Electric Power Systems Research 80 (2010) 723732

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Daily Hydrothermal Generation Scheduling by a new Modied

Adaptive Particle Swarm Optimization technique
Nima Amjady , Hassan Rezai Soleymanpour
Department of Electrical Engineering, Semnan University, Semnan, Iran

a r t i c l e

i n f o

Article history:
Received 11 June 2009
Received in revised form
18 September 2009
Accepted 12 November 2009
Available online 23 December 2009
Keywords:
Hydroelectric system
Thermal plant
DHGS
MAPSO

a b s t r a c t
The fundamental requirement of power system hydrothermal scheduling is to determine the optimal
amount of generated powers for the hydro and thermal units of the system in the scheduling horizon
of 1 day or few days while satisfying the constraints of the hydroelectric system, thermal plants and
electrical power system. Daily Hydrothermal Generation Scheduling (DHGS) is a complicated non-linear,
non-convex and non-smooth optimization problem with discontinuous solution space. To deal with this
complicated problem, a new Modied Adaptive Particle Swarm Optimization (MAPSO) is proposed in this
paper. The inertia weight and acceleration coefcients of the PSO are adaptively changed in the MAPSO
owning tree topology. We split-up the cognitive behavior of PSO into the best and not-best parts. The
proposed not-best cognitive component, unlike recent methods, retains its dynamic behavior throughout
the search process. Personal best position exchanging method is proposed to increase activities of particles to explore broad space. New velocity limiter is also proposed in this paper to enhance exploration
capability and convergence behavior of the MAPSO. The proposed MAPSO is tested on six test systems
and compared with some recent research works in the area.
2009 Elsevier B.V. All rights reserved.

1. Introduction
Daily Hydrothermal Generation Scheduling (DHGS) is an
important issue in economical operation of power systems. The
short-term hydrothermal generation scheduling determines optimal hourly water releases of each reservoir in hydroelectric
plants to generate electrical energy for supporting some part of
the power demand. In order to satisfy the rest of the power
demand, which is not supported by hydropower generation, DHGS
schedules thermal generation units so that the total production cost is minimized during the scheduling time horizon. The
main objective is focused on the optimal use of water resources
for minimizing the production cost of thermal plants considering the practical constraints. In DHGS problem, the constraints
are usually divided into three categories related to hydroelectric
system, thermal plants and electrical power system (satisfying
power demand constraint) [1]. Aside from these constraints, the
cascading nature of hydrosystems causes dependency between
the performances of hydropower plants. Also, the impact of
steam valve loading on operational cost curve of thermal units
intensies non-convexity and non-linearity of the DHGS problem. So, DHGS is a complicated non-linear, non-convex and

Corresponding author. Tel.: +98 021 88889096; fax: +98 021 88880098.
E-mail address: amjady@tavanir.org.ir (N. Amjady).
0378-7796/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2009.11.004

non-smooth optimization problem with discontinuous solution

space.
Several methods, such as dynamic programming (DP) [2], network ow [3], decomposition technique [4], mixed integer linear
programming (MILP) [5], and Lagrangian relaxation (LR) [1,6] have
been proposed to solve DHGS problem in the recent years. Among
the existing methods, DP appears to be the most popular despite the
major disadvantage of drastic growth of computational and dimensional requirements with increasing system size and planning
horizon [7]. The network ow model of DHGS is often programmed
as a linear or piecewise linear one. Linear programming typically
considers that power generation is linearly dependent on water discharge, thus ignoring the head change effect, leading to a solution
schedule with less power generation [8]. Handling of the various constraints increases the number of dual variables and the
complexity of the optimization task in the decomposition technique [9]. Also, the discretization of the non-linear dependence
between power generation, water discharge and head, used in the
MILP to model head variations, augment the computational burden
required to solve DHGS problem. The implementation of LR is complicated and its efciency heavily depends on the size of the duality
gap. Furthermore, solution quality of LR depends on the method to
update Lagrange multipliers [8]. As a result, conventional methods
require models of hydro as well as thermal plants to be represented
as piecewise linear or polynomial approximations of monotonically
increasing nature. However, such an approximation may lead to a

724

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

suboptimal solution, resulting in huge loss of revenue over the time.

Hence, the trend in recent times is to use more realistic models for
hydro and thermal plants [7].
Recently the articial intelligence (AI) methods such as genetic
algorithm [10,11], neural network [12], simulated annealing [13],
differential evolution [14], cultural algorithm [15], evolutionary
programming [7], and particle swarm optimization [1618] have
been presented for the solution of DHGS problem to overcome
these deciencies. Moreover, some combinatorial techniques such
as coevolutionary algorithm (CEA) based on the Lagrangian method
[19] and hybrid simulated annealing-genetic algorithm [20] have
been also proposed to solve this problem. However, the quality
of solution is a concern for the articial intelligence based methods as these methods may trap in local minima or even infeasible
solutions for the complex problem of DHGS. Besides, dependency of these methods on the correct tuning of their adjustable
parameters and initial point is another problem of these methods.
To mitigate the imperfections of the previous AI methods for the
solution of DHGS problem, a new Modied Adaptive Particle Swarm
Optimization (MAPSO) is proposed in this paper. Novel operators
and techniques are incorporated in the MAPSO to enhance its exploration and exploitation capabilities for the solution of complex
non-linear non-convex DHGS problem. Also, the proposed MAPSO
has low sensitivity with respect to both its adjustable parameters
and initial point.
The remaining parts of the paper are organized as follows. In
the second section, the DHGS problem is formulated. The proposed
MAPSO is introduced in the third section. Application of the MAPSO
to solve the DHGS problem is presented in the form of a step by step
algorithm in Section 4. Obtained numerical results are presented
and discussed in Section 5. Section 6 concludes the paper.
2. Problem formulation
2.1. Objective function
Due to negligibility of hydropower production cost, the main
objective of the DHGS is to assess optimal water releases of
hydroreservoirs and power generation of each thermal unit so that
total thermal fuel cost (objective function OF) becomes minimized
[2,21]:
Minimize : OF =

Ns
T 


[fi (Psi,t )]

(1)

t=1 i=1

It is assumed that fi (Psi,t ) is a quadratic function of Psi,t :

2
fi (Psi,t ) = asi + bsi Psi,t + csi Psi,t

(2)

In practice, the fuel cost function has non-differentiable points

due to valve loading effects. To consider these effects of units, a
recurring rectifying sinusoidal term is added to the quadratic cost
function as follows [22]:
2
fi (Psi,t )=asi + bsi Psi,t +csi Psi,t

+ |dsi sin(esi (Psimin

Psi,t ))|

Phj,t is considered to be a polynomial function of water discharge

rate and reservoir storage volume [7,8,21,23]:
Phj,t = Ch1j Vh2j,t + Ch2j Qh2j,t + Ch3j Vhj,t Qhj,t + Ch4j Vhj,t + Ch5j Qhj,t
+ Ch6j

(6)

A more complex function for hydropower generation is presented in [1]. Our proposed solution algorithm (MAPSO) can
consider this hydropower generation function as well, provided
that its data is available.
2.2.2. Prohibited operating zones (POZs)
In practice, thermal units can have prohibited operating regions
as discussed in [22]. The power generations for these units should
be in the non-forbidden operating zones. So, the power generation
for ith thermal unit considering POZ is modeled as follows:

Psimin Psi,t PsiLB,1

PsiUB,k1 Psi,t PsiLB,k

PsiUB,k Psi,t Psimax

k = 2, 3, . . . , NOi

2.2.3. Ramp rate limits

Ramp rate limits of ith thermal generating unit can be described
as:

Psi,t Psi,t1 URi

Psi,t1 Psi,t DRi

if Psi,t Psi,t1
if Psi,t1 Psi,t

(8)

Combining (5) and (8), the ramp rate constrained operating limits of thermal units can be stated as follows:
rmax
Psi,t
= min{Psimax , Psi,t1 + URi }

(9)

rmin
Psi,t
= max{Psimin , Psi,t1 DRi }

(10)

rmin
rmax
Psi,t Psi,t
Psi,t

(11)

2.2.4. Water discharge rate limit

Considering prohibited discharge zones (PDZ) of hydro units [7],
the following constraints for Qhj,t should be imposed:

min
LB,1
Qhj Qhj,t Qhj

QhUB,m1
Qhj,t QhLB,m
j
j
QhUB,m
Qhj,t Qhmax
j
j

m = 2, 3, . . . , NDj

It is noted that thermal unit i with NOi POZs and hydro unit j with
NDj PDZs will have NOi + 1 and NDj + 1 disjoint operating regions,
respectively. These disjoint regions form a non-convex set [22].
2.2.5. Reservoir storage volume limit
Vhmin
Vhj,t Vhmax
j
j

(13)

2.2.6. Water dynamic balance [21]

Vhj,t = Vhj,t1 +Ihj,t Qhj,t Shj,t +

UPj


(Qhr,tr,j + Shr,tr,j )

Phj,t Phmax
Phmin
j
j

(4)

Psimin Psi,t Psimax

(5)

(14)

r=1

2.2.7. Final volume constraint [14]

Vhj,T = Vhj,end

2.2.1. Power generation limits

Each power generation unit has generation capacity boundaries:

(12)

m = NDj

(3)

2.2. Constraints

(7)

k = NOi

(15)

2.2.8. Active power balance

Ns

i=1

Psi,t +

Nh


Phj,t PDt PLt = 0

j=1

PLt can be calculated by the B matrix loss formula [21,22].

(16)

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

725

topology is based on breadth-rst search in which an individual

can move down several levels in the hierarchy but it can move up
at most one level in each iteration [27]. Our proposed tree topology
for the APSO is shown in Fig. 1(b), wherein each level of the tree
has only one node or one particle. In each iteration, all particles are
sorted from the top of the tree to down, based on their Pbest values.
In this topology, each particle can move up and down by any number of levels in each iteration and so the hierarchy is updated in one
iteration. Based on the tree topology, (17) becomes as follows:

Fig. 1. (a) General tree topology and (b) proposed tree topology.

best
Vi,iter+1 =wVi,iter + c1 r1 (Pi,iter
Xi,iter ) + c2 r2 (PARi,iter Xi,iter ) (22)

3. The proposed MAPSO

PSO was introduced by Kennedy and Eberhart as a modern
heuristic optimizer [24]. PSO is a population based search method
which deals with random particles in search space. The particles, i.e.
trial solutions of the optimization problem, share their information
with each other and run toward best trajectory to nd optimum
solution in an iterative process. A velocity vector is dened for
each particle and particle position depends on this velocity. In each
iteration, the velocity and position of particles are updated:
best
best
Vi,iter+1 = wVi,iter + c1 r1 (Pi,iter
Xi,iter ) + c2 r2 (Giter
Xi,iter )

(17)

Xi,iter+1 = Xi,iter + Vi,iter+1

(18)

where Vi,iter and Xi,iter represent the velocity vector and the position
best and Gbest
vector of ith particle at iteration iter, respectively; Pi,iter
iter
are personal best position of ith particle and global best position of
swarm until iteration iter, respectively; w is inertia weight factor
which controls the global and local exploration capabilities of particles; c1 and c2 are cognitive and social coefcients, respectively;
r1 and r2 are two random numbers between 0 and 1. To enhance
the efciency of the PSO, we adjust the inertia weight w to linearly
reduce during the iterations [25]:
w = (wmax wmin )

 iter

max

iter

itermax

+ wmin

(19)

where itermax is the maximum number of iterations. Also, to efciently control the local search and convergence to the global
optimum solution, time-varying acceleration coefcients (TVAC)
are introduced in addition to the time-varying inertia weight
(TVIW) in PSO [26]. A large cognitive component and small social
component at the beginning, allows particles to move around the
search space, instead of moving towards the population best prematurely. During the latter stage in optimization, a small cognitive
component and a large social component allow the particles to converge to the global optima [25]. So, the acceleration coefcients are
adaptively changed as follows [25,26]:
final

c1 = (c1

final

c2 = (c2

c1initial )
c2initial )

 iter 
itermax

 iter 
itermax

final

+ c1initial ,

c1

+ c2initial ,

c2

final

< c1initial

(20)

> c2initial

(21)

where the superscripts initial and nal indicate the initial and
nal values of the acceleration coefcients, respectively. We consider both TVIW as stated in (19) and TVAC as represented in (20)
and (21) in the PSO and call it Adaptive PSO or APSO. Moreover, we
used a specic kind of tree topology for the APSO. In tree topology,
all particles are arranged in a tree and each node of the tree contains
exactly one particle [27] as shown in Fig. 1(a). A particle is inuenced by its own best position so far (Pbest ) and by the best position
of the particle that is directly above in the tree (parent). If a particle
at a child node has found a solution that is better than the best so
far solution of the particle at the parent node, the both particles are
exchanged. In this way, this topology offers a dynamic neighborhood, which enhances the search ability of the PSO. However, tree

best is replaced by PAR

where Giter
i,iter indicating the best position so
far of the parent (directly above particle) of ith particle at iteration
iter.
For the APSO with the tree topology of Fig. 1(b), three new
modications are introduced to construct a novel heuristic search
technique called Modied APSO or MAPSO. These new modications, enhancing the exploration and convergence capabilities of
the proposed method, are described in the following, respectively.

3.1. The split-up of the cognitive part into the best and not-best
components
Selvakumar and Thanushkodi [28] proposed a split-up in the
best and P worst that changed (17) as
cognitive behavior of PSO into Pi,iter
i,iter
follows:
best
worst
Vi,iter+1 = wVi,iter + c1b r1 (Pi,iter
Xi,iter ) + c1w r2 (Xi,iter Pi,iter
)
best
Xi,iter )
+ c2 r3 (Giter

(23)

best is exactly the same as the cognitive component of the

where Pi,iter
worst is personal worst position of ith particle until
basic PSO and Pi,iter
iteration iter; r1 , r2 and r3 are random numbers between 0 and 1.
That is, the particle is made to remember its worst position also.
This modication helps to search the solution space more effectively compared with the classical PSO and has also been used
in some other research works such as [8]. However, our experience with DHGS problem shows that after a few initial iterations,
worst of ith
the swarm particles become better and better and so Pi,iter

worst
particle approximately remains unchanged. In other words, Pi,iter
nearly becomes a static limit and loses its dynamic behavior. So,
the expected enhancement in the exploration capability of the PSO
may not be obtained. To solve this problem, we propose the new
idea of decomposing the cognitive part into the Best and Notbest components instead of splitting up into the Best and Worst
components as follows:
best
not-best
Vi,iter+1 = wVi,iter + c1b r1 (Pi,iter
Xi,iter ) + c1nb r2 (Xi,iter Pi,iter
)

+ c2 r3 (PARi,iter Xi,iter )

best = X
Pi,iter
i,iter ,

best
if AOF(Xi,iter ) < AOF(Pi,iter1
)

best = P best
,
Pi,iter
i,iter1

best
if AOF(Xi,iter ) AOF(Pi,iter1
)

not-best = X
Pi,iter
i,iter ,

best
if AOF(Xi,iter ) AOF(Pi,iter1
)

not-best = P not-best ,
Pi,iter
i,iter1

best
if AOF(Xi,iter ) < AOF(Pi,iter1
)

(24)

(25)

(26)

where AOF is the augmented objective function of the optimization

problem that should be minimized. AOF includes both the objective function and penalty terms. For the DHGS problem, AOF will
best
be introduced in the next section. Equation (25) describes Pi,iter
according to basic PSO. In any iteration that the particle cannot
nd a solution better than its best solution until the previous iterabest
not-best is replaced by X
tion Pi,iter1
, the Not-best component Pi,iter
i,iter

726

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

not-best is continuously
as shown in (26). In this way, the limit of Pi,iter
updated (even in the last iterations) and saves its dynamic behavior
throughout the search process of the PSO enhancing its exploration
capability. It is noted that the both coefcients of the cognitive
part of (24), including c1b and c1nb (related to the Best and Notbest components, respectively), are adaptively changed along the
iterations of the MAPSO based on (20).

3.2. Personal best position exchanging

When a particle position becomes close to the Gbest of the swarm
(here, parent of the particle or PARi,iter ), its personal best position
Pbest will also become close to Gbest after a few iterations and so the
effect of both cognitive and social parts will decrease. Therefore,
this particle becomes lazy in the swarm degrading the exploration
capability of the PSO. This phenomenon is more seen in the latter
stages after the initial part of the search. To remedy this problem
and increase the activity of swarm particles, we propose the idea of
personal best position exchanging. When two particles exchange
their Pbest , their velocity vectors are suddenly changed based on
(24), which makes the lazy particle to move abroad the search
space. So, the lazy particle is renewed, which enhances the exploration capability of the proposed MAPSO. We inspired this idea from
the crossover operator of genetic algorithms (GA) that swaps a part
of information of two individuals. In each iteration, a number of particles of the swarm are randomly selected and grouped in twofold
clusters, like the crossover operator of GA. Then, Pbest of the two
members of each cluster is exchanged. While the diversity of the
search process is enhanced, no information about Pbest of a particle
is lost in this technique (only it is stored in another particle). It is
noted that this technique is different from Small-World method
proposed for random leader selection [29]. Small-World method
is related to the social behavior and is implementable in some specic PSO topologies. On the other hand, the proposed personal best
position exchanging technique is related to the cognitive part and
can be implemented in any PSO. Indeed, we have never seen this
technique in the previous research works on PSO.
3.3. New velocity limiter
In conventional PSO, an upper limit is placed on the velocity in
all dimensions. This upper limit (velocity limiter) prevents particles from moving too rapidly from one region in search space to
another. The maximum velocity allowed actually serves as a constraint that controls the maximum global exploration ability of PSO
[30]. A xed velocity limiter for each dimension is usually considered as a proportion of the allowable position range. However,
selecting a proper velocity limiter is difcult, especially for complex optimization problems [31]. For the initial stages of the search
process, a high maximum velocity allowed is usually required so
that the particles can search different regions of the solution space.
On the other hand, a low maximum velocity allowed may be preferable for the nal stages of the search process so that the PSO can
better converge. A discussion about this matter can be found in [32]
where it has been concluded that the choice of a suitable value for
the velocity limiter can be nontrivial and also very important for
the determination of the overall performance of the algorithm. To
overcome this problem, some modications have been investigated
to determine proper velocity limiter. For instance, a random velocity limiter was introduced in [33] such that the velocity boundary
alters randomly to prevent the velocity of a particle from stopping on a same boundary during the evolution. However, random
selection of maximum velocity allowed may result in inconsistent
velocity limiters along the search process of the PSO. In [31], the
particles are divided into several groups and the maximum veloc-

ity of each group is adaptively controlled such that the maximum

velocity of the worst group is changed to approach the maximum
velocity of the best group if the difference of goodness between the
best and worst group is big. However, the exploration capability
of PSO along the search process is not explicitly considered in this
approach. A review of different velocity limiters of PSO proposed in
the previous research works in the area can be found in [33]. In this
paper, a new velocity limiter is proposed, which adaptively sets the
maximum velocity allowed along the search process of the MAPSO
as follows:
R = Rinitial + (Rfinal Rinitial )

 iter 
itermax

Rinitial > Rfinal

V j,max = R (X j,max X j,min )

(27)
(28)

where R is a dynamic coefcient that adaptively changes along the

search process; Vj,max is velocity limiter for the jth dimension of
the particles (jth decision variable) that is |Vj | Vj,max ; Xj,min and
Xj,max indicate the minimum and maximum allowable position for
the jth dimension. As seen the velocity limiter linearly decreases
from a higher value to a lower one along the search process. By this
velocity limiter the particles can have high velocities in the initial
iterations to broadly search the solution space. In the nal iterations
the particles velocity is more limited to avoid jumping particles. By
the proposed adaptive velocity limiter, the MAPSO can benet from
both good exploration capability and convergence behavior.
4. Application of the proposed MAPSO to solve the DHGS
problem
Application of the proposed MAPSO to solve the DHGS problem,
formulated in section II, can be summarized as the following step
by step algorithm:
Step 1 initialization. The decision variables for the DHGS problem
are hourly reservoir discharges and thermal generations over the
entire scheduling horizon. So, the structure of the position vector
for the particles of the MAPSO to solve the DHGS problem becomes
as follows
X = [Qh1,1 , Qh2,1 , . . . , QhNh ,1 , Ps1,1 , Ps2,1 , . . . , PsNs ,1 , . . . ,
Qh1,T , Qh2,T , . . . , QhNh ,T , Ps1,T , Ps2,T , . . . , PsNs ,T ]

(29)

The swarm of the MAPSO has Npar particles with the structure shown in (29). The initial swarm of the MAPSO is
generated by the reference unit technique [22,34]. Consider
the part [Qh1,t , Qh2,t , . . . , QhNh ,t , Ps1,t , Ps2,t , . . . , PsNs ,t ] of the
position vector X, including reservoir discharges and thermal generations for tth hour (1 t T). The initial reservoir discharges

are randomly chosen in the respective ranges such that Qhmin

j

Qhj,t Qhmax
(1 j Nh ). The initial generations of Ns 1 thermal
j
units are arbitrary selected in the respective ranges based on (11).
The initial generation of one remaining unit, named reference unit,
is so determined that the active power balance constraint of (16)
is satised [22,34]. This initialization process is repeated for all
T scheduling hours to generate one particle with the structure
shown in (29). In this way, Npar particles of the swarm can be initially generated. Now the velocity vectors of the swarm particles
should be initialized. The following strategy is used for creating
the initial velocities:

j,min

Vi,0 = Rand Xi
j

j,max

Xi,0 , Xi

Xi,0

(30)

where Xi,0 and Vi,0 are the initial position (obtained from the previously mentioned initialization process) and initial velocity for ith

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

j,min

j,max

particle in jth dimension, respectively; Xi

and Xi
indicate
minimum and maximum allowable position for the jth dimension of the ith particle, respectively; Rand[a,b] generates a random
not-best = P best = X
number in the range of [a,b]. Also, Pi,0
i,0 and iter = 0.
i,0
Step 2 Evaluation of the objective function. The fuel cost functions
of the thermal generating units fi (Psi,t ) are evaluated according to
(3) based on the current values of Psi,t . Then, the objective function
of the DHGS problem, i.e. OF, is computed according to (1)
Step 3 Construction of augmented objective function. The ramp
rate constrained generation limits of thermal units (except the reference unit), i.e. (11), and the constraint of reservoir discharges
that is Qhmin
Qhj,t Qhmax
(1 j Nh ) are already satised.
j
j
The remaining constraints of the DHGS problem are considered
based on the penalty factor technique. This technique converts
the primal constrained problem into an unconstrained problem
by penalizing constraint violations. The penalty terms are based
on the deviation from the constraints and they are chosen high
enough to make constraint violations prohibitive in the nal solution. The following penalty factors are dened to handle the
remaining constraints of the DHGS problem:
(1) Reference unit constraint penalty

1,t =

rmin

(1 Ps1,t /Ps1,t )Fmax if

rmin
Ps1,t < Ps1,t

rmax 1)F
(Ps1,t /Ps1,t
max

if

rmax
Ps1,t > Ps1,t

else

Fmax =

where 3,j,m,t in (37) is the penalty term for the violation of mth
PDZ constraint of hydro unit j at time interval t. Total penalty
term for the PDZ constraints PT3 is calculated as follows:
PT3 =

(31)

4,j,t =

1,t

(33)

LB,k
i,t /Psi

PsiLB,k < Psi,t Psiave,k

Psiave,k < Psi,t PsiUB,k

1)Fmax

if

(1 Psi,t /PsiUB,k )Fmax

if

else

(34)

(35)

Ns NOi
T 



2,i,k,t

(36)

t=1 i=1 k=1

(3) PDZs constraint penalty

3,j,m,t =

Qhave,m
=
j

if

Vhj,t > Vhmax

j

else

PT4 =

4,j,t

(40)

(41)

(5) Final volume constraint penalty

5,j =

PT5 =

|Vhj,T Vhj,end |
Vhj,end
Nh


(42)

Fmax

(43)

5,j

j=1

PT6 =

(1 Phj,t /Phmin
)Fmax if

Phj,t < Phmin

j

(Phj,t /Phmax
1)Fmax
j

if

Phj,t > Phmax

j

else

Nh
T 


6,j,t

(44)

(45)

1)Fmax
(Qhj,t /QhLB,m
j

(1

Qhj,t /QhUB,m
)Fmax
j

if

QhLB,m
< Qhj,t Qhave,m
j
j

if

Qhave,m
< Qhj,t QhUB,m
j
j

(37)

else

(QhLB,m
+ QhUB,m
)
j
j
2

Considering the total thermal fuel cost OF in (1) and the above
penalty terms PT1 to PT6 , the augmented objective function of the
DHGS problem, i.e. AOF, for each particle in the swarm is computed
as follows:
AOF = OF + PT1 + PT2 + PT3 + PT4 + PT5 + PT6

(PsiLB,k + PsiUB,k )

where 2,i,k,t in (34) is penalty term for the violation of kth POZ
constraint of thermal unit i at time interval t. Total penalty term
for the POZ constraints PT2 is calculated as follows:
PT2 =

(Vhj,t /Vhmax
1)Fmax
j

t=1 j=1

(Ps

Vhj,t < Vhmin

j

Nh
T 


6,j,t =

(2) POZs constraint penalty

Psiave,k

(1 Vhj,t /Vhmin
)Fmax if

Total penalty term 4,j,t for the reservoir storage volume

limits PT4 is calculated as follows:

(32)

t=1

2,i,k,t =

(39)

(4) Reservoir storage volume limit penalty. By means of (14), reservoir storage volume Vhj,t at each time interval t (1 t T) can
be calculated from initial storage volume Vhj,0 . If Vhj,t violates
reservoir storage volume limits, the following penalty term
should be considered:

Total penalty term for this constraint, denoted by PT1 , is

calculated as follows:
PT1 =

3,j,m,t

t=1 j=1 m=1

(6) Hydropower generation limits penalty

fi (Psimax )

i=1

T


Nh NDj
T 



t=1 j=1

where 1,t is in proportion to the reference unit limits violation

and zero in case of no violation of these limits at time interval
t. The rst thermal unit is considered as the reference unit in
(31). Fmax is the penalty factor, which should be chosen high
enough. Here, Fmax is calculated as below [22,35]:
Ns


727

(38)

(46)

The augmented objective function AOF is used for the MAPSO,

e.g. in (25) and (26).
Step 4 Sorting particles in the tree topology. The particles of the
swarm in the MAPSO are sorted in the tree topology, i.e. Fig. 1(b),
best ). For each
based on the AOF of their best positions so far (Pi,iter
particle, the best position so far of its directly above particle is
selected as its parent (PARi,iter ).
Step 5 Updating velocity and position vectors. The velocities of particles are updated based on (24). The proposed adaptive velocity
limiter, based on (27) and (28), is applied to the particles to prevent
them from moving too rapidly. Then, the positions of the particles,
as shown in (29), are updated according to (18). In the new position
of each particle, if Qhj,t and Psi,t (except the reference unit) violate
rmin , Psrmax ], respectheir allowable ranges [Qhmin
, Qhmax
] and [Psi,t
j
j
i,t
tively, they are limited to their respective ranges. For instance, if
Qhj,t becomes greater than Qhmax
, it is limited to Qhmax
. Iteration
j
j
number is incremented, i.e. iter = iter + 1. Then, TVIW (w), TVAC

728

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

Table 2
POZ constraints, ramp-up and ramp-down rate limits and initial state of thermal
units for test case 4.

Fig. 2. Hydrosystem conguration used in the four test cases.

(c1b , c1nb and c2 ) and the dynamic coefcient of the velocity limiter
(R) are updated based on (19), (20), (21) and (27), respectively.
Step 6 Evaluation of the new particle positions and updating the
Best and Not-best components. For the new position of each
particle, the OF and AOF are calculated as described in the steps
2 and 3, respectively. Then, the Best and Not-best components
of each particle are updated based on (25) and (26), respectively.
The personal best exchanging (PBE) technique is applied to the
particles of the MAPSO.
Step 7 Evaluation of the stopping condition. If iter < itermax , go to
step 4. Otherwise, the MAPSO algorithm terminates. It is noted that
due to the execution of the PBE technique, the top particle in the
tree topology, i.e. Fig. 1(b), does not necessarily have the optimum
Best component. So, we search among the Best components of
the particles to nd the Best component owning minimum AOF
value, which is selected as the nal solution of the MAPSO for the
DHGS problem.
Note: The constraints of water discharge rate for hydro units
Qhj,t Qhmax
(1 j Nh ) and the ramp rate constrained
Qhmin
j
j
operating limits of thermal units (except the reference unit) shown
in (11) are satised in the initialization (step 1) and next iterations
(step 5). The constraint handling techniques of steps 1 and 5 are
known as preservation method and solution repair method, respectively [22,35]. The remaining constraints of the DHGS problem are
handled in the step 3 by the penalty function method
5. Numerical results
The effectiveness of the proposed MAPSO to solve the DHGS
problem is evaluated based on four test cases. All test cases include
a multi-chain cascade of four hydroplants, shown in Fig. 2. Water
transport delays between reservoirs are indicated on the gure.

Thermal unit

POZ (MW)
URi (MW/h)
DRi (MW/h)
Psi,0 (MW)

7080
50
40
100

160190
90
70
120

170180
170
120
230

The scheduling period is one day (24 h). Variable natural inow
rate into each reservoir and variable load demand over scheduling
period are considered in the four test cases. The characteristics of
each test case as well as the constraints of the respective DHGS
problem are shown in Table 1. As seen, the complexity of the four
test cases increases step by step. The data of the test cases 1, 2
and 3 can be found in [23,7,36], respectively. The additional data
of test case 4 with respect to test case 3 are shown in Table 2
including POZ constraints, ramp-up and ramp-down rate limits and
initial state of thermal units. Moreover, the same PDZ constraints
of test case 2 are also considered in test case 4. The best results of
the MAPSO and several other methods for the four test cases are
shown in Table 3. In order to illustrate the relative contribution
of the suggested modications in improving the performance of
the proposed MAPSO, the obtained results from the proposed APSO
(with the new version of tree topology), MAPSO(1) = APSO + the rst
modication, MAPSO(2) = MAPSO(1) + the second modication, and
MAPSO(3) = MAPSO(2) + the third modication for the four test cases
are shown in Table 3. The MAPSO(3) is the proposed MAPSO technique. As seen, the proposed MAPSO outperforms all other methods
of Table 3 on all test cases. Also, Table 3 shows that the proposed
modications enhance the performance of the MAPSO step by step.
Obtained results from the MAPSO for the reservoir discharge of
the four hydro units (Qhj,t ) over the 24 h of the scheduling horizon
for the rst test case are shown in Fig. 3. Also, load demand, thermal power generation (Ps1,t ) and total hydropower generation for
this test case are illustrated in Fig. 4. The obtained values for the
decision variables in Figs. 3 and 4 satisfy all constraints of this test
case, including (4), (5), (13), (15), (16) and Qhmin
Qhj,t Qhmax
j
j
(1 j Nh ). Similarly, obtained values for the decision variables
in the three other test cases satisfy the respective constraints of
these cases. In order to also give a graphical view about the convergence behavior of the proposed MAPSO, evolution of classical PSO,
APSO and MAPSO for the fourth test case (the most complex case)
is shown in Fig. 5. In this gure, variation of AOF with the iterations is shown. Fig. 5 shows better convergence behavior of the
MAPSO with respect to PSO and APSO. Also, variation of AOFOF
(total penalty term) with the iterations for the MAPSO is shown in
Fig. 6. As seen, from the iteration 3196, AOFOF becomes zero.
For each test case, the best solution of the proposed MAPSO
among 25 trail runs is shown in Table 3, since the MAPSO begins
from a random initial point (step 1 of the step by step algorithm).
Similar trial runs have been considered for the other methods in
their respective references and their best solutions are reported in
Table 3. In order to further verify the robustness of the proposed

Table 1
The characteristics of the four test cases and the constraints of the respective DHGS problem.
Test case

Characteristics

Constraints of the DHGS problem

Nh = 4, Ns = 1 quadratic cost function, shown in (2), for the single

thermal unit

(4), (5), (13), (15), (16) and

Qhmin
Qhj,t Qhmax
(1 j Nh )
j
j

Test case 1 + valve loading effects of the thermal unit + PDZ constraints
of hydro units
Nh = 4, Ns = 3 cost functions of thermal units include valve loading
effects as shown in (3)

Constraints of test case 1 + (12)

Test case 3 + PDZ constraints of hydro units + POZ constraints of

thermal units + ramp rate limits of thermal units

(4), (7), (11), (12), (13), (15) and (16)

3
4

(4), (5), (13), (15), (16) and

Qhj,t Qhmax
(1 j Nh )
Qhmin
j
j

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

729

Table 3
Obtained results for the four test cases.
Test case 1

Test case 2

Test case 3

Test case 4

Method

OF ($)

Method

OF ($)

Method

OF ($)

Method

OF ($)

EGA [8]
GA [23]
FEP [7]
CEP [7]
IFEP [7]
PSO [8]
BCGA [11]
APSO
RCGA [11]
LWPSO [18]
MAPSO(1)
DE [21]
MAPSO(2)
EPSO [8]
IPSO [16]
MAPSO(3)

934,727.00
932,734.00
930,268.00
930,166.00
930,130.00
928,878.00
926,922.71
926,151.54
925,940.03
925,383.80
924,232.38
923,991.08
923,074.27
922,904.00
922,553.49
922,421.66

NLP [16]
DP [16]
IFEP [7]
DE [16]
HDE [14]
APSO
IPSO [16]
MAPSO(1)
MDE [14]
MHDE [14]
MAPSO(2)
MAPSO(3)

936,709.52
935,617.76
933,949.25
928,236.94
927,895.81
925,991.35
925,978.84
925,963.72
925,960.56
925,547.31
925,054.53
924,636.88

EP [17]
SA [17]
PSO [17]
DE [14]
MDE [14]
HDE [14]
APSO
MHDE [14]
MAPSO(1)
MAPSO(2)
MAPSO(3)

47,306.00
45,466.00
44,740.00
44,526.10
42,611.14
42,337.30
41,858.27
41,856.50
41,241.91
40,875.34
40,225.06

Classical PSO
APSO
MAPSO(1)
MAPSO(2)
MAPSO(3)

47,443.40
41,973.19
41,624.65
41,074.12
40,748.38

Table 4
The best, average and worst OF values ($).
Method

EGA [8]

Test case
Best
Average
Worst

1
934,727
936,058
937,339

Method

LWPSO [18]

Test case
Best
Average
Worst

1
925,383
926,352
927,240

GA [23]
1
932,734
936,969
939,734
EPSO [8]
1
922,904
923,527
924,808

FEP [7]
1
930,268
930,897
931,397

CEP [7]

MAPSO
1
922,421
922,544
923,508

MAPSO, the best, average and worst OF values obtained by the

MAPSO and some other methods of Table 3 are shown in Table 4.
For the other methods of Table 3 not mentioned in Table 4, the
average and worst OF values have not been given in the respective
references. The best OF values in Table 4 are the reported results
in Table 3. As seen, the best, average and worst OF values of the
MAPSO are lower than those of all other methods of Table 4 on the
mentioned test cases.
In order to also evaluate the performance of the MAPSO for
larger DHGS test cases, two test cases 5 and 6 are constructed from
the test case 4 by repeating the thermal and hydro units of this test
case 5 and 10 times, respectively. Also, the load demand of the test

Fig. 3. Reservoir discharge of the four hydro units (Qhj,t ) for rst test case.

IFEP [7]

1
930,166
930,373
930,927

1
930,130
930,290
930,882

IFEP [7]

MAPSO

2
933,949
938,508
942,593

2
924,636
926,496
927,431

PSO [8]
1
928,878
933,085
938,012
PSO
4
47,443
49,238
51,062

BCGA [11]

RCGA [11]

1
926,922
927,815
929,451

1
925,940
926,120
926,538

APSO

MAPSO

4
41,973
42,521
42,874

4
40,748
40,957
41,695

cases 5 and 6 are 5 and 10 times higher than that of the test case 4.
The best, average and worst results of the MAPSO for the test case
5 are 206,834, 214,795 and 223,721, respectively and for the test
case 6 are 415,683, 421,581 and 429,758, respectively. As seen, the
MAPSO obtained good results for these two test cases as well. The
obtained best, average and worst results for the test cases 5 and 6
are about 5 and 10 times higher than those of the test case 4.
The parameters of the MAPSO have been set as follows
based on trial and error (the selected values are the best
initial =
ones among several runs): wmin = 0.4, wmax = 0.9, c1b
final

1.6, c1b

final

final

initial = 0.15, c
= 0.7, c1nb
= 0.1, c2initial = 0.9, c2
1nb

Fig. 4. Load demand, thermal power generation (Ps1,t ) and total hydropower generation for the rst test case.

730

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

Fig. 5. Evolution of classical PSO, APSO and MAPSO for the fourth test case.

1.7, Rinitial = 0.3, Rfinal = 0.1, Npar = 100, itermax = 5000 and
NPBE = 10, where NPBE indicates number of particles participating
in the personal best position exchanging in each iteration. These
deviation (%) =

Fig. 6. Variation of AOFOF with iterations for the MAPSO in the fourth test case.

tively and its results are shown in the last row of Table 5, indicated
by All. In each cell of columns 2, 3, 4, 6, 7 and 8, the number in the
parentheses indicates deviation from the solution with the original
parameters in terms of percentage, computed as follows:

solution (with perturbed parameter) solution (with original parameter)

100
solution (with original parameter)

values of the parameters have been used for all six test cases of
the paper, although these cases have different conditions and sizes.
This matter shows robustness and low sensitivity of the proposed
MAPSO with respect to its parameters (it is not necessary to separately adjust these parameters for each problem). To further verify
the low sensitivity of the MAPSO, we perturbed its parameters
around their original values; however the results of the MAPSO for
the test cases negligibly changed. Sample results of this sensitivity
analysis for the fourth test case are shown in Table 5. In the sensitivity analysis, the initial value of each parameter of the MAPSO is
perturbed in the down and up directions and its perturbed values
are shown in the rst and fth columns of Table 5 titled as Original  and Original + , respectively. Here, the perturbation 
is 20% of the original value. To also evaluate the cumulative effect
of the parameters perturbations, we perturbed all parameters of
the MAPSO simultaneously in the down and up directions, respec-

(47)

For instance, when the initial value of wmin = 0.4 is perturbed

by 20% in the down direction to wmin = 0.4 (0.4 20%) = 0.32,
the best solution becomes 40,798 having 0.122% deviation with
respect to the initial value of the best solution, i.e. 40,748. In
Table 5, negative sign for a deviation means the objective function with the perturbed parameter decreases with respect to its
original value or equivalently the solution with the perturbed
parameter is better than the original solution and conversely for
the positive deviation. It is noted that the selected values for the
parameters of the MAPSO are not the optimum values; they are
only selected after a few tests. However, Table 5 shows that even
with considering the cumulative effect of all parameters perturbations, very small deviations are seen in the results of the proposed
MAPSO.
The computation times of the proposed MAPSO for the test
cases 1, 2, 3, 4, 5 and 6 are 64 s, 83 s, 97 s, 146 s, 898 s and
1827 s, respectively, measured on a simple personal computer Pen-

Table 5
Obtained results from the sensitivity analysis for the fourth test case.
Original 

Best

Average

Worst

Original + 

Best

Average

wmin = 0.32

40,798 (0.122%)

41,014 (0.139%)

41,835 (0.337%)

wmin = 0.48

40,772 (0.059%)

41,004 (0.115%)

41,815 (0.288%)

wmax = 0.72

40,768 (0.049%)

40,975 (0.044%)

41,807 (0.269%)

wmax = 1.08

40,788 (0.098%)

41,034 (0.188%)

41,819 (0.297%)

initial
= 1.28
c1b

40,771 (0.056%)

40,961 (0.010%)

41,842 (0.353%)

initial
c1b
= 1.92

40,760 (0.029%)

40,958 (0.003%)

41,776 (0.194%)

final
c1b

40,749 (0.002%)

40,902 (0.134%)

41,705 (0.024%)

c1b

40,755 (0.017%)

40,979 (0.054%)

41,659 (0.084%)

40,748 (0.001%)

40,939 (0.044%)

41,778 (0.199%)

initial
c1nb
= 0.18

40,764 (0.039%)

40,967 (0.024%)

41,773 (0.187%)

40,755 (0.017%)

40,944 (0.031%)

41,813 (0.283%)

c1nb = 0.12

40,749 (0.002%)

40,988 (0.075%)

41,696 (0.002%)

40,749 (0.002%)

40,959 (0.005%)

41,642 (0.126%)

c2initial = 1.08

40,754 (0.014%)

40,957 (0.001%)

41,603 (0.219%)

40,754 (0.014%)

40,991 (0.083%)

41,809 (0.273%)

c2

NPBE = 8

40,758 (0.024%)

40,975 (0.044%)

41,801 (0.254%)

NPBE = 12

NPar = 80

40,759 (0.027%)

41,030 (0.178%)

41,876 (0.434%)

itermax = 4000

40,757 (0.022%)

40,966 (0.023%)

41,809 (0.273%)

Rinitial = 0.24

40,770 (0.054%)

40,962 (0.013%)

Rnal = 0.08

40,751 (0.007%)

All (down)

40,760 (0.029%)

= 0.56

initial
c1nb
= 0.12
final

c1nb = 0.08
c2initial
final

c2

= 0.72

= 1.36

final

= 0.84

final

final

= 2.04

Worst

40,751 (0.007%)

40,963 (0.015%)

41,855 (0.384%)

40,750 (0.005%)

41,004 (0.117%)

41,804 (0.261%)

NPar = 120

40,744 (0.010%)

40,876 (0.195%)

41,602 (0.223%)

itermax = 6000

40,747 (0.002%)

40,869 (0.213%)

41,693 (0.004%)

41,644 (0.121%)

Rinitial = 0.36

40,759 (0.027%)

40,959 (0.004%)

41,742 (0.112%)

40,974 (0.042%)

41,805 (0.264%)

Rnal = 0.12

40,753 (0.012%)

40,947 (0.023%)

41,726 (0.075%)

41,037 (0.195%)

41,763 (0.163%)

All (up)

40,758 (0.024%)

41,061 (0.255%)

41,749 (0.130%)

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

tium P4 2.8 GHz with 1 GB RAM. These computation times show

low computation burden of the MAPSO to solve the DHGS problem. We did not compare the computation times of the MAPSO
with those of the other methods, since the computation times
of each method have been measured on a different hardware
set.
6. Conclusion
This paper proposes a new MAPSO to solve the DHGS problem,
which is a complicated non-linear, non-convex and non-smooth
optimization problem with discontinuous solution space. Obtaining a good solution is hard for this type of optimization problem.
The proposed MAPSO has TVIW, TVAC and a specic kind of tree
topology. Also, three new modications including the Split-up of
the cognitive part into Best and Not-Best components, personal
best position exchanging and new velocity limiter are incorporated
into the proposed MAPSO to enhance its exploration capability
and convergence behavior. Obtained results from extensive testing of the proposed MAPSO on different DHGS test cases conrm
the validity of the developed approach. The research work is under
way in order to include unit commitment in the DHGS problem, which converts the DHGS into a Mixed Integer Non-Linear
Programming (MINLP) problem. Also, considering the uncertainty
sources of the DHGS problem (such as uncertain load demand and
inow rate of reservoirs), which converts the DHGS into a stochastic optimization problem, can be considered as the future research
work.
Appendix A. List of symbols

OF
T
Ns , Nh

total fuel cost ($)

scheduling horizon (h)
number of thermal generating units and number of hydro
generating units, respectively
asi , bsi , csi , dsi , and esi fuel cost coefcients of ith thermal generating unit with valve loading effects
Psi,t , fi (Psi,t ) power generation (MW) and fuel cost ($/h) of ith thermal unit at time interval t
Phj,t , Qhj,t hydropower generation (MW) and water discharge rate
(m3 /h) of reservoir j at time interval t
Psimin , Psimax minimum and maximum power generation for ith
thermal unit, respectively
rmin , Psrmax ramp rate constrained minimum and maximum
Psi,t
i,t
power generation for ith thermal unit at time interval t,
respectively
Phmin
, Phmax
minimum and maximum power generation for jth
j
j
hydro unit, respectively
Vhj,t
water storage volume (m3 ) of reservoir j at time interval
t
specied storage volume (m3 ) of reservoir j at the end of
Vhj,end
scheduling horizon
Ch1j Ch6j hydropower generation coefcients for jth hydro unit
PsiLB,k , PsiUB,k lower and upper boundaries of kth POZ of ith thermal
unit, respectively
NOi , NDj number of POZs of thermal unit i and number of PDZs of
hydro unit j, respectively
URi , DRi ramp-up and ramp-down rate limits of thermal unit i
(MW/h), respectively
, Qhmax
minimum and maximum water discharge rates of
Qhmin
j
j
jth reservoir, respectively
QhLB,m
, QhUB,m
lower and upper boundaries of mth PDZ of jth
j
j
hydro unit, respectively

731

, Vhmax
minimum and maximum storage volume of jth
Vhmin
j
j
reservoir, respectively
Ihj,t , Shj,t natural inow rate and spillage discharge rate of jth reservoir at time interval t, respectively
 r,j
water transport delay from reservoir r to j
total number of upstream units which are immediately
UPj
above the jth reservoir
PDt , PLt active power demand and total transmission loss at time
interval t, respectively
References
[1] E.C. Finardi, E.L. Dasilva, C. Sagastizabal, Solving the unit commitment problem of hydropower plants via Lagrangian Relaxation and Sequential Quadratic
Programming, J. Comput. Appl. Math. 24 (3) (2005) 317341.
[2] I. Erkmen, B. Karatas, Short-term hydrothermal coordination by using
multi-pass dynamic programming with successive approximation, in: 7th
Mediterranean Electrotechnical Conference, vol. 3, 1994, pp. 925928.
[3] G.G. Oliveira, S. Soares, A second order network ow algorithm for hydrothermal scheduling, IEEE Trans. Power Syst. 10 (3) (1995) 16351641.
[4] H. Habibollahzadeh, J.A. Bubenko, Application of decomposition techniques
to short term operation planning of hydrothermal power system, IEEE Trans.
Power Syst. 1 (1) (1986) 4147.
[5] G.W. Chang, M. Aganagic, J.G. Waight, J. Medina, T. Burton, S. Reeves, M. Christoforidis, Experiences with mixed integer linear programming based approaches
on short-term hydro scheduling, IEEE Trans. Power Syst. 16 (4) (2001) 743749.
[6] M.S. Salam, K.M. Nor, A.R. Hamdam, Hydrothermal scheduling based
Lagrangian relaxation approach to hydrothermal coordination, IEEE Trans.
Power Syst. 13 (1) (1998) 226235.
[7] N. Sinha, R. Chakrabarti, P.K. Chattopadhyay, Fast evolutionary programming
techniques for short-term hydrothermal scheduling, IEEE Trans. Power Syst. 18
(1) (2003) 214220.
[8] X. Yuan, L. Wang, Y. Yuan, Application of enhanced PSO approach to optimal
scheduling of hydro system, Energ. Convers. Manag. 49 (2008) 29662972.
[9] C.E. Zoumas, A.G. Bakirtzis, J.B. Theocharis, V. Petridis, A genetic algorithm solution approach to the hydrothermal coordination problem, IEEE Trans. Power
Syst. 19 (2) (2004) 13561364.
[10] E. Gil, J. Bustos, H. Rudnick, Short-term hydrothermal generation scheduling model using a genetic algorithm, IEEE Trans. Power Syst. 18 (4) (2003)
12561264.
[11] S. Kumar, R. Naresh, Efcient real coded genetic algorithm to solve the nonconvex hydrothermal scheduling problem, Int. J. Electr. Power Energ. Syst. 29
(2007) 738747.
[12] R.H. Liang, Y.Y. Hsu, Scheduling of hydroelectric generations using articial
neural networks, IEE Proc. Gener. Transm. Distrib. 141 (5) (1994) 452458.
[13] K. Wong, Y.W. Wong, Short-term hydrothermal scheduling. Part I. simulated
annealing approach, IEE Proc. Gener. Transm. Distrib. 141 (5) (1994) 497501.
[14] L. Lakshminarasimman, S. Subramanian, A modied hybrid differential evolution for short-term scheduling of hydrothermal power systems with cascaded
reservoirs, Energ. Convers. Manag. 49 (2008) 25132521.
[15] X. Yuan, Y. Yuan, Application of cultural algorithm to generation scheduling of
hydrothermal systems, Energ. Convers. Manag. 47 (2006) 21922201.
[16] P.K. Hota, A.K. Barisal, R. Chakrabarti, An improved PSO technique for shortterm optimal hydrothermal scheduling, Electr. Power Syst. Res. 79 (2009)
10471053.
[17] K.K. Mandal, M. Basu, N. Chakraborty, Particle swarm optimization technique based short-term hydrothermal scheduling, Appl. Soft Comput. 8 (2008)
13921399.
[18] B. Yu, X. Yuan, J. Wang, Short-term hydro-thermal scheduling using particle
swarm optimization method, Energ. Convers. Manag. 48 (2007) 19021908.
[19] R.H. Liang, M.H. Ke, Y.T. Chen, Coevolutionary algorithm based on Lagrangian
method for hydrothermal generation scheduling, IEEE Trans. Power Syst. 24 (2)
(2009) 499507.
[20] S.Y.W. Wong, Hybrid simulated annealing/genetic algorithm approach to short
term hydro-thermal scheduling with multiple thermal plants, Int. J. Electr.
Power Energ. Syst. 23 (2001) 565575.
[21] K.K. Mandal, N. Chakraborty, Differential evolution technique-based shortterm economic generation scheduling of hydrothermal systems, Elec. Power
Syst. Res. 78 (2008) 19721979.
[22] N. Amjady, H. Nasiri-Rad, Economic dispatch using an efcient real-coded
genetic algorithm, IEE Proc. Gener. Transm. Distrib. 3 (3) (2009) 266278.
[23] S.O. Orero, M.R. Irving, A genetic algorithm modeling framework and solution
technique for short term optimal hydrothermal scheduling, IEEE Trans. Power
Syst. 13 (2) (1998) 501518.
[24] J. Kennedy, R. Eberhart, Particle swarm optimization, Proceedings of IEEE Conference on Neural Networks 4 (1995) 19421948.
[25] K.T. Chaturvedi, M. Pandit, L. Srivastava, Self-organizing hierarchical particle
swarm optimization for nonconvex economic dispatch, IEEE Trans. Power Syst.
23 (3) (2008) 10791087.
[26] A. Ratnaweera, S.K. Halgamuge, H.C. Watson, Self-organizing hierarchical particle swarm optimizer with time-varying acceleration coefcients, IEEE Trans.
Evol. Comput. 8 (3) (2004) 240255.

732

N. Amjady, H.R. Soleymanpour / Electric Power Systems Research 80 (2010) 723732

[27] S. Janson, M. Middendorf, A hierarchical particle swarm optimizer and its adaptive variant, IEEE Trans. Syst., Man, Cybernetics, Part B: Cybernetics 35 (6)
(2005) 12721282.
[28] A.I. Selvakumar, K. Thanushkodi, A new particle swarm optimization solution to
nonconvex economic dispatch problems, IEEE Trans. Power Syst. 22 (1) (2007)
4251.
[29] J. Kennedy, Small worlds and mega-minds: effects of neighborhood topology on
particle swarm performance, IEEE Congr. Evol. Comput. 3 (1999) 19311938.
[30] Y. Shi, R.C. Eberhart, Parameter Selection in Particle Swarm Optimization,
Springer, Berlin/Heidelberg, 1998.
[31] T. Takahama, S. Sakai, Solving constrained optimization problems by the
epsilon constrained particle swarm optimizer with adaptive velocity limit control, in: Second IEEE International Conference of Cybernetics and Intelligent
System, 2006, pp. 683689.

[32] S.M. Mikki, A.A. Kishk, Hybrid periodic boundary condition for particle swarm
optimization, IEEE Trans. Anten. Propag. 55 (11) (2007) 32513256.
[33] J. Li, B. Ren, C. Wang, A Random Velocity Boundary Condition for Robust Particle
Swarm Optimization, Springer, Berlin/Heidelberg, 2007.
[34] I.G. Damousis, A.G. Bakirtzis, P.S. Dokopoulos, Network-constrained economic
dispatch using real-coded genetic algorithm, IEEE Trans. Power Syst. 18 (1)
(2003) 198205.
[35] N. Amjady, H. Nasiri-Rad, Nonconvex economic dispatch with AC constraints
by a new real coded genetic algorithm, IEEE Trans. Power Syst. 24 (3) (2009)
14891502.
[36] M. Basu, Interactive fuzzy satisfying method based on evolutionary programming technique for multiobjective short-term hydrothermal scheduling, Elec.
Power Syst. Res. 69 (2004) 277285.