that 5% Placeholder Value for
Wind Capacity is too high– Use different WINTER & SUMMER values for
Wind Capacity in regional Planning Reserve Margin (PRM) Calculation:
• PRM = (∑118 hr regional resources – regional 1 in 2 load)/
∑118 hr regional 1 in 2 load
• Current physical resource adequacy thresholds are:– PRMwinter ≥ 23%– PRMsummer ≥ 24%
October 29, 2009 NW Wind Integration Forum 29
Next Steps
• In the Short-term, refine Wind Capacity Value in PRM Equation– SOME OPTIONS:
• Select 95th percentile wind from summer & winter all hour wind vs. load duration curves ~ Slide 23
• Select 95th percentile wind from summer and winter 6 peak or 18 peak load hour duration curves ~ Slide 24
• Using actual (and possibly synthetic) wind generation data over historical heat wave and cold snap events, calculate average wind capacity contribution over 18 hour sustained peak or 6 peak hour period
• In the Long-term, perform true ELCC evaluation using Monte Carlo picks of temperature-years to point to loads and wind generation– Create additional long-term temperature-correlated synthetic
wind generation records for use in GENSYS