Competitive Procurement Ancillary Services by An Independent System Operator
Competitive Procurement Ancillary Services by An Independent System Operator
Alex Papalexopoulos
Senior Member
499
total cost curves are convex (that is, the bid curves are
convex and if one assumes generators bid their costs then
they would be claiming to have convex total cost curves).
Furthermore, the PX auction model adopted in California
and discussed here recognizes no operational or intertemporal constraints. There may be iterations in the energy
auction to help bidders satisfy their operational
requirements, but they are not significant in describing the
ancillary services procurement model [2,3].
500
The one case where a capacity bid may reflect actual costs
instead of opportunity costs is AGC. AGC requires a
generator to operate at a level higher that its rninirnum
output, in order to allow movement of the unit in downward
as well as upward directions. For example a unit Qith costs
of 5 ckWh would not normally operate at levels higher than
its minimum output when the price in the energy market
falls below 5 c/kWh. In order for it to do so, it is entitled to
bid for a capacity payment to cover the actual costs the unit
may incur.
501
The (x = 1) case:
By setting x=l, the IS0 can determine the maximum extent
of its payments [l]. Under assumptions of cost revealing
bids, the x=l scenario allows a bidder to be indifferent
between providing spinning reserve or energy. Successful
The auctions for reserves differ from the PX auction in one bidders are paid a capacity payment at a price Ciwhere
very important aspect. The goal of the PX auction is to
facilitate trade because there will be many participants on
C i =max {R + E i } - E ,
i;isn
both sides, i.e., buyers and sellers6. Ancillary services are
and
ZZ
is
the
set
of
all successful bidders. They are paid their
fully the ISOs responsibility, so the IS0 acts as the single
purchaser in reserve auctions to meet its reliability energy strike price Ei for any energy they are called upon
obligations. If the purchaser of energy from reserves could to generate in real-time.
be identified in advance it could be forced to schedule
capacity. Because the energy purchase is a random event, The (0 5 x I 1) case:
the IS0 as the agent for all market participants buys the In the design described in [2], the IS0 proposed an auction
reserves according to a preset formula. As a single buyer loosely based on the notion of purchasing call options on
the IS0 can seek to minimize its costs; the formula that the real-time balancing market. In order to pursue a cost
governs the purchase should prevent strategic withholding minimizing bid evaluation, the IS0 would use an unbiased
estimate of the probability x (between 0 and 1). In this case,
of reserve demand.
the bids are ranked according to R, + xEi. Successful
An important property of opportunity costs is that the more bidders are paid at a price
Ci=max {Ri +xE,} - x E i
infra-marginal a unit (the deeper in the loading order), the
i;isfl
greater its opportunity costs in providing spinning reserves. A reserve generator is paid its energy strike price Ei if it is
Thermal units capacity reservation bids should reflect their called upon to generate in real-time, irrespective of the
opportunity costs, which in turn reflect lost profit (market balancing market price.
energy price minus marginal cost).
The capacity
reservation bids R,should be inversely related to the energy The (x=O) case:
bids E;, if energy bids are cost-reflective as they ought to be The (0 I x S 1) case may be criticized as being too complex
in a competitive market. Thus, to minimize the expected and susceptible to being gamed. It is only an approximation
cost of reserves, the IS0 cannot consider only the capacity to the true cost minimization, as the probability of calling
or only the energy bids, but must estimate the probability upon a successful bidder to generate in real-time is likely to
that reserves will be utilized in real-time. It must include a vary according to the energy bids. Presumably, intelligent
probability factor represented by a parameter x in the bid bidders could game this market structure to increase their
evaluation, i.e., bids are ranked on the basis of Ri+XEi. In revenues. One scenario involves bidders lowering their
order to keep the auction transparent, x must be set by the energy bids in order to increase their capacity payments.
IS0 prior to bid submission.
However, this also increases the probability of the bidders
getting called upon to generate in real-time. Ideally, a
bidder would like to lower its energy bid as much as
possible without altering its position in the merit order
stack. Finally, the option-based case does not result in a
The PX energy auction is a double auction while the ISOs uniform price for reserve capacity payments making it less
transparent.
ancillary services auctions are single sided auctions.
502
An alternative design, is to rank bids solely on the capacity Table 3. Table 2 shows the selected quantities and Table 3
reservation: i.e., the IS0 takes bids in order of reservation shows the capacity prices that are paid to the selected
bids, and pays each successful bidder the highest reservation bidders.
bid among those accepted.
Table 2: Spinning reserve auction results
Since the energy bids are ignored in the bid evaluation,
there is less justification to consider them as fixed strike
Selected MW quantities
prices. It might appear reasonable to assume that the
reserve mils are price-takers in the real-time market, i.e.,
they cannot set the real-time price which they must accept.
0.5
80
100
Clearly if all reserve units are price takers (and there is no
1
0
100
80
obligation to replace reserves that are used), the IS0 could
end up dispatching all the reserves without impacting the
Table 3: Capacity payments to successful b,'dders
real-time price.
~~
0.5
1
4
0
10.5
10
12
503
---P
504
and during the hours 7-21, the prices were at or near 9.50
$/MW8. The quantities procured in the hours 7-21 were less
than in the lower priced hours, suggesting insufficient
supply at the capped prices. In cases of such price-cap
induced shortfalls, the IS0 may be forced to rely on the
increased use of generators on long-term Reliability Must
Run (RMR) contracts, at costs that can easily outweigh any
savings that result from the price caps.
Spinning Reserves Market (42/98)
- 10
8 8
--4
I-MW
--
-$/MW/
I
.rt
-0
cu
cu
Disclaimer
This paper does not necessarily reflect the positions of
PG&E. Any errors or omissions are the sole responsibility
of the authors.
References
[ 11 Federal Energy Regulatory Commission (FEW). Joint
application of Pacific Gas and Electric Company, San Diego
Gas and Electric Company, and Southem California Edison
Company for authorization to convey operational control of
designated jurisdictional facilities to an Independcnt System
Operator, (Phase I IS0 filing), FERC Docket No. ER961663-000, April 29, 1996.
[2] Federal Energy Regulatory Commission (FERC).Phase
I1 filings for the California I S 0 and PX, FERC Clocket No.
ER96-1663-001, March 31, 1997.
[3] Federal Energy Regulatory Commission (FERC).
Amended Phase I1 filings for the Califomia I S 0 and PX,
FERC Docket No. ER96- 1663-003, August 15, 1997.
[4] E. Hirst and B. Kirby, Creating Competitivi: Markets
for Ancillary Services, Oak Ridge National L Iboratory,
October 1997.
[5] L.D. Kirsch and H. Singh, Pricing Ancillary Electric
Power Services, The Electricity Journal, October 1995.
[6] Wilson, Robert, Priority Pricing of Ancillary Services,
Report to the Trust for Power Industry Restructuring, May
17, 1997.
[7] Minimum Operating Reliability Criteria (MORC),
Western Systems Coordinating Council (WSCC),
September 1996.