You are on page 1of 7

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

3, AUGUST 2004 1575

Strategic Self-Dispatch Considering Ramping Costs


in Deregulated Power Markets
G. B. Shrestha, Senior Member, IEEE, Kai Song, Student Member, IEEE, and L. Goel, Senior Member, IEEE

Abstract—Ramp rates of generators are generally specified Ramp Rate for Elastic Range (MW/min).
within elastic range of the strength of the shaft to safeguard the Cost rate of the unit for alternative ($/MW).
rotor from fatigue. These limits can, however, be exceeded, albeit Power axis intercept for alternative (MW).
at the risk of reducing the rotor life. Such effects on the rotor
life can be compensated by incorporating appropriate ramping
costs. Power demand and power price in deregulated power
markets have shown a tendency of sudden wide excursions over I. INTRODUCTION
short intervals of time. The ability to respond to such fast change
in demand and price can be quite rewarding. This provides the
motivation to utilize the ramping rates beyond traditional elastic
limits.
R AMP rates are the specified maximum rates at which the
power output of a power producer can be increased (ramp
up rate) or decreased (ramp down rate) at a trading interval.
This paper studies the strategic use of ramping rates beyond When the ramp rate is less than the elastic range, the corre-
elastic limits in a power producer’s self-dispatch in a power market sponding ramping process will not shorten the life of the rotor
with price and demand volatility. A set of ramping processes has
been developed from ramping-cost versus ramping time relation- and therefore no ramping costs are incurred. When the ramp
ships to derive the total operation cost including ramping costs for rate is greater than the elastic range, the economic impact due
various levels of ramping rates exceeding elastic limits. A thorough to rotor fatigue (reduction in rotor life) is expressed in terms of
theoretical analysis has been conducted for the selection of the op- the ramping cost. The ramping cost is obtained by converting
timal ramping process under different conditions. A numerical ex- the fatigue effect of the rotor into factors that can be incorpo-
ample is presented to highlight the potential benefits from the op-
timal use of these ramping processes. It is seen that the benefit from rated in generation scheduling [1].
the strategic use of ramp rates beyond elastic limits not only de- The conservative way to specify the ramp rate is to prescribe
pends on the high price in the power market but also on the initial ramp rates within the elastic range. However, the generation
loading of the generator which limits the capacity available for dis- scheduling based on the conservative ramp rates may not be
patch. optimal as these ramp rates may limit the ability of a unit to
Index Terms—Power dispatch, power market, ramp rates, provide energy and reserve during periods of fast changing
ramping cost, ramping process. demands. This may require operation of the more expensive
units.
NOMENCLATURE Therefore, it may be desirable to widen the range of ramp
rates by considering the ramping process in the generation
Index for the number of ramp rate alternatives.
scheduling so as to realize the tradeoff between the ramping
Index for the number of time intervals (in hours).
costs and the cost of operating the more expensive peaking
Time for one study interval (1 h).
units. As the ramping costs vary with different generation
Maximum output of the generating unit (MW).
schedules, it is more complicated to incorporate the ramping
Minimum output of the generating unit (MW).
costs in the generation schedule. Methods to minimize the
Energy delivery in the interval (MWh).
total system operation costs by considering the ramp costs in
Continuous power level of unit versus time
power scheduling and dispatch are proposed in [1] and [2].
(MW).
In a deregulated environment, power producers have more
Operational cost of the unit with energy ($).
freedom to utilize their generation unit capabilities to enhance
Ramp cost of the unit for alternative at hour
their benefits in the power market. A producer can internally
($).
schedule its generation to operate at its optimal condition
Power price at hour ($/MWh).
provided its capabilities meet the requirements or the level it
Ramping time of the unit for alternative (h or
bids to offer [3]–[5]. Several studies have been carried out on
min).
the impacts of the ramp rate in a deregulated environment. In
Ramp rate of the unit for alternative (MW/min).
[6], the authors introduce the market-based ancillary services
Ramping process duration for alternative (h or
for ramping, which implies that the ramping capability, as
min).
a kind of service, can be directly traded in the market. In
Elastic range of the unit for alternative (MW).
[7], [8], it is pointed out that although ramp rate constraints
are satisfied in hydrothermal scheduling, taking a generation
Manuscript received September 19, 2003. level schedule as an energy delivery schedule may not be
The authors are with the Power Market Research Group, School of Electrical
and Electronic Engineering, Nanyang Technological University, Singapore. realizable. The sufficient and necessary conditions without/with
Digital Object Identifier 10.1109/TPWRS.2004.825891 the consideration of the minimum and maximum outputs of a
0885-8950/04$20.00 © 2004 IEEE
1576 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 3, AUGUST 2004

unit are obtained to check whether an energy delivery schedule


is realizable.
As the ramp rates of a unit limit its rate of change of power
output, these rates can affect the power producer’s ability to
adjust the output quickly during periods of high demand/price
excursions. This may affect considerably the potential benefit
gained from the power market. The purpose of this paper is
to study the possible benefits to a power producer, say an
independent power producer (IPP), if it could dispatch its
own generation strategically using its ramping capabilities in
response to the market information (power price). The effects
of different ramping processes on a power producer’s benefits
are thoroughly analyzed, and a procedure to identify the
optimal ramping process for optimal self-dispatch considering
Fig. 1. Ramping costs for various ramping times.
the ramping costs is presented.
The paper is organized as follows. Section II outlines the fea-
tures of power market used in this study, particularly related to III. RAMPING PROCESS AND ENERGY DELIVERY
ramp rates. Section III explains the ramping processes and the
operation costs including ramping costs. Optimal choice of the A. Ramping Process
ramping process and the corresponding costs are presented in Commonly, the ramping rates are specified within elastic
Section IV. A numerical example is illustrated in Section V be- limits, which do not affect the normal life of the rotor. However,
fore presenting the conclusions in Section VI. higher ramping rates could be adopted provided some rotor
fatigue affecting the rotor life could be tolerated, and such
effects on the rotor life can be expressed in economic terms.
II. FEATURES OF POWER MARKET ENVIRONMENT
A reasonable specification of ramping process will be of fun-
The power market features and operating practices vary damental importance for this analysis. Fig. 1 reproduced here
widely among different systems [9]–[12]. General features from [1] is used in this paper to establish the relationship be-
adopted in this study are outlined here. It is assumed that the tween ramping costs and ramp rates for different ramping dura-
dispatch orders and prices are determined each hour with the tions.
objective of minimizing the power supply cost while satisfying The four curves in the figure show the variation of ramping
balance of supply and demand and other system constraints. costs for 4 alternative ramping processes for
The market-clearing price for every hour is the marginal price ramping excursions , up to 100 MW. The ramping duration
(the highest price of the committed producers). Each producer times of these processes are 60, 40, 20, and 10 min respectively
internally schedules its generating resources and bids to a and is denoted by . As reported in [1] these curves
power pool to sell the energy based on the capacities of its are based on the fatigue index curves provided by the generator
power facilities and the market information such as the power manufacturers. The initial and final statuses of the ramping pro-
price. It is further assumed, for the purpose of this paper, that cesses are steady states. That means the cost curve for a 40-min
during sharp rises in power demand and therefore prices, the ramp can be used to determine the ramping cost only when the
suppliers can increase their sales as permitted by the ramp rates ramping time is 40 min. Therefore, for the proper utilization of
of their generating units. the ramping cost curves, a priori specification of the duration of
Thus, ramp rates are a kind of resource, which can enhance the ramping processes must be done. In this study, the four al-
sales during high price periods. For a power producer, there are ternatives with different ramping times represented by the four
two ways to make use of this resource. First, it may choose to curves are considered.
keep the ramp rate within the elastic range in a conservative way. No Ramping Cost (RC) is incurred if the ramping excursions
This will not shorten the life of the rotor, but the ability to ad- remain below a certain value for each of the ramping processes.
just output quickly will be limited. The other way is to specify For example, it can be seen from Fig. 1 that the ramping cost for
higher ramp rates that will incur ramping costs but can respond 40-min ramp remains zero up to a ramping excursion of 70 MW.
much faster to the change of the power prices. It may be more This value of ramping excursion is termed the Elastic Range and
profitable for a producer to increase the power output quickly is denoted by for alternative . The ramping rate corre-
when the power price is high enough. The benefits gained from sponding to the elastic range is denoted for alternative
the sales during high price period may well exceed the ramping . for 40-min ramp becomes 70 MW in 40 mins, or
cost incurred. A detailed analysis of the ramping processes and 105 MW/h.
the associated ramping costs will be necessary for this evalua- Ramping costs are incurred only when the ramping excursion
tion. Therefore, for an IPP, it would be possible to use ramp rates H or the desired change in power output exceed the elastic range.
strategically to maximize its benefits by properly identifying the The ramp rate above the elastic range is denoted by ,
optimal ramping process for its generating units. which can be calculated as .
Important terms and notations listed above are explained In the following analysis, the ramp curves are simplified by
under the “Ramping Process” in Section III. approximating them by the four dashed straight lines as indi-
SHRESTHA et al.: STRATEGIC SELF-DISPATCH CONSIDERING RAMPING COSTS IN DEREGULATED POWER MARKETS 1577

TABLE I
PARAMETERS FOR THE FOUR ALTERNATIVES OF RAMPING DURATIONS

Fig. 2. Energy delivery at time interval [ k; k + 1].

cated in Fig. 1. The power axis intercept of these straight lines


are denoted by for alternative . The slopes these lines
are the Cost Rates denoted as for alternative .
The parameters for the four ramping alternatives obtained
from Fig. 1 are listed in Table I.
The ramping cost for alternative can now be written as

(1)

These ramping costs are to be incorporated in the operating


cost to develop a strategic scheduling for maximum benefit. Fig. 3. Relationship of energy delivery and ramp rates in [ k; k + 1].
B. Operation Costs As shown in Fig. 3, at time interval , the beginning
The operating cost function of a generating unit in one power level is . is the elastic range. It is clear that
time interval can be expressed as area AFHIGA is the region where no ramping cost is incurred.
Ramping costs are incurred only when the desired change
(2) in power output level or exceed
the elastic range . These conditions are summarized in
Fig. 2 shows the variation of output during the ramping Table II.
process. It consists of the ramping period and a con- When the ramping process is included in the power dispatch,
stant output period hour . The time is denoted 0 at the the effective operating cost of a unit considering the change of
beginning of the hour and 1 at the end of the hour . From the power output may be written as
Fig. 2, the energy delivery during the time interval is

(3)

and the power output is

when
(4)
when .
While considering the ramping process, the distinction be-
tween the power level and energy output must be noted. For
each time interval, the power level during the ramping process
is modeled as a linear function with the ramp rate as the slope.
With the constraints on the maximum and minimum outputs, (5)
the relationship between the generation level and energy output Denoting
in one time interval can be obtained. It is illustrated in Fig. 3
for alternative 2, where the duration of the ramping process is (6)
40 min. (7)
1578 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 3, AUGUST 2004

TABLE II TABLE III


RAMPING COST CONSIDERATIONS UNDER THE FOUR CONDITIONS VALUES OF RR AND B (k) WHEN RAMPING COST IS INVOLVED

The operating cost in (5) can be expressed as where is the production cost and is the ramping
cost involved.
For every choice of ramping alternative , two scenarios are
possible.
i) Ramping cost is involved in the following conditions.
(8) a) and
.
where and depend only on the parameters of the cost b) and
curves of the unit and the operational state of the previous hour. .
It is noted that the ramping time must be no more than the du- From (1), (3), and (8), the benefit function can be de-
ration of the ramping process for alternative 2. When rived as
the ramp rate is greater than , ramping cost is incurred
and the ramping time must be the same as the duration of the
ramping process in order to use the ramping cost curves in Fig. 1
correctly. For example, in Fig. 3, when the power trajectory falls
in the area ABDHFA or area ACEIGA, the ending point can only
be on the segments BF or CG, which means that the ramping
(10)
time is the same as min. In this case, the op-
erating cost may be expressed as (5) where .
For this quadratic expression, the maximum benefit is
On the other hand, when the power trajectory falls in the area
achieved when
AFHIGA, no ramping cost is involved. In this case, the end point
for the ramping process falls in the area AFG, such as point .
The ramping time need not always be the same as . This (11)
may lead to different ramping processes with the same amount
of energy delivery. For simplicity, it is assumed that the ramping and the maximum benefit is
time is the same as .
The above expressions are for alternative 2 with the duration
of the ramping process as 40 mins. Similar expressions can be
obtained for other alternatives with the durations of the ramping (12)
process as 10, 20, and 60 min, respectively. The optimal values of ramp rates, , and the op-
timal values of the benefit, , are obtained using this
IV. OPTIMAL SCHEDULING WITH RAMPING COSTS methodology and are summarized in Table III. It should
Consider scheduling a generating unit with four ramping pro- be noted that when there is a reduction in power
cesses discussed above. Suppose at hour , the initial power output.
output is , and the market price at hour is . If ii) Ramping cost is not involved under the following condi-
an extra energy could be dispatched due to ramping, then the tions.
total benefit derived is a) and .
b) and .
(9) c) and .
SHRESTHA et al.: STRATEGIC SELF-DISPATCH CONSIDERING RAMPING COSTS IN DEREGULATED POWER MARKETS 1579

TABLE IV
VALUES OF E k
( ) AND B (k) WHEN NO RAMPING COST IS INVOLVED

Under these conditions, different ramping processes may re-


sult in the same amount of energy delivery and (2) is used to
obtain the operating cost. Therefore the benefit function of a
unit is

(13)

The maximum value of can be obtained as

(14)

when the energy delivery is

(15)

In addition, the energy delivery is subject to the following


limits when no ramping cost is involved (see (16) and (17) at
the bottom of the page).
As changes in the power output are within the range of elastic
ramp rates, no ramping cost is incurred. The optimal values
of energy delivery and the benefit, , are listed in Fig. 4. Flowchart to determine the ramping process at hour k .
Table IV.
The procedure to determine the ramp rate and ramping time The possible benefit at hour is related to both the market
at hour for maximum benefit at this hour is summarized in a power price at hour : ) and the power output at the be-
flowchart shown in Fig. 4. ginning of hour : ). The maximum benefits are computed
with and without incorporating the ramping considerations for
V. ILLUSTRATIVE EXAMPLE the following range of these factors.
A numerical example is used to (i) demonstrate the method- i) : $ 4/MWh to $54/MWh in $ 0.5/MWh steps.
ology to optimize the benefit from a unit considering the ii) to in 20 MW steps.
ramping cost, and (ii) determine the amount of possible benefit The variation of the benefit through these ranges of operation
vis-à-vis the relevant factors. with/without the consideration of ramping costs are shown in
Consider a 120 MW unit with minimum loading of 20 MW Fig. 5(a) and (b), respectively. Fig. 5(c) shows the difference
(ie, , ). between the two, which is the additional benefit resulting from
Let the cost coefficients be , , the consideration of the ramping process.
. As seen from Fig. 5(c), benefits are higher when the market
Let the ramping cost functions of Fig. 1 and the corresponding price is high and the initial power output is small. When the
data listed in Table I describe the ramping capabilities. price is high, the more energy a unit can supply, the more ben-

when
(16)
when
when
(17)
when
1580 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 3, AUGUST 2004

Fig. 6. Ramping costs due to consideration of ramping process.

Fig. 7. Energy delivery ranges for different initial power outputs.

The extended range of delivery ranges increases the revenue


during high price periods, but it would also incur the associated
ramping costs. The maximum benefit is achieved with optimal
tradeoff between the benefit from increased energy delivery
and the ramping costs incurred.

Fig. 5. Benefits with/without the consideration of ramping process. VI. CONCLUSIONS


Strategic use of ramp rates beyond elastic limits in genera-
efits it can generate. When the ramping process is considered, tion dispatch has been investigated. Such ramp rates will not
the unit will be able to generate more owing to the wider range only incur ramping costs to compensate for the rotor fatigue
of energy delivery. When the market price is about $ 20/MW, due to this process, but they will also widen the possible range
little extra benefit is obtained. When the price is quite low (says of energy delivery, which provides the opportunity to enhance
$ 4/MW), hardly any benefit is obtained by considering the the benefit from the price fluctuations in the market. Ramping
ramping process. capacity, as a resource, should be assessed on the basis of the
Ramping costs incurred for the same range operating condi- overall benefit it can achieve under the market conditions.
tions are shown in Fig. 6. It is seen that higher ramping costs A set of four simplified ramping processes were developed
are incurred when the market price is high and the initial power and used in the analysis. An approach to identify the appropriate
output is small, which indicates that the ramping process is uti- ramping process from the available alternatives to optimize the
lized only when the operating conditions demand it. Despite benefit has been presented. The analysis provides significant
these ramping costs, the overall benefit is enhanced as seen from insight into the ramping process and a valuable understanding
Fig. 5(c). of the effect of the ramping process on the producer’s benefit.
It should be noted that the increase in benefit is realized Only one generating unit and one time interval was considered
because of the ability of the generator to adjust its output in the analysis presented here. Extension of this study to multi-
quickly with the higher ramp rates. Fig. 7 shows the minimum units for a number of intervals would be a good topic for further
and maximum energy outputs possible at different initial unit analysis and investigation.
loadings for the ramping processes. Thus, the area between A numerical example has been presented to illustrate the im-
curves and is the range of energy delivery pact of the ramping process over a wide range of operating con-
with conversational ramp rate limits. The range of energy de- ditions, i.e., power prices and initial power outputs. It is shown
livery with ramping process is extended to the areas between that with optimal choice of ramping processes in the power dis-
and ( , , , and ), respectively. patch, the additional benefit can be quite significant when the
SHRESTHA et al.: STRATEGIC SELF-DISPATCH CONSIDERING RAMPING COSTS IN DEREGULATED POWER MARKETS 1581

power price is high and the initial power output of the unit is [10] I. J. Perez-Arriaga, H. Rudnick, and W. O. Stadlin, “International power
low. It is also seen from the results that the additional benefit system transaction open access experience,” IEEE Trans. Power Syst.,
vol. 10, pp. 554–561, Feb. 1995.
comes from the wider range of energy delivery created by the [11] T. Alvey, D. Goodwin, X. Ma, D. Streiffert, and D. Sun, “Security-con-
higher ramp rates. Therefore, the selection of ramping processes strained bid-clearing system for the New Zealand wholesale electricity
beyond elastic range should be carefully adopted considering market,” IEEE Trans. Power Syst., vol. 13, pp. 340–346, May 1998.
[12] O. B. Fosso, A. Gjelsvik, A. Haugstad, B. Mo, and I. Wangensteen,
both the unit characteristics and the power prices in order to op- “Generation scheduling in a deregulated system: The norwegian case,”
timize the benefits. IEEE Trans. Power Syst., vol. 14, pp. 120–125, Feb. 1999.

REFERENCES
[1] C. Wang and S. M. Shahidehpour, “Optimal generation scheduling with G. B. Shrestha (S’88–M’90–SM’92) received the B.E.(Hon.) degree in elec-
ramping costs,” IEEE Trans. Power Syst., vol. 10, pp. 60–67, Feb. 1995. trical engineering from Jadavpur University, India, in 1975, the M.B.A. degree
[2] , “Ramp-rate limits in unit commitment and economic dispatch in- from the University of Hawaii in 1985, the M.S. degree in electrical power engi-
corporating rotor fatigue effect,” IEEE Trans. Power Syst., vol. 9, pp. neering from Rensselaer Polytechnic Institute, Troy, NY, in 1986, and the Ph.D.
1539–1545, Aug. 1994. degree in electrical engineering from Virginia Tech in 1990.
[3] D. Zhang, Y. Wang, and P. B. Luh, “Optimization based bidding strate- He is an Associate Professor at Nanyang Technological University, Singa-
gies in the deregulated market,” IEEE Trans. Power Syst., vol. 15, pp. pore, and his main area of interest is power system operation and planning.
981–986, Aug. 2000.
[4] H. Song, C.-C. Liu, and J. Lawarree, “Decision making of an electricity
supplier’s bid in a spot market,” in Proc. IEEE Power Engineering Soc.
Summer Meeting, vol. 2, 2000, pp. 692–696. Kai Song (S’99) received the B.E. and M.S. degrees in electrical engineering
[5] S. Hao, “A study of basic bidding strategy in clearing pricing auctions,” from Tsinghua University, China, in 1993 and 1997, respectively. He has re-
IEEE Trans. Power Syst., vol. 15, pp. 975–980, Aug. 2000. cently completed all requirements for his Ph.D. degree in electrical engineering
[6] B. H. Bakken, A. Petterteig, and A. K. Nystad, “Market based ancillary at the Nanyang Technological University, Singapore.
services for ramping,” in Proc. IEEE PES WM2000, Jan. 23–27, 2000.
[7] X. Guan, F. Gao, and A. Svoboda, “Energy delivery scheduling and re-
alizability in deregulated electric energy market,” in Proc. 32nd Hawaii
Int. Conf. System Sciences, 1999. L. Goel (M’92–SM’96) received the B.Tech. degree in electrical engineering
[8] , “Energy delivery capacity and generation scheduling in the dereg- from Regional Engineering College, Warangal, India, in 1983 and the M.S. and
ulated electric power market,” in Proc. 21st IEEE Int. Conf. Power In- Ph.D. degrees in electrical engineering from the University of Saskatchewan,
dustry Computer Applications, 1999, pp. 25–30. Saskatoon, Sask., Canada, in 1988 and 1991, respectively.
[9] H. Outhred, “A review of electricity industry restructuring in Australia,” He is an Associate Professor at Nanyang Technological University, Singa-
Elect. Power Syst. Res., vol. 44, pp. 15–25, 1998. pore.

You might also like