You are on page 1of 6

2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

Comparative analysis of ACOPF and DCOPF based


LMP simulation with distributed loss model
Deepesh Sharma, Naresh Kumar Yadav, Gunjan Bhargava, Anju Bala
Electrical Engineering Department
DCRUST Murthal
Sonepat (Haryana) India

Abstract— This paper introduces the concept of distributed transmission system designing, it is forever necessary and
loss model for calculation of Locational Marginal Price (LMP) in critical task to calculate or forecast LMPs, and if data is
electricity markets on the basis of Optimal Power Flow (OPF). available on generation, transmission, and load, LMPs might
This model has also been presented to address the mismatch issue be obtained using the traditional generation cost optimization
at system reference bus. The DC Optimal Power Flow (DCOPF) model[1], [2]. Zuyi Li et.al [4] introduces the fundamental
model has been evaluated in this paper to calculate the
concepts of market clearing price (MCP) and locational
Locational Marginal Price (LMP). Moreover AC Optimal Power
Flow (ACOPF) model has also been used for comparison
marginal price (LMP) in the electricity markets, and presents
purpose. But to overcome the drawback of lossless DCOPF some interesting observations on MCP and LMP. S.Saroha
algorithm, distributed loss based DCOPF model has been et.al [5] developed various mathematical models for
proposed and compared with other OPF techniques. IEEE 30 bus calculation of market clearing price(MCP) for power trading.
system is used in power world simulator for LMP calculation. Yong Fu et.al [6] gives four different models for LMP
calculations. Authors also give properties and applications of
Keywords—Locational Marginal Price(LMP); Market LMP. F.li et.al [7] shows the LMP simulation algorithms
Clearing Price(MCP); DCOPF; ACOPF; distributed loss; addressing marginal loss pricing based on the dc model.
congestion; power markets.
Generally, due to robustness and speed, LMP simulation or
I. INTRODUCTION forecasting based on generation cost model via linear
programming uses DC optimal power flow (DCOPF). In
The privatization and deregulation of electricity markets has
industrial practice, DCOPF has been used by several software
a very large impact on almost all the power systems around
tools for chronological LMP simulation and forecasting, such
the world. Competitive electricity markets are complex
as ABB’s GriveView™, Siemens’ Promod®, GE’s MAPS™,
systems with many participants who buy and sell electricity. and Power World [8],[9]. K.Purchala et al [10] mentioned in
Much of the complexity arises from the limitations of the
paper that if the line flow is less, the R/X ratio is not more
underlying transmission systems and the fact that supply and
than 0.25 and the voltage profile is sufficiently flat, then DC
demand must be in balance at all times. The Independent
model is used in power flow studies. J.L.M.Ramos et.al [11]
System Operator (ISO) is a regulating entity independent from
shows DC power flow impact on generation scheduling for the
the electric companies and optimizes the overall system
calculation of loss penalty.
operation.The market player generates bids (buy or sell) for E.Litniov et.al [12] gives a new transmission loss model to
bilateral contract, multilateral contract or contract of ISO. The
address the marginal losses and for calculation of loactional
bids consist of specified amount electricity (MW) at a given
marginal price(LMP).T.Overbye et.al[13] conducted a
price of given time. The auctioneer matches the buy and sells
comparison between DCOPF and ACOPF for a 12965- bus
bids subject to the approval to the contract evaluator. The
model of the Midwest U.S transmission grid and shows that
approval is based on available transfer capability (ATC) and
DCOPF is 60 times faster than ACOPF. M.Murali et.al [14]
other contingencies of the network. Seller and buyer agreed
calculated LMP using artificial techniques like genetic
upon a price for a certain number of MWs to be delivered at a
algorithm and bat algorithm. LMP’s can be derived using
specific timing. This price is called market clearing price
either an ACOPF model or a DCOPF model.
(MCP). When no congestion in the system price will be same
To calculate the LMPs in power world stimulator, a
for all the bidders. But in case of congestion the more
distributed loss model is used with DCOPF-based algorithm.
expensive units will have to generate and price will be This is employed to overcome the drawbacks of existing three
different for different areas. That price is called locational
models which are DCOPF model without loss, DCOPF model
marginal price (LMP) for any particular area[1],[2].
with marginal loss and ACOPF model for calculation of LMP.
LMP is the main device for the calculation of nodal prices in
This method gives better results than lossless DCOPF model
power markets and to control congestion of transmission. An
as losses are considered in this method. Losses are distributed
additional cost with marginal price of power to the location
on all the buses as extra nodal demand. Moreover drawback of
without disturbing the limits is called locational LMP. LMP is ACOPF model is overcome in the sense that it takes less
the aggregation of the congestion, marginal loss and costs of
iterations to converge and simple to compute. Section II
marginal energy [3]. From the perspective of generation and
reviews the LMP calculation by delivery factors with and

978-1-4673-9084-2/16/$31.00 ©2016 IEEE


2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

without marginal loss. Section III presents a new distributed total loss of the system i;
loss based DCOPF model to simulate the LMPs. Section IV
net injection at Bus i;
presents the ACOPF model for LMP calculation. LMPs
calculated from different DCOPF models are compared with Loss factor and delivery factor are calculated as given
the LMPs calculated by the ACOPF on IEEE- 30 Bus System below. The loss factor as:
in section V. In Section VI remarks are arranged.
∑ (6)
II. CALCULATION OF LMP
LMP is calculated by the optimal power flow (OPF) models. ∑ (7)
The OPF model considers the fuel curve of the system for
calculating the LMP. There are basically two models of OPF
Where
for calculating LMP which are ACOPF model and DCOPF
kth line power flow ;
model. Earlier models on LMP calculation with DCOPF does
not take into account the line losses. Zero loss price is used for kth line resistance .
calculating energy cost and congestion cost. Non-linear losses
analyses the lossless DCOPF model. After that, marginal loss According to superposition theorem, in the lossless linear
factor and marginal delivery factor supported generation shift dc network, a line flow can be taken into account as the sum of
factor (GSF) are determined. all the contribution from all power sources (generation is
considered as positive source and load as negative source).
A. DCOPF Without Losses This can be shown as
The basic DCOPF model [4] without the attention of losses
can be modelled as the minimization of the full production ∑ ( ) ∑ (8)
value concerned to energy balance and transmission
constraints. The voltage magnitudes are assumed to be unified Equation (8) can used further expand LF as
and the reactive power is neglected. Additionally, it is
assumed that there may be no demand elasticity. This model ∑ ∑
can be written as LP

Min ∑ (1) ∑ (∑ )

∑ =∑ (2) From the above formulation, it is found that, the loss factor
may be positive or negative. It means when loss factor is
|∑ | (3) positive, system losses are increased by increasing the
for k=1, 2, …, M injection at bus and When it is negative, system losses are
decreased with increase in the injection at the bus. As a result
, for i = 1, 2,…,N (4) if loss factor is +ve the Delivery factor is less then 1 and if it is
–ve the delivery factor is greater than 1.
Where,
C. DCOPF Model Considering Marginal Loss
N = Display of buses;
M= Display of lines; As proven in eqn. 9, loss factor counted at the net injection ,
= Generation cost at Bus i ($/MWh); that's the actual dispatch minus the weight at Bus j. On the
= Generation dispatch at Bus i (MWh); other hand, generation dispatch can be moved by loss factors
on account that exclusive generators can be penalized in
, = Max and Min generation output on Bus i;
another way based on their loss factors.
= demand on Bus i (MWh);
In any dispatch before performing is unknown, to solve
= Generation shift factor to line k from bus i;
= Transmission limit of line k. this issue, estimate the dispatch to get the apporximate LF at
every bus. New dispatch results are obtained by projected loss
B. Loss Factor and Delivery Factor factors. This logic reasoning results in the iterative DCOPF
The basic elements for considering marginal loss price are approach. For updation of estimated and result of
the marginal loss factor and the marginal delivery factor. lth iteration is used from (l+1) th iteration. LP-based total
These can be written in mathematical form as DCOPF is solved, in each iteration. Iterative process is
continued till the convergence stop. LMPs can be obtained
(5) effortlessly from the final iteration after convergence.
Where Similarly, the estimated and from the subsequent-to-
marginal delivery factor at Bus i ; remaining iteration may be similar to the very last values.
Primary equation in this model is of lossless DCOPF version
marginal loss factor at Bus i;
2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

in which loss is zero. The algorithm can be formulated as (19)


follows:
III. ITERATIVE DCOPF ALGORITHM WITH DISTRIBUTED
Min ∑ (10) LOSS FOR LOSSES
A. Mismatch at the Reference Bus in the Previous Model
s.t. ∑ ∑ (11) DCOPF model as shown above gives the marginal loss price
by the delivery factors. Still the constraints of line flow in eqn.
∑ (12) 12 takes a lossless network. System reference bus took up
engaged the load of reference bus which results in the
for i ϵ all generators (13) mismatch of the reference bus. If the amount demand is in the
range of few GW or above than it is not possible for the
Where system reference bus to absorb all the system losses because
delivery factor on the Bus from the previous they will be in the range of tens to hundreds of MW.
iteration;
B. Distributed loss Based DCOPF Model
The DCOPF model considering marginal losses gives Line losses should be presented in transmission line to
more accurate results as compared to lossless DCOPF model. resolve the mismatch difficulty at the reference bus. It is very
However the number of iterations are somewhat greater than difficult for line losses due to line flow which is represented
the lossless model. with GSF in LP-based DCOPF without losing the linearity of
With the help of lagrangian function the LMP on each the model.
Bus B may be obtained after calculating the optimal solution Losses of the lines coherent to a bus use the idea of
of generation dispatch. LMP and Lagrangian function can be distributed loss model. Distributed loss method is used right
written as here to distribute losses among each individual line for
decreasing the mismatch on the reference bus. Loss in each
transmission line is split up into two same parts and apply to
(∑ ) each part of line buses. All equal line losses are equal to
proposed nodal demand for every bus.
Here the extra nodal demand at Bus is written as ,
defined as follows:
(∑ ∑ ) ∑ (20)

Where
∑ ∑ (14) lines linked at Bus i.

The line flow can be retrieved from the more nodal demand
(∑ )
calculation in the earlier iteration. Here, is calculated as

∑ (15) ∑ ( ) (21)

Where, In new approach of distributed loss, value of will be


B; different however the calculation of loss factor use the same
=energy price of the eqn. (9). Replacement of eqn. (10)-(13) is done by new
system =price at reference bus; iterative DCOPF formulation as given below
= sensitivity of the kth
transmission constraint. Min ∑ (22)
From eqn. 15, LMP can be easily decomposed into 3
additives: marginal energy rate, marginal congestion rate and s.t. ∑ ∑ (23)
marginal loss rate. Mathematically it can be shown as:
∑ (24)
(16)
for i ϵ all generators (25)
(17)
When the above system gets overlapped the use of the
∑ (18)
generation dispatch of every unit ( ) as the convergence
2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

criterion, other parameters inclusive of the line flows ( , the DCOPF ACOPF Distributed loss
delivery factors , and the device loss will converge as Fig.2. Plot of LMPs for different models
well. Line limit of line 6-10 has been increased by 10% for
evaluation purpose. To maintain the voltage profile reactive
IV. ACOPF BASED LMP CALCULATION power limits are taken as -9900 MVar and +9900 MVar as
In [13], model of ACOPF is not used for marginal cost and ..LMPs are calculated using the DCOPF
simulation, its motto is for observation purpose. The ACOPF model, ACOPF model and distributed loss model. Fig.2 plots
model derogate the entire generation cost concerned to the LMP results obtained from these three models. As seen
balances of nodal real, reactive power, transmission limits, from the results, it is clear that LMPs from ACOPF aren't
generation limits and bus voltage limits[2]. Lagrange identical on each bus even at base load. This change in LMPs
multiplier of equality constraint of nodal real power balance is is due to the congestion present in the system at base load and
identical to LMP at every bus from ACOPF formulation due to losses which are present in the system. These losses are
ignored in DCOPF. LMP results which are obtained from
V.RESULTS AND DISCUSSION lossless DCOPF, distributed loss based DCOPF are compared
In this section comparison of lossless DCOPF and distributed with ACOPF at various load levels. Load is increased in p.u at
loss based DCOPF is given using IEEE 30 bus system all the nodal buses.
simulated in Power World Simulator. Results for LMP 6 1.0 p.u

LMP($/MWh)
calculated from given three models are presented in this
section. 4 1.025 p.u
2 1.05 p.u
A. IEEE 30-Bus System test Results
This section gives the test results for LMP simulation on 0 1.075 p.u
IEEE 30 bus system. There are some assumptions which are
1 4 7 10 13 16 19 22 25 28 1.1 p.u
also discussed.
Bus no. 1.125 p.u
Fig.3: plot of LMPs at different load levels for ACOPF

Figure 3 plots the results of LMP formulation from ACOPF


model. To analyse the effect of congestion on marginal price
load is slightly increased and LMP is analysed at different
buses. As we can see upto load 1.025 p.u, LMPs are nearly
same for both the load cases. Small difference in LMPs is due
to the increased no. of losses when load is increased. But when
we further increase the load LMPs also change.

Fig.1. IEEE 30 bus system simulated in power world simulator

IEEE 30 bus system is slightly modified for the test


purpose. Shunt capacitors are ignored in IEEE 30 bus system
for simplification.

Fig.4. Maximum difference of LMP between each DCOPF model and


4 ACOPF IEEE 30-bus system.
LMP($/MWh)

3.8
3.6
3.4
3.2
3
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Bus No.
2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

Fig.5. IEEE 30-bus system , Average difference of LMP between each model This model has been hired to calculate the LMPs the use
of DCOPF and ACOPF
of most appropriate strength glide (OPF) considering marginal
Figure 4 and 5 show the plot between the Maximum losses which are not considered in lossless DCOPF model.
Difference (MD) and the Average Difference (AD) of nodal
LMPs models. The Maximum difference and Average VI. CONCLUSION
difference of LMP at a given load level given in [14].
A new distributed loss based DCOPF model is laid out in
=max{| | } (32) this paper. This model has been engaged to calculate the
LMPs with the use of optimal power flow (OPF) considering
marginal losses which are not considered in lossless DCOPF
Where i ϵ {1,2…N} model. This method can also be applied to other OPF
techniques which considers contingencies in the system. This
∑ | |
paper initially presents the formulation of loss factors and
delivery factors based on generation shift factors. In this paper
= (33)
also deal with nodal mismatch problem on the reference bus.
Distributed loss primarily based DCOPF model is also
Where = Lossless DCOPF algorithm or the distributed introduced to make the cost of nodal mismatch on the
loss algorithm of LMP, ACOPF algorithm of LMP. reference bus negligible. This is employed by distributing the
Sign of maximum difference is obtained by the sign of losses to every bus in the system. This paper also presents the
( computation of the LMPs using basic lossless DCOPF model
and ACOPF model. A comparison of LMP results acquired
from the lossless DCOPF, the distributed loss-based DCOPF,
As shown in fig.4 and fig.5, difference between the LMPs
and the ACOPF model has been presented in this paper. LMPs
calculated by distributed loss Based DCOPF and ACOPF is
calculated using ACOPF model are accurate but this is not the
lower than the difference between the LMPs calculated by
model which is used in electricity markets for calculation of
lossless DCOPF and ACOPF, except only at one load level
prices owing to its complexity and heavy computational effort.
which is 1.125 p.u. This analysis clearly means that results
It is used only for comparison purposes. LMPs calculated by
obtained from distributed loss Based DCOPF are closely
DCOPF are not accurate results but lossless DCOPF is still is
related to ACOPF than the results obtained from lossless
used for evaluation of LMPs due to its simplicity. LMP
DCOPF. Since the lossless DCOPF ignores the line losses
calculated by distributed loss model in this paper is best
while distributed loss Based DCOPF considers the line losses,
among the on top of models. Results obtained from this model
therefore distributed loss based DCOPF performs better than
clearly indicates that it gives better results as compared to
the lossless DCOPF.
lossless DCOPF model and the results obtained from
B. Discussion of LMP Difference and Critical Load Level distributed loss based DCOPF are better approximation of
ACOPF LMPs results. Another important advantage of this
Here, the causes for LMP difference between distributed model lies in its simplicity. However distribution of losses
loss based DCOPF and ACOPF and critical load level are among various nodes presents some difficulty. Therefore,
discussed. The important LMP difference among DC model future research could explore more easy ways of distributing
and AC model is because of the approximation in the DC losses among the various nodes while preserving the
model, which causes different line flows and different simplicity of the system.
generations at the generating units and leads to different
prices. LMP consists of three components which are energy REFERENCES
price, congestion cost and loss cost. Since there is no
[1] M. Shahidehpour, H. Yamin, and Z. Li, Market Operations in
congestion in the system at base load in lossless DCOPF so
Electric Power Systems. New York: Wiley, 2002.
the second component which is congestion price, will be zero. [2] A. J. Wood and B. F. Wollenberg, Power Generation Operation
As the system is lossless therefore loss component will also be and Control. New York: Wiley, 1996.
zero. Hence marginal price at all buses will be same. In [3] S. Stoft, Power System Economics—Designing Markets For
ACOPF model marginal cost are exclusive due to the presence Electricity. New York: IEEE/Wiley, 2002.
of loss component which is overlooked in DCOPF model. Due [4] Zuyi Li and Hossein Daneshi, “Some observations on market
to congestion, distributed loss based DCOPF model marginal clearing price and locational marginal price,” Proceedings of
prices at base load are different. Due to extra load on the buses IEEE PES General Meeting, pp. 2702-2709, June 2005
which is basically the loss of the system congestion occurs [5] Sumit Saroha and Rohit Verma, “Cross-border Power
Trading Model for South Asian Regional Power Pool”, Electrical
into the system in distributed loss based DCOPF model and
Power and Energy Systems, vol. 44, pp.146–152, 2013.
hence congestion component comes into existence and [6] Yong Fu, Zuyi Li, “Different models and properties on LMP
therefore prices get change at every bus. calculations”, IEEE Power engineering society general meeting,
2006.
2016 International Conference on Control, Computing, Communication and Materials (ICCCCM)

[7] F. Li, J. Pan, and H. Chao, “Marginal loss calculation in [14] M.murali, M.sailaja Kumari, M Sydulu, “Optimal spot pricing in
competitive spot market,” in Proc. 2004 IEEE Int. Conf. electricity markets with inelastic load using constrained bat
Deregulation, Restructuring Power Technol. (DRPT), vol. 1, pp. algorithm,” Electr. Power Syst. Res., vol.62, pp. 897–911, May
205–209. 2014.
[8] J. Yang, F. Li, and L. A. A. Freeman, “A market simulation [15] A. J. Conejo and J. A. Aguado, “Multi-area coordinated
program for the standard market design and decentralized dc optimal power flow,” IEEE Trans. Power Syst.,
generation/transmission planning,” in Proc. IEEE Power Eng. vol. 13, no. 4, pp. 503–509, Nov. 1998.
Soc. General Meeting, 2003, pp. 442–446. [16] L.Liu and A. Zobian, “The importance of marginal loss pricing
[9] R. E. Clayton and R. Mukerji, “System planning tools for the in an RTO environment,” Electricity J., vol. 15, no. 8, pp. 40–45,
competitive market,” IEEE Comput. Appl. Power, vol. 9, no. 3, Oct. 2002.
pp. 50–55, Jul. 1996 [17] P.K.Iyambo, R.Tzoneva, “Transient Stability Analysis of the
[10] K. Purchala, L. Meeus, D. V. Dommelen, and R. Belmans, IEEE 14-Bus Electric Power System”, IEEE conference, Africon
“Usefulness of dc power flow for active power flow analysis,” in 2007.
Proc. IEEE Power Eng. Soc. Annu. Meeting 2005, Jun. 2005, pp. [18] Zechun Hu, Haozhong cheng, Furong Li, “An iterative LMP
454–459. calculation method considering loss distributions”, IEEE Trans.
[11] J. L. M. Ramos, A. G. Exposito, F. J. C. Moron, and S. N. Power. Sytem., vol.25, No.3, August 2010.
Becerra, “On the use of loss penalty factors for generation [19] Luonan chen, Hideki Suzuki, Tsunehisa Wachi, Yukihiro
scheduling,” in Proc. IEEE Power Eng. Soc. Annu. Meeting 2003, Shimura, “Components of Nodal prices for Electric Power
vol. 2, pp. 926–931. Systems”, IEEE Trans. Power. Syst., vol.17, No.1, February
[12] E. Litvinov, T. Zheng, G. Rosenwald,and 2002.
P.Shamsollahi,”Marginal loss modelling in LMP calculation,” [20] Fangxing Li, Rui Bo , “DCOPF-Based LMP Simulation:
IEEE Trans. Power Syst., vol. 19, no. 2, pp. 880–888, May 2004. Algorithm,Comparison With ACOPF, and Sensitivity”, IEEE
[13] T. Overbye, X. Cheng, and Y. Sun, “A comparison of the ac and Trans. Power Syst., vol. 22, No 4, Nov 2007.
dc power flow models for LMP calculations,” in Proc. 37th
Hawaii Int. Conf. Syst. Sci., 2004.

You might also like