You are on page 1of 7

IEEE TRANSACTIONS ON POWER SYSTEMS

Operating Reserve Model in the Power Market


Jianxue Wang, Xifan Wang, Senior Member, IEEE, and Yang Wu, Student Member, IEEE

AbstractMost of the current research on operating reserve


markets assumes that the required capacity is predetermined,
which means that the demand of operating reserve capacity is
inelastic. This paper considers that operating reserve capacity
in a power system is flexible and that one should optimize it by
cost-benefit analysis. Based on the reliability evaluation of the generation system, a clearing model of the operating reserve market is
proposed to determine the optimal reserve capacity and simultaneously clear the operating reserve market. The model is discussed
on both uniform-price and pay-as-bid auction mechanisms. By
using a heuristic method, we suggest an efficient algorithm to
solve the model. Case studies for reliability test system (RTS96)
demonstrate the usefulness and efficiency of the proposed model.
Index TermsCost-benefit analysis, operating reserve capacity,
power market, reliability.

I. INTRODUCTION

N power systems, it is very important to maintain a certain amount of operating reserve capacity to avoid power
shortage due to generator random outages, load fluctuations, etc.
In most research and practices, the amount of operating reserve
capacity is determined by dispatch rules that require operating
reserve is greater than the capacity of the largest generator and a
fraction of the peak load. Although this approach is easy to implement, it cannot maximize the social welfare of the operating
reserve in an unbundled power market.
In power markets, the operating reserve capacity is a commodity like electricity. How to purchase operating reserve as
well as electricity is a hot topic on which many researchers are
working. A model on coordinated scheduling of reserves, contracts, and supplemental energy is proposed in [1]. The suggestion that all reserve markets are cleared together is given in
[2]. A price-based adaptive requirement model of spinning reserve based on the Lagrange relaxation approach is presented in
[3]. Combining the reliability calculation with the unit commitment (UC), paper [4] brings forward a novel model on optimal
scheduling of the spinning reserve. In this research, system operators are supposed to buy predetermined reserve capacity. We
argue that operating reserve capacity should be optimized in the
power market circumstance according to benefits and cost analysis. Obviously, the more expensive the operating reserve capacity, the less operating reserve capacity should be purchased;
the bigger the interrupted energy assessment rate (or the value

Manuscript received October 9, 2003; revised July 21, 2004. This work was
supported by the National Science Foundation of China under Key Project
59937150. Paper no. TPWRS-006342003.
J. Wang and X. Wang are with Xian Jiaotong University,
Xian, Shaanxi 710049, China (e-mail: jxwang@mailst.xjtu.edu.cn;
xfwang@mail.xjtu.edu.cn).
Y. Wu is with Texas A&M University, College Station, TX 77843-3128 USA
(e-mail: wuyang@ee.tamu.edu).
Digital Object Identifier 10.1109/TPWRS.2004.841232

of lost load), the more operating reserve capacity should be purchased. Thus, the proper operating reserve capacity varies with
its cost and the benefit. We call this varied operating reserve capacity flexible reserve. A novel model of flexible reserve and
relative algorithm is discussed in the paper.
The reserve benefit can be evaluated by reduction of Energy
Expectation Not Served (EENS). EENS is a very important and
widely used reliability index in power system planning, which
can be calculated by a reliability evaluation or probabilistic production simulation algorithm [5]. Reserve cost depends on the
payment mode. There are two kinds of payment modes: the uniform-price mode and the pay-as-bid (PAB) mode. Generators
are paid by uniform market-clearing prices in the uniform-price
mode and paid by bid prices in the PAB mode. Reference [6]
points out that the principle of the uniform price is the equal
bidding price criteria and the principle of the pay as bid originates from the equal incremental cost criteria. The comparison
of the two payment modes has been discussed in [7][10]. This
paper proposes the flexible reserve model based on both payment modes. The comparison of the two modes in operating reserve market is given in the case study.
The structure of this paper is as follows: Section II proposes
a clearing model of the operating reserve market. In Section
III, a very efficient algorithm to solve the above model is suggested. The algorithm is demonstrated by a standard test system
in Section IV. Section V contains some conclusions and the emphasis of further research.
II. CLEARING MODEL OF FLEXIBLE OPERATING RESERVE
As mentioned before, a clearing model of an operating reserve market should optimize reserve capacity as well as clear
the market. First, we introduce the model of the flexible operating reserve market.
A. Flexible Operating Reserve Model
Without predetermined reserve capacity, the most important
aspect of the flexible reserve is to find out the optimal operating
reserve capacity that maximizes the social benefit or minimizes the social cost . The objective can be, respectively, expressed as

(1a)
(1b)
where

is the operating reserve capacity to be optimized.


is the decrease of the interruption loss, and
is

0885-8950/$20.00 2005 IEEE

IEEE TRANSACTIONS ON POWER SYSTEMS

the interruption loss when the system has reserve capacity


They can be, respectively, measured by

(2a)
or
(2b)
is the EENS before purchasing operating reserve,
Here,
and
is the systems EENS after purchasing operating
is the Interrupted Energy
reserve capacity . Coefficient
Assessment Rates (IEAR) in North America or Value of Lost
Load (VOLL) in the U.K. VOLL/IEAR varies with different
customer categories, usage time, etc. [11]. In this paper, since
outages due to shortage of generation capacity affect the system
as a whole, it is assumed to be the same for all customers at all
times. As IEAR is an approximate value, the benefit of the operating reserve is difficult to be evaluated accurately. The precision of (2) would be improved with further research on the
IEAR.
in (1a) and (1b) depends on different
The cost function
payment modes. Suppose the bid price of unit is
and the
.
repreclearing price of operating reserve market is
sents the set of selected reserve units. According to PAB mode,
can be expressed as
the purchase cost

Fig. 1. Optimal capacity condition for pay-as-bid mode.

(1b). Taking the partial derivative of (1b), we obtain the necessary condition of optimal reserve capacity
(7)
Because
is not a continuous variable in the purchasing
process, we use the difference equation to replace the above
equation

(3)
(8)
The purchase cost on uniform-price mode

can be defined

or
(9)

as
(4)
The constraints include
(5)
(6)
where
maximum reserve capacity of unit ;
minimum reserve capacity of unit ;
required time of the operating reserve;
ramp speed of unit (in Megawatts per minute).
Equation (5) is the operating reserve capacity constraint of
unit . The ramp speed is considered in (6) to ensure that the
operating reserve capacity can hold on in required time (for
for spinning and nonspinning reserve is ten
example,
minutes).
It must be pointed out that the model above is based on market
structure in which the operating reserve market is independent
from other markets and is cleared right after the energy market.
The correlation among hours in the unit commitment problem
has been solved in the energy market. Hence, in the operating
reserve market, there is no correlation among hours.
B. Condition of Optimal Reserve Capacity
The same condition of optimal reserve capacity can be deduced from (1a) and (1b). The following deduction is based on

where
is the incremental cost of
, and
is the
EENS change that is usually a negative due to increasing oper. Whether or not to purchase
can be deating reserve
termined by (9). When the following inequality holds for increasing

or
(10)
purchasing
is favorable, because the benefit due to decreasing interruption loss
is larger than
of
. Otherwise, one should stop purchasing
the cost
.
The principle of clearing operating reserve capacity under the
PAB mode can be illustrated by Fig. 1. First, we sort the bidding
prices reported by suppliers into a merit order, assuming

From (3), the cost of


sented by

under the PAB mode can be repre-

Thus, we can establish the supply curve


, as shown in
and
in the
Fig. 1.
figure are the cumulative operating reserve capacity and the cumulative cost after purchasing .

WANG et al.: OPERATING RESERVE MODEL IN THE POWER MARKET

Another curve in Fig. 1 is the interruption loss curve


.
At the beginning of the clearing process, the operating reserve
capacity of the system is zero. By generation system reliability
evaluation, we can obtain the respective EENS of the system
and the interruption loss
, as shown in
, the EENS of the system decreased to
Fig. 1. If purchasing
; the benefit is

As shown in the figure,

is larger than the cost of

Therefore, purchasing
is favorable. Next, we will determine
should be purchased. At this point, the system already has
if
. If purchasing , the reserve becomes
reserve

Evaluating system reliability under the condition of reserve ca, we obtain the EENS of the system
. Thus,
pacity
is
the benefit of

When the benefit is larger than the cost of

purchasing
is favorable.
Continuing this process until

Fig. 2. Flowchart of operating reserve clearing process.

and for the uniform-price mode, (9) can be rewritten as


and
becomes the optimal operating reserve capacity. As
shown in Fig. 1, the curve is continuously decreasing until
, where reaches its
the above inequality holds when
minimum.

(13)

III. ALGORITHM OF OPERATING RESERVE MODEL


A heuristic algorithm based on the merit-order method is employed to solve the above operating reserve model. The solution
process of the model for PAB mode is shown in Fig. 2.
In the main loop of the algorithm, the reserve capacity is purchased one unit by one unit according to the merit order.
and
are the systems EENS before and after purchasing
of unit . Thus, the decrease of EENS due
operating reserve
is
to increasing reserve
(11)
For the pay-as-bid mode, (9) can be rewritten as
(12)

Now, we will briefly introduce the solution process in Fig. 2


with the following.
1) Form the merit order according to bidding price.
2) Calculate
by running a reliability evaluation program under the condition that the reserve capacity of the
system is zero.
3) Set initial values.
with the lowest bidding price among the
4) Select reserve
residual units.
5) Check whether constraints are satisfied; if not satisfied, go
to (9), and select the next unit.
to system reserve
, and calculate
.
6) Increase
7) Calculate the decrease of the social cost.
8) Check whether the purchasing condition (10) is satisfied.

IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 3. Test system load curve for a 24-hr period.

Fig. 4. System spinning capacities.

9) When the purchasing condition (10) is satisfied, select the


next unit, and go back to step (4).
is not favorable;
10)When (10) is not satisfied, purchasing
stop purchasing and obtain the optimal reserve capacity.
The main difference between the solution processes of the
two market modes lies in the formulas used to obtain the purchase cost. The incremental social benefit cost in pay-as-bid
mode is calculated by (12), while that in uniform-price mode
is calculated by (13). Thus, replacing (12) with (13) in Fig. 2,
we get the flowchart for the uniform-price mode.
IV. CASE STUDIES
The IEEE test systemRTS96is employed to demonstrate
the proposed operating reserve model. The system has three
areas, and each area has six hydro units, 24 thermal units, and
two nuclear units [12]. The total installed is 10 215 MW. The
peak load is 5860 MW at 10 a.m., and the base load is 3398 MW
at 3 a.m. The unit information (including type, size, bidding, and
selected capacity in energy market) is shown in the Appendix.
The load curve for a 24-hr period is shown in Fig. 3. IEAR is
assumed to be $1/kWh. The spinning reserve market is cleared
by cost-benefit analysis in every period, and the spinning reserve
capacities of 24-hr are given in Fig. 4.
Fig. 4 shows that the spinning reserve curves in two modes
both have the flexible characteristic, i.e., spinning reserve capacities are determined by social benefit analysis in every period
rather than the predefined deterministic criteria. This method
can maximize the social benefit. The shape of SR in Fig. 4 is
similar to the load shape in Fig. 3. It indicates that the peak
capacity of spinning reserve corresponds with the peak load.
The result is reasonable, and the system reliability can be well
guaranteed.
The unit order is fixed in the uniform-price mode, and the
selected unit has the lowest bidding price in the residual capacities. The clearing capacities may vary with the load situation and
sometimes fluctuate abnormally. In the PAB mode, an additional
loop could be introduced in the algorithm, as shown in Fig. 2, to
improve the operating reserve selection strategy. In every loop,

Fig. 5. Comparison of reserve ratios.

the social benefit of all residual reserve units is calculated, and


the unit with maximum social benefit in the current loop is selected. In this way, the unit order is not fixed with the bidding
price but is dynamically adjusted according to their social benefit in each loop. In this approach, the selected unit always has
the highest social benefit, which enhances the stability of the
clearing process.
In order to analyze the feasibility of the flexible reserve, the
ratios of the spinning reserve over the load obtained by the suggested flexible spinning reserve model are compared with those
of the traditional model in other papers. As is shown in Fig. 5,
the scheduled reserve of the traditional model has a smaller ratio
at peak load and larger ratio at base load, while the flexible reserve has a higher ratio at peak load and lower ratio at base load.
Obviously, the clearing capacity of the flexible reserve is more
reasonable.
The optimal spinning reserve capacity is affected by the value
of IEAR. Fig. 6 shows the clearing capacity of the spinning reserve market with different values of IEAR. It can be observed
from these curves that the optimal spinning reserve capacity
varies significantly with the value of IEAR. When the IEAR

WANG et al.: OPERATING RESERVE MODEL IN THE POWER MARKET

Fig. 6.

System spinning capacities with different values of IEAR.

Fig. 8. System spinning capacities versus different values of FOR.

Fig. 7. Social benefit of reserve.

Fig. 9.

Clearing capacity when all plants raise price together.

increases, the spinning reserve benefit becomes larger, and the


system should buy more spinning reserve capacity.
In the view of flexible reserve, the social benefit should be
maximized in each period. The social benefit of two modes
during the peak load is shown in Fig. 7 under the same bids.
The maximum social benefits and optimal spinning reserve capacities are marked in the figure.
Fig. 7 shows that the social benefit increases until reaching
the maximum value, and then, it slowly decreases. In the whole
clearing process, the social benefit in the PAB mode is higher
than that in the uniform-price mode.
It should be pointed out that the advantage of operating reserve flexibility is embodied on the basis of enough capacity.
Under the condition of insufficient capacity, some reserve units
with a high bid are still selected. The clearing process may not
reach the maximum social benefit. So, the flexible operating reserve model is more suitable for large markets.
When the generators of the system are more reliable [they
have lower forced outage ratios (FORs)], less operating reserve

should be required. This adjustment can be automatically realized in the presented model. The required spinning reserve capacities are shown in Fig. 8 for different FORs of the generators.
Comparing with traditional dispatch, the proposed operating
reserve model is more flexible because the information about
the generators robustness can be used in determining optimal
operating reserve capacity. The system with less-robust generators should buy more capacity and vice versa. This will optimize
the usage of our resources.
Furthermore, flexible reserve can also reduce the market
power on the basis of the bidding information in the market.
Consider the worst case when all independent power producers
(IPPs) raise bidding prices together. Under traditional dispatch
rules, all the predetermined reserve capacity should be bought.
In the flexible reserve model, the market could buy less capacity
as countermeasures against rising prices. In order to explain this
effect on the spinning reserve market, Fig. 9 gives the clearing
capacity of spinning reserve as a function of markup rates.
From Fig. 9, we can see that the higher the bidding price,
the less spinning reserve capacity the market should buy. In

IEEE TRANSACTIONS ON POWER SYSTEMS

TABLE I
U100 ENERGY BID

TABLE II
SPINNING RESERVE CAPACITY DURING PEAK LOAD OF AREA 1

TABLE III
SPINNING RESERVE BID

market information and system reliability information when


purchasing the operating capacity. According to the model,
more operating reserve capacity should be purchased when the
system is less reliable, the value of IEAR is higher, and the
bidding price is lower in the power market, and vice versa.
Because the operating reserve capacity of a power system is
not predetermined, the proposed model not only can maximize
social benefit but also can reduce the market power. From the
results of case studies, it seems that PAB mode is superior to the
uniform-price mode in reducing market power and smoothing
the cost for the operating reserve market.
In order to realize the flexible operating reserve market, the
characteristic of IEAR and a specific market structure should
be carefully considered. The PAB and the uniform-price mode
need to be further investigated to judge which is preferable.
APPENDIX
The algorithm was demonstrated on the test system RTS96.
The energy market adopts simple patterns since the emphasis
of this paper is the operating reserve market. The cost of each
unit could be seen in paper [12]. Units submit four-block bids
and keep the bidding curves fixed all day. The bidding curves
contain the following four blocks: The first-block bid is lower
than the cost so that the unit will be chosen to serve. The second
and third block bids are equal to the costs. The fourth bid, which
is higher than the cost, is submitted to game for more benefit.
The energy bid of U100 is given in Table I as an example.
Table II gives the clearing results in the energy market and
the spinning reserve capacities during the period of peak load in
Area 1. The data of Areas 2 and 3 are similar with that of Area
1. The SR capacity is subject to the residual capacity and the
ramp rate simultaneously.
The bid curves of the spinning reserve are founded on the energy bids and the capacities that units can provide in ten minutes,
assuming that the bid is 50% of the energy bid. As an example,
the spinning reserve bid of U100 is given in Table III.
REFERENCES

the flexible reserve model, the adjustment can be carried out


automatically.
V. CONCLUSION
Based on cost-benefit analysis, this paper proposes a novel
operating reserve model. This model makes full use of the

[1] X. Wang, Y.-H. Song, and Q. Lu, A coordinated real-time optimal dispatch method for unbundled electricity markets, IEEE Trans. Power
Syst., vol. 17, no. 2, pp. 482490, May 2002.
[2] J. M. Arroyo and A. J. Conejo, Optimal response of a power generator
to energy, AGC, and reserve pool-based markets, IEEE Trans. Power
Syst., vol. 17, no. 2, pp. 404410, May 2002.
[3] H. B. Gooi et al., Optimal scheduling of spinning reserve, IEEE Trans.
Power Syst., vol. 14, no. 4, pp. 14851492, Nov. 1999.
[4] C.-L. Tseng et al., Price-based adaptive spinning reserve requirements
in power system scheduling, Elect. Power Energy Syst., vol. 21, pp.
137145, 1999.

WANG et al.: OPERATING RESERVE MODEL IN THE POWER MARKET

[5] X. Wang, Equivalent energy function approach to power system probabilistic modeling, IEEE Trans. Power Syst., vol. 3, no. 3, pp. 823829,
Aug. 1988.
[6] E. Yu, J. Zhou, and X. Zhang, Bidding model and bidding principle for
power markets, Automat. Elect. Power Syst., vol. 25, no. 1, pp. 2427,
2001.
[7] C. Vzquez, M. Rivier, and I. J. Prez-Arriaga, If pay-as-bid auctions
are not a solution for California, then why not a reliability market?,
Elect. J., vol. 14, no. 4, pp. 4148, 2001.
[8] A. E. Kahn, P. C. Cramton, and R. H. Porter, Uniform pricing or
pay-as-bid pricing: A dilemma for California and beyond, Elect. J.,
vol. 14, no. 6, pp. 7079, 2001.
[9] T. Mount, Market power and price volatility in restructured markets for
electricity, Decision Support Syst., vol. 30, pp. 311325, 2001.
[10] N. Fabra, N.-H. von der Fehr, and D. Harbord, Modeling electricity
auction, Elect. J., vol. 15, no. 7, pp. 7281, 2002.
[11] K. K. Kariuki and R. N. Allan, Factors affecting customer outage costs
due to electric service interruptions, Proc. Inst. Elect. Eng. Gener.
Transm. Distrib., vol. 143, no. 6, pp. 521528, 1996.
[12] C. Grigg, P. Wong, P. Albrecht, R. Allan, M. Bhavaraju, R. Billinton, Q.
Chen, C. Fong, S. Haddad, S. Kuruganty, W. Li, R. Mukerji, D. Patton,
N. Rau, D. Reppen, A. Schneider, M. Shahidehpour, and C. Singh, The
IEEE reliability test system-1996. A report prepared by the Reliability
Test System Task Force of the Application of Probability Methods Subcommittee, IEEE Trans. Power Syst., vol. 14, no. 3, pp. 10101020,
Aug. 1999.

Jianxue Wang received the B.S. and M.S. degrees in electrical engineering from
Xian Jiaotong University, Xian, China, in 1999 and 2002, respectively. Currently, he is pursuing the Ph.D. degree at Xian Jiaotong University.
His major research interests include power market and system reliability.

Xifan Wang (SM86) received the B.S. degree from Xian Jiaotong University,
Xian, China, in 1957.
He is a Professor in the Department of the Electrical Power Engineering at
Xian Jiaotong University. His major research fields include power market, reliability evaluation, generation planning, system contingence analysis, and stability analysis.

Yang Wu (S05) received the B.S. and M.S. degrees in electrical engineering
from Xian Jiaotong University, Xian, China, in 1999 and 2002, respectively.
Currently, he is pursuing the Ph.D. degree at Texas A&M University, College
Station, TX.
His major research interests include protective relaying and substation
automation.

You might also like