978-1-4244-8417-1/11/$26.00 ©2011
2
C. Reserve Requirements
Both the variability of generation and the variability of the
load impact the ability of a system operator to keep a power
system balanced on a minute-by-minute, hourly, or daily basis.
To the operator, variable renewable generation appears as a
negative load, and thus all reserve calculations must be
completed utilizing the concept of net load (i.e., load-minus-
renewable-generation).
To that end, the utility industry (and BPA in particular)
currently utilizes three different timescales to calculate reserve
requirements. The first, regulation, is defined as the
difference between the minute-to-minute load-minus-
renewable-generation and the 10-minute average load-minus-
renewable-generation. The second timescale, following, is
defined as the difference between the 10-minute average load-
minus-renewable-generation and the hourly average load- Fig. 2. Illustration of differences between following and imbalance reserve
minus-renewable-generation. The final timescale, imbalance, requirements.
3
Fig. 3. Plot of estimated values and 95% confidence intervals for main-effect Fig. 5. Plot of estimated values and 95% confidence intervals for main-effect
beta parameters for following reserve requirement. beta parameters for imbalance reserve requirement.
Fig. 4. Plot of estimated values and 95% confidence intervals for interaction- Fig. 6. Plot of estimated values and 95% confidence intervals for interaction-
effect beta parameters for following reserve requirement. effect beta parameters for imbalance reserve requirement.
ZDYH ȕ3) main-effect terms. To clarify the results, Table 1 and interaction-effect parameters are meaningful on the
presents the estimated values and their confidence intervals for overall following reserve requirement.
HDFK RI WKH ȕi parameters for the following reserve In comparing the interaction-effect parameters, it is
requirement, and Table 2 presents the results for the imbalance LPSRUWDQWWRQRWHWKDWZKLOHȕ4 LVSRVLWLYHWKHRWKHUWKUHHȕ5,
reserve requirement. Please note that the scales for Fig. 3 and ȕ6DQGȕ7) are all negative. Given the model presented in (4),
Fig. 5 are different. this implies that the interaction effects tend to cause a reserve
Both the plot in Fig. 3 and the values in Table 1 requirement that is lower than that which would be found if
demonstrate that the main-effect parameter for wind is much the reserve requirement was predicted from an analysis of the
larger than those for solar and ocean wave (which are both reserve requirement for single sources. For example, if
very similar in value). This can be interpreted to mean that interaction effects are ignored, 15% penetration of wind, 5%
wind has a larger contribution to the reserve requirements than penetration of solar and 5% penetration of ocean wave would
do solar and ocean wave, which correlates well with prior be predicted to require 0.053723 reserve (pu), whereas if the
results in [6]. A similar observation holds true for the interaction effects are included, 0.053709 reserve (pu) is
imbalance reserve requirement scenario as well (Fig. 5 and required. This is a fascinating result, as it hints at the
Table 2), with the difference that in the imbalance reserve possibility for synergistic effects occurring when the various
case, the solar main-effect term has a greater impact than the renewable resources are combined in these specific ways.
ocean wave main-effect term. However, it should also be noted that some of the interaction
In analyzing the results for the following reserve betas are positive, so the total negative interaction contribution
requirements, it can be noted that the confidence intervals for depends on the specific renewable mix. It should also be
each of the eight beta parameters do not include zero. Thus, noted that the interaction-HIIHFW EHWDV ȕ4 WKURXJK ȕ7) are an
the null hypothesis can be rejected in each of these cases. This order of magnitude lower than the main-HIIHFW EHWDV ȕ0
leads to the conclusion that the results for both the main-effect
5
TABLE I TABLE II
RESULTS FOR FOLLOWING RESERVE REQUIREMENTS BETA PARAMETERS RESULTS FOR IMBALANCE RESERVE REQUIREMENTS BETA PARAMETERS
WKURXJKȕ3). This suggests that while a synergistic effect may of allowing a greater combined penetration rate for the same
exist, it also may be small. reserve requirement level. This is an important consideration
A similar analysis can be done for the imbalance reserve for power system operators who are planning for increasing
requirement case. Similar to the following reserve renewable power portfolios in the future.
requirements, the confidence intervals for each of the eight
beta parameters do not include zero. Thus, again, the null IV. MODEL VALIDATION
hypothesis can be rejected in each of these cases, which leads Given the initial results that reject the null hypothesis for
to the conclusion that both the main-effect and interaction- each beta parameter in both the following and imbalance
effect terms are contributing in a meaningful way to the reserve requirement cases, it is useful to further examine the
overall imbalance reserve requirement. validity of the model in (4) by applying it to a new set of wind,
However, unlike in the following reserve requirement case solar, ocean wave, and load data and checking to see how
where three of the four interaction-effect parameters were close the actual calculated reserve requirements are to those
negative, only two are negative in the imbalance reserve predicted by the model. In this case, since only the
requirement case: ȕ5 (wind & wave) and ȕ7 (wind & solar & synthetically-generated sets of wind, solar, and wave data
wave). Of particular interest is the fact that the confidence were used to generate the beta parameters, the real data sets
intervals for the interaction-effect parameters in the imbalance (upon which the synthetic data are based) were not directly
case are generally much larger than those in the following used and thus remain a valid resource with which to perform a
case, indicating a larger spread of results. This is in line with simple check to validate the model.
previous experience that indicates the imbalance reserve Tables 3 and 4 present the model validation results for the
requirement is generally larger and often more volatile than following and imbalance cases, respectively, with the actual
the following reserve requirement [6]. reserve, estimated reserve (from the model (4) using the
The results from both the following and imbalance cases parameters in Tables I and II), and percent error. Two models
demonstrate that there are specific combinations of renewable were verified: a reduced model that uses only the main-effect
resources that could possibly have a beneficial effect in terms betas (ȕ0 through ȕ3 in (4)), and the full model that uses all
TABLE III
MODEL VALIDATION RESULTS FOR FOLLOWING RESERVE REQUIREMENTS
TABLE IV
MODEL VALIDATION RESULTS FOR IMBALANCE RESERVE REQUIREMENTS
6
betas. It is clear that the model is quite accurate, given that [6] D. Halamay, T.K.A. Brekken, “A methodology for quantifying
variability of renewable energy sources by reserve requirement
most of the errors are less than 1%, with a maximum of
calculation,” in Proceedings of the IEEE Energy Conversion Congress
2.451% error. The reduced model is also quite accurate, thus and Expo (ECCE), Atlanta, 2010.
verifying the earlier suggestion that while there can be a [7] S. McArthur, T.K.A. Brekken, “Ocean Wave Power Data Generation for
synergistic interaction between the three different sources, the Grid Integration Studies,” Power and Energy Society General Meeting,
2010 IEEE, Minneapolis, July 2010.
effect can also be quite modest if not optimized. Thus, the
[8] “Total load & wind generation in the BPA control area (balancing
factorial analysis method has provided a useful set of authority area) for 2008,” Bonneville Power Administration, 2008.
parameters with which future analyses can be conducted [Online]. Available:
towards finding an optimal mix of wind, solar, and ocean http://www.transmission.bpa.gov/Business/Operations/Wind/TotalWind
Load% _5Min_08.xls
wave renewable resources.
[9] “2010 BPA Rate Case Wholesale Power Rate Final Proposal Generation
Inputs Study,” Bonneville Power Administration, Tech. Rep., July 2009.
V. CONCLUSIONS [Online]. Available:
http://www.bpa.gov/corporate/ratecase/2008/2010_BPA_Rate_Case/doc
This paper presents a methodology for determining a model s/WP-10-FS-BPA-08_Web.pdf
to predict reserve requirements for balancing authority areas [10] “Reserve Capacity Forecast For Wind Generation Within-Hour
with high penetrations of wind, solar, and ocean wave. This Balancing Service,” Bonneville Power Administration, Tech. Rep.,
model includes interaction effects. The results show that the February 2008. [Online]. Available:
https://secure.bpa.gov/ratecase/openfile.aspx?fileName=WI-09-E-BPA-
interaction effects between different renewable sources can 02_Testimony.pdf&contentType=application%2fpdf
result in a lower reserve requirement than would be required if [11] J. C. Spall, “Factorial design for efficient experimentation,” Control
the sources are considered in isolation, although the reduction Systems Magazine, IEEE, vol. 30, no. 5, October 2010.
may be modest. More study on this to find the best mix is
necessary.
Validation of the presented model with actual data suggest VII. BIOGRAPHIES
the model is accurate. It is also shown that using a reduced
Douglas A. Halamay (SM ‘02, M ’05, GSM ‘08) is a
model that ignores interaction effects may be able to provide graduate student in Energy Systems at Oregon State
an acceptable first-order approximation in cases where the University. He graduated from Gonzaga University
interaction effects are not known. in 2005 with his BSEE (summa cum laude) with an
emphasis in power. After working for three years as
a systems engineer on the P-8A Poseidon at Boeing
VI. REFERENCES Integrated Defense Systems in Renton, WA, he
[1] M. Milligan, K. Porter, E. DeMeo, P. Denholm, H. Holttinen, B. Kirby, returned to graduate school to focus on power
N. Miller, A. Mills, M. O’Malley, M. Schuerger, and L. Soder, “Wind systems and renewable energy integration. He
power myths debunked,” Power and Energy Magazine, IEEE, vol. 7, received his M.S. from Oregon State in 2010 and is now pursuing his Ph.D.
November-December 2009.
[2] “Wind generation capacity in the BPA balancing area authority,”
Bonneville Power Administration, Tech. Rep., October 2010. [Online]. Ted K. A. Brekken (M ‘06) is an Assistant Professor
Available: in Energy Systems at Oregon State University. He
http://www.transmission.bpa.gov/Business/Operations/Wind/WIND_Ins received his B.S., M.S., and Ph.D. from the University
talledCapacity_current.xls of Minnesota in 1999, 2002, and 2005 respectively.
[3] T. Sickinger, “Too much of a good thing: Growth in wind power makes He studied electric vehicle motor design at Postech in
life difficult for grid managers,” The Oregonian, July 17, 2010. Pohang, South Korea in 1999. He also studied wind
[Online]. Available: turbine control at the Norwegian University of
http://www.oregonlive.com/business/index.ssf/2010/07/too_much_of_a_ Science and Technology (NTNU) in Trondheim,
good_thing_growt.html Norway in 2004-2005 on a Fulbright scholarship. His
[4] R. Thresher, W. Musial, “Ocean renewable energy’s potential role in research interests include control, power electronics and electric drives;
supplying future electrical energy needs,” Oceanography, vol. 23, no. 2, specifically digital control techniques applied to renewable energy systems.
June 2010. He is co-director of the Wallace Energy Systems and Renewables Facility
[5] D. Halamay, T.K.A. Brekken, A. Simmons, S. McArthur, “Reserve (WESRF), and a recipient of the NSF CAREER award.
Requirement Impacts of Large-Scale Integration of Wind, Solar, and
Ocean Wave Power Generation,” Power and Energy Society General
Meeting, 2010 IEEE, Minneapolis, July 2010.