You are on page 1of 13

Published in IET Generation, Transmission & Distribution

Received on 21st January 2008


Revised on 6th October 2008
doi: 10.1049/iet-gtd.2008.0308
ISSN 1751-8687
Reliability assessment of restructured power
systems using optimal load shedding
technique
P. Wang
1
Y. Ding
2
L. Goel
1
1
Nanyang Technological University, Singapore
2
Reliability Research Lab, University of Alberta, 4-9 Mechanical Engineering Building, Edmonton, Alberta, Canada, T6G 2G8
E-mail: dingyi@ntu.edu.sg
Abstract: Power systemrestructuring and deregulation has changed the strategy of reliability management of a power
system. Load shedding, and generation and reserve re-dispatch methods used in the existing reliability evaluation
techniques have to be improved to incorporate these changes. An optimisation technique, incorporating those
changes, is proposed in this study to determine load curtailment and generation re-dispatch for each contingency
state in the reliability evaluation of restructured power systems with the Poolco market structure. The problem is
formulated using the optimal power ow (OPF) technique. The objective of the problem is to minimise the total
system cost, which includes generation, reserve and interruption costs, subject to market and network constraints.
A model for the contingency management of a Poolco power market is presented to include generation and
reserve biddings, reliability considerations and transmission network constraints in reliability evaluation. Both
supply side reliability for a generation company (Genco) and demand-side reliability for a customer can be
calculated using the technique. The proposed technique can be used to evaluate both conventional and
restructured power systems, and can provide both economic and reliability information for the independent
system operator to manage system reliability, for Gencos to enhance their reliability, and for customer to select
suppliers. The modied IEEE-RTS with the Poolco market has been analysed to illustrate the techniques. The results
obtained using the proposed technique have been compared with those from the existing load shedding techniques.
Nomenclature
0 normal state index (superscript)
j contingency state index (superscript)
i, k bus index (subscript)
g generation unit index (subscript)
c component index (subscript)
r reserve unit index (subscript)
s customer sector index (subscript)
G Genco index (subscript)
SG set of units in a Genco
N number of buses
NG number of Gencos
F
G
j
set of the failed units in Genco G
NS
i
number of customer sectors
NC number of contingencies
N
c
number of independent components
b number of failed components
CD
G
j
real power capacity not delivered (MW)
CS
G
j
real power capacity shortage (MW)
P
gi
0
real power dispatched (MW)
Q
gi
0
reactive power dispatched (Mvar)
P
is
0
real power load (MW)
Q
is
0
reactive power load (Mvar)
DP
gi
j
real power re-dispatched (MW)
DQ
gi
j
reactive power re-dispatched (Mvar)
628 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org
P
ri
j
real power reserve utilised (MW)
Q
ri
j
reactive power reserve utilised (Mvar)
LC
j
total system real power load curtailment (MW)
C
j
total system cost (k$/h)
TR
j
total reserve utilised (MW)
GC
gi
j
generation costs (k$/h)
RC
ri
j
reserve costs (k$/h)
CC
is
j
customer interruption costs (k$/h)
CP
i
j
, CP
is
j
real power load curtailment (MW)
CQ
i
j
,
CQ
is
j
reactive power load curtailment (Mvar)
jV
i
j
j/u
i
j
bus voltage (kV)
jV
i
j
max
voltage up limit (kV)
jV
i
j
min
voltage low limit (kV)
jS
ik
j
j magnitude of apparent power ow from bus i to
k (MVA)
jS
ik
j
max
apparent power limit of line (MVA)
P
gi
min
minimum real power generation (MW)
P
gi
max
maximum real power generation (MW)
DP
gi
low
ramp down limit (MW)
DP
gi
upp
ramp up limit (MW)
Q
gi
min
minimum reactive power generation (Mvar)
Q
gi
max
maximum reactive power generation (Mvar)
P
ri
min
minimum real power reserve (MW)
P
ri
max
maximum real power reserve (MW)
Q
ri
min
minimum reactive power reserve (Mvar)
Q
ri
max
maximum reactive power reserve (Mvar)
P
Li
0
bus real power load (MW)
Q
Li
0
bus reactive power load (Mvar)
P
Li
max
maximum real power load (MW)
Q
Li
min
minimum reactive power load (Mvar)
Q
Li
max
maximum reactive power load (Mvar)
jY
ik
j
j/d
ik
j
bus admittance
CDF
s
customer damage function ($/kWh)
p
j
state probability
D
j
state departure rate (departures/year)
d
j
state duration (h)
A
c
component availability
U
c
component unavailability
l
c
component failure rate (failures/year)
m
c
component repair rate (repairs/year)
1 Introduction
One of the main objectives of power system restructuring is to
provide the customer more choices on their supply reliability
based on the associated costs [1, 2]. Reliability is not a free
product for customers in the restructured power industry.
The centralised reliability management in the vertically
integrated power systems is moving forward to the
decentralised reliability management in the new environment.
In vertically integrated power systems, similar reliability
policy is usually applied for the same type of customers. The
system operator is concerned more with the system reliability
rather than with the reliability of an individual customer.
Generation re-dispatch and load shedding after a system
contingency state in real-time operation are therefore usually
determined by system operators based on their experience
and judgement on system security. The main objective of
contingency management in real-time system operation is to
solve the system voltage, line overloading, frequency and
power balance problems. The load shedding and generation
re-dispatch techniques for conventional power systems have
been well-developed and applied in real-time operation.
These techniques have also been applied in reliability
evaluation. Most reliability evaluation techniques [35]
perform DC and AC power ow analysis to determine
network violations for each contingency state. Proportional
or priority load shedding are commonly used to alleviate
network violations. The system and load point reliability
indices are determined based on the load curtailments.
Optimisation techniques have been used for load shedding
and generation re-scheduling in the reliability evaluation of
conventional power systems [68]. The objective of load
shedding and generation re-scheduling is to minimise the
total load curtailment subject to system voltage and power
ow constraints. However, different customers have different
interruption costs for the same amount of load curtailment.
The techniques [68] cannot guarantee the minimum
interruption costs although the curtailment is the minimum.
Minimum total interruption cost is used to determine
optimal load shedding in distribution system reliability
evaluation [9]. Generation costs have not been considered
during the generation re-dispatch [69]. The changes
introduced by the restructuring have not been considered in
these techniques.
The most important changes in restructured power systems
are the introduction of electricity and ancillary service markets
which allow customers to participate in reliability
management. Reserve is a product in the ancillary services
market just like power is in the energy market. Customers
can purchase electricity from the energy market and
reliability in the ancillary services market based on their price
and reliability expectations. Some customers may be
prepared to pay more to receive higher reliability and others
may be willing to pay less for lower reliability. Generation
companies also have different offers for providing reserve and
energy. The independent system operator (ISO) will
schedule the reserve considering customer reliability
requirements implicitly by incorporating the customer
interruption cost in the decision making. These changes
make the process of load shedding and generation
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 629
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
re-dispatch more complicated than that used in conventional
power systems. The decision making for load shedding,
reserve and generation re-dispatch in reliability evaluation
should incorporate these changes. The existing techniques
for the reliability evaluation of restructured power systems
[10, 11] utilise the AC power ow analysis. In these
methods, the proportional load shedding and generation re-
dispatch were used in conjunction with a bilateral market
model. Generation cost, reserve costs and customer reliability
choices have not been considered in these techniques.
Goel et al. [12] proposed a framework to implement
supply- and demand-side contingency management in the
reliability assessment of hybrid power markets. The
contingency management model was formulated as a linear
programming problem with the objective of minimising the
curtailment bidding costs from bilateral and spot market
customers, and the reserve bidding costs from each Genco.
All the bidding costs are assumed to be constant. Therefore
the cost models are not practical and accurate. Generation
re-dispatch costs and transmission line contingencies were
not included in the objective function. The main focus of
[13] is to propose a general methodology to evaluate both
reliabilities and prices for the pool and bilateral customers
in a hybrid power market. An optimal power ow (OPF)
technique was used to determine reliability indices and
electricity prices for each contingency state. The reserve
cost from the power pool was not included in the objective
function. The unreliability contributions from supply-side
(Gencos) and the associated demand-side reliability
(customers) were not clearly dened and classied. Goel
et al. [14] focus on the impact of customer responses on
the system and nodal reliabilities. The customer responses
are modelled by the elasticity of the demand for
electricity in the normal state. In these references, the
evaluation of the unreliability contribution from different
market participants including Gencos and transmission
network were not investigated. Total reliability cost has not
been investigated. The reliability-requirement-based load
curtailments to customers in the load shedding process
were not calculated.
This paper investigates the cost-based load shedding and
generation re-dispatch techniques used in real-time market
and system operation for contingency management. A all-
in-one technique for load shedding, generation re-dispatch
and reserve dispatch in the reliability evaluation of
deregulated power systems with the Poolco market
structure has been proposed. The objective of the problem
is to minimise the total system cost including generation
costs, reserve costs and customer interruption costs.
Reliability requirements are considered in an implicit
manner by incorporating the customer interruption cost in
the objective function. It does not mean a customer can
arbitrarily select its reliability level, which is limited by
its interruption cost. More importantly reliability and
economic indices for different market participants are
introduced and calculated to represent their reliabilities
and the associated responsibilities in the restructured
environment. Both supply- and demand-side reliability
indices for each Genco and customer are calculated,
respectively. Therefore the ISO and the power market
administrator cannot only obtain the reliabilities of specic
customers but also know the nal obligation of a Genco to
its energy supply and the associated reliability.
Section 2 classies the existing techniques for load
shedding and generation re-dispatch used in system
reliability evaluation. Customer responses to their
curtailments as customer damage functions are discussed in
Section 3. The model for reliability management of
restructured power systems with the Poolco market
structure is presented in Section 4. Section 5 formulates
load shedding and generation re-dispatch problem for a
contingency based on the minimum total system cost. The
problem has been solved using the sequential quadratic
programming (SQP). Both transmission and generation
failures have been considered in contingency enumeration.
The reliability indices for Gencos and customers are
presented in Section 6. The modied IEEE RTS system
has been analysed using the proposed techniques and the
results are presented in Section 7.
2 Existing load shedding and
generation re-dispatch techniques
Different load shedding and generation re-dispatch
techniques have been implemented for reliability
management of conventional and restructured power
systems [311]. Some techniques [39] have been
developed for security analysis in system operation. The
load shedding techniques used in reliability evaluation [3
11] can be classied into two categories of AC power ow-
based methods [35, 911] and OPF-based approaches
[68]. In the rst category, AC power ow for a
contingency state is calculated to determine the network
violations. The generation is re-dispatched to release
network violations. If the network violations cannot be
alleviated by the generation re-dispatch, loads have to be
curtailed. The proportional and priority load shedding
schemes are usually used in these techniques. The objective
of these techniques is to alleviate power ow and voltage
problems. The proportional load shedding scheme cuts the
load based on the percentage of load at each bus. This
technique ignores the different specic reliability
requirements of customers. The priority load shedding
method considers the importance of different types of
customers. However, it cannot consider the different
reliability requirements for the same type of customers.
Both the techniques usually result in load being over-
curtailed. In order to solve the problem of
over-curtailments, the objective of the OPF-based methods
[68] is to minimise the total load curtailment by
considering network security constraints. For contingency j,
630 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org
the problem is usually formulated as
Min LC
j

N
i1
CP
j
i
(1)
subject to the following constraints:
Power balance constraints

NG
G1

g[G
P
0
gi
DP
j
gi
_ _
P
0
Li
CP
j
i
_ _

N
k1
V
j
i

V
j
k

Y
j
ik

cos u
j
i
u
j
k
d
j
ik
_ _
(2)

NG
G1

g[G
Q
0
gi
DQ
j
gi
_ _
Q
0
Li
CQ
j
i
_ _

N
k1
V
j
i

V
j
k

Y
j
ik

sin u
j
i
u
j
k
d
j
ik
_ _
(3)
Generator limits
P
min
gi
P
0
gi
DP
j
gi
_ _
P
max
gi
(4)
Q
min
gi
Q
0
gi
DQ
j
gi
_ _
Q
max
gi
(5)
Load curtailment limits
0 CP
j
i
P
max
Li
(6)
Q
min
Li
CQ
j
i
Q
max
Li
(7)
Voltage limits
V
i

min
V
j
i

V
i

max
(8)
Line ow constraints
S
j
ik

S
ik

max
(9)
The existing OPF-based methods can only guarantee
minimum load curtailment. However, the total system cost
under minimum curtailment may not be the minimum.
The existing reliability techniques have to be improved in
the new environment to incorporate customer choices,
reserve and generation offers by Gencos. The proposed
technique considers customer choices on reliability with the
objective of minimising the total cost of a restructured
power system with the Poolco model.
3 Customer responses to
curtailment
In a Poolco market, customers can purchase energy and
reliability in term of reserve from the centralised power pool.
The reliability of a customer depends on the cost he or she
wants to pay or has paid. Customers usually pay nodal prices
which are determined by demands and supply bids at system
buses after considering network constraints. When a power
system transfers from the normal operating state to a
contingency state because of random failures, nodal prices
may vary with changes in system operating conditions. If
system generation is inadequate or network is congested
during a contingency state, system operator may commit
reserve or cut demand based on the difference between
customer reliability considerations and reserve prices.
The Federal Energy Regulatory Commission (FERC)
requires public utility load serving entities to establish
demand response mechanisms in which they will identify
the price at which load would be curtailed [15] for a
contingency state. In other words, this price reects the
customer willingness to pay to avoid a service interruption
for a contingency state. It is difcult to directly determine
the price for a customer to pay for its reliability. Customer
response to price increment associated with the interruption
can be indirectly measured by the customer interruption cost.
Customer interruption cost is a function of customers and
the associated interruption characteristics [16]. Customer
characteristics include the diversity of customers, the nature
of customer activities, size of the operation etc. Interruption
characteristics include duration, frequency and time of
occurrence of interruptions etc. The models used in this
paper are the sector customer damage functions, which
reect the diversities of customer sectors. A standard
industrial classication (SIC) [16] has been used to divide
customers into large user, industrial, commercial,
agricultural, residential, government and institutions and
ofce and buildings categories. Postal surveys have been
conducted by the University of Saskatchewan and Canadian
electric power utilities to estimate customer interruption
costs. The survey data have been analysed to give sector
customer damage functions (CDF
s
), which are a function of
the interruption durations. The CDF
s
normalised by the
annual consumption of electricity ($/kWh) for the ve
representative customer sectors are shown in Fig. 1.
4 Model for contingency
management
Load shedding, generation re-dispatch and reserve dispatch
are closely related to power market structures. The power
markets in the world can be classied into bilateral markets,
Poolco markets and Hybrid markets. The method for
contingency management varies from one market to another.
In the Poolco model used in this paper, Gencos bid to sell
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 631
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
electricity and reserve for the maximum prot and customers
bid to buy electricity and reserve based on their electricity
and reliability considerations. Generating units which
provide reserve can be classied into two categories [17] of
Tier 1 and Tier 2 resources. Tier 1 resources include the
generating units are economically dispatched in the normal
operation and can ramp up or down from the normal output
to a new value during a contingency state; Tier 2 resources
include the generating units are served only as spinning
reserve providers, which supply additional reserve not
available from Tier 1. These units usually operate at points
that deviate from economical dispatch in the normal state.
All those biddings have to be included in reliability
evaluation with considering the various system constraints. A
general model for contingency management with the Poolco
market is proposed as shown in Fig. 2. Market settlements,
cost data and network constraints are included in this model.
Generation re-dispatching (Tier 1) and reserve dispatching
(Tier 2) under a given load condition are determined based
on the biddings from Gencos and customers through the
market clearing process. Quadratic equations are used in this
paper to represent generation dispatching and reserve
dispatching costs. The customer damage functions are used
to evaluate customer interruption costs for different customer
sectors [16]. When a system outage or a contingency occurs,
load curtailment, generation re-dispatched and reserve
dispatched will be determined based on the proposed model
using the optimisation technique.
5 Problem formulation
Reliability evaluation of a restructured power system is the
process of predicting reliability of customers, Gencos and
transmission network for a given system operating condition
and market settlement. The system capacity inadequacy has
to be determined when a power system transfers from the
normal operating state to a contingency state because of
generation and/or transmission outages. If there is
generation inadequacy for a contingency state, generation
and reserve has to be re-dispatched and load has to be
curtailed to maintain system operation. The problem is a
complicated multi-objective optimisation which includes
minimise the system operation cost and to maximise
customer reliabilities. The proper solution for multi-objective
optimisation is Pareto optimal. In this paper, customer
reliability requirements are implicitly represented using
customer interruption costs. Therefore the multi-objective
problem becomes the single-objective problem. The objective
is to minimise the total system cost after considering
network and market constraints. The generation re-
dispatched, reserve dispatched and load curtailed under the
minimum total system cost are used to calculate the
reliability indices of customers, Gencos and transmission
network. For contingency state j, the objective function is
Min C
j

N
i1

NG
G1

g[SG
_
GC
j
gi
P
0
gi
DP
j
gi
_ _


r[SG
RC
j
ri
P
j
ri
_ _
_

NS
i
s1
CC
j
is
CP
j
is
_ _
_

_
_

_
(10)
CC
j
is
CP
j
is
_ _
CP
j
is
CDF
s
d
j
_ _
(11)
subject to the following constraints for contingency state j:
Power balance constraints

NG
G1

g[SG
P
0
gi
DP
j
gi
_ _

r[SG
P
j
ri
_
_
_
_

NS
i
s1
P
o
is
CP
j
is
_ _

N
k1
V
j
i

V
j
k

Y
j
ik

cos u
j
i
u
j
k
d
j
ik
_ _
(12)
Figure 1 Customer damage functions for different sectors
Figure 2 Model for contingency management
632 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org

NG
G1

g[SG
Q
0
gi
DQ
j
gi
_ _

r[SG
Q
j
ri
_
_
_
_

NS
i
s1
Q
o
is
CQ
j
is
_ _

N
k1
V
j
i

V
j
k

Y
j
ik

sin u
j
i
u
j
k
d
j
ik
_ _
(13)
Equations (12) and (13) show that the total system supply has
to be balanced with the total system load and transmission
loss after the generation re-dispatch, reserve dispatch and
load curtailment.
Generation constraints
P
min
gi
DP
j
gi
P
0
gi
P
max
gi
(14)
DP
low
gi
DP
j
gi
DP
upp
gi
(15)
Q
min
gi
DQ
j
gi
Q
0
gi
Q
max
gi
(16)
Reserve constraints
P
min
ri
P
j
ri
P
max
ri
(17)
Q
min
ri
Q
j
ri
Q
max
ri
(18)
DP
gi
j
in the above equations is Tier 1 reserve and can ramp up
or down when required in the contingency state, which can
be positive and negative. P
j
ri
in the above equations is Tier
2 reserve dispatched which is positive. Both Tier 1 and
Tier 2 reserve in a contingency state j must be within its
lower and upper limits.
Load curtailment limits
0 CP
j
is
CP
max
is
(19)
Voltage limits
V
i

min
V
j
i

V
i

max
(20)
Line ow constraints
S
j
ik

S
ik

max
(21)
Equations (10) and (11) can be represented as a general
optimisation problem which includes the general objective
function
Min f
j
(x
j
) (22)
where x
j
DP
j
11
, . . . , DP
j
gi
, . . . , P
j
11
, . . . P
j
ri
, . . . ,
_
CP
j
11
, . . . , CP
j
is
, . . . subject to the general equality
constraints (12) and (13)
g
j
(x
j
) 0 (23)
General inequality constraints of (1421)
h
j
(x
j
) 0 (24)
The SQP algorithm [18], which combines the Newton
method with the quadratic programming (QP), has proved
to be a highly efcient method for solving the non-linear
programming problem. Using the SQP algorithm, the
original non-linear programming problem is reduced to a
QP problem
Min
1
2
d
k
_ _
T
H
k
d
k
rf x
j
k
_ _
T
d
k
(25)
subject to
g
j
x
j
k
_ _
rg
j
x
j
k
_ _
T
d
k
0 (26)
h
j
x
j
k
_ _
rh
j
x
j
k
_ _
T
d
k
0 (27)
where k is the iteration number and H
k
is the a positive
denite matrix approximating the Hessian matrix of the
Lagrange function L
j
. H
k
will be updated so as to converge
to the real Hessian matrix of L
j
.
The Lagrange function L
j
of (2224) is
L
j
f
j
(x
j
) rg
j
(x
j
) hh
j
(x
j
) (28)
where r and h the Lagrange multiplier vectors for g
j
(x
j
) and
h
j
(x
j
), respectively.
The new search direction d
k
can be obtained by solving the
QP problem. The new point x
j
k1
can be reset as follows until
it converges
x
j
k1
x
j
k
a
k
d
k
(29)
where a
k
is the optimal step length parameter along the
search direction d
k
.
After solving the problem, the load curtailment, generation
re-dispatched and reserve dispatched can be determined for
each contingency state. These data are used to calculate the
reliability indices in the following section. The random
failures up to the second order are considered in this paper.
Therefore the optimisation technique can converge to a
proper solution in most contingency states. However, it
should be noticed that the optimisation technique cannot
converge for some contingencies with the severe network
isolations. For those states that result in the violations of
physical circuit laws related to generation limits and
network constraints, the corresponding loads and
generators have to be cut and shut down, respectively, to
guarantee the system stable operation. In such special cases,
load curtailments will be determined according to the
physical laws. The procedure of the proposed technique
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 633
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
includes three main steps: to check the network isolations and
to nd the network islands, to determine load curtailments
and generation and reserve re-dispatches for each island
using the optimisation technique if it is possible, otherwise
to determine load curtailments using the circuit laws.
6 Reliability indices
Contingency enumeration and state selection techniques are
used to determine contingency states and the associated state
parameters. Considering a power system with N
c
independent
components, the state probability p
j
, departure rate D
j
and failure duration d
j
for contingency state j with exactly b
failed components can be determined using the following
equations
p
j

b
c1
U
c

N
c
cb1
A
c
(30)
D
j

b
c1
m
c

N
c
cb1
l
c
(31)
d
j

8760
D
j
(32)
The denition of these parameters can be found in [19].
Equation 30 illustrates that the probability of the state j
equals to the multiplication of unavailabilities of failed
components multiplies the multiplication of availabilities of
working components. The reliability indices of customers,
Gencos and transmission network will be determined in
the following section using the proposed technique.
6.1 Customer reliability indices
Customer choices on price and reliability make non-uniform
reliability possible in restructured power systems, where
customer satisfaction is the main concern of reliability
evaluation. Customer reliability indices can be calculated
from the load curtailment for each contingency state
obtained from the load shedding technique.
The total customer interruption cost for state j (TCOST
j
)
and the total customer load curtailment for state j (LC
j
) can
be calculated as
TCOST
j

N
i1

NS
i
s1
CC
j
is
CP
j
is
_ _
(33)
LC
j

N
i1

NS
i
s1
CP
j
is
_ _
(34)
The expected energy not supplied (EENS
i
) and expected
customer interruption cost (ECOST
i
) for customer at bus i
can be calculated using the following equations
EENS
i

NC
j1
8760 p
j

NS
i
s1
CP
j
is
_ _
(35)
ECOST
i

NC
j1
8760 p
j

NS
i
s1
(CC
j
is
CP
j
is
_ _
_ _
(36)
6.2 Genco and transmission network
reliability indices
In a restructured power system the ISO and power market
administrator usually want to know the nal obligation of a
Genco to its energy supply and the associated reliability. In
this paper, the expected energy not supplied (EENS) and
the expected interruption cost (ECOST) for each Genco
and the transmission network are introduced and calculated
using the proposed technique. In order to calculate these
two indices, the real power capacity not delivered from each
Genco for each state has to be determined in the load
shedding process.
The capacity not delivered CD
j
G
for Genco G is the
difference between the generation dispatched in the normal
state and one in the state j. This index can be calculated as
CD
j
G

N
i1

g[SG
P
0
gi

g[SG
gF
j
G
P
max
gi
_
_
_
_
_
_
_
_
_
_
(37)
If CD
j
G
is less than or equal to zero, Genco G has enough
capacity to meet its supply obligation. If CD
j
G
is greater
than zero Genco G has not enough capacity to meet its
supply obligation. The capacity shortage CS
j
G
can be
calculated as
CS
j
G
max(CD
j
G
, 0) (38)
In this case, the reserve from Gencos will be dispatched or/
and the loads will be curtailed based on the costs to balance
the generation and load for system stable operation.
If there is load curtailment, the energy not supplied ENS
j
G
caused by Genco G for state j can be calculated based on the
total reserve dispatched TR
j
and the capacity not delivered
CD
j
G
using (39) and (40)
ENS
j
G
d
j

CS
j
G

N
G
G1
CS
j
G

N
G
G1
CD
j
G
TR
j
_ _
(39)
TR
j

N
i1

NR
j
i
r1
P
j
ri
(40)
where

N
G
G1
CD
j
G
TR
j
_ _
is the total load curtailment
634 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org
caused by the generation system and CS
j
G
=

N
G
G1
CS
j
G
is the
percentage of load curtailment contribution from Genco G.
The interruption cost COST
j
G
(k$) caused by Genco G for
state j can be calculated as
COST
j
G
d
j

CS
j
G

N
G
G1
CS
j
G

N
G
G1
CD
j
G
TR
j
_ _
LC
j
TCOST
j
(41)
where

N
G
G1
CD
j
G
TR
j
_ _
=LC
j
is the percentage
contribution of the generation system to the total customer
load curtailment.
Considering all the system states, expected customer
energy not supplied EENS (MWh/year) and interruption
cost ECOST (k$/year) for Genco G can be dened and
evaluated based on the state departure rate D
j
and state
probability p
j
for state j using
EENS
G

NC
j1
D
j
p
j
ENS
j
G
_ _
(42)
ECOST
G

NC
j1
D
j
p
j
COST
j
G
_ _
(43)
In (42) and (43), multiplying D
j
with p
j
we can obtain the
frequency (occ/year) of encountering the state j using the
frequency and duration technique [19]. Multiplying the
frequency of state j with the ENS
j
G
(MWh) and COST
j
G
(k$), we can obtain the EENS and ECOST caused by
Genco G for state j. Considering all the system states, we
can obtain the EENS (MWh/year) and ECOST (k$/year)
caused by Genco G. The EENS and ECOST of the
transmission network can be calculated from the EENS
and ECOST of the whole system minus the EENS and
ECOST of the generation system using the following
equations, respectively
EENS
T
EENS
S

NG
G1
EENS
G
(44)
ECOST
T
ECOST
S

NG
G1
ECOST
G
(45)
Considering all the states, the system indices EENS
S
and
ECOST
S
can be calculated based on the total load
curtailment LC
j
, total interruption cost TCOST
j
and state
probability p
j
for state j using the following two equations,
respectively
EENS
S

NC
j1
8760 p
j
LC
j
_ _
(46)
ECOST
S

NC
j1
8760 p
j
TCOST
j
_ _
(47)
The denitions and theoretical justications of the reliability
indices in this section can be found in [19].
7 The IEEE RTS studies
The proposed technique was used to analyse IEEE-RTS
[20] shown in Fig. 3. The system is modied as a
restructured power system with Poolco market. It is
assumed that all generators at bus i belong to a single
Genco i. The system has 10 generation companies and 17
bulk load point customers. The bidding curve for energy or
reserve utilisation from each Genco is assumed to be a
quadratic equation, which is illustrated in Table 4. The
customer damage functions for different customer sectors
are shown in Fig. 1. Parameter values for customer damage
functions of different sectors and customer sector
allocations at load buses have also been supplemented in
Tables 5 and 6, respectively. The random failures up to the
second order, which include the failure of a generating unit
or a transmission line, and the failure of two units or two
Figure 3 IEEE RTS
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 635
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
lines or the combination of one unit and one line, are
considered in the illustrative example. For each system state
considered, load shedding and generation re-dispatch are
evaluated by using the proposed OPF technique and
corresponding reliability parameters are also calculated. The
annual constant peak load is used in the analysis. After
considering all the contingency states, the annual reliability
indexes can be calculated.
A Pentium IV 2.4 GHz personal computer is used in the
simulation studies. The computing time considering all the
system states is about 2 h. The technique is proposed for
off line calculation. Therefore the computing time was not
optimised when we wrote the program. The computing
time can be signicantly reduced to minutes when high-
speed computer is used and the program is optimised.
The results obtained from the proposed technique are
compared with those from the existing methods in the case
studies. The conventional AC power ow is used to
calculate the reliability indices in the rst two cases. The
proportional and priority load shedding techniques are
applied in Cases 1 and 2, respectively. In Case 2 it is
assumed that load points 1 and 2 have higher service
priority than other load points. The OPF-based reliability
technique with the minimum total system load curtailment
is used in Case 3. In Case 4, the proposed OPF-based
reliability technique with optimal generation re-dispatching
and load shedding considering customer interruption cost
is used to calculate the load point reliability indices. In
Case 4, it is supposed that the two 20 MW units at bus 1
and two 20 MW units at bus 2 are served as reserve
providers (Tier 2 resource). In the normal state, these units
are not economically dispatched by the system operator.
Other generating units are economically dispatched in the
normal state (Tier 1 resource).
Table 1 shows the expected energy not supplied (MWh/
year) of each bulk load point for the four cases. The
expected interruption costs (k$/year) of each bulk load
point for the four Cases are shown in Table 2. Tables 1
and 2 also show the total system EENS and ECOST for
the four cases.
It can be seen from Table 1 that the priority load shedding
(Case2) has the largest total system EENS (39 687 MWh/
year) followed by Case 1 (39 128 MWh/year) with
proportional load shedding and Case 4 (30 625 MWh/
year) with minimum total system cost. The total system
EENS calculated using the minimum load curtailment
technique (Case 3) is the minimum (30 417 MWh/year).
The difference between Cases 3 and 4 is only 0.68%. The
difference between Case 3 and other two cases is over 28%.
This means that the proposed technique also has the lower
EENS.
It can be seen from Table 2 that the total system ECOSTs
(k$/year) for Cases 4, 3, 1 and 2 are 97 101, 747 221,
367 290 and 364 650, respectively. The difference between
Case 4 and other cases is over 73%.
The total systemcost including generation costs, reserve costs
and customer interruption costs (k$/year) for the four cases are
875 890, 873 190, 1239 000 and 588 930, respectively.
Tables 1 and 2 also show that optimal load shedding and
generation re-dispatching in reliability management using
the proposed technique (Case 4) guarantee both the lower
total system EENS and the minimum total system ECOST.
Although the system EENS for Case 3 is the lowest among
the four cases, the corresponding total system ECOST is the
highest because the minimum load curtailment ignores the
customer willingness to pay for its reliability. It can be seen
from the results that the customer response to its reliability
is a very important factor in reliability evaluation of
restructured power system with Poolco market and therefore
should be considered in the evaluation.
The results in Tables 1 and 2 are calculated based on
2850 MW system load. However, the proposed technique
can be easily implemented to calculate reliability indices for
different operating conditions and assumptions. The results
for the high load level of 2992.5 MW has been calculated
Table 1 The EENS for four cases
Bus Case 1 Case 2 Case 3 Case 4
1 1482.7 0 0 327.16
2 1331.7 0 0 1047.3
3 2471.2 2701.1 6974 7329.7
4 1015.9 1110.3 3392.6 2227.1
5 974.78 1065.3 0.021 0.021
6 1867.3 2040.6 11576 5642.4
7 1716.1 1875.6 0 0
8 2347.7 2565.8 0 1269.9
9 2402.6 2625.8 191.7 3641.9
10 2677.1 2925.9 2634.3 3876.3
13 3638.2 3976.2 0 413.42
14 2663.5 2911.3 3760.4 1637.6
15 4352.2 4756.5 0 854.16
16 1372.9 1500.4 0 414.27
18 4571.7 4996.5 1888.2 1426.9
19 2484.9 2715.8 0 102.5
20 1757.3 1920.6 0 414.27
Total 39 128 39 687 30 417 30 625
636 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org
and compared with those for 2850 MW using the proposed
technique (Case 4). In this case, the system EENS and
ECOST for 2992.5 MW are 116 530 MWh/year and
322 940 k$/year, respectively, which increase 280.5 and
232.6% comparing with those for 2850 MW. The total
expected system cost is 861 890 k$/year, which increases
46.35% comparing with those for 2850 MW. That means
system reliability and economical indices are very sensitive
to the system demand.
The proposed technique can also provide the reliability of
each Genco and the overall transmission system. Table 3
shows the EENS (MWh/year) and the associated ECOST
(k$/year) for each generation company. The EENS and
ECOST caused by the Gencos at buses 1, 2, 7 and 22 are
zeros because these Gencos have more generating units and
capacity to meet the obligation. The EENS caused by the
Gencos at buses 18 and 21 are 12 133 and 11 988,
respectively, which are 39.62 and 39.15% of the EENS
caused by the generation system, respectively. The ECOST
caused by the Gencos at buses 18 and 21 are 38 840 and
38 539, respectively, which are 40 and 39.7% of the
ECOST caused by the generation system, respectively. In
those companies, the high EENS and ECOST values is
due to the less generating units each with high capacity.
Therefore those Gencos should be responsible for customer
interruptions and should purchase additional reserve to
increase their reliabilities. It can be concluded that the
Gencos with less units each with high install capacity may
have large contribution to system unreliability. The EENS
caused by Gencos is 30 623.72 MWh/year, which is
slightly less than the 30 625 MWh/year considering both
Table 2 The ECOST for four cases
BUS Case 1 Case 2 Case 3 Case 4
1 15 541 0 0 588.31
2 17 295 0 0 1883.2
3 9324.3 10 215 164 120 34 176
4 22 675 24 840 111 510 6575.5
5 23 629 25 884 0.1721 0.1721
6 30 547 31 725 284 840 24 349
7 13 421 14 703 0 0
8 19 614 21 486 0 2049.7
9 48 117 52 710 5258.3 5849.8
10 8606.1 9427.7 45 558 6315
13 34 319 37 595 0 743.44
14 6852.9 7506.8 94 060 6020.3
15 40 229 44 069 0 1536
16 16 777 18 378 0 744.97
18 26 854 29 418 41 876 5340.6
19 16 718 18 314 0 184.32
20 16 777 18 378 0 744.97
Total 36 7290 364 650 747 221 97 101
Table 3 The EENS and ECOST of Gencos for case 4
Bus EENS
G
ECOST
G
1 0 0
2 0 0
7 0 0
13 53.56 75.367
15 209.09 271
16 206.97 268.26
18 12 133 38 840
21 11 988 38 539
22 0 0
23 6033.1 19105
Total 30623.72 97098.63
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 637
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
generation and transmission. Therefore the EENS caused by
the transmission network is very small because of the strong
grid connection.
8 Conclusion
An optimisation problem to determine load shedding,
generation re-dispatching and reserve utilisation for
reliability evaluation of restructured power systems with the
Poolco market model has been formulated in this paper.
The objective of the problem is to minimise the total
system cost including generation costs, reserve utilisation
cost and customer interruption costs. Customer reliability
requirements in terms of interruption costs are implicitly
incorporated in the decision process when a system
deciency occurs. The OPFtechnique based on SQP has
been used to solve the problem. Reliability indices for
Gencos, bulk load point customers and overall system have
been introduced for the new market environment. Different
existing techniques have been compared with the proposed
technique. The proposed technique can be used to evaluate
the reliability of both conventional and restructured power
systems. The modied IEEE RTS has been analysed to
illustrate the technique. The results show that the proposed
technique can provide useful information for reliability
management of restructured power systems. The evaluated
reliability indices can be used in power markets to
determine the penalty payments of each Genco.
9 References
[1] SHAHIDEHPOUR M., ALOMOUSH M.: Restructured electrical
power systems: operation, trading, and volatility (CRC,
New York, 2001)
[2] BILLINTON R., SALVADERI L., MCCALLEY J.D., ET AL.: Reliability
issues in todays electric power utility environment, IEEE
Trans. Power Syst., 1997, 12, (4), pp. 17081714
[3] ALLAN R.N., BILLINTON R., BREIPOHL A.M., GRIGG C.H.:
Bibliography on the application of probability methods in
power system reliability evaluation: 19871991, IEEE
Trans. Power Syst., 1994, 9, (1), pp. 4149
[4] ALLAN R.N., BILLINTON R., BREIPOHL A.M., GRIGG C.H.:
Bibliography on the application of probability methods in
power system reliability evaluation, IEEE Trans. Power
Syst., 1999, 14, (1), pp. 5157
[5] BILLINTON R., FOTUHI-FIRUZABAD M., BERTLING L.: Bibliography
on the application of probability methods in power
system reliability evaluation 19961999, IEEE Trans.
Power Syst., 2001, 16, (4), pp. 595602
[6] BILLINTON R., KHAN E.: A Security based approach to
composite power system reliability evaluation, IEEE Trans.
Power Syst., 1992, 7, (1), pp. 6572
[7] MEIO A.C.G., MELLO J.C.O., GRANVILLE S.: The effects of
voltage collapse problems in the reliability evaluation of
composite systems, IEEE Trans. Power Syst., 1997, 12, (1),
pp. 480488
[8] DORNELLAS C., SCHILLING M., MELO A., SOUZA J.C.S., DO
COUTTO FILHO M.B.: Combining local and optimised
power ow remedial measures in bulk reliability
assessment, IEE Proc. Gener. Transm. Distrib., 2003,
150, (5), pp. 629634
[9] WANG P., BILLINTON R.: Optimal load-shedding technique
to reduce the total customer interruption cost in
distribution system, IEE Proc. Gener. Transm. Distrib.,
2000, 147, (1), pp. 5156
[10] WANG P., BILLINTON R.: Reliability assessment of a
restructured power system considering the reserve
agreements, IEEE Trans. Power Syst., 2004, 19, (2),
pp. 972978
[11] WANG P., BILLINTON R.: Reliability assessment of a
restructured power system using reliability network
equivalent techniques, IEE Proc. Gener. Transm. Distrib.,
2003, 150, (5), pp. 555560
[12] GOEL L., APARNA V.P., WANG P.: A framework to implement
supply and demand side contingency management in
reliability assessment of restructured power systems, IEEE
Trans. Power Syst., 2007, 22, (1), pp. 205212
[13] DING Y., WANG P.: Reliability and price risk assessment of
a restructured power system with hybrid market structure,
IEEE Trans. Power Syst., 2006, 21, (1), pp. 108116
[14] GOEL L., WU Q., WANG P.: Reliability enhancement of a
deregulated power system considering demand response,
IEEE Power Eng. Soc. Gen. Meet., 1822 June 2006
[15] FERC.: Commission Establishes Prospective Mitigation
And Monitoring Plan For California Wholesale Electric
Markets, http://www.ferc.gov/
[16] TOLLEFSON G., BILLINTON R., WACKER G.: A Canadian customer
survey to assess power system reliability worth, IEEE Trans.
Power Syst., 1994, 9, (1), pp. 443450
[17] The PennsylvaniaNew JerseyMaryland
interconnections http://www.pjm.com/
[18] BERTSEKAS D.P.: Nonlinear programming (Athena
Scientic, Belmont, MA, 1999)
[19] BILLINTON R., ALLAN R.N.: Reliability Evaluation of Power
Systems (Plenum Press, New York and London, 1996)
[20] IEEE Task Force.: IEEE reliability test system, IEEE
Trans. Power App. Syst., 1979, 98, pp. 20472054
638 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org
10 Appendix
Table 4 Bidding parameters for energy and reserve utilisation of RTS
generating units
Unit no. Bus no. Installed capacity (MW) Bidding function
coefcients
a ($/MWh) b ($/MW
2
h)
11 1 20 103.6 0.4
12 1 20 103.6 0.4
13 1 76 15.30 0.5
14 1 76 15.30 0.5
21 2 20 103.6 0.4
22 2 20 103.6 0.4
23 2 76 15.30 0.5
24 2 76 15.30 0.5
31 3 100 52.8 0.3
32 3 100 52.8 0.3
33 3 100 52.8 0.3
131 13 197 50.6 0.4
132 13 197 50.6 0.4
133 13 197 50.6 0.4
151 15 12 63.3 0.7
152 15 12 63.3 0.7
153 15 12 63.3 0.7
154 15 12 63.3 0.7
155 15 12 63.3 0.7
156 15 155 12.44 0.3
161 16 155 12.44 0.3
181 18 400 6.3 0.1
211 21 400 6.3 0.1
221 22 50 0.5 0.1
222 22 50 0.5 0.1
223 22 50 0.5 0.1
224 22 50 0.5 0.1
225 22 50 0.5 0.1
226 22 50 0.5 0.1
231 23 155 12.44 0.1
232 23 155 12.44 0.1
233 23 350 12.10 0.1
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640 639
doi: 10.1049/iet-gtd.2008.0308 & The Institution of Engineering and Technology 2009
www.ietdl.org
Table 6 Customer sector allocations at load buses for the RTS
Bus Peak load (MW) Large users (MW) Agriculture (MW) Residential (MW) Industrial (MW) Commercial (MW)
1 108 23.83 0 34.03 36.94 13.20
2 97 32.26 0 50.05 0 14.69
3 180 80.0 6.33 52.50 33.25 7.92
4 74 20.22 0 34.52 0 19.26
5 71 0 0 22.83 28.10 20.07
6 136 38.1 8.38 49.70 29.34 10.48
7 125 25 18.24 38.44 31.92 11.4
8 171 87.67 0 55 11.67 16.66
9 175 91.13 19.54 23.46 0 40.87
10 195 116.92 8.77 41.54 20.46 7.31
13 265 74.84 17.91 82.15 60.95 29.15
14 194 84.55 0 62.89 40.74 5.82
15 317 228.0 0 54.84 0 34.16
16 100 59.85 0 25.90 0 14.25
18 333 207 0 63.27 39.96 22.77
19 181 111.2 0 55.60 0 14.20
20 128 59.85 0 53.9 0 14.25
Table 5 Parameter values for customer damage functions of different sectors
Failure duration (min) Interruption cost ($/kWh)
Large users Industrial Commercial Agricultural Residential
1 0.073 0.46 0.129 0.027 0.0004
20 0.111 1.332 1.014 0.155 0.044
60 0.163 2.99 2.951 0.295 0.243
240 0.291 8.899 10.992 1.027 2.235
480 0.604 18.156 28.02 2.134 6.778
640 IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 7, pp. 628640
& The Institution of Engineering and Technology 2009 doi: 10.1049/iet-gtd.2008.0308
www.ietdl.org

You might also like