Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Point-of-connection (POC) scheme of transmission pricing in decentralized markets charges the partici-
Received 21 November 2006 pants a single rate per MW depending on their point-of-connection. Use of grossly aggregated postage
Received in revised form 16 September stamp rates as POC charges fails to provide appropriate price signals. The POC tariff based on distribution
2008
of network sunk costs by employing conventional tracing assures recovery of sunk costs based on extent
Accepted 18 October 2008
of use of network by participants. However, the POC tariff by this method does not accommodate eco-
nomically efficient price signals which correspond to marginal costs. On the other hand, the POC tariff,
if made proportional to marginal costs alone, fails to account for sunk costs and extent of use of network.
Keywords:
Power transmission
This paper overcomes these lacunae by combining the above stated desired objectives under the recently
Point-of-connection transmission charges proposed optimal tracing framework. Since, real power tracing problem is amenable to multiple solu-
Power flow tracing tions, it is formulated as linearly constrained optimization problem. By employing this methodology, con-
sideration of extent of network use and sunk cost recovery are guaranteed, while objective function is
designed such that the spatial pattern of price signals closely follows the pattern of scaled locational mar-
ginal prices. The methodology is tested on IEEE 30 bus system, wherein average power flow pattern is
established by running various simulation states on congested and un-congested network conditions.
Ó 2008 Elsevier Ltd. All rights reserved.
0142-0615/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2008.10.007
60 A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66
based tariff should charge the user of the system with the relevant Looking at the key features demanded by an ideal POC scheme,
marginal cost of transmission. A consequence of the adoption of a the following options of calculating the POC tariff can be tried out
non-transaction based tariff is that the geographical distance be- by combining conventional mechanisms under different
tween buyers and sellers of power does not affect transmission principles:
charges. However, as the charges differ between various points of
connection to the system, the transmission tariff may induce a gen- Use of flat rate throughout the system: This is the postage
erator to choose a certain location [14]. stamp method and has obvious disadvantages as mentioned ear-
In [15,16], ex-ante transmission pricing scheme has been devel- lier. Thus, it neither provides efficient price signals, nor it
oped based on the real power tracing. The method determines the accounts for the usage of network by various entities.
share of every network user in a particular line’s average power Combining schemes under guidelines ‘1’ and ‘2’: This option
flow based on flow tracing and the statistics of the user’s previous involves carrying out conventional proportionate tracing either
network usage. The advantage of ex-ante pricing scheme is that the on a number of power flow simulation states or on the historical
participants of the power market know the price before trading be- data. The power flow tracing enables one to find out usage of
gins. Moreover, since the calculations are based on distributing network element by each participant and its further participa-
sunk costs, the methodology can be used as a financial instrument tion in sunk cost of that element. This type of ex-ante pricing
for recovery of transmission investments. The method provides an scheme is proposed in [15,16]. Though this scheme satisfies
innovative way of determining POC tariff by assessing the network the first two principles stated above, nothing can be claimed
usage of each participant in ex-ante manner. about the principle of providing economically efficient signals.
As mentioned in [17], a transmission pricing scheme should re- So long as loss cost allocation is considered, it is shown in [24]
cover the sunk cost of the transmission system in an equitable that the spatial variation of marginal pricing and tracing based
manner, while minimizing impact on the efficiency of the short- pricing follows almost the same pattern. However, same may
term markets. Also in [18] it is stated that the transmission price not hold true for network element sunk cost allocation. It is well
system has to be defined that does not alter the market decisions, known fact that sunk cost of each network element varies on
related with the operation of the existing capacity. According to various parameters like interest on loan, return on equity,
[8], the transmission price system based on ‘sunk’ costs tends to al- debt-equity ratio, O&M charges, interest on working capital,
ter the economic dispatch. Hence, it is mentioned that the trans- depreciation, etc. Thus, spatial variation of loss allocation and
mission prices must compromise between signalling short run sunk cost allocation may not match for a realistic system. The
marginal costs and offering reasonable assurance of cost recovery results for sunk cost allocation based on conventional propor-
over the long run. In [19], it is mentioned that the spatially variate tionate tracing may not follow the pattern of efficient price
usage based access charges in two part tariff may strengthen the signals.
price signals provided by marginal costs, if the method used to cal- Scheme based on marginal prices: In this scheme also, the
culate access charges gives similar signals to those of marginal point charges are calculated using number of simulation states.
charges. It is shown that there is a high similarity between mar- However, the difference is that the optimal power flow is carried
ginal charges and loss allocation based on tracing. However, when out with generator cost minimization as an objective function.
it comes to sunk cost distribution based on tracing, its pattern may This enables calculation of nodal marginal prices. And then,
or may not match with that of loss allocation or marginal cost. This the sunk cost distribution is done in proportion to the marginal
is because, the sunk costs of transmission are of no concern while prices, which subsequently decides the point charges. However,
obtaining the optimal power flow solution. In the light of the this scheme does not accommodate the second guideline stated
above, the lacuna associated with the methods proposed in above, as ‘extent of use’ of the network by participants is not
[15,16] is that they tend to alter the price signals based on market quantified. Calculation of ‘extent of usage’ is important because
decisions. Moreover, the signals under network congestion situa- it relates the sunk costs of the transmission network to the point
tion do not get reflected in the ex-ante prices. This is because only charges. It is worthwhile to note that the network sunk costs do
consideration given to decide the point charges is the sunk cost of not figure in the calculation of nodal marginal prices.
network elements.
The above discussion provides the following essential guide- It can be seen that none of the above conventional choices can
lines for the design of the point charges: satisfy all three requirements of point charges. This paper aims at
developing a POC tariff for decentralized market participants by
1. The point charges should be decided so that they recover the accommodating all of the above mentioned guidelines. To accom-
sunk or embedded costs of network [2,4,7,17]. modate the third condition, we make use of generalized framework
2. However, these point charges should be based on ‘extent of use’ of tracing rather than conventional proportional tracing. To achieve
of the network by participants [15,16,20,21]. this, we exploit the multiplicity of solution space in real power
3. The spatial variation of point charges should be in tune with tracing. Most of the tracing algorithms and applications developed
that of economically efficient price signals, to the maximum so far, employ proportionate sharing principle [26–28] to choose an
possible extent [8,11,22]. answer from the large solution space. However, recently a novel
approach to tackle the large solution space of real power tracing
The principle of recovery of embedded costs calls for use of based on optimization technique was proposed in [29,30]. This
any of the methods under rolled-in pricing paradigm. The extent framework enables us to freeze to a tracing solution with an objec-
of usage can be quantified by means of power flow tracing. To in- tive function that depicts the requirements. A couple of applica-
duce efficient use of transmission grid and the generation re- tions of the proposed approach are presented in [29,31], while
sources by providing efficient price signals, the spot price theory performance of the proposed approach in the presence of circular
was developed in [23]. The objectives achieved by means of effi- flows is presented in [32]. In this paper, we define an objective
cient pricing are enlisted in [11,25]. Therefore, the third guideline function for this optimal tracing problem such that the point-of-
demands that the point charges should be in tune with the spatial connection charges calculated considering network sunk cost dis-
variation of marginal costs. The discussion on the efficiency of the tribution employing real power tracing, try to follow spatial varia-
price signals provided by marginal costs is out of scope of this tion of marginal prices, as closely as possible. Thereby, it ensures
paper. conformation to the above guidelines. It should be noted that every
A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66 61
feasible solution within the generalized framework of tracing is a the transmission usage cost at ith bus for load PLi , of the line lm, is
L;ðsÞ i;ðsÞ
valid tracing solution. The solution based on conventional propor- given by ðclm ylm PLi Þ. Thus, the total transmission system usage
tional tracing represents one of the feasible solutions among these. cost for a load i is given by
This paper is organized as follows: The paper starts with the L;ðsÞ
X i;ðsÞ L;ðsÞ
concept of locational transmission price (LTP) in Section 2. In Section
TC i ¼ P Li ylm clm ð2Þ
8lm
3, brief account of optimal tracing framework is provided. The
objective function that satisfies goals of this paper is established The locational transmission price (LTPi) for a load bus in state s is
in Section 4. This section also introduces concept of pseudo-LMP obtained by dividing the above by P Li .
which mimics the pattern of spatial variation of locational mar- X i;ðsÞ L;ðsÞ
LTPi;ðsÞ ¼ ylm clm $=MW ð3Þ
ginal prices. Section 5 presents an algorithm for calculating the
8lm
point-of-connection charges. Results on IEEE 30 bus system are
provided in Section 6. Section 7 concludes the paper. If the simulation is carried out for Nt states, then the average LTPi is
given as
r;ðsÞ
X
nL
i;ðsÞ
P lm ¼ ylm PLi The vectors x and y represent the decision variables associated with
i¼1 generation tracing and load tracing problems, respectively. The set S
62 A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66
represents the set of all possible tracing solutions and a specific set 3.2. Generic form of optimal tracing problem
of x and y vectors represents a solution to unified optimal tracing
problem. In [29,30], it is shown that set S can be characterized by A generic formulation of the optimization problem for real
a set of linear equality and inequality constraints. In fact, set S is power tracing is given as follows [30]:
both compact and convex. This leads to a linear constrained optimi-
min f ðx; yÞ ð12Þ
zation problem. The objective function models the relationship be-
tween the flow entities and associated costs. In this paper, details Ad 0 x bd
¼ ð13Þ
about the formulation of optimal tracing problem are not provided 0 Au y bu
due to space restriction. These details can be found in [29,30]. How- yik PLi xki PGk P 08k 2 f1; 2; ; nG g
ever, brief account of various constraints of optimal tracing problem
8i 2 f1; 2; ; nL g ð14Þ
is given next.
½0; 0 0T 6 ½x 6 ½1; 1 1T ð15Þ
3.1. Optimal tracing problem constraint modeling ½y P ½0; 0 0 T
ð16Þ
For the power flow tracing problem, two versions are devel- The constraints [Ad][x] = [bd] represent three classes of equality con-
oped: Generation tracing and load tracing. For each version, equal- straints for generation tracing formulation, discussed earlier. Simi-
ity constraints are grouped into following categories: larly, the constraints [Au][y] = [bu] represent corresponding
constraints for load tracing formulation. Constraints (14) model
1. Flow specification constraints: These constraints are developed non-negativity of loss characteristics. Inequality constraints (15),
for series branches i.e., lines and transformers. For generation (16) model the limits on x and y variables. This generalized tracing
tracing problem, these constraints mean nothing but expressing framework is referred to as optimal tracing problem throughout this
xklm as component of total injection of generator Gk on line lm. paper. The objective function can take various forms depending on
This set of constraints is given as follows: the application and the fairness requirements. In [29], the objective
function is proposed such that the sum of overall deviations of per
X
nG unit transmission prices of all loads is minimized, leading to a least
Plm ¼ xklm :PGk 8 set of lines ð8Þ absolute value (LAV) problem, which further can be converted into
k¼1 Linear programming (LP) problem.
The above optimal tracing formulation provides a generic
Similar constraints for load tracing problem are developed as framework to solve the real power tracing problem. It provides a
follows: modular plug and play kind of tool box which can be employed to
X
nL solve various tracing related problems with appropriate objective
Plm ¼ yilm :PLi 8 set of lines ð9Þ function for the optimization problem, that depicts fairness. Next
i¼1 section discusses the objective function that satisfies the goals of
this paper.
2. Source and sink specification constraints: These constraints
pertain to shunts, e.g., generators and loads. For generation
tracing problem, these constraints express contribution of gen- 4. Objective function
erators in loads. This set of constraints can be represented as
follows: Let LMPi,(s) denote the short run marginal cost or the locational
marginal price of node i, in simulation state s. As mentioned in Sec-
X
nG
tion 2, the objective function of Eq. (12) should be such that the
P Li ¼ xki PGk for i ¼ 1; . . . ; nL ð10Þ spatial variation pattern of LTP i;ðsÞ obtained by tracing results
o
k¼1
should match as closely as possible, with that of LMPi,(s).
Similar constraints for load tracing problem are developed as Let total transmission cost of the network to be paid by loads be
follows: TCL. This is known a priori and is given as
X
X
nL TC L ¼ bTC lm ð17Þ
P Gk ¼ yik PLi for k ¼ 1; . . . ; nG ð11Þ 8lm
i¼1
Let, pLMPi,(s) represent the pseudo-LMP of load bus i in state s, such
3. Conservation of commodity flow constraints: The conserva-
that,
tion of commodity flow enforces that in a multi-commodity
network with zero storage capacity of nodes, the conservation X
nL
of flow constraint also holds for all commodity flow pLMPi;ðsÞ PLi ¼ TC L ð18Þ
components. i¼1
obtained should match with pLMPi(s) for all load buses. Hence, the 4. Calculate pLMPi,(s) using Eqs. (19) and (20).
objective function of Eq. (12) can be written as follows: 5. Solve optimal tracing problem presented in Section 3 with an
objective function of Eq. (21). Also, solve tracing problem by
X
nL
min jpLMPi;ðsÞ LTPi;ðsÞ
o j ð21Þ proportional sharing principle.
i¼1 6. Calculate LTPiðsÞo and LTP iðsÞ
p for all i.
7. Store LTP o and LTPi;ðsÞ
i;ðsÞ
. iter iter + 1.
where, LTP i;ðsÞ
o is given as: o
l m
TC lm / jLMP LMP jP lm ð24Þ
where, Plm represents the real power flow (MW) for standard data
$/MW
3
and LMPl, LMPm represent nodal locational marginal prices across
the line. Algorithm of Section 5.1 represents a generic step-by-step
methodology to calculate the point-of-connection charges. By using 2
the same algorithm, both LTP io;c and LTPio;uc can be calculated.
1. Initialize iter 0.
2. Create a state (s) by randomly sampling a load condition. 0
3. Carry out OPF for this state (s). The result of OPF will give gen- 5 10 15 20 25 30
Bus No.
eration dispatches, power flows and locational marginal prices
(LMPi,(s)) for all i. Fig. 1. The POC charges with b = 1; a1 = 0.2; a2 = 0.8.
64 A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66
Table 1
i Mean and standard deviation of dio and dip for all cases.
5
LTPo,uc
i
pLMPuc b a1 a2 dio dip
i
LTPp,uc l r l r
4 1 0 1 36.8072 41.5132 44.3792 49.7151
1 0 23.6121 22.7633 32.9127 29.0573
0.2 0.8 34.1681 37.76322 42.0859 45.58354
0 1 44.8831 42.8281 55.1711 51.9840
$/MW
3
0.75 1 0 34.5716 32.3932 49.6391 40.2358
0.2 0.8 42.8208 40.74112 54.0647 41.5872
0 1 45.0080 42.8383 55.0156 51.9693
2 0.5 1 0 33.5358 33.5496 50.0550 45.3009
0.2 0.8 42.7135 40.9805 54.0234 50.6356
jLTP ip pLMPi j
0
5 10 15 20 25 30 dip ¼ 100% ð27Þ
Bus No.
pLMPi
Fig. 2. Plots of LTPio;uc , pLMPiuc and LTPip;uc with b = 1, a1 = 0, a2 = 1. Table 1 shows the mean and standard deviations of dio and dip for
various values of b, a1 and a2.
It can be seen from Fig. 2 that under un-congested case, the
pseudo-LMPs for all buses are restricted within a tight band. How-
ever, sufficient distortion can be witnessed in case of congested
Fig. 2 shows the plots of LTP io;uc , pLMPiuc and LTP ip;uc . This is sim- network situation (Fig. 3). It is important to note that whatever
ilar to the case when a1 = 0 and a2 = 1. Similarly, Fig. 3 shows the may be the network operating state and whatever may be the
plots of LTP io;c , pLMP ic and LTP ip;c . This is similar to the case when set of sunk costs of transmission network, the mean associated
a1 = 1 and a2 = 0. with dio will always be lesser than the mean associated with dip .
It should be noted that the following will hold true always: This is apparent from Table 1. This means that the spatial pattern
X
nL X
nL of point charges obtained by optimal tracing will be more close to
LTPio PLi ¼ LTPip PLi ð25Þ that of marginal prices, than that obtained by proportional
i¼1 i¼1 tracing.
In other words, increase or decrease in LTP io of a particular bus i will Another important point to be noted is that the tracing re-
be at the cost of corresponding decrease or increase in LTP io of other sults are topology dependent even in the case of optimal tracing
buses. framework. However, they are set within the boundaries of con-
To find out the nearness of the LTPio with the pLMPi, the cluster strained region so as to meet the objective. For example, load on
analysis for the deviation of LTPio and LTPip from that of pseudo-LMP bus 2 makes very less use of the network (line 1–2) and thus
is done. Let dio denote the deviation of LTP io from that of pseudo- has least POC charge obtained by both, optimal and proportional
LMP. Then, tracing. On the other hand, load on bus 30 makes maximum use
of network and has highest POC charges obtained by both the
jLTPio pLMPi j methods. However, LTPio will be adjusted (within tracing con-
dio ¼ 100% ð26Þ straints) such that overall deviation from pseudo-LMP will be
pLMPi
minimum.
i 4
LTPo,c LTPo
i
5 i i
pLMPc 3 pLMP
i i
LTPp,c LTPp
$/MW
2
4
1
$/MW
0
3 5 10 15 20 25 30
Bus No.
1.2
i
2 LTPo
1 i
pLMP
0.8 i
$/MW
LTPp
0.6
1
0.4
0.2
0 0
5 10 15 20 25 30 5 10 15 20 25 30
Bus No. Bus No.
Fig. 3. Plots of LTP io;c , pLMPic and LTP ip;c with b = 1, a1 = 1, a2 = 0. Fig. 4. The POC charges with b = 0.75; a1 = 0.2; a2 = 0.8.
A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66 65
1.5 LTP
these studies. The POC tariff thus obtained can be employed to
1
charge the transaction based and non-transaction based trades.
0.5
0
References
5 10 15 20 25 30
Bus No.
[1] Hunt S, Shuttleworth G. Making competition work in the electricity
market. Academic Press; 1981.
Fig. 5. The POC charges with b = 0.5; a1 = 0.2; a2 = 0.8. [2] Green R. Electricity transmission pricing: An international comparison.
Utilities Policy 1997;6:177–84.
[3] Szkuta BR, Sanabria LA, Dillon TS. Electricity price short-term forecasting using
6.2. Results with b = 0.75 artificial neural networks. IEEE Trans Power Syst 1999;14(3):851–7.
[4] Arriaga IJP, Rubio FJ, Puerta JF, Arceluz J, Marin J. Marginal pricing of
transmission services: An analysis of cost recovery. IEEE Trans Power Syst
In this case, the loads are charged 75% of the network costs 1995;10(1):546–53.
while the generators are charged 25%. The results with a1 = 0.2 [5] Rubio-Oderiz FJ, Arriaga IJP. Marginal pricing of transmission services: A
comparative analysis of network cost allocation methods. IEEE Trans Power
and a2 = 0.8 are shown in Fig. 4.
Syst 2000;15(1):448–54.
Bus numbers 2,5 and 8 have both generators and loads on it. [6] Rudnick H, Palma R, Fernandez JE. Marginal pricing and supplement cost
Table 2 enlists the POC charges for both generators and loads on allocation in transmission open access. IEEE Trans Power Syst 1995;10:
these buses, for b = 0.75 and b = 0.5. 1125–42.
[7] Lima JWM. Allocation of transmission fixed charges: An overview. IEEE Trans
Power Syst 1996;11:1409–18.
6.3. Results with b = 0.5 [8] Shuttleworth G. Electricity transmission pricing: The European perspective,
Technical Report. National Economic Research Associates; 1999.
[9] Oloomi-Buygi M, Salehizadeh MR. Considering system non-linearity in
The POC charge allocation when both generators and loads transmission pricing. Int J Electr Pow and Energ Syst 2008;30:455–61.
share half of the cost is shown in Fig. 5. [10] Shirmohammadi D, Gribik PR, Law ETK, Malinowski JH, O’Donnell RE.
From Table 2, it may be noted that the POC charges are in tune Evaluation of transmission network capacity use for wheeling transactions.
IEEE Trans Power Syst 1989;4:1405–13.
with the degree of deficiency or adequacy of power on that bus. For [11] Tabors RD. Transmission system management and pricing: New paradigms
example, on bus 2, with b = 0.75, average generation is 97 MW, and international comparisons. IEEE Trans Power Syst 1994;9:206–15.
while average load is 21.7 MW. Hence, the loads are provided [12] Website: <www.nordpool.com>.
[13] Kristiansen T, Wangensteen I. Effect of losses on area prices in the Norwegian
incentive by charging them lesser than the generators. Exactly
electricity market. Int J Electr Pow and Energ Syst 2006;28:43–7.
opposite phenomenon can be observed on bus 8. For the same case, [14] Bergman L. The Nordic electricity market: Continued success or emerging
the average generation on this bus is 5 MW, while average load problems? Swedish Econ Policy Rev 2002;9:51–88. Available from: <http://
www.ekradet.konj.se/sepr/LBergman_9_2.pdf>.
amounts to 30 MW. Hence, in this case, generators are given incen-
[15] Pantos M, Gubina F. Ex-ante transmission service pricing based on load flow
tive by charging them less as compared to the loads. Bus 5 repre- patterns. IEEE Trans Power Syst 2004;19:796–801.
sents a case in between these two extreme cases. [16] Pantos M, Gubina F. Ex-ante transmission service pricing via power-flow
tracing. Int J Electr Pow and Energ Syst 2004;26:509–18.
[17] Concept paper on open access in inter-state transmission, Technical Report.
7. Conclusion Central Electricity Regulatory Commission, India; 2003. <http://
www.cercind.org/open_access_full.pdf>.
[18] Danitz F, Rudnick H, Zolezzi J, Watts D. Use based allocation methods for
This paper provides a real power tracing based methodology to
payment of electricity transmission systems. In: Proceeding of the Power
determine the ex-ante point-of-connection (POC) rates for the par- System Technology PowerCon; 2002. p. 907–11.
ticipants of decentralized market. The desired qualities of POC [19] Al-rajhi AN, Bialek JW. Marginal and tracing pricing of transmission: An
empirical comparison. In: Proceeding of the 14th PSCC, Sevilla; 2002. p. 1/27–
charges include consideration of network usage, recovery of sunk
7/27.
costs and provision of price signals, all of which are hard to achieve [20] Bialek J. Allocation of transmission supplementary charge to real and reactive
in a single scheme, if the point charges are devised by conventional loads. IEEE Trans Power Syst 1998;13:749–54.
66 A.R. Abhyankar, S.A. Khaparde / Electrical Power and Energy Systems 31 (2009) 59–66
[21] Strategies in tariff setting for transmission, Technical Report. Central [30] Abhyankar AR, Soman SA, Khaparde SA. Real power tracing: An optimization
Electricity Regulatory Commission, India; 2003. <http://www.cercind.org/ approach. Int J Emerg Electr Power Syst 2005;2(1088):1–31. Available from:
chapter6_2.htm>. <http://www.bepress.com/ijeeps/vol3/iss2/art1088>.
[22] Kinnunen K. Network pricing in nordic countries – An empirical analysis [31] Abhyankar AR, Soman SA, Khaparde SA. New paradigm of tracing algorithms:
of the local electricity distribution utilities’ efficiency and pricing, Ph.D. thesis. Application to fair loss allocation in Indian system. In: Proceeding of the
Carl Von Ossietzky University: Finland; 2003. <http://docserver.bis.uni- International Conference on Future Power Systems. Amsterdam; 2005. p. O11–
oldenburg.de/publikationen/dissertation/2003/kinnet03/pdf/kinnet03.pdf>. 10/1–O11–10/6.
[23] Schweppe FC, Caramanis MC, Tabors RD, Bohn RE. Spot pricing of [32] Abhyankar AR, Khaparde SA, Soman SA. Multiple solutions approach to tackle
electricity. Kluwer Academic Publisher; 1988. circular flows in real power tracing. In: Proceeding of the IEEE PES General
[24] Bialek JW, Kattuman P. Real and reactive power tracking: Proof of concept and Meeting, Montreal; 2006. p. 1–8.
feasibility study, Technical Report. TR112416, Electric Power Research [33] Bialek JW. Elimination of merchandising surplus due to spot pricing
Institute; 1999. of electricity. IEE Proc -Gener Transm Distrib 1997;144(5):399–405.
[25] Chen L, Suzuki H, Wachi T, Shimura Y. Components of nodal prices for electric [34] Strbac G, Kirschen D, Ahmed S. Allocating transmission system usage on the
power systems. IEEE Trans Power Syst 2002;17:41–9. basis of traceable contributions of generators and load flows. IEEE Trans Power
[26] Bialek JW. Tracing the flow of electricity. IEE Proc -Gener Transm Distrib Syst 1998;13:527–34.
1996;143(4):313–20. [35] Golub GH, Loan CFV. Matrix computations. The Johns Hopkins University
[27] Kirschen D, Allan R, Strbac G. Contribution of individual generators to loads Press; 1996.
and flows. IEEE Trans Power Syst 1997;12:52–60. [36] Bertsekas DP. Network optimization: Continuous and discrete models. Athena
[28] Wu FF, Ni Y, Wei P. Power transfer allocation for open access using graph Scientific; 1998.
theory-fundamentals and applications in systems without loopflow. IEEE [37] Comparison of transmission pricing in Europe: Synthesis 2004, Technical
Trans Power Syst 2000;15:923–9. Report. European transmission system operators tariff task force; 2005.
[29] Abhyankar AR, Soman SA, Khaparde SA. Optimization approach to real power <http://www.etso-net.org/upload/documents/080405%20Synthesis%202004%-
tracing: An application to transmission fixed cost allocation. IEEE Trans Power 20FINAL%20%20.pdf>.
Syst 2006;21(3):1350–61.