You are on page 1of 7

1

A DECENTRALIZED CONGESTION MANAGEMENT


USING INTERIOR POINT METHOD
Brijesh Singh, R. Mahanty and S.P. Singh, Senior Member, IEEE

Abstract- This paper presents a framework to carry out an In multiple market environments, the seams issues lead to
optimal power flow in a coordinated multi area decentralized market inefficiency in the operation of the combined
system. An AC optimal power flow model is used for independent markets. Hence, a decentralized approach is needed to
optimal dispatch of each area while the global economic optimal
facilitate economically efficient energy trading between
of whole electric energy system is achieved. This is achieved by
means of the interior point optimization procedure. In this regional electricity markets [6].
approach, each participant tries to maximize their own profit In [7], a decentralized model partitioned by tie-line
with the help of system information announced by the operator. between individual markets has been proposed for
The developed algorithm can be run in parallel either to carry coordinating trading between regional electricity markets.
out numerical simulations or in an actual multi area electric Tie-line information is exchanged at the end of each
system. The simulation is conducted on a three area modified iteration until final convergence is achieved. Different
IEEE-30 bus system. The results show clearly the effectiveness of decomposition methods for dividing the
the suggested method. It is shown that the suggested
Interconnected electricity markets into individual markets
decentralized approach can yield the same optimal dispatch
solution as the centralized approach. are introduced in [8].
After decomposition of a single joint market, a
Index Terms – AC-OPF, power market, open access, security decentralized approach to congestion management in
constrained optimal power flow, decentralization, optimal interconnected markets is proposed in [9]. Reference [10]
pricing. presents a decentralized approach for congestion
management in a spot market. An efficient economic
I. INTRODUCTION dispatch for competitive electricity markets is analyzed to
solve seams problems. The theory for algorithms of duality
An important goal of the restructuring of the electric
and decomposition in mathematical programming is
introduced in [11].
power industry is to move from ‘command and control’ As the push for deregulation of electric utilities continues,
decision making for the planning and operation of definition of objective function for optimization of
generation and transmission to a decentralised market- economic system operations is becoming critical. The
driven approach with the goal of achieving total lowest cost conventional benefit optimizations have been mainly based
to end-users without sacrificing reliability. After the on economic dispatch which enables us to achieve cost
restructuring of the electric power industry, profit minimization in single utility environment [12, 13].
generating companies have been developed to deliver Currently many of these issues are being played out in
electric energy in a competitive market and independent real time with the privatization of power system. The model
regulated transmission system operators (TSOs) manage the makes it possible the independent optimal dispatch of each
operation of the transmission system. The congestion area while the global optimum of the whole system is
management is one of the central issues for centralized achieved. This is particularly relevant in multi-utility or
optimal power flow reported in different literatures [1-5]. multi-country setting. In order to achieve an economical
The recent trend is towards large multinational electricity optimal dispatch of whole electric energy system, it is not
markets, such as the internal electricity market (IEM) in necessary to set up a common control centre. It is sufficient
Europe. However, there are technical and economic to interchange a small amount of information among the
challenges in the operation of a single joint market by involved utilities or countries [14].
combining different regional electricity markets. If an Decentralised OPF algorithm is an iterative algorithm in
individual market optimizes its own electricity market which the TSO of each region iteratively solves a modified
without coordinating with its neighboring markets, seams OPF subproblem for its own region and exchanges tie-line
issues arise between regional electricity markets. information with the TSOs of neighboring regions [15, 16].
Brijesh Singh, Ph.D. Research Scholar is with Department of Electrical In [17], the transmission congestion management and
Engineering, Institute of Technology, B.H.U., Varanasi. discovering the congestion prices in decentralized approach
(e-mail: brijeshsingh81@indiatimes.com). to maximize the profit independently for each participant is
S.P. Singh and R. Mahanty are with the Department of Electrical
discussed. The focus is given on the operation of multi
Engineering, Institute of Technology, B.H.U., Varanasi.
(sps5957@indiatimes.com, mahantyr@yahoo.co.in) utility system with objective of benefit maximization.
The current investigation being reported in this paper
deals with congestion management using decentralized and
centralized approaches. The centralized and decentralized
power systems in the multi-utility environment are defined
as under:

978-1-4577-1002-5/11/$26.00 ©2011 IEEE


2

• Centralized power system: This system consists of The objective in (1) is to maximize total social welfare. The
several electric utilities which are controlled by a equality constraints in (2) denote the power balance for each
centralized dispatching control like a tight pool. transaction. The inequality constraints in (3) denote the
output (or input) limits of generators (or consumers).
• Decentralized power system: In this system, each Inequality constraints in (4) denote the line capacity limits.
utility acts independently and has its own For simplicity the transmission lines is supposed to be
dispatching control. congested only in one direction.
The formulation of the problem is similar to that reported in
[17] but the solution is obtained by interior point method. B. Mathematical Model for Decentralized based OPF
AC load flow has been conducted unlike DC load flow in
The COPF model (1)–(4) is converted into DOPF model
[17]. Accuracy of decentralized approach is authenticated
below. Firstly we rewrite (1)–(4) to a simpler form. The
by comparing the results with centralised approach. IEEE decision variables of contract K are defined as
30 bus system has been used to demonstrate the
performance of method proposed. The test results reveal
that the proposed method produced superior results as The social welfare related to transaction k is rewritten as
compared to that reported in [17].
II. MATHEMATICAL FORMULATION

A. Outline of Centralized OPF Model Forward Markets

Forward contract market for real power is mainly focused, Equation (2)-(4) is local constraints of individual contracts
where dc load flow is used. In the perfect competitive that are not interactive among transactions. The original
market, the ISO adjusts the contracts to maximize the social DOPF model is written to a simpler form as
welfare to achieve efficient operation with all constraints
satisfied. This COPF is given by

(1)

This equality condition is only true for those lines which


are congested, means the lines are not overloaded, but
whole transfer capability has been utilized.

C. Solution by Interior point approach

This method is also known as Karmarkar interior point


Where method [18, 19]. The conventional optimization needed the
Nb number of bus; number of iterations to reach the optimum solution can grow
T set of transactions in the market exponentially. In fact there are extreme points visited before
T = {1,2,....K}; the optimum is reached in linear programming (LP). In
K index of each transaction, for all K T; interior point based optimization a polynomial-time
g(K) set of generators in transaction K, for all K T; algorithm that cuts across the interior of the solution space.
d(K) set of consumers in transaction K, for all K T; The algorithm is effective for extremely large LPs. The
Pgi output of generator i of transaction K, also an element main idea of the Karmarkar method is introduced here and
of generator output Pgi vector of transaction K, for all i then is provided the computational details of the algorithm.
g(K) and K T;
Pdi demand of consumer i of transaction K, an element of Basic idea of the Interior-point Algorithm:
demand vector Pdi of transaction K, for all i d(K)
and K T; For the solution by this technique, the problem is formulated
Bi benefit function of consumer i of transaction K, as
for all i d(K) and K T; min f(x)
Ci cost function of generators i of transaction K, s.t. h (x) = 0
for all i g(K) and K T; gl ≤ g(x) ≤ gu (7)
m index of transmission line involved in congestion where
management; f(x) is the objective function
Pij(m) load flow caused by transaction k on line m, in which h(x)=0 is a set of nonlinear equality constraints
Generator sensitivity factor are used, for all K T; (Power balance)
Pij max power flow limit in transmission line m in MW; gl ≤ g(x) ≤ gu is a set of nonlinear inequality constraints
x is a vector consists of state variables and
control variables
3

In interior point OPF, firstly construct a Lagrange function Fig.2 [17]. In order to demonstrate the multi-area example
for NLP eq. (7). the whole system is partitioned in three area A, B and C. it
Lg = f(x)-yTh(x)-zT[g(x)-l-gl]-wT[g(x)+u-gu] is assumed that each area have three load and two generator.
This way there are nine demand and six generation bidders
altogether. The marginal cost functions of generators and
marginal benefit functions of consumers are listed in Table
Where 1 and the electricity output q in MW.
y, z and w are Lagrange multipliers for equality and
inequality constraints respectively
li and ui are slack variables
µ is the barrier parameter.
So, the Karush-Kuhn-Tucker (KKT) conditions for NLP (7)
are
Lx = Jf - Jhy – Jg (z+w) = 0
Ly = h(x) = 0
Lz = g(x) – l – gl = 0
Lw = g(x) + u – gu = 0
Ll = LZE - µE = 0
Lu = UWE + µE = 0 (8)
Here
L = diag (l1, l2,…….lr)
U = diag (u1, u2, …….ur)
W = diag (w1, w2, …….wr)
Z = diag (z1, z2, …………….zr)
E = [1,1,……..1]T Fig. 2 Example system (IEEE 30 bus 3 area system)
Jf,, Jh and Jg are Jacobian matrixes of f(x), h(x) and g(x)
respectively. Table: 1
Using reduced Newton method to solve the nonlinear eq.
Assumed market demand and
(8) then can get the following three decomposed linear supply functions for example system
equations: Load
Generation Marginal cost bus Marginal
Utilities bus no. of supply no. benefit
(9) 3 3.88-0.03q
and 1 0.02q+2 4 3.9-0.05q
A 2 0.0175q+1.75 7 3.8-0.05q
12 3.8-0.04q
(10) 13 0.025q+3 15 3.95-0.02q
and B 23 0.025q+3 17 3.7-0.02q
24 3.6-0.03q
(11) 22 0.625q+1 26 3.8-0.04q
where C 27 0.00834q+3.25 27 3.7-0.02q
Lx! = Lx+ Jg [L-1(Ll!+ZLz)+(Lu!-WLw)] Suppose generators 1 and 2 belong to utility A, generators
H = -Hf+Hh+Hg-Jg[L-1Z-U-1W]Jg 13 and 23 belong to utility B, while generators 22 and 27
Ll! = LZE-µE-ΔwΔu belong to utility C. Here each utility has only its own
Hf , Hh and Hg are Hessian matrixes of f(x), yTh(x) and cT resources information and it all has the right to access the
g(x) respectively. information base to obtain the biddings of all the demands
here c = z + w. for optimized its own generations. In this system topology
The following are the main steps for interior point OPF the possible congested branches are line 10-17 and line 22-
based solution 24 with transfer limits of 28 MW and 16MW, respectively.
Step 1) Initialization: give initial values, and formulate Where 32.47 MW and 17.05 MW respectively are
admittance matrix Y = G + jB. overflowing. Using this proposed method the Table 2 shows
Step 2) Formulate Jacobian and Hessian matrix of objective the test results in centralized OPF for two cases: without
function, equality and inequality constraints: Jf, Jh, line constraint and with line constraints.
Jg, Hf, Hh and Hg. For optimizing individually its own generators to meet
Step 3) Formulate linear system (9)-(11), using predictor- the demand biddings, the suggested decentralized interior
corrector strategy to slove this linear system. If convergence point optimization approach is applied and the results
criteria are satisfied, then stop; otherwise, go to step 2. obtained are shown in Table 2 and 3 for two cases. These
two cases are without imposition of line limit and
III. TEST RESULT imposition of line limits. It is observed that the proposed
method succeed in managing the congestion as the line flow
The algorithm proposed is implemented in MATLAB on after application of proposed method produced line flow of
Intel (R), Core 2 Duo, and 2.66 GHz processor. Its 7.8230 MW and 12.034 MW in congested lines 10-17 and
performance is studied on modified IEEE-30 Bus system in 22-24 respectively in centralized optimal power flow.
4

The results of these cases were obtained by centralized [17] is respect of generator outputs and costs for both the
formulation and solution by proposed interior point method cases of constrained and unconstrained solutions. In [17],
in order to authenticate the results of decentralized for the line limit satisfaction the total demand on system has
approach. Therefore the results of centralized approach for been reduced, rather than in proposed method the total
above mentioned two cases have also been obtained and demand has been unchanged. Only for the congestion
illustrated in Table 4. It is observed from Table 3 and Table management the generator rescheduling has been adopted
4 that the sum of demand and generation are same in case of by OPF solution. The losses term is also not included with
centralized and decentralized approaches. However the the results of [17]. Then if the cultured load and the losses
generator outputs are different in two approaches. In are incorporate with these results, then the cost of
proposed method losses are also incorporated with generation has been more in compared with the proposed
generators outputs. In decentralized approach the first method. These results are depicted in Table 5. It can be
generator of utility A is considered as a slack bus. The observed from this table that the cost obtained by
losses due to other utility have been incorporated with this centralized and decentralized approaches are very close.
first generator. The additional cost of loss is incorporated However these costs obtained by proposed method are
with this generator. This is shown in table 5. Further, the lower than that reported in [17] using both the approaches
results obtained by proposed method for two cases have and cases.
been compared with those obtained by method reported in

Table: 2. Results of Decentralized approach without line constrains

GENERATIONS (MW)
Ref [17] Using DCLF Proposed Using ACLF
Utility Generator PG PG losses PL
1 60.3187 75.0455
A 2 83.2210 80.0000 7.2230
13 8.2553 12.5000
B 23 8.2553 12.5000 1.9300
22 35.3021 15.0000
C 27 0.0000 0.0000 1.6000
195.0455 10.7530
Total generation 195.3524 205.7985
DEMANDS (MW)
Ref [17] Using DCLF Proposed Using ACLF
Load A B C A B C
3 17.3529 1.6055 3.4955 18.4307 4.2500 3.0305
4 10.7929 0.9831 2.0965 11.2584 2.7500 2.0183
7 21.9920 2.4459 5.2431 10.2584 1.7500 1.0183
12 10.9881 1.2318 2.6209 12.8230 2.1875 1.2729
15 29.4926 2.4454 5.2431 29.3960 8.1250 6.2957
17 16.9916 2.4462 5.2431 23.1460 1.8750 0.0458
24 7.9500 1.6748 3.4958 13.7640 0.0000 0.0000
26 10.9881 1.2317 2.6210 12.8230 2.1875 1.2729
27 16.9916 2.4462 5.2431 23.1460 1.8750 0.0458
143.5398 16.5106 35.3021 155.0455 25.0000 15.0002
total 195.3525 195.0457
5

Table: 3. Results of Decentralized approach with line constrains

GENERATIONS (MW)
Ref [17] Using DCLF Proposed Using ACLF
Utility Generator PG PG losses PL
1 57.2912 45.5321
A 2 79.7305 69.5002 5.1700
13 7.3135 12.0000
B 23 12.0889 12.0000 1.9500
22 32.1607 22.6353
C 27 2.7192 33.9878 2.3400
195.6554 9.4600
Total generation 191.3040 205.1154
DEMANDS (MW)
Ref [17] Using DCLF** Proposed Using ACLF
Load A B C A B C
3 18.4845 2.1200 3.8201 13.9507 4.2500 7.7217
4 11.4826 1.2758 2.2901 8.5705 2.7500 4.8330
7 23.9327 3.1773 5.7271 7.5705 1.7500 3.8330
12 10.9338 1.6365 2.8566 9.4631 2.1875 4.7913
15 28.2749 3.1753 5.7272 22.6761 8.1250 13.3326
17 18.7965 3.1794 5.7272 16.4261 1.8750 7.0825
24 5.4212 0.0958 0.1471 9.2841 0.0000 3.0551
26 7.2382 1.5602 2.8574 9.4631 2.1875 4.7913
27 12.4573 3.1821 5.7272 16.4262 2.1875 7.0826
137.0217 19.4024 34.8800 113.8304 25.3130 56.5231
Total 191.3041 195.6660

Table: 4. Results Centralized approach

GENERATIONS (MW)
Generation without line constraints Generation with line constraints
Ref [17] Proposed Ref [17]** Proposed
Generator PG (DCLF) PG (ACLF) PG (DCLF) PG (ACLF)
1 60.3201 66.5455 57.2872 87.1750
2 83.2254 77.8500 79.7324 26.6553
13 8.2530 12.5000 7.3184 15.0000
23 8.2530 12.5000 12.0863 12.6430
22 35.3019 25.6500 32.1598 41.8801
27 0.0000 0.0000 2.7202 12.0000
Loss -- 9.6700 -- 9.8941
Total 195.3534 204.7155 191.3043 205.2475
DEMANDS (MW)
Demands without line constraints Demands with line constraints
Load Ref [17] Proposed Ref [17] ** Proposed
3 22.4562 25.7112 24.4241 25.9224
4 13.8726 16.0267 15.0499 16.1535
7 29.6812 13.0267 32.8394 13.1535
12 14.8402 16.2834 15.4268 16.4419
15 37.1812 43.8167 37.1765 44.1337
17 24.6811 25.0668 27.7035 25.3836
24 13.1196 13.7640 5.6626 12.3392
26 14.8402 16.2834 11.6562 16.4419
27 24.6811 25.0668 21.3652 25.6963
Total 195.3534 195.0457 191.3042 195.666
6

Table: 5. Comparison of generation cost ($/MW) of two cases

Centralized OPF
Without constraints With constraints
Ref [17] Proposed Ref [17] Proposed
567.2568 566.0951 546.5729 572.8055
Decentralized OPF
Without constraints With constraints
Utilities Ref [17] Proposed Ref [17] Proposed
A 401.1183 451.2022 377.6660 314.3436
B 52.9393 82.8125 63.1979 79.2000
C 113.1920 29.0625 105.7042 174.3698
Total 567.2496 563.0772 546.5681 567.9134
** Load has been cultured in ref. [17] for the line limit satisfaction but in proposed method load is not cultured. The total
system demand is fixed for both cases.

IV. CONCLUSIONS network of computers”, IEEE Trans. Power Syst., vol.


20.no. 3, pp. 25-33. Aug. 2005.
This paper presents a multi-area OPF algorithm which [8] A.J. Conejo, J.A. Aguado, “Multi area coordinated
makes possible the independent dispatch of each area while decentralized DC optimal power flow”, IEEE Trans.
achieving a multi-area optimum. The algorithm is based on on Power Systems, vol. 13, No.4,Nov. 1998,pp. 1272-
the Interior point based optimization procedure. This 1278.
algorithm is of particular interest in a multi-utility setting [9] S. Hao, “Decentralized approach to inter congestion
where coordination achieves substantial saving but where management”, IEEE Trans. Power Syst., vol. 20. no.2,
dispatching independence has to be maintained and its own pp. 675-683. May. 2005.
profits independently, with the help of system information [10] S. Hao and D. Shirmohammadi, “Congestion
announced by the operator. The results of IEEE 30-bus management with Ex Ante Pricing for decentralized
system have demonstrated the effectiveness of the suggested electricity markets”, IEEE Trans. Power Syst., vol. 17.
method. Therefore, the new method is physically more no.4, pp.1030-1036, Nov. 2002.
transparent and attractive than the conventional OPF [11] Kai Liu, Yixin Ni, F.F. Wu and T.S. Bi “Decentralized
approach the comparison with an existing method reveals congestion management for mulilatral transactions
that the proposed technique produced higher benefit. based on optimal recourse allocation ”, IEEE Trans.
Power Syst., vol. 22. no.4, pp. 1835-1842. Nov. 2007.
V. REFRENCES [12] F.Nishimura, Richard D. Tabors, M.D. llic,
J.R.Lacalle-Melero, “Benefit optimization of
[1] Yong-Hua Song and Xi- Fan Wang, “Operation of centralized and decentralized power systems in a multi
market-oriented power system”, Springer-Verlag area environment”, IEEE Trans. Power Syst., vol. 8.
London limited, 2003. no. 3, pp. 1180-1186. Aug. 1993.
[2] X. Wang, and Y.H. Song, “Advanced Real-Time [13] P.R. Gribik, G. A. Angelidis, and R. R. Kovacs,
Congestion Management through both Pool Balancing “Transmission access and pricing”, IEEE Trans. Power
market and Bilateral market”, IEEE Power Syst., vol. 14. no. 3, pp. 865-876. Aug. 1999.
Engineering Review, Feb 2000, 20(2), pp.47-49. [14] P. N. Biskas and A.G. Bakirtzis, “Decentralized
[3] Richard D. Christil, “Transmission management in congestion management of interconnected power
deregulated environment”, pro. IEEE, Feb 2000, 88(2), systems”, Proc. Inst. Elect. Eng. Gen., Trans., Distrib.,
pp.170-195. vol. 149, no. 4, pp. 432-438, Jul. 2002.
[4] H. Singh, S. Hao, and A. papalexopoulos, [15] X. Wang, Y. H. Song, and Q. Lu., “Lagrangian
“Transmission congestion management in competitive decomposition approach to active power congestion
Electricity markets”, IEEE trans. on Power System, vol management across interconnected regions”, Proc.
13, No. 2 May 1998, pp, 672-679. Inst. Elect. Eng. Gen., Trans., Distrib., vol. 148, no. 5,
[5] Sudipta Dutta and S. P. Singh, “Optimal rescheduling pp. 497-503, Sep. 2001.
of generators for congestion management based on [16] A. Huang, Sung-Kwan J., K. Bin Song, Jin- Ho Kim
particle swarm optimization” , IEEE trans. on Power and K. Lee, “Asynchronous decentralized method for
System, vol 23, No. 4 Nov. 2008, pp, 1560-1569. interconnected electricity markets”, Electric power
[6] Bakirtizs A.G., Biskas P.N., “A decentralized solution system research, vol.30, pp.283-290, 2008.
to the DC-OPF of interconnected power systems”, [17] P. Wei,Y.Ni and F.F. Wu, “Decentralized approach for
IEEE Trans. Power Syst., vol. 18. no. 3, pp. 1007-13. congestion management and congestion price
Aug. 2003. discovering”, IEE Proc. Gener. Transm. Distrib., vol.
[7] Bakirtizs A.G., Biskas P.N., “A decentralized 149, no.6,Nov.2002.
implementation of DC optimal power flow on a
7

[18] Hamdy A. Taha, “Operation Research: An


Introduction”, Pearson Education (Singapore) Inc.,
2002.
[19] N. Karmarkar, “A new polynomial-time algorithm for
linear programming”, Combinatorica,vol. 4, no. 4, pp.
373-395, 1984.

You might also like