Engineering, 2013, 5, 121-126
doi:10.4236/eng.2013. 51b022 Published Online January 2013 (http://www.SciRP.org/journal/eng)
The Research about the Trans-provincial Centralized
Bidding Trading Market of East China Power Grid --II：
Model Analysis
Bin Zou 1, Jie Fan1, Xiao-gang Li2, Li-bing Yan g2
1School o f Mechatronic En gineering and Automation, S hangha i Univer s ity, Shanghai, C hina
2East China Power Grid Co., Ltd. Shanghai, China
Received 2013
Abstract
In this paper a novel cost function based on the relationship between operation cost of unit and generation load rate is
employed in an agent-based model of Trans-provincial Centralized Bidding Trading Market of East China Grid.
Simulation results are compared to real data to prove that the model is correct. Further analysis on simulation results
point out the wa y to achieve a n all-win game for power market member s : generation companies impro ve their average
load rates of the units by selling their electricity in the market, which makes units' cost drop and settlement price stay
lo wer tha n b e nc h mar k p r ice . Consequently electricity-dema nd pr o vinc es save d e xp ens es, and units increase their profits.
In co ncl usion, the tra ns-pro vincial electricity market of East China Po wer Grid is a successive case which improves the
efficiency of the electricity industry b y market-oriented measures.
Keywords: Power Market; Agent-based Simulation; East China Grid; Trans-provincial Centralized Trade; Partial
Electricity Compe tition
1. Introduction
East China power grid is an interconnected power grid,
which contains five provincial power grids, and each
province power grid is a control zone. Each provincial
power grid company is responsible for their respective
provinces electricity and power balance. In order to meet
the load demand, about 1% of the total energy needs to
be exchanged among the provinces. The Trans-provincial
Centralized Bidding Trading Market of East China
Power Grid（TPM-ECPG）is designed to take advantages
of competition to form energy exchange programs
between provinces.
About 9 9% of t he gener ating energy is arranged using
so-called planned schedule method. In the method, each
unit was allocated a certain amou nt of generating energy
during the next year according to the installed capacity of
the un it, t he uni t type and t he for ecast ing lo ad. T hen, t he
po wer system operation sched ules the unit outp ut to meet
the actual load and guarantees annual generating energy
of the unit equal to the value to be assigned. And
Electricity is purchased by the grid company with the
benchmark price that is validated by the National
Devel opme nt and Refo rm Co mmi s sio n.
The TPM-ECPG that was put into operation in Dec.
2008, is the only one power auction. The result of last
three-year operatio n indicates that market trading volume
increased year after year, and to most degree electricity
demand has been satisfied. The trans-provincial
transaction platform has become the one of the main
ways of the tra ns-provincial transaction in East China
Power Grid.
In this p a pe r agent-based model is proposed as the tool
for the modeling and analysis of the trans-provincial
centralized auction transactio n p latfor m of East C hina. In
a lot of literatures, the agent-based model not only was
employed to estimate market characteristic with different
pricing method, namely uniform clearing pricing,
mid-point pricing and pay-as-bid pricing[5-7], but also
used to discuss the influence on the power market
exerted by blocking, collusionbid and emission trading
of CO2[11-18]. The method is able to show the basic
features that market members bid as the behavior of
maximizing their benefits, and its extensibility allows
taking more practical market factors into consideration.
The important feature of the TPM-ECPG is a part of
the electricity competitio n, part of electricity is
scheduled using traditional cost-based method, and

B.ZHOU ET AL.
122
another is sa le by competition. What is the impact of the
non-competitive part on the competitive market? In fact,
the market shows some interesting phenomenon. For
example, power generation companies are willing to
participate in market competition, even if the market
price is lower than the benchmark price.
This paper is one of the two-part series. This article
will use an agent-based simulation method to model the
TPM-ECPG. By doing so, it can help to analyze
simulated cases based on realistic data of East China
trans-provincial centralized power transaction.
2. Introduction of the TPM-ECPG
The TPM-ECPG functions as a double-auction market,
basic rules are as follows: buyers and sellers initially
report their respectively offered quantities of electricity
(MWh) a nd bid prices (￥/ MWh).
The approach of market clearing applies matchmaking
tradeoff model. The highest bid of demand side would
match the lowest ask price of supply side. Then the
market clearing price is their average value of the two
offered prices. The matchmaking procedure goes on by
the price sequences of sellers and buyers.
The transmission fee and loss compensation fee
belongs to demand-side power grid, while the
trans-provincial transmission fee and loss compensation
fee belongs to the East China Power Grid. The loss
compensation fee is 1% of benchmark price, and the
transmission fee is the difference between demand price
and supply price in principle. But if the sum of
transmission fee and loss compensation fee is more than
0.03￥/KWh, then the transmission fee is 0.03￥/KWh.
More detailed market rules Re ferences [1].
3. Agent-based simulation model
Unlike the model of the conventional electricity market,
the model of the East Chi na po wer mar ket must take into
the planned generation. The proposed model in this paper
employs the characteristics as followed: 1) Market
members in the TPM-ECPG always bid for profit
maximization; 2). Generated electricity and price of
planned generation is described through a cost function,
which can embody the influence exerted on the power
auction market. 3). Transaction rules are set up in
accordance with practical regulations of the TPM-ECPG.
To sum up, the model possesses the most market
characteristic s of TPM-ECPG, and is expected to explore
the inner mar ket mechanism b y simu lation.
3.1 Cost fun c tion mod e l
The generator's cost model presented is divided into two
parts of the operating cost and capacity cost. Capacity
cost is a constant during the modeling and simulation.
The main operation cost is fuel cost, in order to represent
the impact of the planned generation, the power
generation load rate is employed to calculate the
operation cost of the generating u nit in this paper.
For any one of the power supplier
i
, its planned
generating quantit y of electric it y is divided into t wo parts:
electricity on peak periods and electricity on valley
periods, the accepted quantity of electricity is classified
into peak and valley part as well.
=+
i ip iv
ipipp ips
ivivp ivs
QQ Q
QQQ
QQQ

= +

= +

(1)
Where
i
Q
denotes total generating electricity of
Unit
i
;
ip
Q
,
iv
Q
denote generating electricity of Unit
i
on peak period and valley period respectively;
ipp
Q
,
ivp
Q
denote planned generation of Unit
i
on peak period
and valley period respectively;
ips
Q
,ivs
Qdenote accepted
bid of Unit
i
on peak period and valley period
respectively.
In this paper one month is set up as observation period.
The average load rates of Unit
i
on peak/valley period
are as followed:
,
,ma x
,
,ma x
30 14
30 10
ip
i lp
i
iv
i lv
i
Q
rP
Q
rP
=
××


=
××

(2)
Where
,i lp
r
，
,i lv
r
denote average load rate of Unit
i
on peak/valley period.
,maxi
P
is unit’s maximum output.
30×10, 30×14 represent valley-period of 10 hours per
da y, peak-period of 14 hours per day. Each month has 30
days on an average.
The avera ge output of U nit
i
on peak/valley perio d is
as followed:
,
,ma x
,
,ma x
30 14
30 10
ip
i lp
i
iv
i lv
i
Q
rP
Q
rP
=
××


=
××

(3)
Where
,
ip iv
PP
denote average output of Unit
i
on
peak/valley pe r iod
Coal consumption of Unit
i
can be calculated by
coal consumption curve in Figure.1, which is expressed
as quadratic function:
2
=
iiiii i
raPb Pc×+×+
(4)
where
,,
iii
abc
are coefficients for different types of
generator units,
i
P
is unit’s output. Reference [19] shows
that the way to estimate coal consumption can be
accepted.
The running cost of power generation companies
expressed as the product of the coal consumption,

B.ZHOU ET AL.
123
electricity gener ation and coal p r ices
,
,
=
=
ipi pcip
ivi vciv
CrQ p
CrQ p
××


××


c
c
(5)
Where
ip
C
,
iv
C
denote operation cost of unit
i
on
period/valley period.
C
p
denotes coal price，
,i pc
r
，
,i vc
r
can be
obtain from (4).
3.2 Decision - making model
The decision-making model of the power supply
company is just maximizing profit based on its objective
function, and decision variable is the ask price
,
bid
i sell
ρ
,
( )
,
, ,,,,, ,,
max+-+ +
bid
i sell
isellibmi planibmi plani peakivalleyicap
RpQpQC CC
ρ
=××
(6)
where the planned quantity of generating electricity is
scheduled by provincial regulators and its price is
benchmark price of the respective provinces. Owning to
the proposed model, which is based on load rates, the
impacts of generating plans can be involved in decision
making smoothly.
In terms of electricity-demand provinces, they
purchase insufficient power through the trans-provincial
auction after implementing generating plans, so
insufficient power can also be regarded as a known
constant in the model. The power-demand provinces
expect to bid to satisfy their demand. Similarly, the
decision-making model o f such provinces is maximizi ng
profit of their obj ective function, and dec ision variable is
the bid pric e
,
bid
i buy
ρ
:
,
, ,, ,
max -
bid
i buy
i buyisi buyi buy ir
Rp QpQ
ρ
=××
(7)
Where,
,i buy
R
,
,ir
p
,
,i buy
p
and
,i buy
Q
denote profit, retail
price, closing bid price, accepted quantity of bid of
electricit y-demand province
i
respectively.
3.3 Agent-based simulation
Both the power supply company's decision-making
model (6) and power-demand province's
decision-making model (7) cannot be directly solved by
optimization methods, for they are all game procedures.
The agent-based simulation is appropriate to deal with
the game problems.
According to historical data, planned peak/valley
period generating electricity, quoted quantity, quoted
price, probable quotation set and other arguments can be
set up before simulation, reference [7] introduced this
method in detail. Simple agent-based simulation
procedure is as followed:
Step 1: In according to the roulette-like method, which
acquires selective prob a b ility, quotations can be pick up
for both units and buyers.
Step 2: After executing market-clearing, profits of units
and buyers can be calculated by ( 6) and (7).
Step 3: Units and buyers amend their selective
probability of quotation sets based on profit changes. The
target is to increase the probabilities of one quotation
whic h make s hi ghe r pr o fit.
Step 4: Iteration pluses one. If any quotation in the two
sets does not converge, go to Step 1.
The selective probability learning process, i.e., step 3,
using the Roth-Ere algorithm [7][8][9], detailed discussion
of the algorithm exceeds the scope of this article, please
refer to the literature [7][8][9].
4. Case Study
4.1 Data setting
In order to keep accordance with actual operating data of
the trans-provincial centralized bidding transaction
Table.1 Data of generators’ units
NO.
Installed
Capacity
MW
Province
Offered
Quantity
￥/MWh
Planned
Generation
MWh
1 300 A
6480
179280
2 600 A
12960
358560
3 1000 A
21600
597600
4 600 A 12960 358560
5 300 A
6480
179280
6 1000 A
21600
597600
7 1000 B 21600 597600
8 300 B
6480
179280
9 600 B
12960
358560
10 600 B 12960 358560
11 1000 B
21600
597600
12 300 B
6480
179280
13
600
B
12960 358560
Figure.1 coal consumpt ion curves of 3 ty pes of units

B.ZHOU ET AL.
124
platform of East China, the experimental environment
for power supply companies and their units in this paper
are set up and present in Tab.1. These power companies
with the sole unit are a ffiliate d to two di fferent pro vince s
of A and B, the electricity-demand provinces are C and D.
It is assumed that the coal price of province A is 720
Yuan/ton, and the coal price of B is 750 Yuan/ton.
4.2 Analysis on market equilibrium
Based on the data setting, market equilibrium res ults can
be obtained by agent-based simulation. Table.2 shows
the market clearing results of electricity-demand
provinces and units. The average bid prices of C and D in
Tab.3 is close to the actual bid of demand provinces in
2010 transactions(422.26￥/MWh vs 421.00￥/MWh),
relative error is about 0.3%. In addition the average
units’ ask prices(335.82￥/MW h) also approximate to
those in 2010(335.82￥/MWh), which leads to a relative
error of 4.6%.
Table.2 Simulat ion results on the o ffer and bid
Market
players
Quotation
￥/MWh
Winning
Bid
￥/MWh
Accepted
Quantity
MWh
Province C
416
400
79000
Province D
426
389
79000
300MW Unit
368
367
3400
600MW Unit
330
359
12960
1000MW Unit
326
358
21600
Furthe r ana lysi s on the res ults of Table.2 show that the
average winning bid prices of units and
electricit y-demand provinces compare to their actual
transaction data respectively, relative errors are 1.6% and
1.8%(absolute errors are 6.23 ￥/MWh and 6.51
￥/MWh). All the results and analysis figure out that
market equilibriu m of the model in this paper is close to
real market operation, and the model can embody most
of trans-provincial elec tr ic auction market.
According to Table.2, average ask-prices of 300 MW
units, 600MW units and 1000MW units are sorted in
descending order, which is as same as their operation
cost sorted order. It is indicated that generation
companies who owe auction units always bid on the
basis of units’ cost. The market characteristic makes
low-cost units take a dvantages in market auction.
The simulated winning bid prices of province C and
D(416.00￥/MWh and 426.00￥/MWh) are lower than
their benchmark prices(430.00 ￥/MWh and 457.00
￥/MWh) respectively, that saves expenses for both
demand-p rovince C and D.
Simultaneou sly the average winning bid price of units
in province A and B (359.27 ￥/MWh and 359.42
￥/MWh) are also lower than their benchmark prices
(410.00￥/MWh and 400.00￥/MWh) respectively.
To find out why generation companies are willing to
bid at a low price, average profit per unit (total profit
divide total ge neratio n) a nd average load rate are put into
consideration as shown in Table.3.
As it presents in Table.3, load rate increases after
generation companies participate in auctio n market. T hat
makes running cost go down. Moreover, when total
generation increases by 3%, the settlement price of
incremental part decreases nearly by 10% comparing to
benchmark price. Whereas the cost drops, total profit
grow by 9% and unit profit also goes up.
Table.3 the profi ts a nd load rate of the power s uppliers
part icipating or not in i nter-provinvial t r ans ac tio n
Total
generation
MWh
Total
profit
￥
Profit
per
unit
￥/M
Wh
Load rate
P V
Without
auction
transaction
4900320 2209618
65 45.09 75% 89%
Wit h bid
transactions 5058320 2414121 93 47.72 77% 91%
In order to further study the characteristics of the
TPM-ECPG, in this paper 11 trading scenarios on
different market scales, of which total installed is fixed,
are set up. Owing to a small wiggle room for buyers’
quoted prices, units’ average winning bid price can
mostly indicate market trading status(As shown in
Figure.2, where the horizontal axis represents the
percentage of maximum offer of electricity in market
biding to total installed capacity, namely market scale).
As can be seen from the simulation results of
Figure.2, When the TPM-ECPG maintain in a small
transaction volume, namely the competitive part of
electricit y accounted for less than 25%, units average
clearing price overall keeps a downward trend. It is
pointed that the market within a c e r ta in small siz e has the
ability to reduce the generation costs continually, to
promote genera tion compa nies to participate in bid ding，
and possess es the conditions to expand transaction scale.
When the transaction scale goes up, average clearing
electricity price rebounded, and presents a substantia l
rebound trend constantly. As the market expands to the

B.ZHOU ET AL.
125
scale of trans-provincial trading electricity accounting
for more than 75%, then, the total capacity which can be
put into auction is large enough to approach to
Figure.2 Units’ average bid pri ce under different market
scale
deregulated electricity competition scenario, and the
average clearing price is slightly lower to medium-sized
market, but still higher than smaller one.
In order to study the root cause of this situation, the
study focused on the units’ average unit profit (total
revenue / (planned generation + accepted bided
generation)) under different market size when the
simulated market get an equilibrium, as showed in Figure.
2.
Figure.3 Units’ marg in pr of i t under different market scale
As can be seen from Figure.3, the units' average unit
profit decrease constantly when the market scale
gradually expands. On the one hand it is indicated that
the planned power generation is still as dominant
position in both supply and demand side, and generation
companies which participate in electricity auction cannot
get enough profit caused by coal consumption reduction
to compensate the loss of planned generation. On the
other hand, before the trans-provincial transactions
accounts for 25% of total capacity, the unit profit is
greater than zero, so the power companies' profitable
opportunitie s still e xist u ntil p ro fit is less tha n zero as the
scale of market transactions expand. That is to say
generation companies have to bid at a high price level in
a larger auction market with current rules. It is also
explained why there is a sudden rebound of price in
Figure.2.
Combining all above simulation analysis, the planned
generation as the main method of electricity supply and
demand in the current market environment cannot be
shake n, b u t T PM-ECPG mak es a better market efficiency,
and to some extent be on the potential for further
expansion.
5 Conclusion
In this paper, the trans-provincial centralized bidd ing
trading market of East China is modeled by an
agent-based method, a novel cost function and cases
based on actual data are employed to simulate the
operation of the market. The simulation results are used
to analyze market characteristics on the equilibrium
status. Here I summarize what we can learn from the
modeling and simulations:
1) The proposed model in this paper reflects basic
characteristics of the trans-provincial centralized
transaction platform, in which its market
equilibrium accord with practical statistics of the
power auction market.
2) The generation companies that participate in the
trans-provincial auction platform possess restricted
strategy space, they bid base on cost, so that is
possible for e fficient units to a cquire more available
electricity to generate. It helps to achieve the goals
of energy saving and emission abatement
3) Trans-pr ovincial tra nsactions make the
electricit y-demand side save purchasing cost, the
supply side cut down operating cost, and it is a well
designed power market which satisfy the society,
electricity generation enterprise and grid co mpa nies.
The essence of the all-win game is that the
generation companies bid base on their cost, and the
transactions make the generating cost cutoff by
increasing the units' load rates. For all the reasons
above, both buyers and sellers can gain profit when
the market closing prices are lower than benchmark
prices.
4) Simulated results indicate that the trans-provincial
centralized bidding transaction platform can
preserve basic market characteristics after
expanding its tr a nsaction sca le within constraints.
5) Trans-provincial centralized auction platform ca n
further expand on the existing basis, but planned
generation is still the main profitable part for
generation companies, to expand the scale is not
easy to be too large.

B.ZHOU ET AL.
126
Referen ces
[1] ZOU Bin, WANG Xiaojun, The Research about the
Trans-provincial Centralized Bidding Trading Market of
East China Power Grid --I：Empirical Analysis, Power
and Energy Engineering Conference, 2013
[2] Ross Baldick,UdiHelman,Benjamin F. Hobbs, Richard P.
O’Neill,Design Of Efficient Generation Markets,
Proceedings Of The IEEE , 2005, 93(11):1198-012.
[3] National Electricity Regulatory Commission, The
instruction that accelerating the development of the
trans-provincial centralized bidding transactions in East
China[R], 2009(in Chine s e ) :
http://www.ecerb.gov.cn/publicJsp/detail.JSP?inforId=8586
[4] East China Electricity Regulatory Commission, East
China Trans-provincial Centralized Auction Power
Market Achieved Remarkable Success on Energy Saving
and Emission Reduction[R], 20 10 ( i n Chi ne s e ) :
http://www.serc.gov.cn/jgdt/pcjg/2010004/t20100402_1
2861.htm
[5] Eric Guerci, Mohammad Ali Rastegar, and
SilvanoCincotti. Agent-based Modeling and Simulation
of Competitive Wholesale Electricity Markets[J],
Handbook of Power Systems II, Energy Systems,
Springer-Verlag Ber lin Heidelberg, 2010, Part3:241-286
[6] Anke Weidlich, Daniel Veit. A critical survey of
agent-based wholesale electricity market models[J],
Energy Economics 2008, 30:1728–1759
[7] Zou Bin, Yan Maosong, Xie Guangqian, Comparisons
Among Pricing Methods in Pool-based El ectrici ty Mar ket
by Agent-based Simulation Part One: Model and
Algorithm [J], Automation of Electric Power Systems,
2004, 28 ( 15 ) :7-14.
[8] Nicolaisen J, Petrov V, Tesfatsion L. Market Power and
Efficiency in aComputational Electricity Market with
Discriminatory Double-auction Pricing. IEEE Trans on
Evolutionary Computation, 2001, 5(5): 504~523
[9] Ro th Alvin. Erev Ido . Learni ng in Ext ensive-for m Games:
Experiment Data and Simple Dynamic Model in
the Intermediate Term. Game and Economic Behavior,
1995(8): 164~212
[10] Yan Maosong, Modern Power Market and Electricity
Value Equivalent Pricin g System[M], Beijing: China
Power Press, 2008(in Chinese)
[11] LIU Guozhong, WEN Fushuan, XUE Yusheng. Impacts
of Emissions Trading and Renewable Energy Sources
Development Schemes on Electricity Market
Operati on[J]. Automation of Electric Power Systems,
2010, 34(7):21-25.
[12] WANG Yeping, JING Zhaoxia, CHEN Haoyong1, et al.
A Met hod for Designing Agent-based Electricity Market
Simulation Experiments[J], Aut omat ion of Electric Power
Systems, 2009,33(17):57-60,85.
[13] Eric Guerci, StefanoIvaldi, SilvanoCincotti,Learning
Agents in an Artificial Power Exchange: Tacit Collusion,
Market Power and Efficiency of Two Double-auction
Mechan isms[J]. Comput Econ, 2008, 32:73–98
[14] AudunBotterud, Matthew R. Mahalik, Thomas D.
Veselka,Heon-Su Ryu, Ki-Won Sohn,Multi-Agent
Simulation of Generation Expansion in Electricity
Markets[C]. Power Engineering Society General Meeting,
IEEE,2007:1-8
[15] AnkeWeidlich, Frank Sensfuß, Massimo GenoeseI,
Daniel Veit, Studying the effects of CO2 emissions
trading on the electricity market: A multi-agent-based
approach [J]. Emissions Trading,Springer, 2008, Part
B:91-101
[16] ZHOU Hai-ming, WANG Hai-ning, SHI Shu-hong, et al,
Development of Electricity Market Simulation System
and Its Application[J]. Power System Technology, 2010,
34(1):117-121
[17] TAN Zhong-fu, LI Li, WANG Jian-jun, et al. A Fuzzy
Bayesian Learning Model in Agent-based Electric Po wer
Bilateral Negotiation[J]. Proceedings of the CSEE, 2009,
29(7):106-11 3.
[18] TIAN Jian-wei, HU Zhao-guang, WU Jun-yong, et al.
Dynamic Economy and Power Simulation System Based
on Multi-agent Modelling[J]. P roceedin gs of the CSEE[J],
2010, 30 ( 7):85-91.
[19] Liu Fu guo, Jiang xuexia, Li zhi, Investigation on Affects
of generator Load on Coal Consumption Rate in Fossil
Power Plant. Power System Engineering, 2008，24（4）：
47-49.
[20] Yan Maosong, Modern Power Market and Electricity
Value Equivalent Pricin g System[M], Beijing: China
Power Press, 2008(in Chinese)