20091030Comment, Protest.pdf

PETER J. RICHARDSON
ISB #3195
RICHARDSON & O'LEARY, PLLC
P.O. BOX 7218,83707
515 N 27th
Boise, Idaho 83702
[phone]
peterca richa rdsona ndolea ry. com
RECEI 0
inOq OCT 30 PM 4: 39
GREGORY M. ADAMS
ISB #7454
RICHARDSON & O'LEARY, PLLC
P.O. BOX 7218,83707
515 N 27th
Boise, Idaho 83702
[phone]
9 regca richardsonandoleary. com
Attorneys for Renewable Northwest Project
BEFORE THE IDAHO PUBLIC UTILITIES COMMISSION
IN THE MATTER OF THE PETITION OF )
ROCKY MOUNTAIN POWER FOR AN )
ORDER REVISING THE WIND )
INTEGRATION RATE FOR WIND- )
POWERED SMALL POWER GENERATION)QUALIFYING FACILITIES )
)
)
)
No. PAC-E-09-07
COMMENTS AND PROTEST OF
THE RENEWABLE NORTHWEST
PROJECT
COMES NOW, the Renewable Northwest Project (RNP) by and through its attorney of
record and pursuant to that Notice of Petition and Modified Procedure issued dated September
23,2009 and hereby lodges its Comments and Protest. In support thereofRNP says as follows:
SUMMARY
RNP requests the Commission deny Rocky Mountain Power's Petition for the reasons
stated below. In the alternative, RNP urges the Commission not to proceed pursuant to modified
RNP Comments and Protest 1
procedure and schedule an appropriate proceeding to resolve the issues associated with Rocky
Mountain Power's Petition. Rocky Mountain Power's wind integration analysis results in a rate
that is likely twice the actual cost. This has both ratemaking implications for Idaho customers as
well as regional implications for wind development. The following technical comments are
based upon RNP's Staffs detailed review ofPacifiCorp's 2008 IRP, more specifically Appendix
F thereto entitled "Wind Integration Costs and Capacity Planing Contributions." RNP has
extensive experience with the derivation of and flaws in that document which forms the basis for
its curent fiing.
ROCKY MOUNTAIN'S WIND INTEGRATION ANALYSIS IS FLAWED
Overview
The source of Rocky Mountain Energy's wind integration recommendation is its 2008
Integrated Resource Plan ("IRP") which can be found in docket No. P AC-09-06. The IRP
analysis had the explicitly stated purpose of developing". .. a methodology to support the costs
associated with resource portfolio analysis for the IRP1..." The analytical techniques used in the
study represent a radical deparure from earlier Rocky Mountain Energy studies, those of other
Northwest utilties, and studies across the US and internationally. The new methodology was
developed without involving outside experts or stakeholders. Flaws in the methodology are
relatively basic and easily addressed, but are so fundamental that Rocky Mountain Energy's
stated results canot be accepted as a serious approximation of their wind integration costs. The
present study doubles Rocky Mountain Energy's earlier projections of wind integration costs,
and probably overstates the costs by a factor of two or more.
1 Rocky Mountain Energy 2008 IRP Appendix F page 269.
RNP Comments and Protest 2
Specific Concerns
1. The most fudamental shortcoming in Rocky Mountain Energy's methodology is that the
variability and uncertinty introduced by wind is considered separately from the variabilty
and uncertainty already on the power system due to load. The reason this is importt is that
the forecast errors and short-term (less than one hour) variability of wind and load are not
normally correlated with one another. This is a crucially important issue, but not a new one.
Every wind integration study of which RNP is aware has netted wind against load in deriving
the reserve requirement, including Rocky Mountain Energy's previous analyses dating back
to 2003, as well as the analyses ofIdaho Power Company, Avista and Portland General
Electric. The point is well documented in wind integration literatue: "The requirement of
extra reserves is quantified by looking at the variations of wind power production, hourly and
intrahour, together with load variations and prediction errors." Ackerman, "Wind Power in
Power Systems", p. 158, Wiley, 2005. Exhibit 1. Similarly from another source, "The
increase in short term reserve requirement is mostly estimated by statistical methods
combining the variability or forecast errors of wind power to that of load and investigating
the increase in the largest variations seen by the system." IEA Task 25 Final Report, page
13,2009. Exhibit 2.
In cases where the wind variability is small in absolute (MW) terms compared with load,
the incremental need for reserves is not significant. On the other hand, if the load and wind
variability (or uncertainty) are comparable, the incremental need for reserves is not double,
but approximately 40% higher. This is due to the fact that the reserve requirement is
proportional to the standard deviation of the variability.
Rocky Mountain Energy's implicit assumption that the reserve requirement is
RNP Comments and Protest 3
independent of the load would only hold if the load variability and uncertinty are small
fractions of the wind variability and uncertainty. Rocky Mountain Energy offers no
information suggesting this is the case. Rocky Mountain Energy offers no basis at all for
failing to net load and wind to determine reserve requirements.
2. Rocky Mountain Energy's representation of wind generation from new wind projects-
especially on the east side significantly overstates the reserve requirement. In past wind
studies, Rocky Mountain Energy represented new wind projects as time-shifted time series
from existing projects. The time shift preserved some correlation between existing and new
projects, without assuming 100% correlation between the fleet additions and the existing
fleet. Representing the new wind projects using a multiplicative constat establishes a higher
correlation between the new wind projects and the existing ones than is reasonable.
This effect may not be significant for the west side projects, because the capacity ofthe
incremental projects is a relatively small fraction of the existing projects. On the east side
however, the nameplate additions are on the same order as the existing projects and the effect
is very significant--specially on the shorter timescale (ten minutes) where correlations
among even relatively nearby projects tend to be relatively smalL.
RNP understands that new projects wil be sited close enough to existing projects that the
correlations among existing projects can be used to help specify correlations between
existing and new projects. RNP agrees that the existing correlations are useful, but Rocky
Mountain Energy did not use them to help specify correlations between the new projects and
existing ones; instead, they implicitly assumed 100% correlation between matched pairs of
existing and new projects. The company apparently believes that the uncertainty in the
locations of new projects renders their assumption reasonable. RNP believes the opposite is
RN Comments and Protest 4
tre. No two existing projects exhibit 100% correlation, and only an extraordinar and
almost unmaginable set of circumstances would cause the correlations to be any higher than
Rocky Mountain Energy has assumed.
The reason this is important is that diversity in wind project output (the extent to which
output levels are not correlated) tends to reduce both the variability and uncertainty in wind
generation. Persistence forecasts are more accurate for uncorrelated projects than for
correlated projects. In underestimating wind generation diversity, Rocky Mountain Energy's
methodology essentially ensures the resulting reserve requirements are overestimated by a
significant amount. At a minimum, Rocky Mountain Energy should represent new project
output by time-shifted levels at existing projects, adjusting the time shift until the correlation
with a selected nearby project reaches a level similar to that seen between existing adjacent
projects.
3. Rocky Mountain Energy's assumption that all balancing purchases entail market
transactions and market transaction costs is not supported by, and is not consistent with,
previous Rocky Mountain Energy studies, or other utility studies. Earlier studies by Rocky
Mountain Energy assumed a constant $0.50IMWh cost to all market transactions. The
present study deviates significantly from the previous assumptions, substatially increasing
the resulting estimated inter-hour balancing costs.
Rocky Mountain Energy's assumption about higher hour-ahead trading costs is not well
documented. However, even given the new higher costs, it is not correct to assume that all
imbalances are settled in the markets. For example, if the wind generation is unexpectedly
high during a heavy load hour, Rocky Mountain Energy wil likely have the abilty to reduce
generation somewhere on its system (e.g., on a hydro or fossil unit) without incurng a
RNP Comments and Protest 5
market transaction cost.2 Similarly, if a shortfall of wind from the expected amount occurs
on a light load hour it would be unusual for Rocky Mountain Energy to need to rely on
markets to make up such a difference. The net effect of Rocky Mountan Energy's
assumption is an overall overestimate of the inter-hour costs by a significant amount.
Rocky Mountain Energy apparently believes that adjusting resources for the realities of
load, generation, and market prices for the next hour is never economic because it results in
an uneconomic dispatch. This suggests that the Company can never save more by changing
operations at its own projects than incurng a transaction cost to operate resources on
another system. The fallacy of this argument is easily ilustrated with an example:
Assume that the prevailing market price is $501MWh and Rocky Mountain Energy has
optimized its resources such that the most expensive resource operating has a marginal
cost of $50IMWh. If the wind generation exceeds the plan for that hour and Rocky
Mountain Energy finds itself with an additional 150 MWh for that hour, Rocky Mountain
Energy may choose to back down its $501MWh resource to save $5,000 on that hour, or
sell the 100 MWh into the market for $42.50IMWh ($50IMWh less 15% transaction cost)
and ear $4,250. In this case it is clear that the economic choice is to back down the
owned resource and not transact in the market.
There may be times when the conditions of the example do not hold (Rocky Mountain
Energy has no resource operating at the marginal rate, or is at minimum generation levels),
but operating on the assumption that it is always more economic to transact in the market is
simply not credible, nor supported by Rocky Mountain Energy's study.
2 It is possible that there might be some cost (or benefit) associated with operating a thermal or hydro unit at a
different point in its power efficiency curve. However, this would not be the same order of magnitude as the 10-
25% market transaction costs assumed in Rocky Mountain Energy's analysis.
RNP Comments and Protest 6
4. Rocky Mountain Energy's insistence that day-ahead balancing needs should always be
rounded up is difficult to understand and is not supported. If Rocky Mountain Energy seeks
to balance the system as closely as possible, it should round off the day-ahead purchase and
sales requirements. Rounding up would make Rocky Mountain Energy routinely surlus
when purchasing, and routinely deficit when sellng (assuming that's what they mean3). For
reliability puroses, it would be reasonable to round up purchase requirements over heavy
load hours to minimize the chance of being short over critical hours. Conversely, it might be
reasonable to set the balance slightly short on light load hours to reduce the risk of ruing
into minimum generating requirements. However, making the system over-long and over-
short randomly serves only to unecessarily add to the need for, and cost of, hour-ahead
balancing services.
5. All other wind integration studies RNP is aware of, including Rocky Mountain Energy's
earlier analysis show an increasing cost (on a per megawatt-hour basis) of wind integration as
wind generation is added to the system. Rocky Mountain Energy only examined the most
extreme level of wind penetration, reached in 2021, and uses it to justify the wind integration
cost ascribed throughout the study horizon. Rocky Mountain Energy should either derive a
levelized cost reflecting the increasing cost through time or pick a cost level based on an
average level of wind development through the study horizon. Using a level determined for
2021 throughout the study is in itself a considerable overstatement of the wind integration
3 In conversations with RNP Staff, Rocky Mountain Energy could not explain precisely what rounding up their
market transactions meant. As stated in the IRP document, it appears that Rocky Mountain Energy rounded up
purchases (making them more surlus) and rounded up sales (which would make them more deficit). However in
conversation, Rocky Mountain Energy staff suggested that the operation was necessar to ensure the company is not
short. This implies rounding up purchases and rounding down sales. In either case, the assumption improves
neither the economics nor reliabilty. For example, systematic over-purchasing risks CPS 2 violations in light load
hours. For reliabilty puroses, Rocky Mountain Energy should err on the over-purchase side during heavy load
hours and under-purchase on light load hours. This does not seem to be what they meant and would not be
RNP Comments and Protest 7
costs actually incurred.
6. The wind forecast assumption for the analysis was apparently based on the average wind
generation level between one and two hours prior to the beginning of each operating hour.
The IRP Appendix F states that the curent state of forecast accuracy is a 40-45 minute
persistence forecast. The difference between Rocky Mountain Energy's stated forecast
capability and that relied upon in its analysis leads to a significant overestimate of the hour-
ahead forecast error, and the corresponding intra-hour reserve requirement. Furher, it is
clear from experience in the BP A rate case that the state of the art for wind forecasting of
Northwest wind projects is closer to the 30-minute persistence levels achieved by employing
meteorologists to forecast wind in real time. Given the expense Rocky Mountain Energy
associates with the forecast error, it seems only prudent to employ more sophisticated
forecasting techniques to minimize costs to ratepayers. In any event, if Rocky Mountain
Energy is setting schedules on 40-45 minute persistence forecasts, the wind integration cost
analysis should reflect that.
Conclusion
For the technical reasons noted above the Commission should deny Rocky Mountain
Energy's petition for an order increasing its wind integration rate. RNP stands ready to fully
participate in any investigation into the continued reasonableness of the existing wind
integration rate should the Commission initiate such a docket.
DATED this 29th day of October 09.
economic in any case.
RNP Comments and Protest 8
RICHARDSON & O'LEARY PLLCpi-d~P er J. Richardson
ISB #3195
Attorney for Intervenor,
Renewable Northwest Project
RNP Comments and Protest 9
RNP Exhbit 1
Editor
Ackermann
~_.:i
Q."
~
to~_Il:i~
~
(D.,
U\
1
to
3'"
158 Power System Requirements
for freau1mcy control (load following), if the penetration of wind power is large enough
to increse the total variations in the system.
Prediction tools for wind pOWer production play an important role in integration. The
system operator has to increase the amount of reserves in the system because, in
addition to load swil1gs, it has to be prepared to compensate un predicted variations in
production. The accuracy of the wiiid forecasts can co'Gtributeto risk reduction. An
açurate forecst allows the system operator to count on wind capacity, thus reducing
costs without jeopardising system reliability.
The requirement of extra reserves is quantified by looking at the variations of wind
pøwer pro¡iuction, hourly and intrahour, together with load variations and prediction
errors. The ei-tra reserve requirement of wind power, and the costs associated with it,
Øf!. be estiinated either by system models or by analytical methods using time series
of wind power production together with system vaiiables. Wind power production is
not sti-ai~'tforwatd to model in the existing dispatch models, because of the uncer-
tainty of forecast errors involved on several time scles, for instance (Dragoon and
Miligan, 20(3). Below, we wil briefly describe analytical methods with statistical
mesures.
TIie effect of the variations can be statistically estimated using standard deviation.
What the system sees is net ioad (load minus wind power production). If load and wind
PQwer pro¡uction are uncorrelated, the net load variation is a simple root mean square
(RMS) COmbination of the load and wind power variation:2 2 2( (fiotiil) = (O'load) + (O',,'¡iid) ,(8. I)
wliel'e O'IOI"1i O'loud and O',,'¡nd are the standard deviations of the load, net load and wind
power production time series, respectively.
The Larger the area in question and the larger the inherent load fluctuation in the
systeni the larger the amount of wind power that can be incorporated into the system
withoii.t Ìli-creasing variations. The reserve requirement can be expressed as three times
the standard dèviation (30' covers 99 % of the variations of a Gaussian distríbution).
The incremental increase from combining load vaiiations with wind variations is
3 tIines (alOUd - O'load). More elaborate methods allocating extra reserve requirements
for wind power can be used, especially with nonzero correlations and any nuinber
of individual loads and/or resources (Huds.on, Kirby and Wan, 2001; Kirby and
Hirst, 2000). .
On the time s.cale of seconds and ininutes (primary control) the estimates for increased
reserve requirements have resulted in a very small impact (Ernst, 1999; Smith et al.,
iO(). Ths is beause of the smoothing effect of very short variations of wind power
productiQn; as they are not correlated, they cancel out each another, when the area is
large enough.
For the time scale of 15min to I h (secondary control) it should be taken into account
that load variations are rl10re predictable tban wind power vaiiations. For this, data for
load and wind predictions are needed. Instead of using time series of load and wind
power variations, the time series of prediction errors one hour ahead are used and
standard deviations are calculated from these. The estimates for reserve requirements as
a result of use of wind power have resulted in an increasing impact if penetration
RNP Exhibit 2
Final report,
Phase one 2006-08
I ind Task 25
Hannele ~Qlttin, Peter Meiom, An1je Or, Fran van Hulle, Bernd Lange,
Mark O'Maley, Jan Pierik, Bar Urnmel, John Olav Tande, Ana Estanqu,ero. Manuel
Matos, Emilo Gqmei, Lennar Seder, Gora Strac, Anr Shakor, JoãQ Rio, J.
Chares SirtJ, Michael Milg~ & Eri EIa
Design and operation of power
systems with large amounts of
wind power
VI
tbatany storae should be operated according to the needs of aggrgated system
balacing. It is not cost effective to provide dedicated back-up for wind power in
lare power systems where the variabilty of al loads and generators are
effectvely reduced by aggegating, in the same way as it is rtt effective to have
cliCáed storage for outages in a ceriinthèrmal power plant, or having specific
Plâìts forlowing the variation of a certain load.
Integtion cost of wind power: Many studies address integraton costs.
Intøgi()ncost is the extra cost of the design and operation of the non-wind par
of me. p()wer system when wind power is integrated. Integraton cost ca be
dividèdinto different components arising from the increase in the operatiQJal
balançing)G:ost and grid reinforcement cos. It is inportt to note whether a
maret. cost ha been estated or the results refer to technical costs for the
power system. A "market cost" include trsfer of money from one iict(). to
another actor, while "technical costs" inplies a cost for the whole systeii Nl(lt
studies so far håye concentrated on the costs of integrating wind inrothe Pøwer
system whieal.s cost-benefit analysis work is emerging. There is alo bøefit
when adding wind power to power systems: it reduces the totalopratingcQs
and emissions aswi4 replaces fossil fuels. Integration costs of wind power nee
to be compard to something, like the production costs or market .valueöf wind,
power, or integriÇ)ncöst of other production forrs. To enable.fai cOmPàin
between power systms with differing amounts of windpO\er,th~systenis
should in pnnciple have same CO2 emissions, reliabilty, etc. The value of the
capacit crdit ofwindpøwer~an also be stated.
Increase in short termr_rve requirements duetò wind power: Wind
generåtion may requir sYstem operaors to car addionaoperatig/ reserves.
From both the experience aid results from studies peormed, . a .. significat
challenge is the varibilty of wind power within 1-6 hrs. Frequency control
(timescale of second) and inertial response are not crcial probleinswhen
integttñ Wid. power into large systems at the present time, but can be a
challenge for small systemS.ánd vvll beome more of a challenge for systems
With high penettion in the futue. The increase in short terr reserve
requirment is mostly estinated by statistical methods combining the varabilty
or forecast errors of wind power to that of load and investigatig the incrse in
the larges vaations seen by the syste. The impact of windpower is.mosly
seen in the 10 minutes to some hours tie scale, and only little in the seond to
second autoinaticrrquencycontrl time scale. The estIated Ílcrasein short
term reserve requirements in the studies summarised in this report has a lare
13
CERTIFICATE OF SERVICE
I HEREBY CERTIFY that on the 29th day of October, 2009, a true and correct copy of the within
and foregoing COMMENTS AND PROTEST OF THE RENEWABLE NORTHWEST PROJECT
was served in the manner shown to:
Ms. Jean Jewell
Commission Secretay
Idaho Public Utilities Commission
472 W. Washington (83702)
PO Box 83720
Boise, ID 83720-0074
lL Hand Delivery
_U.S. Mail, postage pre-paid
Facsimile
Electronic Mail
datarequest~pacificorp.com
Data Request Response Center
PacifiCorp
825 NE Multnomah, Suite 2000
Portland, OR 97232
Via electronic mail only
_ Hand Delivery
_U.S. Mail, postage pre-paid
Facsimile
X Electronic Mail
=eß).~Peter Richardson