692 IEEE TRANSACTIONS ON POWER SYSTEMS VOL. 19, NO.

1, FEBRUARY 2004

paper], may contain flat regions. This is consequence of the fact that REFERENCES
each generating unit is supposed to have a constant marginal cost; this [1] J. Villar and H. Rudnick, “Hydrothermal market simulator using game
implies that the offer curve of the price-taker firms contains flat regions. theory: Assessment of market power,” IEEE Trans. Power Syst., vol. 18,
As a result, the residual demand curve will also contain flat regions; pp. 91–98, Feb. 2003.
[2] R. Palma, O. Moya, and H. Rudnick, “Fondecyt Research Project
therefore, the demand curve faced by each unit (in the game by units) 1 020 801,” Interim Rep., Santiago, Chile, Mar. 2003.
or by each firm (in the game by firms) will also present them.
The problem is that these flat regions imply discontinuities in the
marginal benefit curve faced by each agent (unit or firm). For a given
unit or firm, this may mean the existence of multiple local maximum.
However, this is not a problem by itself, since the solution methods
used for the decoupled problems in both game strategies assure that Discussion of “Unbundled Reactive Support Service: Key
the global maximum benefit for the unit or firm is obtained as a result.1
Characteristics and Dominant Cost Component”
Nevertheless, this fact may still lead to multiple equilibria. This is
because small variations in the generation of other units or firms may P. Marannino and F. Zanellini
result that one of them, in particular, reacts with comparatively large
jumps in its own production.
The authors [1] have to be congratulated for the very interesting paper
It must be remembered that this problem arises due to the attempt
where they clearly outline the key characteristics of the reactive power
of explicitly modeling the discontinuities of the functions of the price-
support service in the emerging competitive electricity market environ-
taker firms; a problem that is often ignored in most of the studies on
ment. The incoming liberalization is pushing power system engineers
the matter, including those that consider supply curve models instead
to develop new methodologies for the evaluation of the contributions of
of Cournot models. There is not a manageable form to consider this
different reactive power sources to the system security and stability [2].
aspect in supply function models. Therefore, it is possible to eliminate
Voltage and reactive power control played a very important role in the
this problem by making simplifications with respect to the behavior of
vertically integrated electric systems, aimed to minimize real power
the price-taker firms or, simply, forgetting their existence and assuming
losses, to maintain adequate voltage profiles, even in the presence of
that all of the firms act strategically.
system component outages, and to avoid voltage collapse or instability
Experience has demonstrated that in those cases where market power
phenomena. In the new deregulated and unbundled structure, the reac-
is potentially more intensive, usually the cases where the price-taker
tive support is a fundamental ancillary service at disposal of the ISO,
firms have a small effect in the price, the multiple equilibria problem
responsible of the maintenance of the security and quality levels im-
is not present in the Cournot model. This is precisely the situation of
posed by the central government authority. In addition, the particular
the study case, the Chilean Central Interconnected System, where the
characteristics of the service need to save a certain degree of centralized
price-taker firms are comparatively very small, and the potential of ex-
management with a reduced possibility of a system-wide competitive
ercising market power on the part of the other market players seems
market due to the local nature of reactive power. In the paper the anal-
very important. This phenomenon is consistent with the fact that when
ysis on the reactive support cost is addressed to specify the so-called
these firms are small, the flat regions of the demand curves are smaller,
“dominant component” of variable cost. It is the opportunity cost con-
and the jumps that take place in the offers of the generators are not so
sequent to the obligation given by the ISO to the producer to curtail
meaningful.
part of the real power production (revenue losses) when the generator
Therefore, in the Chilean case, it is better not to make simplifications
hits the capability limits in order to satisfy additional reactive power
in the treatment of the price-taker firms, because of their small size and
requirements. The authors put in evidence the dominant position of the
because the multiple equilibria problem does not seem to arise; at least it
ISO in posting decision consequent to his definition of the system op-
was not evident in the multiple simulations and revisions that were made.
erating security constraints.
However, it should be recognized that the obtained results represent,
Our comments are related to the role of the ISO in the voltage and
in general, one of potentially several equilibria. Nevertheless, it is prac-
reactive power control and to the computational tools it can employ
tically sure that the equilibrium with higher prices is the most beneficial
for the definition of a nondiscriminatory reactive support. Considering
for each strategic firm. In a repetitive exercise such as this, it is rea-
only the support given by the synchronous generators, the control
sonable to expect that the firms will move toward the most beneficial
schema adopted assumes a paramount importance. The decision
break-even point (of the possible equilibria). The experience of several
making variables are the terminal voltages of the generators, if only
researchers that have analyzed the multiple equilibria in the Cournot
the primary voltage regulation is active, the voltage setting points of
model seems to indicate that the equilibrium obtained with an iterative
the so-called “pilot nodes” and the area reactive levels, if the secondary
algorithm ([1], Figs. 1 and 2) is that of the higher price of all the pos-
and the tertiary voltage regulation are present [3]. The settlement of
sible multiple equilibria (first reference in the paper [1]).
the allowable voltage levels in the different busses of the system, as
Thanks again to Dr. Hogan and Dr. Barroso for their contribution to
shown in the paper, has a deep impact on reactive power output of the
this research.
generators. It has to be determined in order to maintain the system at
1In the case of the game strategy by units, the price ranges where demand the needed security level, avoiding the risk of hard interventions of
faced by each unit does not contain flat regions are analyzed separately; there- the ISO, like the curtailment of specified transactions. In the opinion
fore, discontinuities in the marginal benefit function do not appear. Thus, all of of the discussers the use of a Security Constrained Optimal Reactive
the local optimum can be found analytically and the global optimum may be se-
lected by comparison. In the case of the game strategy by firms, as the decoupled
problems are very complex because of the high number of variables (real and
integer), a search method by computational inspection was used; it also assures Manuscript received April 25, 2002.
to find the global optimum. This is based on the fact that the solution is sought The authors are with the Department of Electrical Engineering, Università
proving a discreet number, but sufficiently large, of possible clearing prices; degli Studi di Pavia, Pavia 27100, Italy (e-mail: mara210@unipv.it).
then the price that implies the greater benefit is selected (global optimum). Digital Object Identifier 10.1109/TPWRS.2002.804965

0885-8950/04$20.00 © 2004 IEEE

IEEE TRANSACTIONS ON POWER SYSTEMS VOL. 19, NO. 1, FEBRUARY 2004
Power Flow (SCORPF) procedure should be recommended for the
solution of the problem. Besides the constraint settlement, the choice
of the objective function is a point of needed discussion.
The maximization of the loadability margin [4] to maintain adequate
reactive reserves for additional transactions but also the traditional min-
imization of real power losses to reduce the transmission service rates
are still now proposed [5]. The SCORPF solution gives important indi-
cations, by means of the Lagrange multipliers of the active constraints,
on the possible pricing of the reactive support at operation level and
on an efficient allocation of new reactive resources at planning level.
Combining technical and economical requirements, SCORPF seems to
be able to give information also on the nondominant component of the
variable cost. The discussers would like to know the opinions of the au-
thors about the suitability of the use of the OPF tools for the definition
of reactive power support costs (paying attention also to the nondomi-
nant terms).
REFERENCES
[1] G. Gross, S. Tao, E. Bompard, and G. Chicco, “Unbundled reactive sup-
port service: Key characteristics and dominant cost component,” IEEE
Trans. Power Syst., vol. 17, pp. 283–289, May 2002.
[2] W. Xu, Y. Zhang, C. P. da Silva, P. Kundur, and A. Warrack, “Valuation
of dynamic reactive power support services for transmission access,”
IEEE Trans. Power Syst., vol. 16, pp. 719–728, Nov. 2001.
[3] S. Corsi, P. Marannino, N. Losignore, G. Moreschini, and G. Piccini,
“Coordination between the reactive power scheduling and the hierar-
chical voltage control of the EHV enel system,” IEEE Trans. Power
Syst., vol. 10, pp. 602–608, May 1995.
[4] P. Marannino, F. Zanellini, M. Merlo, S. Corsi, M. Pozzi, and G.
Dell’Olio, “Evaluation of load margins with respect to voltage collapse
in presence of secondary and tertiary voltage regulation,” in Bulk
Power System Dynamics and Control—V, Onomichi City, Japan, Aug.
26–31, 2001.
[5] K. Bhattacharya and J. Zhong, “Reactive power as an ancillary service,”
IEEE Trans. Power Syst., vol. 16, pp. 294–300, May 2001.
Closure on “Unbundled Reactive Support Service: Key
Characteristics and Dominant Cost Component”
George Gross, Shu Tao, Ettore Bompard, and Gianfranco Chicco
We thank Prof. Marannino and Mr. Zanellini for their interest in and
discussion of, our paper [1]. We appreciate their complimentary and
insightful comments on the paper and are most pleased that they share
our views on the key physical characteristics and the dominant cost
component of the generator-based reactive support as an unbundled
ancillary service in the restructured electricity industry.
The discussion raises issues concerning the role of the Independent
Grid Operator (IGO) in voltage support and reactive power control and
the computational tools deployable for the acquisition and pricing of
Manuscript received May 28, 2002.
G. Gross is with the Department of Electrical and Computer Engineering,
University of Illinois at Urbana-Champaign, Urbana, IL 61801 USA (e-mail:
gross@uiuc.edu).
S. Tao is with the ABB Energy Information Systems, Santa Clara, CA 95050
USA (e-mail: shu.tao@us.abb.com).
E. Bompard and G. Chicco are with the Dipartimento di Ingegneria Elettrica
Industriale, Politecnico di Torino, corso Duca degli Abruzzi 24, 10129 Torino,
Italy (e-mail: bompard@polito.it; chicco@polito.it).
Digital Object Identifier 10.1109/TPWRS.2002.804966
0885-8950/04$20.00 © 2004 IEEE
693
the reactive support service. The discussers refer to a Security Con-
strained Optimal Reactive Power Flow (SCORPF) procedure as a pos-
sible tool for this purpose. They state that such a tool will be useful in
“the definition of reactive power support costs (paying attention also
to the nondominant terms).” Unfortunately, we do not share the view
expressed in this statement since the SCORPF tool has, in reality, no
capability to provide cost information. The OPF may provide appro-
priate pricing signals when the economics of the situation are properly
represented. But, the discussion on OPF application requires that a few
aspects of OPF be clarified.
The OPF is a powerful centralized decision-making tool developed
for the vertically integrated utility (VIU) structure. Its effective applica-
tion in the competitive environment requires the appropriate represen-
tation of market economics and the introduction of meaningful metrics
[2], [3]. A key complication in the OPF formulation lies in the arbi-
trariness of the optimum due to the wide discretionary powers of the
IGO in the formulation of the constraint set and the definition of the
set point values of certain control variables [2]. These considerations
make the choice of an appropriate objective very challenging since the
economics of the situation must be appropriately reflected.
The OPF framework provides the setting for the evaluation of dual
variables that have the economic interpretation of prices. For example,
the relatively simple model for the reactive loads used in [4] can pro-
vide effective price signals for congestion relief. The OPF framework
permits the incorporation of contingencies; however, in the presence
of contingencies, the value—not the cost—of reserves must be explic-
itly represented. This value is linked to the effective limit of a reactive
power source, which depends on the amount of reactive power that can
be actually provided under specified system conditions [5]. This effec-
tive reactive limit is different than the generator reactive power limit
since the system conditions under one of the specified contingencies
may prevent the generator from reaching its reactive power limit.
This brief discussion leads to the conclusion that the deployment
of an OPF tool for the reactive power support service acquisition and
pricing requires the careful consideration and addressing of some very
challenging issues. Two issues are particularly critical. These are de-
tailed below.
1) As correctly pointed out by the discussers, the selection of the
objective function for an OPF tool is not a straightforward issue.
Due to the local nature of reactive power, the optimization to
minimize the total costs of reactive power support services based
on the price bids submitted by the generators is unfortunately
not appropriate. Moreover, the local monopoly power of certain
generators makes a centralized IGO-operated structure for
reactive support both impractical and unsuitable for competi-
tion. This has been clearly illustrated and strongly emphasized
in the paper. The discussers refer to the maximization of the
loadability margin to maintain adequate reactive reserves for
additional transactions and the traditional minimization of
real power losses as two possible candidates for the objective
function. While the loadability margin maximization and the
loss minimization can be thought of as a common good enjoyed
by all the market participants, i.e., all the transmission network
customers, such objective functions may conflict with each
individual generator’s self interest. Moreover, since these ob-
jective functions do not involve either the costs incurred by the
individual generators or the values to the transmission customers
in providing/acquiring the reactive support requirements, they
are unable to provide a direct and clear price signal for reactive
support services to individual generators. The lack of such a
price signal results in no appropriate incentive to the generators
to supply the necessary reactive support services. Thus, while
these objective functions may be appropriate choices in the VIU