Вы находитесь на странице: 1из 10

IEEE Transactions on Power Systems, Vol. 12, No.

2, May 1997

539

Unbundling of Transmission and Ancillary Services


Part I: Technical Issues

Assef Zobian
Marija D. IliC
Laboratory for Electromagnetic and Electronic Systems
Massachusetts Institute of Technology
Cambridge, MA 02139, USA

Abstract
This paper presents derivations of basic formulae for computing the contributions of each economic transaction to the
network power flows throughout the system in steady-state operation of interconnected electric power systems. It is shown
that the net system power imbalance caused by each transaction can be obtained as a function of all transactions present
on the system. In addition, formulae are proposed for calculating the contributions of every ancillary generation unit t o each
transaction. This generation is needed to balance the system
in response to economic transactions. Formulae supporting
this are based on reformulating the load flow problem in terms
of distributed slack bus.

I.

Introduction

This paper is concerned with the basic allocation of transmission system use and some of the eneration-based anciliary services required in support o competitive electric
power markets. Accurate decomposition of power flows
into flows contributed by individual transactions on the
power system network is essential for cost-based pricing
of transmission and ancillary services. However, this appears to be complex mainly because of the interaction of
flows caused by various transactions. This issue is usually
referred to as the loop flow or parallel flow problem.
An economic transaction between an arbitrary number of buses is specified in terms of net power injection Si into these buses. The most common scenario of
such transaction is an economic agreement to sell a given
amount of power from bus i to bus j over a fixed contractual period and at a fixed price for ancillary and transmission services.
The underlying technical contribution of this paper
is two-fold. The first contribution is defining the impact
of simultaneous economic transactions on the real and reactive power flows on transmission lines and other transmission system equipment. Formulae are derived for computing the contributions of each economic transaction to
the network power flows throughout the system.
The second contribution is a method and formulae
to calculate the contribution of each generating unit participating in balancing the system to a given transaction.
This generation is needed t o balance the system under
known time-varying transactions and unknown transmission losses. Formulae to calculate the generation contributions are derived by reformulating the load flow problem in
terms of distributed slack bus. The slack bus-based power
contribution to balance the system is distributed into a set
of generatin units, referred to here as economic dispatch
IED) units
The generation scheduling to compensate
or the power imbalance can be done in variety of ways.
We suggest in this paper t o use participation factors-based
economic dispatch.

E].

It is shown that the net system power imbalance resulting from each transaction is a function of all transactions present on the system. The power imbalance caused
by a particular transaction can be compensated for by
allocating the net power imbalance among different ED
generating units according to the participation factors.
An approach is taken to decompose the flows into
components associated with each particular transaction,
and an interaction component, based on the superposition of all transactions on the network. The interaction component accounts for the non-linearity of the
power flow equations. The interaction component of all
transactions on the system cannot be associated with a
single transaction. Moreover, only a small percentage of
a given transaction contributes to this interaction component. This decomposition forms an essential part of
our framework for recovery of fixed cost based on actual
network power flows.
Part I of this paper is organized as follows: First,
mathematical formulae for decomposition of power flows
are derived assuming a single slack bus. Next, the concept
of distributed slack bus [l]is introduced to account for the
fact that many generators are participating in ED generation when balancing the power on the system. The derived
decomposition formulae are generalized to account for the
ramification of distributed slack bus on our proposed approach. Using the derived decomposition formulae, the
power imbalance at the ED generating units is calculated.
Part I1 of this paper introduces a systematic framework for cost-based pricing of transmission service based
on the proposed power flow decomposition approach and
presents an accurate scheme for pricing of generationrelated ancillary services based on the proposed allocation
of power imbalances approach.

11.
0

Notation

Bus enumeration: i = 1 , 2 , ...,n nodes, node # n is


the slack bus.

E,: Complex valued voltage at node i, and E =


[ E l ,E2, E3, ...,&IT
is a vector of all node voltages in

the system. Note that boldface font is used for vector


symbols. Ei = V,ej61, V,, Si ; i = 1 , 2 , . . . , ( I g ] :
Load and generator voltages and angles respective y.

I,: Complex valued current at node i, and I =


[I~,12,13,
...,I,]* is a vector of all node currents in
the system.

Si = Pi j Q i : Net complex valued power generated


at node i, P; is the real power, Qi is the reactive
power part, and S = [SI,
S2,S3, ...,S,IT is a vector
of all nodes complex valued power in the system.

96 WM 319-4 PWRS A paper recommended and approved by the IEEE


Power System Engineering Committee of the IEEE Power Engineering
Society for presentation at the 1996 IEEE/PES Winter Meeting, January 2125, 1996, Baltimore, MD. Manuscript submitted August 1, 1995; made
available for printing January 15, 1996.
0885-8950/97/$10.00 0 1996 IEEE

540
0

yz3 = Gz,j- 3 Bz,3: Admittance of a branch connecting


nodes z and 1 . For a typical branch, G,,j,B,,j > 0.
Y is the Y-bus admittance matrix.

After eliminating the last equation which corresponds


to the slack bus, Equation ( 3 ) can be re-written as:

Shunt admittance at node i with b, > 0 for capacitive shunts.

e jb,:

Economic transaction, number k , between any


number of buses.

e Tk:

I M : Complex valued current flow matrix, whose elements are the complex valued current flows on the
transmission lines (symmetric matrix).

SM : Complex valued power flow matrix, whose elements are the complex valued power flows on the
transmission lines in both directions (not symmetric).

111. Contribution of individual


transactions t o the network flows
The derivations are presented in the context of a general power system with the notation introduced above.
To decompose flows on the network into components contributed by specific transactions, consider the basic equations governing the flow on the network. These equations
are simply the load flow equations, which are written in a
matrix form:

I = diag(E*)-'S = Y * E

(1)

where E* represents the complex conjugate of E, and


diag(E*) is a matrix whose diagonal elements are those
of the vector E*'.
The elements of the bus admittance matrix, Y, are

YZJ

Y,E, = Y,, * E, I,
(5)
The Y, matrix is now an invertible matrix, whose inverse is the impedance matrix Z used for load flow studies,
Le., Y;' = Z .
Complex valued voltage E, is computed from equation ( 5 ) as:

E, = Z * Y,,

j=1

For simplicity of analysis we only consider the case


of bi = 0 for all i '. Recall that the admittance matrix
plays a fundamental role in computing the nodal voltages
Ei for specified power inputs Si. To start with, it is not
possible to specify power inputs into all buses because,
mathematically, the bus admittance matrix is a singular
matrix and can not be inverted '. This requires the concept of a slack bus, a bus with generation output which
can only be specified in response to the economic transactions on the system. Its role is to balance the losses in
the system, which can not be specified a priori.
Equation (1)written explicitly in terms of individual
nodal voltages and currents takes on the form :

Generalization of these results is straightforward with this assumption relaxed.


31t is relevant t o keep this fact in mind whenever making the
case about the qualitatively different role of ancillary generation in
supplying competitive generation/demand market, i.e., its reactive
rather than competitive role.

* E, + Z * I,

(6)

Claim 1: Z * Y e n= 1, where 1 i s a vector of all ones.


Proof:

Y,, is a column vector which can be expressed as a linear


combination of vectors of Y,. Actually, it is an exact
summation of the vectors of Y,; Z * Y, = I , identity
matrix, so Z * Y , , should give a vector of all ones4.0
Equation (6) becomes:

E,

-Yq

'Here the convention used for t h e complex power is


diag(E*)I

or

E, = Z *I,

(7)

Let V = E, - E:, and E l = 1 * E,; a (n-1) vector


whose elements are all equal to E,.
Without loss of generality, consider only two simultaneous transactions on the system, T k and Tm. The nodal
currents corresponding to all individual transactions can
be specified for any two arbitrary nodes, except at the
slack bus.
For example, the injected current into the transmission system caused by the first transaction is specified by
the following vector:
ITk =

[ X , 0 , 0 ,..., -Y,0, 0 ] T ,

(8)
Note that this vector has (n-1) components only. Assume
the injected currents caused by the second transaction are
given by:
I T m = I, - I T k
(9)
The complex valued current injections are defined to
correspond to the power injections computed using the
complex voltage prevailing at the actual operating point.
The operating point is determined using all scheduled
transactions on the system, i.e., the current injections
should be computed at the load flow solution for all scheduled transactions. Thus computing major power flow components caused by individual transactions requires the
knowledge of all transactions on the system. The prior
knowledge of all transactions on the system is essential
for other purposes like stability and security studies.

4This result does not hold when b, # 0, but still this vector can
be computed as a function of admittance matrix data.

541

Equation (7) can be written replacing I, by the summation of currents introduced by both transactions:

V = E,

- EL = ZI, = Z ( I T ~ 1

~ (10)
)

The specified complex valued power injections according to equation (1) are:

S, = diag(E*,) * I,
In order to decompose power flows into components
contributed by individual transactions, multiply equation 10) by ITL.This leads to the decomposition given
as fol ows:

The interaction component becomes larger for large


transactions but generally remains a small percentage for
admissible operating conditions. It becomes a large percentage of the transaction for stressed systems. While
relatively small, the interaction component should not be
ignored in the design of pricing mechanisms. This is potentially an effective way of overcoming the nonlinearity
feature of the flows; i.e., its dependence on everything
else on the system, without causing harm or unfairness
to any participant. This result is a very important guide
in separating the effects of a known transaction from all
future uncertainties on the system. The above formulae
are straightforward to generalize to an arbitrary number
of transactions.

A.

Computation of the major component s

Using the above Claim 2, it is proposed here that the


major contribution of each individual transaction to the
network flows can be computed separately from its interactions with the other transactions present on the system.
The derived formulae follow directly from (12) and
are expanded here to explicitly state the contribution of
any transaction T k specified in terms of its injected curk the power line flows throughout the
rent vector 1 ~ on
entire power grid.
For the transaction TI,,the complex valued voltage
vector is computed as follows:

En,~lc
=

ZIT^ + EL

(15)

The complex valued power flow ( n x n ) matrix is


computed as:
Here ST^ and ST^ are defined as the vectors of complex valued power injected into the system by the first
and second transaction, respectively. The complex valued
vectors S,k and S k m are defined to be the interaction
components for the first and second transaction, respectively, and S k m , t o t a l is the vector of complex valued power
caused by the interaction between the two transactions.
Formulae (12) and (13) lead to our main technical
finding stated as:
Claim 2: Under normal operating conditions the interaction component S k , is a small percentage of the transaction itself, ST^.
Proof:
The major component diag(Ek) is not present in the
interaction component. This component is of the order of
one. Therefore, the S k m and S,I, components are of the
order of voltage differences between buses and the slack
bus multiplied by the values of injected currents according to scheduled transactions, or the order of V T *~1 ~
V T ~= ZITI, is of the order of 0.05, since usually voltages on power systems are allowed to vary within 5% of
nominal values, and phase angle differences are usually
small during normal operating conditions. 1~~ is of the
order of the transaction itself. The total interaction component is of the order of 5% of the transaction. In order
to formalize the proof in mathematical terms, one can use
induced matrix norms to measure the relative size of different matrices. As described above, for all transactions
during normal operating conditions, the following condition is always satisfied:
IIZITrn112

<<

IlEh

+ ZITk112

(14)

That is, the sum of the square of the terms in vector ZIT,
is much smaller then the sum of the square of the terms
in the vectors E: + ZIT^. 0

SMT~
= diag(ETk)*IMTk

(16)

where ET^ = [ E n , ~ k: E,] is the augmented voltage vector, and I M T ~is the complex valued current line flow
matrix in response to the current injection IT^ caused by
transaction, Tlc, computed as:

PMTk(i,i) = 0
IMTk(G3)

(J%k(i)

* Y ( i ! j )f o r i f(b77)

- ETIe(j))

Note that I M ~ k ( z , i=
) 0 under the assumption that
bi = 0. When this assumption is relaxed the shunt current
P M ~ k ( i , i=
) bi * E ~ k ( ican
) be used to identify the major
flows of a transaction T k on that shunt element b,. This
element can be a shunt capacitor or reactor, static VAR
compensator, on any other shunt device on the transmission system.
~ .The vector of complex valued nodal power is the summation of the line power flows:

ST^ = S M T *~ 1

(18)
Note that the last component in this vector defines the
contribution of the slack bus to this particular transaction
T k . This observation is significant since the total power
imbalance and losses on the system are reflected at the
slack bus. We return to further interpretation of this in
context of the actual implementation of the slack bus by
means of several generators participating in the economic
dispatch.
The above formulae hold for each particular transaction T k , independently of the number and location of
different transactions. This procedure is not computationally intensive. If the load flow solution is given, it involves
simple multiplication of matrices and vectors.

542

B.

Computation of the interaction component

Starting from formulae (12) and (13) and following a procedure analogous to the one used to derive the major impact of transactions on network flows, one obtains formulae for computin the minor impact of the interactions
between any num er of transactions on the matrix of network flows throughout the grid. These derivations, neglecting the slack bus elements, result in:

fl

SMkm,total

= diag(ZIm)*IMT,

+ diag(ZITm)*IMTk

(19)
Here SMkm,totalrepresents an (n - 1 x n - 1) matrix
of complex valued power line flows throughout the grid
reflecting interactions between the two injected currents
IT^ and 1 ~ ~ .
The complex valued power vector is summation of
line power flows caused by these interactions, i.e.,
Skm,total

= diag(ZITk)*ITm

Skm,total

SMkm,total

+ diag(ZITm)*ITk

*1

(20)

Note that the element in (20) which corresponds to


the slack bus is missing. Later we derive formulae to compute losses on the system for a single and distributed slack
bus.
The calculation of the component S k m , t o t a l requires
simultaneous information about all transactions on the
system, and it is hard to use in real-time for ex-ante pricing, particularly when setting fixed price contracts. It is
for this reason that Claim 2 above is so significant. The
major power flow components are defined at the scheduled
voltage operating point and are fixed, while the interaction component is defined to account for all the power flow
variations on the system.

IV.

Distributed slack bus

The theoretical analysis presented above is dependent


upon the concept of a single slack bus. Recall that the
slack bus issue arises primarily because of lack of prior
knowledge of system losses. Because of this, to maintain
the real power balance on the system one cannot specify
the real power generated at all generators. Mathematically, the need for slack bus is seen in the singularity of
the load flow real power Jacobian.
The computed losses on the transmission system are
a function of the choice of a slack bus. For example, computations using a slack bus electrically distant from the
points of a certain transaction mi ht indicate either infeasibility of the transaction or hi& losses associated with
it. In the actual operation of electric power systems there
is no single slack bus, instead there are many generators
distributed geographically throughout the system which
take on the function of a slack bus. To account for this a
distributed slack bus must be used. The pricing for losses
or imbalances on the system should correspond to actual
operation for the pricing scheme to be fair.
Many researchers recognized the inadequacy of a single slack bus and its diverse effects on the steady state calculations in power systems. However, few have attempted
to remedy this problem. For example, Zaborszky et al.
in [5] used a linear transformation to distribute the real
power imbalance reflected at the slack bus by self balancing pairs. In other words, for each pair of buses, one
can balance the net real power by increasing power a t the
generator or reducing it at the load bus. This approach
utilizes the minimum spanning tree algorithm assuming
a lossless system. Because it neglects losses, this approach

is not satisfactory for our objectives. A second approach


to solve the slack bus problem was proposed by CaloviC et
a1 in [l].A similar variation of this solution can be found
in [4k This approach is based on participation factors,
whic resembles actual operation of power systems. We
use this approach as the basis for our decomposition of
network power flows.
In the next sections we present the mathematical formulation for distributed slack bus formulations of the load
flow problem, and study the implications of distributed
slack bus on our proposed power flow decomposition approach.

A.

Load flow formulation using a distributed slack bus

As mentioned above, the slack bus is a mathematical artifact that does not fully reflect actual operation of electric power systems. Instead, there exist a relatively small
number of generating units designated as load following
units. These units participate in the load frequenc control (LFC) and automatic generation control (AGC?; with
the purpose of balancing the interconnected system in response to demand uncertainties. Effectively they manage
to maintain system frequency at 60 Hz. These units are
distributed at various geographic locations in the system.
They are necessary in order to make a transaction feasible without degradin the quality of supply and reliability of the system. Tfle amount of real power imbalance
in the system is distributed among these units based on
participation factors, k a . These are determined based on
combined cost and reliability criteria. They all add up to
unity.

IC, = 1.00

(21)

While the units used for steady state loss compensation may not be the same as the units participating in
AGC, their role of a distributed slack bus can be modeled similarly to the way participation factors are used for
AGC purposes [ 5 ] . In particular, the load flow problem
can be restated in terms of a distributed slack bus by setting the participating units initially at scheduled values,
Pto. It is recognized that, absent load changes, losses create net power imbalance, I B , that are not known a priori.
For this reason the load flow problem formulation with
distributed slack bus is posed as the problem of solving
for the total imbalance in addition to the voltage magnitudes and phase angles. The percentage of imbalance is
distributed to participating generating units as
Pi = Pi0

+ kiIB,

i E E D g e n e r a t i o n units

(22)

Only the real power part of the system Jacobian is


modified according to the following equation:

(LA)

With this minor modification of the last column only,


the problem can be solved iteratively, using standard
Newton-Raphson method. The elements in the last column are simply the participation factors of units participating in the ED and zero otherwise.

543

Note that after each iteration 1, not only the voltage


magnitudes and phase angles get updated, but also the
real power injection at the units participating in the ED.
These units generation set points are recomputed according to the participation factors.

P:"

B.

= P:

+ ki AIBz

i E E D g e n e r a t i o n units (25)

Implications of distributed slack bus


formulation on decomposition of network flows

The above mathematical formulation of the load flow


problem in terms of distributed slack bus can be used
in our formulation to decompose the flows on the system.
This formulation reflects much more realistically system
operation than single slack bus formulation.
Consider a system in which there are G generators
that participate in the ED, for which we use the g subscript; the remaining L = n - G independent generators
and loads, are denoted by the 1 subscript. Based on this,
equation (3) can be partitioned into:

Note that both Yll and Y,, matrices are invertible matrices.
Solving for El using equation (26), we obtain:
El

= -YilYlgEg + Yi'Il

E, = -YTtY,lEl+ YitI,

V Z , T=~ &

- El = YG1Ii,rlc

(34)

Similarly, the voltage difference associated with


transaction T m is given as:
Vl,Tm

= &,Tm - E;

YilIl,~m

(35)

The corresponding complex-valued power vectors of


the major components, for transaction T k and T m are
given in equation (36) and (37), respectively.
SZ,Tk = diag(Ez,n)* I l , T k

(36)

2. Interaction component
The complex-valued power vectors for the interaction
components are given as:

(29)

Smk,total = smk + S k m
(40)
The interaction component flows are function of the
voltage difference only for each transaction. Recall that
this is the main fact behind our Claim 2.

C.

then equation (26) can be written

E, = E; + YitI,
Augment all the buses into one vector to obtain:

, ~ k

(27)

Let E; = -YilYl,Eg, then equation (27) can be written


as :
El = E;
Yi'Il
(28)
Similarly, one can solve for E,, to obtain:

Let E$ = -Y;;Y,lEl,
as :

1. Calculation of major flow component


caused by a given transaction
The major component of power associated with a
given transaction independent of other transactions on the
system can be identified as follows.
Assume, as before, there are two transactions on the
system, T k and Tm. The voltage difference, for nonparticipating generators and loads, associated with transaction T k is given as:

(30)

Allocation of imbalances to individual


transactions

Let the total imbalance at the ED units be the vector IB,


and S i , a vector of scheduled power, and K is a vector of
participation factors (zero elements for non-participating
units).

K = [O,O,O, ...k l , kz, ....kcIT

(41)

and

or

V = E - E' = Z,,,I

(32)

and

E' = [El,E$IT
(33)
Note that equation (32) is similar to equation (7),
in which E: is replaced by E'. Using this similarity in
structure one can use the same equations for the flows as
derived before with minor modifications. Moreover, same
reasoning can be used to prove Claim 2 for the case of
distributed slack bus.
The formulae for the calculation of the major component of power associated with a given transaction and
the interaction component differ from the case of a single slack bus. In the following subsections we derive the
modified formulae based on the above distributed slack
bus formulation.

IB = K * I B
The current can be expressed as current for scheduled
transactions plus another term to meet the imbalance on
the system. The additional term, IB, is zero for all buses
except those participating in the ED.

I = diag(E*)-'Si,

+ diag(E*)-lIB

(43)
The second term in equation (43) is the imbalance
current, caused by all transactions on the system.

I = Y E diag(E*)-lIB
(44)
Claim 3: T h e n e t current imbalance associated with
each transaction is unique and can be identijed exactly.
However, there exists a degree of freedom in assigning percentage of this imbalance t o different units participating in
the ED.

544

Proof:
One can write equation (29) for the case of two current injections as:

YggEg = YggEb + ( I g , T k + Ig,Tm)


(45)
Replacing E$ by its value, and using equation (27)
YggEg

= YgiYil [YigE,

Numerical example

V.

In this section we present numerical results that illustrate


our theoretical analysis. First, we study a simple five bus
system with a single slack bus. Second, we study the
same system with a new generator acting as a second slack
bus. We show the major flows and interaction (minor)
component for the two cases.

+ (Ii,~le+Ii,~m)]+yL'
(Ig,~le
+I!7,~m)
A.
(46)

Single slack bus

In this section a simple five bus system given in Figure 1 is


used as a numerical example for the above mathematical
Solving equation (46) for current injections at the ED
formulation using a single slack bus. This example helps
generation units,
in understanding different complex issues pertaining to
network interaction. The following results are achieved
I g , T k + I g , T m = ( Y g g -Y,IY1llYI,)Eg+YgIY~l(I1,Tk+II,Tm)
using a MATLAB program. Data for the five bus system
(47)
is given in Appendix A.
Knowing that Y , Y-bus admittance matrix, is singular it follows that:

l T ( Y g-gYglY,lY1g)= 0
(48)
This observation implies that the net current innbalance, C I B , for any transaction is a function of that transaction only:

CIBT,

1T Ig,Tm

lTYglYilI1,Tm

(50)

This proves the first part of Claim 3.


There exists a degree of freedom in choosing individual components of I g , T k , so we choose them according to
the participation factors. In other words, we distribute the
total imbalance current injection in the system caused by
a given transaction according to the participation factors.
This distribution is subject to the condition that the sum
of all losses at any generation bus participating in the ED
is equal to the load flow solution value.

* CIBTI,

=K

(51)
This proves the second part of Claim 3 . 0
Note that allocating the imbalance current at fixed
complex valued voltage is equivalent to allocating the
power among different generators.
Claim 4: T h e power imbalance associated with each
transaction can be identified exactly as a f u n c t i o n of all
transactions o n the system.
The voltage difference at ED units is:
Ig,Tk

+
Figure 1: One line diagram of a five bus system
The load flow system variables are shown in Table 1.
All values are expressed in per unit values. Bus # 5 is
the slack bus. There are two scheduled transactions on
the system. The first is from generator bus#4 to load bus
#2, the contract is for 100 MW, 1.0 per unit. The second
is from generator bus#3 to load bus #1, the contract is
for 120 MW, 50 Mvar, 1.20, 0.5 in per unit. The system
variables with all the scheduled transactions taking place
are shown in Table 1. In this case the solution is achieved
using a simple load flow since there is only one ED unit,
i.e., the slack bus. Note the real power imbalance at the
slack bus reflects only the losses since there is no mismatch
between load and generation in both transactions.

and the voltage is given by equation (53). The vo1tag;e E$


is function of El, which is function of all the transactions
on the system.
E$
The power vector at ED units is given as:
Eg,Tk = V g , T k f

(53)

S g , T k = d@(Eg,Tk)*Ig,Tk
(54)
which is function of E1.O
The objective here is to compute the imbalances, including losses for each scheduled transaction; these imbalances are a function of all scheduled transactions on
the system. Only the impact of scheduled transactiolns is
considered here.
Finally, the total imbalance on the system for transaction T k is the summation of all imbalances.

IBTk = 1

* Sg,Tk

(55)

i o.02+ 0.05i I

0.02

+ 0.05i I

1.0000

Table 1: Vectors of complex valued power, current and


voltage for all transactions
Table 2 shows the system variables for transaction
one only, while Table 3 shows the system variables for
transaction two only. Table 4 shows the line power flows
for transaction one, two and the total flows. The interaction component between the two transactions, S k m , t o t a l ,
is given in Table 5. Note that the slack bus interaction
component value is zero.

545

Table 2: Vectors of complex valued power, current and voltage corresponding to the first transaction only

bus #
1

ST2

1.20 - 0.501
0

IT2

ET^

-1.20 - 0.561
0

0.9762 - 0 ,03801
.. . 3 - 0.00121

Table 3: Vectors of complex valued power, current and voltage corresponding to the second transaction only

bus #
1
2

521

s
1
2

-0.0002 + 0.0018i
0

0.0026

+ 0.0014i

Sl2,total

-0.0002 + 0.0018i
0.0026 + 0.0014i

Table 5: Vectors of complex interaction components

bus

Table 6: Vectors of complex valued power, current, voltage, and distribution factors for all transactions

S'T 1

1 ;

0.0073
0.0027

+ 0.0175i
+ 0.0056i

j 0.0082 + 0.0172i j 0.9978


I 0.0027 + 0.0057i I 0.9848

+ 0.0481i

+ 0.0002i

Table 7: Vectors of complex valued power, current and voltage corresponding to the first transaction only

546
T1

T2

Total F l o w s 1 1

Table 8: Line flows for all transactions on the system

S12,totaljl

s12

Table 9: Vectors of complex valued power corresponding


to the interaction between the two transactions

B.

Distributed slack bus

The same transactions and transmission system parameters are used here as in the previous example, however we
have added another generator between buses 1 and :2, as
in Figure 2. This new bus is considered an ED generator
with zero scheduled generation.

Figure 2: One line diagram of a six bus system

systems with distributed slack bus or many generation


units participating in the ED. Second, net power imbalance caused by each transaction can be also identified as
a function of all transactions on the system. However,
there exists a degree of freedom in dividing this imbalance among different generation units participating in the
ED.
Using the derivations introduced in this paper, one
can calculate the power flows and imbalance on the network subject to each given economic transaction. These
formulae are an important part of our proposed framework for cost-based pricing of transmission and ancillary
services in competitiGe elesric power markets proposed in
Part I1 [6], [7].

References
M. S. CaloviC, V. C. Strezoski, Calculation of
Steady-State Load Flows Incorporating System Control Effects and Consumer Self-regulating Characteristics, Intl Journal on Electrical Power & Energy
Systems, Vol. 3, No. 2, April 1981, pp 65-74.
D. Hayward, et al Operating Problems with Parallel
Flows, 91 WM 226-1 PWRS, IEEE Power Winter
Meeting New York, Feb. 1991.
M. IliC, J. Zaborszky, Dynamics and Control of the
Large Electric Power Systems, MIT Course Notes,
Course 6.686, 1995.
L. S. Luen, The Load Flow Problem Without Slack
Bus, McGill University, Masters Thesis 1979, Montreal, Canada.
J. Zaborszky, G. Huang, S. Y. Lin, Reactive
and Real Power Control For Computationally Effective Voltage and Thermal Management, IEEE/PES
Summer Meeting Seattle, Washington, 1984, paper
no. 84 SM 618-5.
A. Zobian, M. IliC, A Simple Solution to the Loop
Flow Problem in Pricing Transmission Services,
LEES Technical Report TR95-001, MIT, LEES,
Cambridge, January 1995.
A. Zobian, M. IliC, A Framework for Cost-Based
Pricing of Transmission and Ancillary Services in
Competitive Electric Power Markets Proc. of the
American Power Conference, April 1995, Chicago, Ill.
A. Zobian, An Efficient Computer Package for Analysis and Control in Large Scale Electric Power Systems, American University of Beirut, Masters Thesis 1990, Beirut, Lebanon.

Appendix A: Data for the five bus system


Table 6 shows the load flow solution for the system
using participation factors. Generator # 5 assigned a participation factor of 75% while generator # 6 assigned a
participation factor of 25%. Note that the system real
power losses are less than in the case of a single slack bus,
as expected. This means less operating cost and higher
efficiency but it does not effect fixed cost recovery.
Table 7 shows the system variables for transaction
one only. Table 8 shows the line power flows for transaction one, two and the total flows. The interaction component between the two transactions, Skm,total, is given in
Table 9.

VI.

Conclusions

In this paper an engineering-based approach is proposed


to decompose power flows on the transmission system into
many components, each associated with a given transaction and an interaction component. This approach is further generalized to resemble the actual operation of power

The transmission lines parameters are shown in Table 10.

from bus

5
5

# to bus #
1
3

K.
0.01
0.01

x 1
0.10
0.10

Table 10: Transmission Lines parameters

541

Discussion
Celso Gonzalez and Hugh Rudnick (Universidad
Cat6lica de Chile, Santiago, Chile): Congratulations to the
authors for their valuable contributions, particularly in
presenting a decomposition of complex power flows in
relation to individual transactions, separating them into a
major component (independent of other transactions) and an
interaction component (dependent on everything else). It
seems that the latter one copes for most of the non-linear
behavior of the transaction, hence in practice it may not be so
small.
We would appreciate a comment from the authors in
relation to the most relevant assumptions which the approach
requires, specially under a deregulated competitive
environment. In particular, it is not all clear to us that Claim
2 will still hold under normal operating conditions
independent of the network topology, not to mention heavily
used grids or even decoupled (due to line overloading) power
systems. Hence, along this line of argument, the setting of
ex-ante pricing can be jeopardized because of the importance
of the interaction component. Do the authors have any
experience of how large can the interaction component
become in stressed systems?. All in all, we think the
approach could be used in setting long-range pricing schemes
whereby an average use-of-system approach is needed.
The use of participation factors seems an interesting way
to deal with actual power balancing [A]. We wonder, how in
practice, considering that an optimal dispatch is made, are
these values set in a competitive environment. Most probably
these values are constantly changed (Le. along a day), thereby
altering the value and significance of not only the actual
imbalance but also the interaction component. In fact this
takes us to Claim 3. Can this degree of freedom significantly
be altered due to different economic assignment of generating
units, maintenance or contingencies?.
Finally, it is stated that ancillary generation is needed to
balance generation with demand in an economic way. How do
the authors think that services such as load frequency control
units, operational/spinning reserve and other ancillary
services, be separated according to their impact on each
transaction?. Furthermore, would the authors consider that an
approach using participation factors (or similar), could be
developed for reactive power flow balance?.
The authors comments on these points will be
appreciated.
[A] Rudnick, H., Palma, R., FernBndez, J. Marginal pricing
and supplement cost allocation in transmission open access.
IEEE Transacthns on Power Systems, Vol. 10, NQ2,May
1995, pp. 1125-1132.
Manuscript received February 26, 1996.

A. A. Zobian and M. D. 11% :


We thank the discussors for their kind comments on our
contributions.

The question regarding relative contribution of the interaction


component is an important one. It is discussed in response to
Mr. Maragonon Lima (Part 11). We agree with these
discussors that the interaction component could be significant
under unexpected system changes. It is for this reason that we
suggest to incorporate it into the transmission price. However,
since it cannot be determined independently from other
transactions, and further supported by Claim 2 that at least
under normal conditions and typical network topologies the
interaction component is relatively small, we propose an exante pricing formula which applies the same rule for
allocating the interaction component cost to all scheduled
transactions. Moreover, when the interaction component
becomes large, it will not jeopardize the pricing scheme
simply because the limit of this case is a postage stamp
method. When the major flow component is zero, the fixed
transmission cost is divided equally among all users on a MW
basis.

The discussors comments regarding use of participation


factors are well taken.
The use of the proposed approach in practice, if an optimal
dispatch is made, is simple. The same participation factors
that are used in the economic dispatch are used in computing
the share of each unit in meeting the power balance for a
given transaction. Definitely these factors change along the
day as the operating condition change. If prices are to be set
ex ante then an expected average values of these participation
factors should be used the same as with the marginal costs of
the units participating in economic dispatch. If the price are
to be computed or adjusted ex post then the actual values can
be used to compute the costs on a very short term basis, and
then all the costs incurred over the lifetime of the contract
should be added up to make the charge for using that service.
These factors can be set based on

Regarding pricing details for other ancillary services, in


particular load frequency control and operationalhpinning
reserve, we point out that this paper provides a general
framework for usage-based pricing of all ancillary services.
Our preliminary response is that it is very difficult to attribute
the operationalhpinning reserve with a given transaction or a
given load even at the margin, instead we believe that this is a
system service and a fixed rate per MW is a reasonable way to
allocate the costs incurred in providing it. Much more
detailed work is under way in our group which is devoted to
the specifics of the ancillary services. The discussors may
wish to see our recent monograph on this topic [ 11.
The use of participation factors for reactive power flow is an
interesting point, especially knowing that reactive power is
not dispatched directly according to production cost or any
other performance criteria, rather the voltages at generation
units are set either to minimize the real power losses on the

548

transmission system or to meet voltage constraints at load


buses. Thus, unlike real power, participation factors are not
used in dispatching reactive power. However, reactive power
particiaption factors can be computed for a given dispatch
and used in ex post pricing of reactive power. Whether such
an approach is feasible or relevant for pricing purposes
requires more research.

[l]M. Ilid, X. Liu, Hierarchical Power Systems Control: Its


Value in a Changing Industry, London Limited SpringerVerlag Series on Advances in Industrial Control, May 1996.

Manuscript received April 23, 1996.

Вам также может понравиться