2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 H o n g K o n g

Trading in Power Exchanges with Consideration

of the Cost of Generation Scheduling under
the New Electricity Trading Arrangements
Somporn Sirisumrannukul and Goran Strbac, Member IEEE

Abstract-This paper presents an optimisation method for offer number corresponding to offer index d.
the trading in power exchanges (PXs) with the consideration price of o n ( d ) at period t in the single market.
of the cost of generation scheduling under the New
Electricity Trading Arrangements (NETA) in the electricity quantity in h4W to be purchased of o ~ ~ ( dat)
market of England and Wales. The power exchange market, period t in the single market.
running parallel to NETA since its implementation, allows number of bids at period t in the block market.
market participants to adjust their contract positions in real
time to avoid any excess or shortfall in the balancing number of bids in the block market.
mechanism (BM) by submitting bids or offers. With bids and bid number corresponding to bid index q.
offers present in PXs, generating companies have price of brt(g) at period t in the block market.
opportunities to gain profit from this market. A proposed
methodology is able to optimisel power exchange and quantity in M W to be sold of bit(q) at period t
generation scheduling problems simultaneously, thereby in the block market.
assisting generating companies to decide which bids or offers number of offers at period t in the block
should be carried out, and to commit and dispatch their units market.
to meet their net contract positions. The methodology will be
demonstrated by a 5-unit power system. number of offers in the block market.
offer number corresponding to offer index r.
Index Terms-The New Electricity Trading Arrangements, price of ori(r) at period t in the block market.
Power Exchanges, Generation Scheduling.
quantity in MW to be purchased of on(r) at
I. NOMENCLATURE period t in the block market.
output power of unit i at period t.
e total profit defined as the profit from PXs
generation cost of unit i when its output is & f .
minus the cost of generation scheduling.
number of units in the system.
half-hourly index.
minimum stable generation of a unit i.
study periods.
maximum output of unit i.
bid index in the single market.
start-up cost of unit i.
offer index in the single market.
hot start-up cost of unit i.
bid index in the block market.
cold start-up cost of unit i.
offer index in the block market.
thermal time constant of unit i.
number of bids at period t in the single
duration unit i has been off-line.
market.
4 number of bids in the single market. X L duration unit i has been on-line.
bn(c) bid number corresponding to bid index c. ~~e,co,t~uct net contract position at period t.
s;(r,bn(c)) price of bfz(l:)at period t in the single market. D' original contract position at period t.
fi(t,bfl(c)) quantity in MW to be sold of ~ I I ( C ) at period t pAaxrrade maximum net traded power in MW
in the single market. at period t.
Nit?) number of offers at period t in the single f2.maX(hn(c)) maximum quantity of h 4 c ) in the single
market. market.
number of offers in the single market. $.max(onid)) maximum quantity of U J @ ) in the single
Ni
market.
f F m a ( b n ( q ) ) maximum quantity of h ( q ) in the block
This work was supported by Energy Policy and Planning Office of
Thailand. market .
Sompom Sirisumrannukul and Goran Strbac are with the p,8.max(on(r)) maximum quantity of on(r) in the block
Department of Electrical Engineering and Electronics, UMIST, PO Box
88, Manchester, M60 lQD, UK (e-mail: market.
[email protected], G.Strbac @umist.ac.uk).

0-7803-8237-4/04/$17.0002004IEEE
184
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

binary-number indicator of bid bn(q) at period 11. INTRODUCTION

t, i.e. if @(t,bn(q)) = I , bid brt(q) is scheduled to
be sold and @(r,bn(q)) = o otherwise. S ince the introduction of NETA, the short-term power
exchanges (PXs) have been established and
considerably developed. The trading in the short-term PXs
binary-number indicator of <Jfl(r) at period t,
has become one of daily-business activities among
i.e. if 6;(t,on(r))=i, on(r) is scheduled to be generating companies under the NETA environment. The
purchased, and 6 z ( r , 0 n ( r ) )= o otherwise. purpose of PXs is to allow market participants to adjust
first period of bn(L) in the single market. their contract positions in respect of each half hour trading
length in half-hourly period(s) of b t ~ ( c )in the period by purchasing or selling electricity in order to
avoid any excess or shortfall prior to the opening of the
single market.
balancing mechanism (BM) [ 1,2], which National Grid
first period of or@) in the single market.
Company (NGC) as a system operator employs to keep the
length in half-hourly period(s) of on(d) in the system balanced.
single market.
first period of bn(4) in the block market. The deregulation of electric power industries is leading
length in half-hourly period(s) of bn(q) in the generating companies to schedule their own resources
block market. using price-based criteria with the aim of maximising
first period of on(r) in the block market. profit. A question encountered by generating companies is
length in half-hourly period(s) of on(r) in the how to find optimal dispatch and unit commitment
block market. decisions not only meeting their contract positions but also
bid index in the single market. benefiting from trades available in PXs. In finding such
decisions, an optimisation tool needs to be able to
first period of bid g in the single market.
accommodate generation scheduling and power exchange
counter of bid g in the single market. trading simultaneously, taking into account the discrete
binary number. blocks of the standard products in PXs as well as system
binary number. and individual unit constraints.
offer index in the single market.
The highlight of this paper is given to an optimisation
first period of offer h in the single market.
framework for profit maximisation in the short-term
counter of offer / I in the single market. trading in PXs with consideration of the cost of generation
binary number. scheduling based on generating companies’ perspective.
binary number. The contents of this paper are organised as follows.
bid index in the block market. Section I11 briefly details the background of the short-term
offer index in the block market. trading in PXs, the standard products, and the trading
system. Section IV presents an optimisation formulation.
first period of bid 11 in the block market.
A case study is described in section V. Section VI
first period of offer L in the block market. summarises the paper.
sufficiently large positive number.
sufficiently small positive number. 111. POWEREXCHANGES
IN NETA
maximum number of trades. Power exchanges are operated as a computerised
unit commitment status of unit i at period t, market. Market participants may submit bids (for those
i.e. if 11: =I, 11; is committed and U : = o who want to buy) and offers (for those who want to sell) in
otherwise. a term of the standard products that determine how much
minimum up time of unit i. quantity they are willing to purchase or sell at what prices.
minimum down time of unit i. Bid and offer prices for all of the products are quoted at a
duration unit i has been on-line. flat rate throughout a specified delivery period, and can be
modified or withdrawn at any time until they are
ramp-up rate limit of unit i. electronically accepted one and a half hour before the
ramp-down rate limit of unit i. period of delivery.
spinning reserve at period t.
maximum spinning reserve of unit i. There are two power exchanges currently in operation
initial power of unit i. in the electricity market in England and Wales: 1) the UK
initial condition of unit i. Power Exchange (UKPX), and 2) the UK Automated
coefficient of a polynomial cost function Power Exchange ( M X ) [ 2 ] . The power exchange of
of unit i. interest in this paper is UKPX, in which a majority of
trades take place [2]. Trading in the short-term UKPX
based on a half-hourly basis is specified in a discrete

185
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

block. The discrete block of the electricity products may

be associated with one half-hour or span over several half-
hours. The typical products can be found in [3].
B. Constraints
The standard products can be traded in two different
ways. The first one, called single market, is the place The objective function given in (1) is subject to
where bid and offer quantity has to be posted in multiple constraints of the power exchange and the generation
of 1 MW only. Therefore, in this market, power can be scheduling problems as follows:
purchased and sold in any multiple quantity of 1 MW up
to the full quantity. The other, called block market, is the B.lNet contract position: a net contract position after
place where bid and offer quantity has to be posted in purchasing and selling power fromlto PXs for both single
multiple of 50 M W only, i.e. 50 MW and 100 MW etc. and block markets is given below.
Trading in the block market is operating based on a “make
or break” principle. Either the entire blocks of bids and
offers are sold or purchased, or they are not fulfilled [4],
i.e. bids and offers can be matched only in the full
quantity.

B.2 Maximum net traded power: net total purchased and

IV. OFTMISATION
FORMULATION sold power has to be less than maximum net power that
can be traded.
A. Objectivefunction
(4)
The objective function is to maximise the profit from
PXs minus the cost of generation scheduling over study
Vr = 1.2,...,T
periods and can be written mathematically as follows.

B.3 Bid and offer quantity: power to be purchased and

sold has to be less than or equal to maximum quantitiy of
bids and offers repectively.

B.3.1 Single market

The above objective function states that based on

generating companies’ perspective, bids and offers
correspond to the price and quantity at which generating
companies are willing to sell and purchase respectively.
Therefore, bids produce revenue whereas offers create
cost. It is assumed that revenue from bilateral contracts is
constant and not included in the above objective function.
B.3.2 Block market
A. 1 Generation cost
Piece-wise linear cost functions with two elbows are
used to represent polynomial cost functions of generating
units in form of F ~ ( c ) = o ; ~+ *h q + C . These two elbows can
be obtained by dividing equally a range between minimum
stable generation and maximum output of generating units.

A.2 Start-up cost

B.4 Discrete block: two markets are modelled: the single
The start-up cost of generating units can be and block markets.
represented by the exponential function of the time that a
unit has been shut down [ 5 ] . B.4. I Single market

186
2004 E E E International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

This constraint can be therefore expressed as:

B.6 Power output limits: generating units have to be

dispatched when committed within their minimum stable
generation and maximum output capability.
B.4.2 Block market

B.7 Minimum-up time: once generating units are

committed, they should remain on-line up to minimum
periods of time, qi'p .

B.8 Minimum-down time: once generating units are de-

committed, they should not be re-started until after
minimum periods of time, T / ,~ have
~ elapsed.

B.9 Ramp-rate capabilit): power output level of on-line

generating units cannot be increased or decreased to any
desired value within a short period of time.
B.5 Maximum numbers of trades: in practice, with more
of less the same amount of profit from two different
solutions, generating companies would prefer the one with
a smaller number of trades to be carried out. In the block
market, &r,bn(y)j and @(t,on(r)) are used as the counters of
B. 10 Power balance: power in each period has to be equal
trades. In the single market, the additional counters have
to the net contract position.
to be introduced to indicate that if Pj(r.bn(cj) or $ ( r , o n ( d ) ) is
greater than zero, it will be equal to one and zero
otherwise. This can be achieved by using the folowing
equations [6].
B. 11 Spinning reserve requiretnent: spinning reserve
B.5.1 For bid g provided by part-loaded units is given as follows [7].

The standard solver [8] is employed to solve the above

mixed-integer programming problem.

v. CASESTUDY
The test system given in table V of the Appendix
B.5.2 For offer h consists of 5 generating units with a total installed
capacity of 462 MW and a 48 half-hourly contract
position with peak demand of 400 M W . The operational
characteristics of the generating units were collected from
[9], with the modification in the unit of currency and with
some additional data defined by the authors.

Suppose that all of the 5 generating units are selected

to participate in PXs and that the first period of the

187
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

contract position starts on 11:OO p.m. today and the last with the different products. The committed capacity for
one ends at 11:OO p.m. on the next day. Only frequently the net contract position is shown in Fig.4.
seen half-hour and two-hour products are expected to be
purchased and sold. Bid and offer prices are based on the
UKPX fixing price, which is the average price for all
trades in each settlement period on the UKPX [lo]. Bid 40
and offer prices are, therefore, considered to be the same h

in each period even though in general, offer prices are $ 30

higher than bid prices. However, a small positive number F

?2 20
(0.0001) is added to an offer price in each period to avoid
the unfavourable situation that a bid and an offer for the 10
same product are selected with the same amount of
quantity in an optimal solution, resulting in zero net profit
with two trades. The UKPX fixing prices used in this case 1 7 13 19 25 31 37 43
study are assumed, given in Fig.1. The prices for the two- Period (half hour ) -
hour product are approximated by averaging the UKPX Fig.1.UKPX fixing price.
fixing prices in the appropriate half-hours.
TABLE I
POWER SOLD IN THE SKGLE MARKET
There are 60 bids and 60 offers in the single market
Period Product Power sold (MW) Revenue ( E )
whereas in the block market, 30 bids and 30 offers are
2 1-24 two-hour 50.00 3,269.50
supposed to be available only in peak periods, i.e. period
29-32 two-hour 42.00 2,183.97
17 to 40. The quantity of all bids and offers is S O M W in 33-36 two-hour 31.00 1,526.44
the single and the block markets. It should be emphasised 37-40 two-hour 32.00 2,611.36
that generating companies practically have to decide 41-44 two-hour 36.00 1,847.52
which bids and offers are likely to be available and then
use them as input data. A number of bids and offers are TABLE II
POWER PURCHASED Lpi THE SKGLE MARKET
not therefore necessarily as many as the one presented
Period Product Power purchased (MW) Cost ( E )
here. The other relevant input data is given as follows:
21 half-hour 14.00 200.06
44 half-hour 23.00 271.86
1-4 two-hour 50.00 1.209.01

TABLE III
POWER SOLD IN THE BLOCKMARKET
Period Product Power sold (MW) Revenue ( E )
20 half-hour 50.00 524.25
The simulation results for the profitable trades are 25-28 two-hour 50.00 2,815.25
shown in Table I to IV. The computation time for the
simulation is about 90 seconds, tested on PC Pentium 4 TABLE IV
POWER PURCHASED IiV THE BLOCKMARKET
1.5 GHz. There are S bids and 3 offers selected in the
Period Product Power purchased (MW) Cost (E)
single market. Only 2 bids are taken from the block
market. Without participation in PXs, the optimal solution
is -&123,895.09. In the other words, the cost of generation
scheduling is 2123,895.09. With participation in PXs, the
optimal solution becomes -El 18,114.55, f13,697.36 and
f131,811.91 of which are the profit from PXs and the cost
of generation scheduling respectively. This means the
total profit is 25,780.54 (-El 18,114.55-(-2123,89S.O9)).
3
The original contract position is graphically compared E
200 -
with the net contract position shown in Fig.2. The net
traded power in each period is demonstrated in Fig.3,
where the negative and the positive values represent the
100 - -
-Original contract position
Net contract position

net power purchased and sold respectively. It can be seen 0

that the power is sold from period 20 to 44, in which the
prices are high whereas the power is purchased during
period 1 to 4, in which the prices are low. In periods 21
and 44, the power is purchased and sold at the same time

188
2004 IEEE Intemational Conference o n Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 H o n g K o n g

60
VI. CONCLUSION
1
This paper presents the optimisation framework to
40
determine output of generating units owned by generating
20 companies with the objective to maximise the profit from
the short-term power exchanges under NETA with
consideration of the cost of generation scheduling over the
-20 one-day time span. The presented formulation optimises
the power exchange and the generation scheduling
-40
problems simultaneously taking into account the various
-60 I i constraints in both problems.
Period (half -hour )
The case study is performed using the 5-unit system. It
Fig.3. Net traded power in each period is shown from the case study that the profit obtained from
PXs is significant. However, other profit-influencing
factors such as prices, maximum number of trades, and
maximum net traded power have to be taken into
consideration. It is also demonstrated that the developed
algorithm is computationally efficient. The methodology

100 -
- Net contract position
-Committed capacity
presented in this paper would not be only useful for profit
maximisation for generating companies but also for cost
minimisation if their contract positions could not be
0 delivered. Consequently, their contract positions could be
met as economically as possible by PXs while reflecting
the cost of generation scheduling.

VII. APPENDIX
TABLEV
THE GENERATIKG
UNIT CHARACTERISTICSOF THE 5-UNIT SYSTEM
Unit $‘“n emax U b c ai p; 7; T? ~ f ” R;bl ”., x:
(MW) (MW) (f/MWh’) (f/MWh) (f) (E) (h) (h) (h) (MWIhl2) (MW/h/2) (MW) (MW) (h)

1 2.40 12.00 0.0253 25.547 24.389 0.00 0.00 1.00 0.00 0.00 24.00 30.00 8 0 -I
2 12.00 50.00 0.0031 9.407 23.626 68.00 30.00 2.00 1.00 1.00 12.50 12.50 4 0 -1
3 40.00 100.00 0.0043
’ 13.600 400.000 160.00 40.00 3.00 2.00 3.00 20.00 20.00 7 0 -2
4 45.00 120.00 0.0070 19.200 219.775 80.00 80.00 4.00 4.00 3.00 22.00 22.00 7 100 3
5 40.00 180.00 0.0056 12.700 40.000 80.00 80.00 3.00 4.00 3.00 32.50 32.50 II IS0 6

VIII. REFERENCES IX. BIOGRAPHIES

The New Electricity Trading Arrangements Volume 1, The Office Somporn Sirisumrannukul is currently a PhD student at the Electrical
of Gas and Electricity Markets, July 1999. Energy and Power System Research Group, Department of Electrical
The Review of the First Year of NETA, Volume 1, The Office of Engineering and Electronics, University of Manchester Institute of
Gas and Electricity Markets, July 2002. Science and Technology (UMIST), UK. His main research interests are
http://www.ukpx.com. in the area of power system economics, operation and reliability.
http://www.eex.de/ info-center/faq/spot2-e.asp.
Allen J Wood and Bruce F. Wollenberg, “Power Generation Goran Strbac (M 1995) is professor of electrical power engineering at
Operation and Control”, John Wiley & Sons, 1996. Department of Electrical Engineering and Electronics, University of
Wayne I Winston “Operations Research Application and Manchester Institute of Science and Technology (UMIST), UK. His
Agorithms”, Third Edition, Wadsworth Publishing Company 1994. main research interests are in the area of economics of power systems
Arthur I Cohen, “Modeling Unit Ramp Limitation in Unit centred on pricing of network and ancillary services, regulation and the
Commitment”, Proceedings of the Tenth Power Systems economics of dispersed generation.
Computation Conference, 1990,pp 1107-14.
[LOG CPLEX Optimization Library Version 9.0,2003.
Orero S 0,Irving M R., “Large scale unit commitment using a
hybrid genetic algorithm”, International Journal of Electrical Power
& Energy Systems. Vol.19. No. I , January 1997,OD 45-55.
[lo]McclaqC, Bliar V, Alcock D and Bake; R, “The’lmpact of NETA
Imbalance Prices on Contracting Strategies for Generation and
Demand”, Power System Management and Control (PSMC), 2002,
17-19April 2002,pp 13-15.

189