Sunteți pe pagina 1din 11

www.ietdl.

org

Published in IET Generation, Transmission & Distribution


Received on 29th April 2009
Revised on 21st December 2009
doi: 10.1049/iet-gtd.2009.0221

ISSN 1751-8687

Optimal flexible alternative current


transmission system device allocation under
system fluctuations due to demand and
renewable generation
S.J. Galloway1 I.M. Elders1 G.M. Burt 1 B. Sookananta2
1
Department of Electronic and Electrical Engineering, The Institute for Energy and Environment, The University of Strathclyde,
204, George Street, Glasgow G1 1XW, UK
2
Department of Electronic and Electrical Engineering, Ubonratchathani University, Ubonratchathani 34190, Thailand
E-mail: stuart.galloway@strath.ac.uk

Abstract: This study proposes two methods for the optimal placement of flexible alternative current transmission
system (FACTS) devices considering variations in demand and renewable generation output. The basic
optimisation technique utilised is the differential evolution algorithm and the objective is to minimise the cost
of generation. The static performance of the FACTS device is considered here. Simulation shows that with
renewable generation present in the network, the system state at peak demand is not always the most
suitable state to use for the determination of the optimal FACTS allocation. From this, techniques based on
the Monte Carlo simulation are proposed to determine the location for which the operation of FACTS device
gives highest benefit in terms of saving cost of conventional generation. These techniques collectively are
called renewable uncertainty-based optimal FACTS allocation techniques. This study shows the effectiveness of
the techniques in the determination of the optimal FACTS placement for networks with a high penetration of
renewable generation.

Nomenclature xij reactance of the line connecting bus i to bus j


xTCSC reactance of the TCSC
CV coefficient of variation
r a random number uniformly distributed
DE differential evolution under [0, 1]
GA genetic algorithm m mean
IEEE- IEEE reliability test system s standard deviation
RTS
P active power
MCS Monte Carlo simulation
Si the state of the generating unit at bus i
OPF optimal power flow
FR failure rate of the corresponding generating unit
RTS- reliability test system for educational purposes
EP p.d.f. probability density function
TCSC the thyristor-controlled series capacitor DT total demand
i, j buses at the ends of the line being monitored GT total generation
c, k observed parameter for Weibull distribution LT total losses
d demand CG cost of electricity generation

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 725
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010
www.ietdl.org

Gi power output of generation at bus i generation technologies. This paper also recommends the
Gimin minimum power output of generation at bus i suitable system state for which the FACTS allocation
solution should be derived. This paper utilises two
Gimax maximum power output of generation at bus i
techniques for determination of the optimal FACTS
MVAij the apparent power flowing along line ij placement considering variations in demand and renewable
MVAmax
ij maximum allowable transaction amount from bus generation output. These techniques are based on the use
i to bus j of Monte Carlo simulation (MCS). The first technique is
Vi bus voltage at bus i for estimating the system state, which the corresponding
optimal solution for FACTS allocation problem would give
the highest annual benefit in terms of saving cost of
conventional generation. The second technique is for
estimating the optimal placement of FACTS devices. This
1 Introduction paper also recommends the suitable system state for which
Recently, transmission line utilisation has increased the FACTS allocation solution should be derived.
substantially because of the marketing of power from
merchant (non-utility-owned) power plants under
deregulated environments. When the transmission capacity
2 Background
is constrained, an electrical utility must offer to enlarge its The numerical example in this paper is only for the thyristor-
transmission capacity to provide transmission services under controlled series capacitor (TCSC). However, the discussion
the rules of certain regulators [1]. However, obtaining and methods presented in this paper can be applied to
approval to site and build new transmission capacity is FACTS devices in general. The following sections provide
becoming more difficult due to the social and especially some background to the problem domain.
environmental concerns. Therefore increasing transmission
capability along existing routes and making maximum use
of existing transmission system assets through upgrades is
2.1 FACTS device modelling
an attractive alternative, and the use of flexible alternative The configuration of a typical TCSC from a steady-state
current transmission system (FACTS) devices is one of the perspective is the fixed capacitor with a thyristor-controlled
options available. reactor [6] as shown in Fig. 1. The total equivalent
impedance of the compensator, xTCSC is a function of the
To determine the optimal allocation of FACTS devices in capacitive and inductive reactances and the firing angle of
order to obtain the highest possible benefit from them, a thyristor valves. For simplicity, the compensation of TCSC
suitable optimisation technique should be utilised. A variety for the line connecting bus i to bus j is represented as
of optimisation techniques, including linear programming shown in Fig. 2.
[2], sequential quadratic programming [3], Tabu search [4]
and particle swarm optimisation [5] have been applied to The control limits on the thyristor firing angle are
the allocation problem. Some of these methods require converted and simply represented by line compensation
modification of power flow equations and are applied to limits. To not over compensate, the maximum value of the
convex and continuous functions. With more flexible capacitance is fixed at 70% compensation and the
characteristics, heuristic optimisation methods can be maximum value of the inductance is fixed at 20%
applied more generally. A frequently used technique for the compensation [6, 10]. These limits are taken as one of the
FACTS placement problem is the genetic algorithm (GA), constraints for the TCSC allocation problem.
which is an evolutionary optimisation technique that has
been shown to be able to accommodate the problem
complexity [6, 7] even though they often perform slowly.
2.2 Optimisation procedure – DE
An evolutionary optimisation technique, called differential DE is an evolutionary optimisation technique. The key parts
evolution (DE), is employed in this paper. This technique of the method are a scheme for generating trial parameter
can be used to provide a faster and more computationally vectors and the use of a weighted difference between two
tractable solution compared to the GA [8]. population vectors to form a third candidate vector. The
concept of the technique is illustrated in Fig. 3 as presented
There is an increasing trend to replace existing thermal in [8].
generation capacity with low carbon sources, including
renewable generation [9]. When some types of renewable
generation are introduced to the power system, the problem
of intermittency is also added. This paper investigates the
problem of allocation of FACTS devices while taking
account of system state variability due to uncertain demand
and intermittent renewable generation. Wind generation is
selected in this paper to represent the impact of renewable Figure 1 TCSC configuration

726 IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725– 735
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-gtd.2009.0221
www.ietdl.org

with better fitness values are entered to the new population.


The best member with smallest (largest for maximisation)
fitness value is maintained and updated. The procedure
is repeated until a stopping criterion is met. For
minimisation, the searching process would stop when the
best fitness found is less than a target value or it would
Figure 2 Steady-state equivalent of TCSC stop when the number of iterations exceeds a specified
number.

Open source code of the DE [8] is utilised in this study. In this paper, the source code [8] is modified to stop the
The searching process starts with a random initial solution space search when the best solution found within a
population for which the fitness value for each population specified number of iterations is changed less than a
member is then evaluated. For optimal minimisation, the specified tolerance. The number of iterations for each
population member with the smallest fitness value is evaluation of change in best solutions is found to be 40 and
recorded. Three members are then randomly picked from tolerance is 0.01. The cross-over fraction (probability) and
the initial population. The operation of the DE algorithm population size are 0.5 and 20, respectively. These
is used to introduce mutation into subsequent populations particular DE parameters are selected after a number of
by adding the weighted difference between two members different empirical trials and give better performance
(population vectors) to form a third vector (base). The against simulation test problems.
resultant population is called a mutant population. The old
population members are replaced by the random mutant
population members at a rate based on a cross-over fraction
2.3 Monte Carlo simulation
(probability). The fitness values of the new candidate MCS is the general designation for stochastic simulation
population members obtained after cross-over are then using random numbers and can be used to solve not only
compared with those of the old populations and the ones stochastic but also deterministic problems [11]. It can be

Figure 3 Differential evolution concept

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 727
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010
www.ietdl.org

used to develop the system state including demand and wind † Simulation stops when CV is less than the pre-specified
generation output for each trial through random sampling. In tolerance value.
this paper, it is used because a stochastic output of the wind
generation is involved. † Simulation pauses at a given number of samples and it is
checked to see if CV is acceptable. If not, the number of
2.3.1 Demand estimation: To generate random samples can be increased.
demand values, a normal distribution is used to represent
the probability density function (p.d.f.) of demand (d ). In this paper, MCS is used to develop the hourly state of
Hence, the mean (m) and standard deviation (s) of the set the system, including demand and wind generation output
of demand values in a year are required to represent annual for each trial through random sampling. It is recognised
demand profile. The normal distribution is presented as that the availabilities due to outages of the conventional
generators and other components such as transmission lines
  have important stochastic characteristics, but they have not
1 (d − m)2
f (d ) = √ exp − (1) been included at this stage of the research as the non-
s 2p 2s2 network elements are the focus. Future research will see
network determinants brought to this problem, with the
2.3.2 Wind generation output estimation: To improved problem characterisation and the accompanying
generate random values of the wind generation output, a computational overhead.
distribution function is utilised, assumed to be the same
form as the wind speed distribution. The Weibull
distribution is commonly used to model wind speed [12]. 3 Problem formulation
Since all wind generation capacity is aggregated at a single
In the regulated or power pool-based power system, the
network node in the studies presented in this paper, a
system is operated with optimal generation dispatch in
Weibull distribution can therefore be used to represent
which the cheapest possible production is obtained within
wind generation output without further consideration of
standard system security and without violation of circuit
geographical diversity effects. Thus, the observed
flow and voltage constraints. As the cost of production is
parameters (c and k) for Weibull distribution are required.
one of the key considerations by the system operator, it is
Note that the wind generation provides a large number of
also suitable to be used as an objective function for the
zero outputs which in this paper, are regarded as plant
optimal FACTS placement problem. In this case, cost of
outage. The Weibull distribution is
system losses is included in the cost of generation.
  k−1   
k P P k Under steady-state conditions, when line flows reach the
fw (P) = exp − (2)
c c c thermal limit or causes violation of the voltage stability
limit, more expensive generation is dispatched to keep the
By using MCS, energy demand and wind generation network away from these constraints. The FACTS devices
output are randomly drawn by using their distribution would be used to relieve these constraints and consequently
characteristics. The availability of a wind generating unit result in a lower cost of operation as the system is operated
can be estimated based on corresponding failure rates (FRs) in the remaining expanded constraints. To evaluate this
[11]. First, a random number is generated uniformly. If it benefit, cost of generation can be used. Thus, the problem
is less than the unit FR, the generating unit is inactive. In to be solved is to locate FACTS devices in the network so
contrast, if it is greater than the FR, the unit is active. as to minimise the cost of generation. An optimally located
device will, therefore, allow the network to be operated
The state of the generating unit at bus i is closer to the economic dispatch. Formally, the problem can
be stated as follows

0, if r , FR
Si = (3) min. f = CG (P) (4)
1, if r . FR
such that GT = DT + LT (5)
The estimated demand and wind generation output are
taken for system states that have a high likelihood of MVAij ≤ MVAmax
ij (6)
occurrence. Optimal placement of FACTS in these states is
expected to see the highest savings benefits across the full Gimin ≤ Gi ≤ Gimax (7)
range of systems states.
Vimin ≤ Vi ≤ Vimax (8)

2.3.3 Convergence criterion: The coefficient of 0.3xij ≤ xij − xTCSC ≤ 1.2xij (9)
variation (CV) is often used as the convergence criterion of
the MCS. The following two stopping rules can be utilised The constraints (5) – (8) are real power balance, branch
with normalised mean and standard deviation: thermal limits, generation limits and voltage stability,

728 IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725– 735
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-gtd.2009.0221
www.ietdl.org

respectively. Constraint (9) is on FACTS device operation, Table 2 Generation and cost functions of the modified
presented in terms of rated capacity, as discussed in Section RTS-EP
2. The numerical example will be given for only series
compensation, and hence, the limit shown here is the Buses 1 2 3 4 5 6
sizing limit of the series compensator device which is Pmax (MW) 130 170 0 170 0 20.3
allowed to operate both in inductive and capacitive regions.
The solution technique for the problem is presented in the Qmax (MVAr) 75 87 0 106 0 210.25
next section. Qmin (MVAr) 243 250 0 257 0 210.25
cost function 70P 15P 0 15P 0 0
4 Hourly optimisation of FACTS
placement with varying demand
and generation only 2987 h from the total 8760 h that the TCSC provides
compensation and only 639 h that the TCSC provides
This section provides an initial set of case studies for system large cost savings of generation.
state that should be used to obtain the solution with the
corresponding highest annual benefit. The hourly optimal From Fig. 4, the most frequently found solution is branch
allocations of TCSC over a year are determined. The 9. From a study, the optimal location for the peak demand
optimisation problem is performed to achieve the minimum state of the system is branch 7. Note that for this case, at
cost of optimal active power generation as explained in this peak loading state, there is no output from the wind
Section 3. However, the benefit or saving which is a generating unit. Even though branch 9 is the most
difference of the generation cost before and after placing frequently found solution, TCSC at branch 7 gives higher
FACTS device is considered in this numerical example. annual benefits in terms of cost saving of generation.
Normalised demand and wind generation profiles given in Specifically, at branch 7, TCSC provides about £0.7m while
[13] are used to adjust the peak demand and wind generating at branch 9, it provides about £64. Therefore in this case,
output obtained from test data. The optimal FACTS branch 7 should be taken as an optimal allocation for the
allocation is performed by using the DE algorithm in which TCSC. From this study, the device often operates optimally
the MATPOWER 3.0.0 [14] is used to solve the optimal in capacitive region and mostly near its maximum rating.
power flow (OPF) problem and returns the fitness to the DE.

4.1 Case study I 4.2 Case study II


The hourly optimal placements of a series compensator over a
The hourly optimal placements of a series compensator over a
year are determined for the modified IEEE 14-bus test
year are determined for the modified reliability test system for
system [13] and a histogram of the solutions is shown in
educational purposes (RTS-EP); the network data presented
Fig. 5. Wind generation is added on bus 14 with 34 MW
in [15] is modified by adding conventional generation to bus
peak output at 0.8944 pf lagging on the initial state. The
4 and wind generation to bus 6. Demand is increased to stress
cost functions of generation at bus 1, bus 2 and other buses
the system and is given in Table 1. Generation output and
are given as
cost functions are given in Table 2. The demand profiles
are assigned for the system loads as shown in Table 3.
C(PG1 ) = 0.2P 2 + 50P + 1500 (10)
A histogram of the optimal location for TCSC over a year
is shown in Fig. 4. This test network is operated under C(PG2 ) = 0.1225P 2 + 20P + 4000 (11)
unconstrained merit order except when the demand is peak
or near peak. From the simulation, under high demand, C(PG3,G6,G8 ) = 0.085P 2 + 12P + 7300 (12)
the output from wind generation at bus 6 helps to reduce
system congestion. Therefore TCSC operates and gives where PGi is the power output (MW) of the ith generator.
benefits in terms of saving cost of generation only when The demand profiles are assigned for the system loads as
demand is near peak and the output from wind generation
shown in Table 4. The network is stressed by increasing
is too small to release transmission congestion. There are load as shown by the values presented in Table 5.

From Fig. 5, the first and second most frequently utilised


Table 1 Peak demand of the modified RTS-EP locations of a TCSC for this test system are branch 10 and
branch 15, respectively. Annual cost saving from TCSC at
Buses 1 2 3 4 5 6
branch 15 is approximately £0.8m. At branch 10, the
MW 0 50 102 128 76 50 TCSC provides approximately £1.8m, which is the highest
annual benefit in terms of generation cost saving. This
MVAr 0 10 20.4 25.6 15.2 10
location is also optimal for the peak loading state.

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 729
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010
www.ietdl.org

Table 3 Type of demand for the modified RTS-EP

Type of demand Domestic economy Domestic unrestricted Commercial Industrial


Buses 2 6 4, 5 3

Therefore for this case, branch 10 should be selected for the £53 498.45, respectively. Therefore by considering the
installation of the device. saving values, TCSC on branch 16 gives the highest saving
and it should be taken as the optimal solution.

4.3 Case study III From case studies, the annual optimal allocations of the
The hourly optimal placements of a series compensator are TCSC for the modified IEEE-14 bus and the modified
further determined for the modified IEEE-RTS, for which IEEE-RTS are the locations that are most frequently found
network data is given in [13]. The system has 24 buses and over year. These locations may also be optimal for peak
38 branches. It is modified by adding wind generation at demand state depending on the renewable generation output
bus 4. The installed capacity of the wind generation is at at the state. However, for the modified RTS-EP the optimal
12% of total network demand. This is found to be the location of TCSC is the optimal location found under the
maximum capacity of wind generation that can be accepted peak demand period, not the most frequently found over year.
at this node without causing system instability in this case. This small test case has different features from the modified
The demand profiles are assigned for the system loads as IEEE test cases. It has limited number of conventional
shown in Table 6 and the histogram of optimal solutions generation connections and the wind generation is located
found in a year are shown in Fig. 6. such that its output helps reducing network congestion
during the peak period. This network normally operates with
From Fig. 6, branch 16 is the most frequently found merit order, rather than allowing self-dispatching. For this
solution and branch 23 is the second most frequently found case, the FACTS device provides benefits only during the
solution. At the optimal location, the device usually peak period with low output from wind generation.
operates in the capacitive region for almost all states. In
this case, at the peak demand state the fraction of output
from wind generation is 100% and the optimal location of 5 Proposed techniques based on
TCSC for this test network is branch 4. Fig. 6 indicates MCS
that bus 16 is the most likely location for highest cost
savings, but each of the buses is likely with a decreasing From the studies in Section 4, the peak demand state is not
probability. Therefore the annual savings when the device always suitable for implementation with the FACTS
is optimally operated at each location are determined. allocation solver, as it does not provide the optimal
solution. Techniques based on MCS are proposed to avoid
The generation cost savings obtained from the TCSC the calculation for every changing state of the system in the
when it provides optimal compensation on branches 4, 16 optimal FACTS allocation problem. This section proposes
and 23 are calculated. It is found that the annual saving two techniques based on MCS for determination of the
obtained from branch 16 is £66 538.81. The annual savings FACTS device allocation considering variations in demand
obtained from branch 4 and 23 are £2147.15 and and renewable generation output. The first technique is to
estimate the system state and determine the solution of the

Figure 4 Histogram of the hourly optimal TCSC allocations Figure 5 Histogram of the hourly optimal TCSC allocations
in a year for the modified RTS-EP in a year for the modified IEEE 14-bus test system

730 IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725– 735
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-gtd.2009.0221
www.ietdl.org

Table 4 Type of demand for IEEE-14 bus test system

Type of demand Domestic economy Domestic unrestricted Commercial Industrial


Buses 6, 13, 14 5, 10, 11, 12 3, 4 2, 9

Table 5 Demand for IEEE-14 bus test system

Buses 2 3 4 5 6 9 10 11 12 13 14
MW 56 245 125 19 29 75 24 9 16 35 40
MVAr 11.2 45 225 3.8 5.8 15 4.8 1.8 3.2 7 8

Table 6 Type of demand for the modified IEEE-RTS

Type of demand Domestic economy Domestic unrestricted Commercial Industrial


Load 1, 16, 20 2, 4, 5 3, 6, 7, 14, 19 10, 13, 15, 18

problem for this estimated state. This technique is called the Step 2: Determine c and k parameters of Weibull distribution
renewable uncertainty-based optimal FACTS allocation for the yearly wind generation profile.
estimated state (RUFA-ES). The second technique is to
estimate the solution of the problem and is called the Step 3: Generate random load and wind generation output by
renewable uncertainty-based optimal FACTS allocation using the profile characteristic from steps 1 and 2.
estimated FACTS location (RUFA-EL). The proposed
techniques are tested on the case studies used in Section 4 Step 4: Run OPF; if it converges, record the random state. If
and compared with the optimal locations found in Section 4. it does not converge, let the OPF constrain the output from
wind generation and record the state.
The following subsections provide algorithm for the
proposed techniques and tested results. Step 5: Repeat steps 3 and 4 for n iterations and check if CV
for random values are less than the pre-specified tolerance
value.
5.1 MCS for an estimated system state to
be utilised in the optimal FACTS allocation Step 6: Determine the average load and wind generation and
problem take as an estimated state.
The algorithm for this proposed technique is as follows:
Step 7: Determine optimal TCSC placement for the
Step 1: Determine mean and standard deviation of yearly load estimated state from step 6.
profile.
The number of iterations in step 5 can be selected by
considering CV of the random values in step 3. For the
numerical examples in this paper, the initial number of
iterations is empirically chosen to be 100, 500 and 1000. In
this numerical study, it is checked that the chosen number
provides an acceptable CV. The optimal FACTS
allocations for three case studies are shown in Table 7.

From Table 7, by using this proposed technique, the


optimal allocations of FACTS devices for the modified
IEEE-14 bus and the modified IEEE-RTS are given.
However, this technique cannot provide annual optimal
location of the device for the modified RTS-EP as this
network requires compensation only during peak period
Figure 6 Histogram of the hourly optimal TCSC allocations with low output of wind generation. Therefore this
in a year for the modified IEEE-RTS proposed technique is not recommended to be used for the

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 731
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010
www.ietdl.org

Table 7 Optimal allocation of TCSC at the estimated system state

Cases MC
100-iteration 500-iteration 1000-iteration
Location Capacity (MVAr) Location Capacity (MVAr) Location Capacity (MVAr)
I 9 70 9 69.42 9 50.28
II 10 69.36 10 70 10 69.98
III 16 69.85 16 67.23 16 66.07

Table 8 Average difference of CV for demand and wind generation output

MC iterations Cases
Modified IEEE-14 bus Modified IEEE-RTS
Demand Wind generation Demand Wind generation
100 0.0662 0.042 0.0511 0.1601
500 0.0651 0.0925 0.0536 0.146
1000 0.0582 0.1825 0.0543 0.1321

network, which usually operates in unconstrained merit networks, the 100-iteration MC provides least different
order, that is, one with a centralised dispatch based values for wind generation. Therefore higher MC iteration
predominantly on cost. It is recommended for the network does not always provide better performance.
with usually constrained dispatch operation.
For the random wind generation output, the difference
The average values of the CV for demand and wind between CV of the data and the CV of the randomly
generation output from 100, 500 and 1000 iterations generated value cannot be used to consider whether the
Monte Carlo (MC) for the modified IEEE-14 bus and the MC provides a good estimation. As sometimes the random
modified IEEE-RTS are given in Table 8. output of wind generation causes the OPF to fail to converge,
the recorded wind generation which is used to determine the
For the demand, the value of CV is more consistent and as estimated state is determined from the OPF. However, in
such the MCS represents well the characteristic of demand. this case the proposed technique does provide the
From Table 8, although the 1000-iteration MC provides optimal placements of FACTS device for both test cases II
least different values for CV of demand across the two and III.

Figure 7 Histogram of the optimal TCSC allocation for the Figure 8 Histogram of the optimal TCSC allocation for the
modified IEEE 14-bus test system, determined by using modified IEEE-RTS, determined by using the MCS-based
the MCS-based technique with criterion 1 technique with criterion 1

732 IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725– 735
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-gtd.2009.0221
www.ietdl.org

Figure 9 Convergence of CV for the optimal TCSC allocation


a For modified IEEE-14 bus test system
b For modified IEEE-RTS, determined by using the MCS-based technique with stopping criterion 1

5.2 MCS for estimated optimal location Alternatives of stopping criteria are:
of FACTS devices Criterion 1. The number of simulations reaches a target value.
The algorithm for this proposed technique is as follows:
Criterion 2. The number of identical solutions (i.e. same
location) reaches a target value.
Step 1: Input p.d.f. of demand at each bus and wind
generation output; By using RUFA-EL with stopping criterion 1, a possible
target can be that the acceptable CV is achieved from the
Step 2: Generate random demand and wind generation random states in step 2. For this numerical example, the
output; target number of simulations is set to be 100. It is found
that in these case studies a value of 100 provides an
Step 3: Determine the optimal location and rating of the acceptable CV. The number of solutions found for each
FACTS device for the random state; location is shown in Fig. 7 for the modified IEEE 14-bus
test system and in Fig. 8 for the modified IEEE-RTS. The
Step 4: If the OPF of a random state does not converge for most frequently found solutions for both the test cases are
every trial of FACTS device location and rating within the same as the solutions found for the corresponding cases
a specified maximum generation of the optimisation in Section 4. Similar to the RUFA-ES, the RUFA-EL
technique, ignore this state and repeat step 1 to step 3 again; technique proposed in this paper is not recommended for
networks with largely unconstrained operation.
Step 5: Repeat steps 1 – 4 until stopping criterion is met.
Convergence of CV for the estimated location of the TCSC
using criterion 1 is given in Fig. 9a for the modified IEEE-14

Figure 10 Histogram of the optimal TCSC allocation for the Figure 11 Histogram of the optimal TCSC allocation for the
modified IEEE-14 bus test system, determined by using the modified IEEE-RTS, determined by using the MCS-based
MCS-based technique with criterion 2 technique with criterion 2

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 733
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 12 Convergence of CV for the optimal TCSC allocation


a For modified IEEE-14 bus test system
b For modified IEEE-RTS, determined by using the MCS-based technique with criterion 2

bus test system and in Fig. 9b for the modified IEEE-RTS. usually operated under constrained dispatch, the studies on
The convergence shows that the number of MCS iterations test networks show that a FACTS device gives benefit not
is acceptable. Tolerance for convergence is 0.0001. By using only when the demand is at peak or near peak but also at
RUFA-EL with stopping criterion 2, a target number of all demand levels. Optimal placement of the device is
identical solutions for the FACTS location is initially set to determined for every hour in a year, and it is found that
be 50. When the target is reached, CVs of the random the operation of FACTS device at the most frequently
values are checked for acceptability. If not acceptable, the found location gives higher benefit than at the optimal
target number will be increased. However, for these case location found for peak demand. This conclusion is for
studies the CVs are found to be acceptable when the networks not operated with constrained dispatch which is
frequency of the most frequently found solution reaches 50. common in practice.
The simulation results for the modified IEEE14-bus test
system and for the modified IEEE-RTS are shown in Techniques based on MCS are proposed to avoid
Figs. 10 and 11, respectively. calculation for every possible state of the system in the
optimal FACTS allocation problem. The RUFA-ES
The optimal locations for both test systems obtained using technique first estimates the system state which should be
these proposed techniques are the same and are about the used and the RUFA-EL technique is then used to estimate
same as those found in Section 4. Convergence of CV for the placement. It is found that the solution obtained by
the estimated location of the TCSC by using criterion 2 is using these techniques is the solution that gives the highest
given in Fig. 12a for the modified IEEE-14 bus test benefit in comparison to the uncompensated case.
system and in Fig. 12b for the modified IEEE-RTS. The
convergence shows that the number of MCS iterations is
acceptable. Tolerance for convergence is 0.0001. From the 7 References
tests, these proposed techniques can provide optimal
allocations of FACTS devices for the considered test cases. [1] FULDNER A.H.: ‘Upgrading transmission capacity for
wholesale electric power trade’, a feature article [Online].
Available at: http://www.eia.doe.gov/cneaf/pubs_html/
feat_trans_capacity/w_sale.html, accessed 2010
6 Conclusions
The increasing deployment renewable generation makes the [2] DE OLIVEIRA E.J., MARANGON J.W., LIMA , DE ALMEIDA K.C.:
problem of optimal placements FACTS devices in ‘Allocation of FACTS devices in hydrothermal systems’,
transmission networks more challenging. The original IEEE Trans. Power Syst., 2000, 15, (1), pp. 276 – 282
contribution of this paper is to propose two techniques for
determination of the optimal placement of FACTS devices [3] GITIZADEH M. , KALANTAR M.: ‘A new approach for
under uncertainty. Furthermore, the paper has shown that congestion management via optimal location of FACTS
for networks with renewable generation the system state devices in deregulated power systems’. Third Int. Conf.
which must be examined to determine optimal FACTS Electric Utility Deregulation and Restructuring and Power
device location is not necessarily the system demand peak. Technologies, 6 – 9 April 2008, pp. 1592– 1597

This paper examines the system state which should be used [4] MORI H., MAEDA Y.: ‘A hybrid meta-heuristic method for
when determining the optimal location of FACTS device for optimal allocation of UPFCs’. IEEE Int. Symp. on Circuits
networks including renewable generation. For networks and Systems, 24 – 27 May 2009, pp. 1705– 1708

734 IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725– 735
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-gtd.2009.0221
www.ietdl.org

[5] HASHEMZADEH H., HOSSEINI S.H. : ‘Locating series [10] LIE T.T., DENG W.: ‘Optimal flexible AC transmission system
FACTS devices using line outage sensitivity factors and (FACTS) devices allocation’, Electr. Power Energy Syst., 1997,
particle swarm optimization for congestion management’. 19, (2), pp. 125– 134
IEEE Power & Energy Society General Meeting, 26 – 30
July 2009 [11] BILLINTON R., LI W.: ‘Reliability assessment of electric
power systems using Monte Carlo methods’ (Plenum
[6] GERBEX S., CHERKAOUI R., GURMOND A.J.: ‘Optimal location of Press, New York and London, 1994)
multi-type FACTS devices in a power system by means of
genetic algorithms’, IEEE Trans. Power Syst., 2001, 16, (3), [12] CELIK A.N.: ‘Energy output estimation for small-scale wind
pp. 537– 544 power generators using Weibull-representative wind data’,
J. Wind Engng. Ind. Aerodyn., 2003, 91, pp. 693–707
[7] RASHED G.I. , SHAHEEN H.I. , CHENG S.J.: ‘Evolutionary
optimization techniques for optimal location and [13] Power systems test case archive – UWEE- [Online].
parameter settings of TCSC under single line contingency’. Available at: http://www.ee.washington.edu/research/
IEEE Power & Energy Society General Meeting, 20–24 July pstca/, accessed 2010
2008
[14] ZIMMERMAN R.D. , MURILLO-SÁNCHEZ C.E., GAN D. :
[8] STORN R., PRICE K. : ‘Differential evolution – a simple ‘MATPOWER: A MATLAB power system simulation
and efficient heuristic for global optimization over package’, [Online]. Available at: http://www.pserc.cornell.
continuous spaces’, J. Glob. Optimisation, 1997, 11, (4), edu/matpower/, accessed 2009
pp. 341– 359
[15] BILLINTON R., KUMAR S., CHOWDHURY N., ET AL .: ‘A reliability
[9] Department of Energy and Climate Change, Digest of test system for educational purposes – basic data’, IEEE
UK Energy Statistics, DECC, London, 2009 Trans. Power Syst., 1989, 4, (3), pp. 1238 – 1244

IET Gener. Transm. Distrib., 2010, Vol. 4, Iss. 6, pp. 725 – 735 735
doi: 10.1049/iet-gtd.2009.0221 & The Institution of Engineering and Technology 2010

S-ar putea să vă placă și