Electric Power Systems Research 79 (2009) 1047–1053

Contents lists available at ScienceDirect

Electric Power Systems Research

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

An improved PSO technique for short-term optimal hydrothermal scheduling

P.K. Hota a,∗ , A.K. Barisal a , R. Chakrabarti b
a
Department of Electrical Engineering, U.C.E., Burla, Orissa 768018, India
b
Department of Electrical Engineering, Jadavpur University, Kolkata, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a new approach to the solution of optimal power generation to short-term hydrother-
Received 4 July 2008 mal scheduling problem, using improved particle swarm optimization (IPSO) technique. The practical
Received in revised form 1 November 2008 hydrothermal system is highly complex and possesses nonlinear relationship of the problem variables,
Accepted 6 January 2009
cascading nature of hydraulic network, water transport delay and scheduling time linkage that make
Available online 6 February 2009
the problem of finding global optimum difficult using standard optimization methods. In this paper an
improved PSO technique is suggested that deals with an inequality constraint treatment mechanism called
Keywords:
as dynamic search-space squeezing strategy to accelerate the optimization process and simultaneously,
Dynamic search-space squeezing strategy
Hydrothermal scheduling
the inherent basics of conventional PSO algorithm is preserved. To show its efficiency and robustness, the
Particle swarm optimization proposed IPSO is applied on a multi-reservoir cascaded hydro-electric system having prohibited oper-
Practical constraints ating zones and a thermal unit with valve point loading. Numerical results are compared with those
Multichain reservoirs obtained by dynamic programming (DP), nonlinear programming (NLP), evolutionary programming (EP)
Differential evolution and differential evolution (DE) approaches. The simulation results reveal that the proposed IPSO appears
to be the best in terms of convergence speed, solution time and minimum cost when compared with
established methods like EP and DE.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction can solve such problems in different formulations [3,4]. However,

this method suffers from computational overburden for its large
Short-term hydrothermal scheduling as one of the constrained dimensionality when applied to a practical sized power system.
power system optimization problems, has complex and nonlinear The nonlinear programming with network flow method [5] has
characteristics with various types of constraints and nonlinear rela- difficulties in dealing with constraints of system.
tionship of problem variables that make the problem of finding the In the past decades, stochastic search algorithms like simulated
global optimum difficult using any standard optimization meth- annealing (SA) [6], genetic algorithm (GA) [7–9] and evolutionary
ods and programming techniques. The fundamental requirement programming (EP) [10,11] have been proved to be very efficient
of power system hydrothermal scheduling is to generate the opti- in solving complex power systems problems but, these heuristic
mal amount of generated powers for the hydro and thermal units methods do not always guarantee the globally optimal solution.
in the system to meet the load demands in the scheduling hori- Rather, quite often they can produce suboptimal or near globally
zon of 1 day or few days while satisfying various constraints on optimal solution. The recent research has identified few drawbacks
the hydraulic and power system network. Usually, the objective of the stochastic methods like GA, EP, and SA out of which the most
function to be minimized in a hydrothermal scheduling problem critical disadvantage is their slow convergence towards optimal
is the total fuel cost of thermal units. The optimal hydrothermal solution. The particle swarm optimization (PSO), first introduced
scheduling of hydrothermal power system is basically a nonlinear by Kennedy and Eberhart [12–14] is one of the recent additions
problem involving nonlinear objective function and a combination to modern heuristic algorithms. It was developed through simula-
of linear and nonlinear constraints. Classical methods [1,2] have tion of a simplified social system, and has been found to be flexible
been successfully employed in solving most economical hydrother- and robust in solving complex power system problems than con-
mal generation schedule under the practical constraints. However, ventional methods. Recently, PSO has been successfully applied to
these methods have difficulties in handling various constraints and various fields of power system optimization such as power sys-
are time consuming methods and hence, not suitable to address tem stabilizer design [15], reactive power and voltage control [16],
such types of problems. A dynamic programming (DP) method dynamic security border identification and ELD problems [17,18].
In this paper, a new approach to the inequality constraint
treatment mechanism called as dynamic search-space squeezing
∗ Corresponding author. Tel.: +91 6632431026; fax: +91 6632430204. strategy is devised to optimize the nonlinear scheduling hydrother-
E-mail address: p [email protected] (P.K. Hota). mal problem. Moreover, dynamic search-space squeezing strategy

0378-7796/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2009.01.001
1048 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053

differs from the dynamic search-space reduction strategy [18] in voir storage limits, water discharge rate limits, hydraulic continuity
two aspects. Firstly, the limits of position are determined from the constraints and initial and final reservoir storage limits. These con-
relative distance measured from the best position of the group and straints are discussed as below.
position boundaries, but not from the individual distance as in case
of reduction strategy. Secondly, the margins of reduction in case (a) Demand constraints: This constraint is based on the principle
of proposed squeezing strategy are considered from both the sides of equilibrium between the total generation from hydro and
of position simultaneously unlike one side consideration at a time thermal plants and the total system demand plus the system
[18] and also it varies adaptively instead of remaining constant. losses in each hour of scheduling j.
The position of any individual or particle is adjusted between their
limits which are calculated based on the relative distance between 
n

PTj + PHij = PDj + PLossj , for j = 1, 2, . . . , m (2)

the ever best position of the group and the inequality boundaries
to enhance the global exploration abilities. More recently, another i=1

evolutionary optimization technique called as differential evolu- (b) Thermal generator constraints: The operating limit of equivalent
tion (DE) [19] has emerged as a promising tool compared to GA, thermal generator has a lower and upper bound so that it lies
EP and SA techniques. Its applications to few power system prob- in between these bounds.
lems have also been reported in the literature that are found to be
exciting. Therefore, to investigate the potential of the proposed IPSO PTmin ≤ PTj ≤ PTmax (3)
approach, one example of hydrothermal system [8,11] has been con-
(c) Hydro generator constraint: The operating limit of hydro plant
sidered and their simulation results are compared to those of recent
must lie in between its upper and lower bounds.
approaches reported in the literature including the DE approach.
min max
PHi ≤ PHij ≤ PHi (4)
2. Problem formulation
(d) Reservoir capacity constraint: The operating volume of reservoir
The prime objective of the short-term hydrothermal schedul- storage limit must lie in between the minimum and maximum
ing problem is to minimize the total thermal cost such that the capacity limits.
load demands PDj supplied from hydro plants and a thermal plant Vimin ≤ Vij ≤ Vimax (5)
in the intervals of the generation scheduling horizon can be met
and simultaneously, all the equality and inequality operation con- (e) The water discharge constraint: The variable net head operation
straints are satisfied. Fig. 1 depicts the schematic diagram of is considered and the physical limitation of water discharge of
hydrothermal system. turbines, qij , must lie in between its maximum and minimum
operating limits, as given by
2.1. Objective function and constraints
qmin
i
≤ qij ≤ qmax
i
(6)
The total fuel cost for running the thermal system to meet the (f) Hydraulic continuity constraint: The storage reservoir volume
load demand in scheduling horizon is given by F. The objective limits are expressed with given initial and final volumes as
function is expressed mathematically, as

Ru
 

m
  Vi(j+1) = Vij + qu(j−) + su(j−) − qi(j+1) − si(j+1) + ri(j+1)
minimize F = fj PTj (1)
u=1
j=1
for j = 1, 2, . . . , m (7)
Subject to the following various system constraints:
This nonlinear constrained hydrothermal scheduling optimiza- where  is the water delay time between reservoir i and its up-
tion problem is subjected to a variety of constraints depending upon stream u at interval j and Ru is the set of upstream units directly
practical implications like the varying system load demand, the above hydroplant i.
time coupling effect of hydro subsystem, the cascading nature of (g) The hydro power generation, PHij , is assumed to be a function of
the hydraulic network, the time varying hourly reservoir inflows, discharge rate and storage volume
thermal plant and hydro plant operating limits, system losses, reser-  
PHij = c1i Vij2 + c2i q2ij + c3i Vij qij + c4i Vij + c5i qij + c6i (8)

where c1i , c2i , c3i , c4i , c5i and c6i are the coefficients, and the
electric loss PLossj is a function of power generations.

3. Particle swarm optimization

The particle swarm optimization (PSO) is an optimization tech-

nique inspired by swarm intelligence and theory in general such
as bird flocking, fish schooling and even human social behavior. It
provides a population-based search procedure in which individu-
als, called particles fly in a m dimensional search space, where each
dimension corresponds to a parameter in a function being opti-
mized. The position mechanism of the particle in the search space is
updated by adding the velocity vector to its position vector as given
in Ref. [18]. Let X and V be a particle position and its corresponding
velocity in a search space respectively. The best position achieved by
a particle is recorded and denoted by Pbest. The best particle among
Fig. 1. Schematic diagram of hydrothermal system. all particles in the population is represented as gbest. The updated
 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053 1049

randomly selected to satisfy the constraints of initial and final

reservoir storage limit. Let, Qp = [q11 , q12 , . . . , q1(d−1) , q1(d+1) ,
. . . , q1m ; q21 , q22 , . . . , q2(d−1) , q2(d+1) , . . . , q2m , . . . ; qi1 , qi2 , . . . ,
qi(d−1) , qi(d+1) , . . . , qim ], be the trial vector designating the pth
individual of the population. The hydro discharge at the dependent
interval qid is calculated from (7) with zero spillage, as

m

m
 
Ru m

qid = Vi1 − Vi25 − qij + rij + qu(j−) (15)

j=1 j=1 u=1 j=1

j=
/ d

Fig. 2. The search-space squeezing mechanism of the pth particle during activation.
After knowing the water discharges, the reservoir volumes at
different intervals are determined. Then, the hydro generations are
calculated from (8). Knowing the calculated hydro generations, PHj
velocity and position of a particle can be calculated as shown in the
and the given load demand PDj , for j = 1, 2, . . ., m, thermal genera-
following formulae.
   tions PTj can be calculated as
V ijp
k+1
= w × V ijp
k 
+ c1 × rand() × Pbest k
ijp − Xijp + c2 × rand() 
n
  PTj = PDj + PLossj − PHij (16)
k
× g bestij1 − X (9)
ijp i=1

The cost function, which is the sum of the thermal cost and viola-
 k+1 = X
X  k + V k+1 , i = 1, 2, . . . , n; j = 1, 2, . . . , m (10) tion of final storage of reservoirs with a penalty multiplier of 10,000,
ijp ijp ijp
is taken in this problem.
where k is the pointer of iterations (generations), w the inertia The volume of water in the reservoirs at the end of each interval
weight factor, i the number of hydro units, p the number of particles is then calculated using Eq. (7). All the generation levels, discharges,
in a group, j the number of intervals, c1 and c2 the acceleration con- reservoir water volumes and initial and final reservoir storage vol-
stant, in general, and rand() is the random number in the range[0,1]. umes must be checked against their limiting values as per Eqs.
(3)–(8).
3.1. Dynamic search-space squeezing strategy
3.3. Stopping rule
Judicious choice of search space of the particle not only improves
the speed of convergence but also ensures the algorithm to be less The iterative procedure of generating new solutions with min-
susceptible for being trapped on local optima, i.e., the solution of imum function value is terminated when a predefined maximum
IPSO is highly optimal. When there are no significant improvements number of iterations (generations) is reached.
in the performance of solution achieved, the dynamic search-space
4. Improved PSO algorithm for hydrothermal scheduling
squeezing strategy is activated. In this case, the search space is
problem
dynamically readjusted (i.e., squeezed) based on the relative dis-
tance between gbest and lower and upper limits of discharge of The computational processes of IPSO technique can be described
ith hydroplant at jth interval denoted by Lij and Hij , respec- in the following steps.
tively. Both the relative distances are variables, not always equal
and constant, which are represented as follows: • Step 1. Input parameters of the system and specify the upper and
lower boundaries of each variable.
gbest k − qij,min qi,j max − gbest k
kLij = , kHij = (11) • Step 2. Initialize randomly the particles of the population.
qij,max − qij,min qi,j max − qij,min • Step 3. Let, Qp = [q11 , q12 , . . . , q1m ; q21 , q22 , . . . , q2m , . . . , qn1 , qn2 ,
kLij + kHij = 1 (12) . . . , qnm ], be the trial vector denoting the particles of population
to be evolved. The elements of qij are the discharges of turbines
At iteration k + 1, the adjusted limits of discharge of turbine of of reservoirs at various intervals subjected to their capacity con-
ith hydroplant at interval j are determined as follows: straints in (6). qid , be the dependent discharge of ith hydroplant at
  dth interval is randomly selected from among the committed m
qk+1
ij,min
= qij,min + gbestpk − qij,min × kLij (13)
intervals. Then, knowing the hydro discharges, storage volumes
  of reservoirs Vij are calculated by (7). Then PHij is calculated from
qk+1 = qij,max − qij,max − gbestpk × kHij (14)
ij,max (8) for all the intervals.
The limits of discharge of turbine at intervals are varying in itera- • Step 4. Compare each particle (4 × 24) evaluation value with its
tion but always dependent on the location of gbest in the boundary. Pbest. The best evaluation value among Pbest is denoted as gbest.
The updated maximum and minimum limits are described in (13) • Step 5. Update the iteration as k = k+1; inertia weight, velocity by
and (14) and always satisfied by (6). The activation of dynamic space (9), and position by (10).
squeezing process is illustrated in Fig. 2. • Step 6. Each particle is evaluated according to its updated position,
only when satisfied by all constraints. If the evaluation value of
3.2. Improved PSO-based hydrothermal scheduling each particle is better than the previous Pbest. The current value
is set to be Pbest. If the best Pbest is better than gbest, the value is
Taking the number of particles to be N, the number of set to be gbest.
scheduling intervals as m and the number of hydro units as • Step 7. The dynamic search-space squeezing strategy is activated
n, each initial trial vector Q(i,j,p) denoting the particles of to adjust the upper and lower boundaries of the particles relative
population to be evolved, for p = 1, 2, . . ., N, is selected. The dis- to latest gbest using (13) and (14).
charge of ith hydroplant at jth interval is randomly generated as • Step 8. If the stopping criterion is reached, then print the result
qij ∼ U(qmin
i
, qmax
i
). Let, qid be a dependent hydro discharge rate and stop; otherwise repeat steps 2–7.
1050 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053

5. Numerical results

5.1. Test system

To verify the applicability and performance of the proposed IPSO,

a test system has been adopted from Ref. [8]. It consists of an equiv-
alent thermal power plant and a multichain cascade of four hydro
plants. The schedule horizon is 1 day with 24 intervals of 1 h each.
The hydraulic sub-system is characterized by the following:

(a) a multichain cascade flow network, with all of the plants in one
stream;
(b) river transport delay between successive reservoirs;
(c) variable head hydro plants;
(d) variable natural inflow rates into each reservoir;
(e) prohibited operating regions of water discharge rates;
Fig. 3. Convergence characteristics of IPSO and DE algorithms for case 1.
(f) variable load demand over scheduling period.

The data of the test system considered here are the same as in
[8] and the additional data with valve point loading effect and with
prohibited discharge zones of turbines are also same as in Ref. [11].
The fuel cost function of the equivalent thermal unit with valve
point loading is
  2
f PTj = 5000 + 19.2PTj + 0.002PTj
   
+ 700 sin 0.085 PTj
min
− PTj  (17)

The lower and upper operation limits of this unit are 500 and
2500 MW, respectively.
The spillage rate for the hydraulic system is not taken in to
account (for simplicity) and further, the electric loss from the hydro
plant to the load is taken to be negligibly small. The lower and upper Fig. 4. Hydro reservoir storage volumes (without prohibited discharge zones).
operation limits of hydraulic system are 0 and 500 MW, respectively.
To demonstrate the effectiveness of the proposed IPSO, the fol- 5.2.1. Case 1: quadratic cost curve without prohibited discharge
lowing three different cases have been considered. zones
This case does not consider the prohibited discharge zones. Fig. 3
• Case 1. For the purpose of comparison with the reported results, shows the nature of convergence of Improved PSO and DE algo-
the system is considered without valve point loading effect and rithms in solving the hydrothermal problem with quadratic cost
no prohibited discharge zones. and without considering prohibited hydro discharge zones. The
• Case 2. Quadratic cost with prohibited discharge zones is consid- best hydrothermal schedule out of 25 different initial trial solu-
ered. tions with quadratic cost and without prohibited discharge zones
• Case 3. Here, the systems with valve point loading effect and with obtained from the proposed IPSO are reported in Figs. 4–6. The
prohibited operating hydro discharge zones are also considered. optimal hydro discharge and the hydro power generations with
minimum cost obtained by the proposed IPSO method are reported
in Tables 1 and 2, respectively.
5.2. Results

The present work has been implemented in command line of

Matlab-7.0 for the solution of hydrothermal scheduling. The pro-
gram was run on a 3.06 GHz, Pentium-IV, with 256 MB RAM PC.
After a number of trials of run with different values of PSO parame-
ters tuning, such as inertia weight, number of particles, maximum
allowable velocity, the details of key parameters selected are:
wmax = 0.9, wmin = 0.4, Z = 20, c1 = c2 = 2.0, N = 8. When dynamic
search squeezing strategy is activated the parameters selected are:
wmax = 0.01, wmin = 0.0, c1 = c2 = 2.0. Similarly, the differential evo-
lution method [19] with the control parameters of population
size (N) = 300, scaling factor (F) = 0.8, crossover factor (CR) = 1.1 has
been implemented on same processor. The improved fast evolu-
tionary programming (IFEP) method as described in Ref. [11] was
also implemented on same machine for comparison of computa-
tional efficiencies of proposed IPSO method along with DE and IFEP
methods. Fig. 5. Hourly hydro plant power generations (without prohibited operating zones).
 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053 1051

Table 2
Hydro plant power outputs and total thermal generation (case 1).

Hour Hydro power generation (MW) Thermal

generation (MW)

Plant 1 Plant 2 Plant 3 Plant 4

1 84.8718 57.0885 0.0000 200.0937 1027.9460

2 89.0272 53.1593 0.0000 187.7553 1060.0582
3 79.5621 52.7828 0.0000 173.7333 1053.9218
4 74.6655 55.8625 0.0000 156.7917 1002.6804
5 75.0358 55.3782 24.9935 178.7412 955.8514
6 76.1394 56.2478 25.6474 198.9585 1053.0069
7 75.5769 59.9913 27.8000 217.4442 1269.1875
8 77.3927 64.3957 36.4021 234.1880 1587.6215
9 71.7988 64.4425 36.5727 239.0194 1828.1666
10 82.0797 68.9404 35.6260 243.5498 1889.8041
11 83.6198 70.8904 35.8123 247.4587 1792.2188
12 70.5678 74.0332 37.0829 250.6104 1877.7057
13 74.7289 70.1897 38.3569 262.2292 1784.4954
Fig. 6. Hydro plant discharge (×104 m3 ) without prohibited discharge zones.
14 81.1808 71.2667 36.4771 272.4351 1738.6404
15 80.2920 71.7134 37.0921 273.9726 1666.9299
Table 1 16 77.7657 74.8203 38.8893 274.3680 1604.1567
Hourly plant discharge(×104 m3 ) without prohibited zones (case 1). 17 78.6181 74.4294 40.4080 277.9838 1658.5607
18 77.9269 70.9925 43.6449 285.3427 1662.0930
Hour Plant 1 Plant 2 Plant 3 Plant 4 19 74.0479 75.2151 46.2129 288.5131 1756.0110
20 79.8611 76.2096 50.0281 282.6284 1791.2728
1 9.7411 7.1169 29.9993 13.0000
21 57.9151 76.2314 50.8169 295.5735 1759.4631
2 10.8630 6.3861 30.0000 13.0000
22 74.1956 66.4645 53.1627 299.8488 1626.3284
3 8.8952 6.1000 30.0000 13.0000
23 62.5884 68.1963 54.6092 296.5827 1368.0234
4 8.1116 6.3320 29.9999 13.0000
24 61.9915 64.2009 56.1989 289.6533 1117.9554
5 8.3078 6.1261 18.2585 13.0000
6 8.6060 6.1899 18.1079 13.0001 Total cost of thermal generations ($) 922553.49
7 8.5422 6.7856 17.5518 13.0005
8 8.8659 7.4970 14.7774 13.0000
9 7.7622 7.4589 14.8775 13.0002
fying the reservoir end-volume constraints. From Fig. 3, it is quite
10 9.5013 8.1092 15.5348 13.0000
11 9.6368 8.3829 15.5803 13.0002 apparent that IPSO has the faster convergence than the DE method.
12 7.2715 9.0405 15.6428 13.1777 Moreover, the proposed IPSO, DE and IFEP methods when imple-
13 7.7602 8.3956 16.0838 14.3160 mented on the same machine converged to their corresponding
14 8.6500 8.5386 16.9949 15.4274
optimal solutions in 38.46, 51.26 and 287 s, respectively.
15 8.4013 8.5765 17.1510 15.6069
16 7.9434 9.2683 16.8523 15.6539
17 8.0435 9.4643 16.6387 16.0798 5.2.2. Case 2: quadratic cost with prohibited discharge zones
18 7.9383 9.1650 15.7577 16.9917 The optimal hydrothermal schedule with minimum thermal
19 7.3904 10.4917 15.0469 17.4428
production cost as 923443.17 ($) is obtained without valve point
20 8.3197 11.2565 13.1868 16.6592
21 5.3882 11.7910 10.0000 18.6965 loading effect and with prohibited hydro discharge zones by the
22 7.4140 9.5642 10.0002 20.1176 proposed IPSO method. The convergence characteristics of IPSO and
23 5.8784 10.2858 10.0089 20.6332 DE algorithms are shown in Fig. 7 and the reservoir storage volumes
24 5.7680 9.6774 10.0748 21.0386 for this case obtained from proposed IPSO method is shown in Fig. 8.
The optimal hydro discharge obtained from proposed IPSO method
is reported in Table 4. From Fig. 7, it is quite apparent that IPSO has
Table 3 provides comparison of the optimal system costs the faster convergence than the DE method in obtaining the optimal
obtained from the proposed IPSO with that of DP, nonlinear pro- solution for this case also.
gramming (NLP), IFEP and differential evolution. The proposed
approach yields better result than DP, NLP, IFEP and DE while satis-
5.2.3. Case 3: cost curve including valve point loading effect and
with prohibited discharge zones
The convergence characteristics of IPSO and DE for this case are
shown in Fig. 9. For this case also the proposed IPSO method has
the faster convergence than the DE method in obtaining the optimal
solution. The optimal hydrothermal schedule out of 25 different ini-
tial trial solutions with hourly hydro plant discharges obtained from
the proposed IPSO method is reported in Table 5. Figs. 10 and 11
show reservoir storage volumes and hydro power generations,

Table 3
Comparison of optimal costs for test system with quadratic cost and no prohibited
discharge zones (case 1).

Method Minimum cost ($)

DP 928919.15
NLP 924249.48
IFEP 930129.82
DE 922555.44
Fig. 7. Convergence characteristics of IPSO and DE algorithms for case 2. IPSO 922553.49
1052 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053

Fig. 8. Hydro reservoir storage volumes with quadratic cost and prohibited dis-
charge zones. Fig. 9. Convergence characteristics of IPSO and DE algorithms for case 3.

Table 4
Hourly plant discharge (×104 m3 ) with prohibited discharge zones only (case 2).

Hour Plant 1 Plant 2 Plant 3 Plant 4

1 9.2868 8.3460 29.7008 13.0000

2 10.3916 6.3853 30.0000 13.0000
3 9.1902 6.0592 29.7270 13.0000
4 7.5796 6.3585 29.7318 13.0000
5 9.2045 6.0000 19.0572 13.0000
6 9.6307 6.2722 17.4694 13.0000
7 9.2709 8.4179 17.2118 13.6820
8 9.0652 8.0441 14.7220 13.0000
9 6.3745 6.8097 15.8388 13.0000
10 9.4701 8.1676 14.3730 13.0000
11 9.2573 8.0816 15.3120 14.4835
12 7.2626 9.0283 15.8715 13.1031
13 7.8521 8.4019 16.4117 13.0000
14 9.3427 8.5189 16.0289 14.3084
15 9.0101 8.5843 17.1513 14.9682
Fig. 10. Hydro reservoir storage volumes with VPL and PDZ.
16 7.2009 9.1666 16.6262 15.2973
17 7.9364 9.2655 16.0502 18.1436
18 7.5351 8.2932 16.7235 18.0408
respectively when obtained from proposed IPSO method. The
19 6.8853 10.0073 15.1547 15.4242
20 9.0066 10.4849 14.3833 18.2754 result, i.e., the optimal cost obtained by the proposed IPSO is com-
21 5.3773 12.0070 10.9653 18.0271 pared with those obtained from DP, NLP, IFEP and DE in Table 6 for
22 7.4140 9.4914 10.0000 19.0167 case 3.
23 5.9352 10.2115 10.0148 20.4558 From Table 6 it is observed that the minimum cost found to
24 5.5203 9.5971 10.0193 21.1190
be 925978.84 ($) by the proposed IPSO method and 928236.94
Table 5 ($) by the DE method as compared to the previous minimum
Hourly plant discharge (×104 m3 ) with VPL and PDZ (case 3). cost reported in the literature obtained by IFEP method [11], i.e.,
Hour Plant 1 Plant 2 Plant 3 Plant 4
933949.25 ($). Similarly, the execution time needed for getting opti-
mal solution are 31.11, 36.82 and 203.02 s by the proposed IPSO,
1 9.9637 8.2736 29.6589 13.4198
DE and IFEP methods, respectively. It is observed that the solution
2 11.7735 6.8989 29.9833 13.1410
3 9.2522 6.0236 29.7248 13.0446 time decreases with valve point loading (VPL) and prohibited dis-
4 10.9340 8.3357 29.1396 13.0459
5 7.3742 6.9743 13.2820 13.0698
6 7.2507 6.1087 28.6740 13.0469
7 5.3273 9.1245 16.2364 13.0009
8 6.5324 6.3395 12.4942 13.0124
9 9.6014 6.0421 12.2137 13.0113
10 10.1383 10.2283 12.9690 13.1431
11 9.9104 6.0998 11.7375 13.3492
12 5.3277 6.1669 13.6945 13.0118
13 9.9914 6.3674 17.3425 13.3196
14 6.4250 6.3360 18.7418 13.0067
15 7.9234 8.2533 14.0969 13.4098
16 6.7321 6.6511 15.4041 13.0627
17 9.6811 11.2585 17.1248 15.1722
18 7.5019 11.3390 16.7130 18.3091
19 9.4432 6.4588 15.1047 13.7489
20 5.0691 13.1986 12.9095 18.7675
21 7.2319 14.2906 10.0935 15.7652
22 7.3055 12.4959 11.7400 21.2703
23 7.9151 12.3818 10.7841 19.7107
24 6.3945 6.3531 13.8968 21.2058
Fig. 11. Hourly hydro power generations with VPL and PDZ.
 P.K. Hota et al. / Electric Power Systems Research 79 (2009) 1047–1053 1053

Table 6 Vij volume of water stored in ith reservoir at interval j

Comparison of optimal costs for test system with VPL and PDZ (case 3).
Vi1 initial reservoir storage of hydro plant i
Method Minimum cost ($) Vi25 final reservoir storage of hydro plant i
pro
DP 935617.76
qi prohibited discharge zone of ith hydro unit
NLP 936709.52 Z a constant used to limit the velocity of particles
IFEP 933949.25
DE 928236.94
References
IPSO 925978.84

[1] A.J. Wood, B.F. Wollenberg, Power Generation Operation and Control, John Wiley
and Sons, New York, 1984.
charge zones (PDZ) as compared to case 1. The simulation results [2] M.F. Carvalloh, S. Soares, An efficient hydrothermal scheduling algorithm, IEEE
for all the three cases reveal that the proposed IPSO gives cheaper Trans. PWRS 4 (1987) 537–542.
generation schedule within smaller execution time when com- [3] L. Engles, R.E. Larson, J. Peschon, K.N. Stanon, Dynaming programming applied
to hydro and thermal generation scheduling, in: IEEE tutorial course text,
pared with DP, NLP, IFEP and DE methods. Therefore, the proposed 76CH1107-2-PWR, IEEE, New York, 1976.
IPSO technique appears to be the best in terms of convergence [4] S. Chang, C. Chen, I. Fung, P.B. Luh, Hydroelectric generation scheduling with an
speed, solution time and minimum cost among other established effective differential dynamic programming, IEEE Trans. Power Syst. 5 (1990)
737–743.
methods. [5] Q. Xia, N. Xiang, S. Wang, B. Zhang, M. Huang, Optimal daily scheduling of cas-
caded plants using a new algorithm of nonlinear minimum cost network flow,
6. Conclusion IEEE Trans. Power Syst. 3 (3) (1988) 929–935.
[6] K.P. Wong, Y.W. Wong, Short term hydrothermal scheduling: Part 1. Simulated
annealing approach, IEE Proc. C 141 (1994) 497–501.
This paper presents a new approach to short-term hydrothermal [7] Y.-G. Wu, C.-Y. Ho, D.-Y. Wang, A diploid genetic approach to short-term schedul-
scheduling while preserving the inherent properties of PSO algo- ing of hydrothermal systems, IEEE Trans. Power Syst. 15 (November (5)) (2000)
1268–1274.
rithm. The proposed dynamic search-space squeezing strategy is
[8] S.O. Orero, M.R. Irving, A genetic algorithm modeling framework and solution
used to satisfy the inequality constraints and to reduce the search- technique for short-term optimal hydrothermal scheduling, IEEE Trans. Power
space margin judiciously to make convergence speed faster. The Syst. 13 (2) (1998) 501–518.
IPSO has provided the best solution for the hydrothermal system [9] P.-H. Chen, H.-C. Chang, Genetic added scheduling of hydraulically coupled
plants in hydro-thermal coordination, IEEE Trans. Power Syst. 11 (May (2))
as compared to other heuristic methods like IFEP and DE. The pro- (1996) 975–981.
posed approach has produced results better than those reported by [10] P.K. Hota, R. Chakrabarti, P.K. Chattopadhyay, Short-term hydrothermal
other algorithms and the solutions obtained, have better solution scheduling through evolutionary technique, EPSR 52 (November (2)) (1999)
189–196.
quality and good convergence characteristics. The IPSO approach [11] N. Sinha, R. Chakrabarti, P.K. Chattopadhyay, Fast evolutionary technique for
can easily be extended to other complex optimization problems short-term hydrothermal scheduling, IEEE Trans. Power Syst. 18 (1) (2003)
faced by the utilities. 214–220.
[12] J. Kennedy, R. Eberhart, Particle swarm optimization, in: Proceedings of the IEEE
International Conference on Neural Networks, vol. 4, Perth, Australia, 1995, pp.
Appendix A. List of symbols 1942–1948.
[13] Y. Shi, R.C. Eberhart, A modified particle swarms optimizer, in: Proceedings of
the IEEE International Conference on Evol. Comput., May, 1988, pp. 69–73.
[14] R.C. Eberhart, Y. Shi, Comparison between genetic algorithms and particle
F total generation cost swarm optimization, in: Proceedings of the IEEE International Conference on
fj (PTj ) fuel cost of PTj during jth interval Evol. Comput., May, 1998, pp. 611–616.
[15] M.A. Abido, Optimal design of power system stabilizers using particle swarm
PTj power generation of thermal unit at interval j optimization, IEEE Trans. Energy Conv. 17 (3) (2002) 406–413.
m number of scheduling intervals [16] H. Yoshida, K. Kawata, Y. Fukuyama, S. Takayama, Y. Nakanishi, Particle swarm
n number of hydro plants optimization for reactive power and voltage control considering voltage secu-
rity assessment, IEEE Trans. Power Syst. 15 (2000) 1232–1239.
PHij power generation of ith hydro unit at interval j [17] Z.L. Gaing, Particle swarm optimization to solving the economic dispatch con-
PDj system load demand at interval j sidering the generator constraints, IEEE Trans. Power Syst. 18 (August (3)) (2003)
PLossj system transmission losses at interval j 1187–1195.
[18] J.B. Park, K.S. Lee, J.R. Shin, K.Y. Lee, A particle swarm optimization for economic
qij water discharge rate of ith hydro unit at interval j dispatch with non-smooth cost functions, IEEE Trans. Power Syst. 20 (1) (2005)
rij inflow rate into the storage reservoir of ith hydro plant at 34–42.
interval j [19] K. Price, R. Storn, Differential evolution: numerical optimization made easy, Dr.
Dobbs J. 22 (4) (1997) 18–24.
sij spillage of ith reservoir at interval j