Available online at www.sciencedirect.

com

Applied Soft Computing 8 (2008) 1392–1399

www.elsevier.com/locate/asoc

Particle swarm optimization technique based short-term

hydrothermal scheduling
K.K. Mandal *, M. Basu, N. Chakraborty
Department of Power Engineering, Jadavpur University, Kolkata 700098, India
Received 14 September 2006; received in revised form 13 October 2007; accepted 16 October 2007
Available online 22 October 2007

Abstract
Particle swarm optimization is applied to determine the optimal hourly schedule of power generation in a hydrothermal power system. A multi-
reservoir cascaded hydroelectric system with a nonlinear relationship between water discharge rate, net head and power generation is considered.
The water transport delay between connected reservoirs is taken into account. In the present work, the effects of valve point loading in the fuel cost
function of the thermal plants are also taken into consideration. The developed algorithm is illustrated for a test system consisting of four hydro
plants and three thermal plants. Cost characteristics of individual thermal units are considered. The test results are compared with those obtained
using evolutionary programming and simulated annealing technique. It is found that the convergence characteristic is excellent and the results
obtained by the proposed method are superior in terms of fuel cost and computation time.
# 2007 Elsevier B.V. All rights reserved.

Keywords: Hydrothermal scheduling; Cascaded reservoirs; Particle swarm optimization

1. Introduction make the optimization problem more tractable. Some of these

solution methods are dynamic programming [1], functional
Optimum scheduling of power plant generation is of great analysis [2], network flow [3] and mathematical decomposition
importance to electric utility systems. With the insignificant [4]. Recently simulated annealing technique [5], evolutionary
marginal cost of hydroelectric power, the problem of programming technique [6] and genetic algorithm [7–9] have
minimizing the operational cost of a hydrothermal system been applied separately for optimal hydrothermal scheduling
essentially reduces to that of minimizing the fuel cost for problem and circumvented the above-mentioned weakness.
thermal plants under the various constraints on the hydraulic Sinha et al. used fast evolutionary programming to solve short-
and power system network. The main constraints include: the term hydrothermal scheduling problem and presented promis-
time coupling effect of the hydro sub problem, where the water ing results [10]. Yuan and Yuan discussed the application of
flow in an earlier time interval affects the discharge capability at cultural algorithm to the generation scheduling of hydrothermal
a later period of time, the cascaded nature of the hydraulic systems [11].
network, the varying hourly reservoir inflows, the physical Particle swarm optimization (PSO) happens to be a
limitations on the reservoir storage and turbine flow rate, the comparatively new combinatorial metaheuristic technique
varying system load demand and the loading limits of both which is based on the social metaphor of bird flocking or
thermal and hydro plants. fish schooling [12,13]. This algorithm has come to existence in
The hydrothermal scheduling problem has been the subject mid 90s and has gained prominence from late 90s. Early works
of investigation for several decades. Most of the methods that on PSO have shown the rich promise of emergence of a
have been used to solve the hydrothermal co-ordination relatively simple optimization technique which is easier to
problem make a number of simplifying assumptions in order to understand compared to other evolutionary computation
techniques presently available e.g. genetic algorithm, evolu-
tionary programming. Another advantage of PSO can be the
* Corresponding author. Tel.: +91 33 23355813; fax: +91 33 23357254. possibility of tuning smaller number of free, tunable parameters
E-mail address: [email protected] (K.K. Mandal). to arrive at the desired goal. The PSO technique has been

1568-4946/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.asoc.2007.10.006
 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399 1393

reservoir inflows, the physical limitations on the reservoir

Nomenclature storage and turbine flow rate and the loading limits of both
thermal and hydro plants.
Psim output power of ith thermal unit at time m
The success of any evolutionary optimization technique is
Pmin
si ; P max
si lower and upper generation limits for ith ther-
largely dependent on the methods of constraints handling. Most
mal unit
of the evolutionary algorithms were originally developed to
asi, bsi, csi, dsi, esi cost curve coefficients of ith thermal
solve unconstrained optimization problems. However, over the
unit
last few decades, several methods have been proposed to handle
PDm load demand at time m
constraints in evolutionary algorithms, such as PSO. Micha-
Phjm output power of jth hydro unit at time m
lewicz et al presented a complete review of constrained
Pmin
hj ; P max
hj lower and upper generation limits for jth hydro
optimization problems in evolutionary algorithms [17]. In this
unit
paper, the strategy to generate and keep control variable in the
Qhjm water discharge rate of jth reservoir at time m
feasible region is used for constraints handling [17]. This paper
Vhjm storage volume of jth reservoir at time m
also considers the cost characteristics of individual thermal
Qminhj ; Q max
hj minimum and maximum water discharge rate
units instead of considering an equivalent thermal unit. To
of jth reservoir
validate the PSO-based hydrothermal scheduling algorithm, the
Vhmin max
j ; Vh j minimum and maximum storage volume of jth
developed algorithm has been illustrated for a test system [18].
reservoir
The same problem has also been solved by evolutionary
C1j, C2j, C3j, C4j, C5j, C6j power generation coefficients of
programming (EP) and simulated annealing technique (SA) and
jth hydro unit
the results are compared. The performance of the proposed
Ihjm inflow rate of jth reservoir at time m
method is found to be quite encouraging as compared with
Ruj number of upstream units directly above jth
other methods
hydro plant
Shjm spillage of jth reservoir at time m
2. Problem formulation
tlj water transport delay from reservoir l to j
Ns number of thermal generating units
Due to zero incremental cost of hydro generating units, the
Nh number of hydro generating units
hydrothermal scheduling problem is aimed to minimize the fuel
m, M time index and scheduling period
cost of thermal plants, while making use of the availability of
hydro power as much as possible. The objective function and
associated constraints of the hydrothermal scheduling problem
are formulated as follows.
applied to various fields of power system optimization. Gaing
used PSO to solve economic dispatch problem considering 2.1. Objective function
generator constraints [12]. Yoshida et al. applied PSO to control
reactive power and voltage considering voltage security The fuel cost function of each thermal generating unit
arrangement [13] and have presented promising results. Abido considering valve-point effects [18] is expressed as the sum of a
proposed a revised PSO technique for optimal design of voltage quadratic and a sinusoidal function. The total fuel cost in terms
stabilizer [14]. Park et al. presented a method for solving of real power output can be expressed as
economic dispatch with non-smooth cost functions [15]. Yu 
X Ns 
M X 
et al. applied particle swarm optimization technique to solve
short-term hydrothermal scheduling [16] with an equivalent f ¼ asi þ bsi Psim þ þ d si
csi P2sim
m¼1 i¼1
thermal unit having smooth cost functions connected to hydel   

systems. Here the constraints were handled by penalty function min
 sin esi  Psi  Psim  (1)

method [16]. But the performance of PSO to short-term
hydrothermal scheduling for interconnected individual thermal
units with non-smooth cost function has not been tested yet. 2.2. Constraints
This paper proposes PSO method for short-term optimal
scheduling of generation in a hydrothermal system which (i) Power balance constraints
involves the allocation of generation among the multi-reservoir The total active power generation must balance the
cascaded hydro plants and thermal plants with valve point predicted power demand, at each time interval over the
loading so as to minimize the fuel cost of thermal plants while scheduling horizon
satisfying the various constraints on the hydraulic and power
system network. The main constraints include: the time X
Ns X
Nh

coupling effect of the hydro sub problem, where the water Psim þ Ph jm  PDm ¼ 0; m2M (2)
i¼1 j¼1
flow in an earlier time interval affects the discharge capability at
a later period of time, the varying system load demand, the The hydroelectric generation is a function of water dis-
cascade nature of the hydraulic network, the varying hourly charge rate and reservoir water head, which in turn, is a
1394 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399

function of storage. for the particle i is represented as vi ¼ ðvi1 ; vi2 ; :::; vid Þ. The
modified velocity and position of each particle can be
Ph jm ¼ C 1 j Vh2 jm þ C2 j Q2h jm þ C 3 j V h jm Qh jm þ C4 j V h jm calculated using the current velocity and the distance from
þ C5 j Qh jm þ C6 j ; j 2 Nh m 2 M (3) pbestid to gbestd as shown in the following formulas:
(ii) Generation limits ðkþ1Þ ðkÞ
vid ¼ w  vid þ C 1  randðÞ  ðpbestid  xkid Þ þ C 2
Pmin max
h j  Ph jm  Ph j ; j 2 Nh m 2 M (4)
 randðÞ  ðgbestd  xkid Þ (9)
and
ðkþ1Þ ðkþ1Þ
xid ¼ xkid þ vid ; i ¼ 1; 2; :::; N P ; d ¼ 1; 2; :::; N g (10)
Pmin max
si  Psim  Psi ; i 2 Ns; m 2 M (5)
(iii) Hydraulic network constraints where, NP is the number of particles in a group, Ng the number
of members in a particle, k the pointer of iterations, w the inertia
The hydraulic operational constraints comprise the water
weight factor, C1, C2 the acceleration constant, rand() the
balance equations for each hydro unit as well as the bounds on ðkÞ
uniform random value in the range [0,1], vi the velocity of
reservoir storage and release targets. These bounds are
a particle i at iteration k, vd  vid  vd , and xki is the current
min k max
determined by the physical reservoir and plant limitations as
position of a particle i at iteration k.
well as the multipurpose requirements of the hydro system.
In the above procedures, the parameter vmax determined
These constraints include:
the resolution, with which regions are to be searched between
the present position and the target position. If vmax is too
(a) Physical limitations on reservoir storage volumes and
high, particles might fly past good solutions. If vmax is too
discharge rates,
small, particles may not explore sufficiently beyond local
Vhmin max
j  V h jm  Vh j ; j 2 Nh; m 2 M (6) solutions.
The constants C1 and C2 represent the weighting of the
Qmin max
h j  Qh jm  Qh j ; j 2 N h; m 2 M (7) stochastic acceleration terms that pull each particle toward the
pbest and gbest positions. Low values allow particle to roam far
(b) The continuity equation for the hydro reservoir network
from the target regions before being tugged back. On the other
Ru j
X hand, high values result in abrupt movement toward or past,
V h jðmþ1Þ ¼ V h jm þ I h jm  Qh jm  Sh jm þ ðQhlðmtl j Þ target regions. Hence, the acceleration constants C1 and C2
l¼1 were often set to be 2.0 according to past experiences.
þ Shlðmtl j Þ Þ; j 2 Nh; m 2 M (8) Suitable selection of inertia weight w provides a balance
between global and local explorations, thus requiring less
iteration on average to find a sufficiently optimal solution. As
3. Particle swarm optimization originally developed, w often decreases linearly from about 0.3
to 0.2 during a run. In general, the inertia weight w is set
Particle swarm optimization [12,13] is one of the most according to the following equation:
recent developments in the category of combinatorial
wmax  wmin
metaheuristic optimizations. This method has been developed w ¼ wmax   iter (11)
under the scope of artificial life where PSO is inspired by the itermax
natural phenomenon of fish schooling or bird flocking. PSO is where itermax is the maximum number of iterations and iter is
basically based on the fact that in quest of reaching the the current number of iterations.
optimum solution in a multidimensional space, a population of
particles is created whose present coordinate determines the 4. Development of the proposed algorithm
cost function to be minimized. After each iteration the new
velocity and hence the new position of each particle is updated In this section, an algorithm based on particle swarm
on the basis of a summated influence of each particle’s present optimization method for solving hydrothermal scheduling
velocity, distance of the particle from its own best performance, problem is described below.
achieve so far during the search process and the distance of the Let pk ¼ ½Ps1 ; Ps2 ; :::; Psi ; :::; PsN s ; Qh1 ; Qh2 ; :::; Qh j ; :::; Qh
particle from the leading particle, i.e. the particle which at N h T be a trial matrix designating the kth individual of a
present is globally the best particle producing till now the best population to be evolved and Psi = [Psi1, Psi2, . . ., Psim, . . .,
performance i.e. minimum of the cost function achieved so far.
Let x and v denote a particle position and its corresponding Table 1
velocity in a search space, respectively. Therefore, the ith Comparison of cost and CPU time
particle is represented as xi = (xi1, xi2, . . ., xid) in the d
Method Cost ($) CPU time (s)
dimensional space. The best previous position of the ith particle
is recorded and represented as pbesti = (pbesti1, pbesti2, . . ., PSO 44,740 232.73
EP 47,306 9879.45
pbestid). The index of the best particle among all the particles in
SA 45,466 246.19
the group is represented by the gbestd. The rate of the velocity
 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399 1395

Table 2
Hydrothermal generation (MW) schedule using PSO
Hour Ph1 Ph2 Ph3 Ph4 Ps1 Ps2 Ps3
1 78.4789 58.9698 51.5360 190.0776 104.5031 213.8706 52.5640
2 98.7472 51.3930 35.5855 215.9900 23.8279 124.9101 229.5464
3 59.1493 62.7712 0 200.5065 110.5834 126.5341 140.4554
4 73.2112 82.4517 40.7559 157.4580 20.8114 135.5511 139.7606
5 94.5033 75.1698 0 86.8169 172.6196 124.8379 116.0525
6 73.3762 56.9390 46.7195 214.4880 42.1109 47.0870 319.2795
7 53.2180 83.6461 49.0503 214.1431 28.2230 292.1999 229.5196
8 71.0535 46.8312 36.8237 218.7056 102.6407 124.9065 409.0387
9 59.9540 55.7740 50.0264 112.9635 102.6688 209.8155 498.7978
10 57.1109 57.9154 30.6000 208.1164 102.5504 124.9079 498.7989
11 53.4066 64.8689 43.2134 225.4627 95.8228 119.2332 497.9924
12 72.2461 65.1559 24.9751 248.1211 102.6756 295.7209 341.1052
13 70.2261 45.2693 49.9454 225.5011 102.3307 124.5046 492.2228
14 55.8106 64.4608 39.7461 253.3341 174.3252 124.3798 317.9433
15 89.6494 47.9316 51.6412 233.9556 54.6200 212.9228 319.2794
16 56.0929 71.8105 52.7541 217.8425 174.5684 209.8153 277.1162
17 67.6928 48.0426 48.2623 254.5290 102.3788 209.8158 319.2787
18 102.5615 52.9784 54.5686 270.9449 174.9935 144.6741 319.2791
19 77.2032 42.7816 48.4496 265.5154 106.7311 210.0397 319.2794
20 106.3199 47.5832 25.5896 273.9399 157.6725 209.3763 229.5186
21 89.2563 43.3299 53.6018 262.9318 21.5424 209.8180 229.5198
22 72.2180 59.6648 56.3912 216.4961 102.6498 123.0606 229.5194
23 71.9877 44.0812 53.8873 225.1673 102.0467 123.3101 229.5196
24 94.0653 73.0342 56.0243 281.2591 25.2289 40.8680 229.5203

PsiM], Qhj = [Qhj1, Qhj2, . . ., Qhjm, . . ., QhjM]. The elements Psim equation:
and Qhjm are the power output of the ith thermal unit and the
discharge rate of the jth hydro plant at time m. The range of the X
Ns X
Nh

elements Psim and Qhjm should satisfy the thermal generating Psdg m ¼ PDm  Psim  Ph jm ; m2M (14)
i¼1 j¼1
capacity and the water discharge rate constraints in Eqs. (5) and
(7), respectively. Assuming the spillage in Eq. (8) to be zero for i 6¼ d
simplicity, the hydraulic continuity constraints are
Table 3
X
M M X
X Ru j X
M Hourly plant discharge (104 m3) using PSO
V h j0  V h jM ¼ Qh jm  Qhlðmtl j Þ  I h jm ; Hour Qh1 Qh2 Qh3 Qh4
m¼1 m¼1 l¼1 m¼1 (12)
1 8.5555 7.4992 18.1231 10.8422
j 2 Nh 2 14.3348 6.2990 21.3399 14.5900
3 5.9070 7.9036 28.7327 14.4767
To meet exactly the restrictions on the initial and final 4 7.7853 11.8477 17.5560 11.2848
5 12.8422 10.4999 29.2430 6.0004
reservoir storage in Eq. (8), the water discharge rate of 6 8.2472 7.3769 12.7561 19.4849
the jth hydro plant Qhjd in the dependent interval d is then 7 5.4138 13.4997 12.3870 18.8234
calculated by 8 7.7772 6.6029 19.9857 17.1731
9 6.1193 8.0078 16.0197 6.0037
X
M Ru j
M X
X 10 5.6243 8.3886 21.1618 12.3119
Qh jd ¼ V h j0  V h jM þ I h jm þ Qhlðmtl j Þ 11 5.0484 9.6550 18.3446 14.1618
m¼1 m¼1 l¼1
12 7.2286 9.8157 21.9390 17.4442
13 6.8873 6.5185 11.2568 14.0701
X
M
14 5.1092 9.7192 18.8320 17.4400
 Qh jm; j 2 Nh (13) 15 9.5568 6.8461 13.9569 14.3529
m¼1 16 5.1020 11.1883 14.4914 12.1962
17 6.3903 6.9757 16.7958 15.1253
m 6¼ d
18 11.9721 7.7824 11.9124 17.8042
19 7.5847 6.3569 17.4819 16.8772
The dependent water discharge rate must satisfy the constraints 20 13.4012 7.0271 22.7354 18.5884
in Eq. (7). 21 9.5754 6.2713 10.6628 17.5893
Also to meet exactly the power balance constraints in 22 7.0465 8.6510 13.0717 12.1497
23 7.0015 6.0775 16.8621 13.0607
Eq. (2), the thermal generation Psdg m of the dependent thermal
24 10.4896 11.1899 10.2428 20.0000
generating unit dg can then be calculated using the following
1396 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399

Table 4
Hydrothermal generation (MW) schedule using EP
Hour Ph1 Ph2 Ph3 Ph4 Ps1 Ps2 Ps3
1 59.6971 76.1105 58.2978 136.2429 20.0000 40.0000 359.6518
2 98.1400 54.2710 51.1128 133.9707 20.0000 300.0000 122.5055
3 74.2949 59.9475 0 125.0788 175.0000 41.7699 223.9089
4 52.5262 48.6082 0 123.0997 144.1324 201.2153 80.4182
5 78.9486 61.1910 43.2268 131.5164 110.0883 127.4410 117.5880
6 55.7000 50.5689 35.8694 185.3684 142.0702 279.5960 50.8271
7 55.2889 66.5631 47.7782 141.9622 20.0000 300.0000 318.4076
8 53.2603 81.5457 41.5613 277.5724 20.0000 300.0000 236.0603
9 53.9010 64.9030 0 256.9061 20.0000 300.0000 394.2899
10 54.5130 66.8582 44.6110 266.1023 24.8543 264.7809 358.2803
11 64.3684 79.2279 45.0740 298.5612 20.0000 294.1676 298.6010
12 102.2673 69.3926 46.3391 287.0284 20.0000 300.0000 324.9727
13 107.8884 66.5722 49.4810 285.6861 165.9274 300.0000 134.4448
14 88.2555 40.2006 49.8349 270.4288 20.0000 231.1308 330.1494
15 96.5846 50.7539 44.5557 249.5699 34.8315 271.3955 262.3090
16 74.8178 44.8537 22.5062 196.3611 175.0000 300.0000 246.4611
17 99.0936 63.7453 55.0776 261.2482 175.0000 206.4108 189.4245
18 96.6291 59.6989 56.4442 261.8895 175.0000 281.8639 188.4743
19 88.7707 41.3290 54.2310 262.4874 20.0000 182.7914 420.3905
20 56.8347 69.5851 19.3848 280.7036 91.6543 199.6279 332.2096
21 54.8278 39.8686 30.6204 224.9936 85.1912 300.0000 174.4984
22 54.9523 45.3750 51.4514 282.6280 73.7683 40.0000 311.8250
23 95.2658 43.6099 55.7569 270.8351 20.0000 116.4898 248.0425
24 90.9315 73.8509 57.5457 203.8075 20.0000 300.0000 53.8645

The dependent thermal generation must satisfy the constraints dimensions, searching points and velocities. These
in Eq. (5). initial particles must be feasible candidate solutions
The cost function f is to be minimized. Now the entire that satisfy the practical operating constraints.
algorithm can be described as follows. Step 2. Calculate the cost function f of each individual pk in the
population.
Step 1. Initialize randomly the particles of the population Step 3. Compare each particle’s cost value with that of its
according to the limit of each unit including individual pbest. The particle with the best cost value among the
pbests is denoted as gbest.
Table 5 Step 4. Modify the member velocity of each particle according
Hourly plant discharge (104 m3) using EP to Eq. (9).
Hour Qh1 Qh2 Qh3 Qh4 Step 5. Modify the member position of each particle according
1 5.8525 10.7864 14.6691 6.3251 to Eq. (10).
2 13.1523 7.0291 17.6645 6.3937 Step 6. If the new cost value for any kth particle is less than its
3 7.8700 7.8456 30.0000 6.0000 previous value, the new coordinates for that particle
4 5.0000 6.0000 28.5065 6.1642 will be stored as its pbestk. Also compare the cost
5 8.5242 7.6300 16.7988 7.3122
values of all the pbestk for each particle k and determine
6 5.3807 6.0000 19.0743 11.4903
7 5.2913 8.3973 10.0650 7.0841 gbest.
8 5.0000 11.9606 18.0863 19.1168 Step 7. If the number of iterations reaches the maximum, then
9 5.0000 8.9380 29.1460 14.9476 go to Step 8. Otherwise, go to Step 4.
10 5.0000 9.4423 10.5976 15.8222 Step 8. The individual that generates the latest gbest is the
11 6.0529 12.6007 14.1040 20.0000
solution of the problem.
12 12.2523 10.6156 10.0000 19.8768
13 15.0000 10.4054 12.3968 20.0000
14 9.4450 6.0000 10.0000 16.2756 5. Simulation results
15 11.0065 7.2962 19.0175 14.4597
16 7.3368 6.2104 23.9402 9.4208 The proposed method has been applied to a test system
17 11.4684 9.1742 10.0000 15.8038
[18] which consists of a multi-chain cascade of four hydro
18 11.0076 8.6902 10.0000 16.3224
19 9.5464 6.0000 17.9785 17.3262 units and three thermal units. The scheduling period is 24 h
20 5.2249 11.1026 25.1126 20.0000 with 1 h time interval. The hydro sub-system configuration
21 5.0000 6.0000 23.1910 12.2730 and network matrix including the water time delays are
22 5.0000 6.5063 18.4159 20.0000 shown in Fig. 4. The load demand, hydro unit power
23 10.6965 6.0000 10.0000 20.0000
generation coefficients, reservoir inflows, reservoir limits
24 9.8917 11.3692 16.0716 11.5431
are given in Tables 8–11, respectively. The generation
 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399 1397

Table 6
Hydrothermal generation (MW) schedule using SA
Hour Ph1 Ph2 Ph3 Ph4 Ps1 Ps2 Ps3
1 78.3903 74.1385 56.3559 162.3815 24.2116 125.0026 229.5195
2 97.7979 50.1356 30.8119 128.7342 38.0349 294.7246 139.7610
3 63.3947 60.3407 9.1223 194.7561 102.7251 40.1375 229.5238
4 97.3098 75.6792 0 113.9027 21.5228 112.0660 229.5194
5 54.1858 80.9770 50.4874 205.1282 96.7997 42.6614 139.7605
6 71.0311 57.3973 51.6163 201.8774 116.9828 40.1512 260.9438
7 57.9326 47.2780 54.5377 227.3499 123.1722 210.2103 229.5193
8 58.9483 61.4541 42.7358 202.9926 29.8131 294.7759 319.2800
9 77.3869 45.1242 39.9115 237.6656 20.3957 260.4757 409.0404
10 80.7412 53.2376 51.6567 136.7474 102.1729 156.6664 498.7778
11 93.7264 57.8925 54.8321 256.1628 108.1426 209.9641 319.2795
12 61.7638 73.3491 0 248.8246 148.7360 208.2874 409.0392
13 57.8838 61.9687 53.3924 174.7167 159.2448 283.5146 319.2790
14 55.0856 56.2009 14.0878 262.4338 139.0348 273.6384 229.5187
15 68.5416 65.8840 54.0030 253.7398 128.4827 209.8242 229.5203
16 59.0146 47.4124 18.3823 234.6083 171.4950 209.8156 319.2718
17 56.8386 50.1895 53.4710 265.5626 90.4497 124.4824 409.0063
18 84.3849 72.4405 52.3568 267.5360 106.0124 125.7594 411.5100
19 64.5248 57.3962 47.2861 237.1270 122.4115 42.4553 498.7992
20 93.8746 43.8015 15.4593 269.0357 99.6856 124.8930 403.2503
21 100.6623 73.4125 51.2289 229.7599 20.4354 294.7417 139.7594
22 105.1393 69.7763 51.6995 250.5860 103.0717 50.2075 229.5197
23 65.6108 51.2800 53.3198 271.5090 44.7271 134.0334 229.5200
24 63.0104 48.7320 56.3684 279.9623 106.8715 105.2957 139.7597

limits, cost coefficients of thermal units are given in this test system. In case of SA, for the same test system, initial
Table 12. temperature, scaling factor for Gaussian probability distribu-
In applying the proposed PSO algorithm for the test system, tion and the maximum trials per iteration are set to 50,000,
the appropriate values of population size NP and maximum 0.00001 and 30, respectively.
iteration number Nmax are set to the values of 30 and 100, Table 1 shows the total cost and CPU time obtained from
respectively. In case of EP population size NP and maximum PSO, EP and SA technique. The determined hydrothermal
iteration number Nmax are set to 50 and 1000, respectively for generation schedules and water discharge rates using proposed
PSO method are shown in Tables 2 and 3. The determined
Table 7 hydrothermal generation schedules and water discharge rates
Hourly plant discharge (104 m3) using SA using EP are shown in Tables 4 and 5. The determined
Hour Qh1 Qh2 Qh3 Qh4 hydrothermal generation schedules and water discharge rates
1 8.5403 10.3345 16.1051 8.3866 using SA are shown in Tables 6–12. Figs. 1–4 show the cost
2 13.6043 6.3644 22.3596 6.1407 convergence obtained from PSO, EP and SA, respectively. The
3 6.4308 7.7943 24.8594 12.3084
4 14.3844 10.6079 26.6646 6.0693
5 5.4685 12.2736 11.8965 16.1777
6 7.7936 7.8035 12.0580 15.7032
7 5.9511 6.3001 14.3439 18.4727
8 6.0008 8.6488 19.5957 13.7264
9 8.6089 6.1708 20.4401 16.2504
10 9.1343 7.2219 16.4284 6.6782
11 12.2123 7.7924 12.1285 18.9804
12 6.1475 10.6618 28.7205 18.6905
13 5.5601 8.6806 13.4009 9.6938
14 5.1280 7.7454 24.6328 18.6024
15 6.6141 9.3579 11.8211 17.6093
16 5.4370 6.2803 23.9945 15.7715
17 5.1819 6.5286 13.9335 18.1180
18 8.6081 10.4912 15.4033 19.3797
19 6.0256 8.1135 17.7509 14.2797
20 10.1989 6.0791 23.8518 19.0581
21 11.8616 11.3723 13.4090 13.2003
22 13.9035 10.8862 14.5428 15.4845
23 6.2784 7.4808 10.7278 18.4074
24 5.9260 7.0101 12.6499 19.9997
Fig. 1. Cost convergence from PSO method.
1398 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399

Table 8 Table 9
Load demand Hydro power generation coefficients
Hour PD (MW) Hour PD (MW) Hour PD (MW) Plant C1 C2 C3 C4 C5 C6
1 750 9 1090 17 1050 1 0.0042 0.42 0.030 0.90 10.0 50
2 780 10 1080 18 1120 2 0.0040 0.30 0.015 1.14 9.5 70
3 700 11 1100 19 1070 3 0.0016 0.30 0.014 0.55 5.5 40
4 650 12 1150 20 1050 4 0.0030 0.31 0.027 1.44 14.0 90
5 670 13 1110 21 910
6 800 14 1030 22 860
7 950 15 1010 23 850
8 1010 16 1060 24 800

Table 10
Reservoir inflows (104 m3)
Hour Reservoir Hour Reservoir Hour Reservoir
1 2 3 4 1 2 3 4 1 2 3 4
1 10 8 8.1 2.8 9 10 8 1 0 17 9 7 2 0
2 9 8 8.2 2.4 10 11 9 1 0 18 8 6 2 0
3 8 9 4 1.6 11 12 9 1 0 19 7 7 1 0
4 7 9 2 0 12 10 8 2 0 20 6 8 1 0
5 6 8 3 0 13 11 8 4 0 21 7 9 2 0
6 7 7 4 0 14 12 9 3 0 22 8 9 2 0
7 8 6 3 0 15 11 9 3 0 23 9 8 1 0
8 9 7 2 0 16 10 8 2 0 24 10 8 0 0

software is written in Matlab language and executed on a Table 11

Pentium IV personal computer. The optimality of the solutions Reservoir storage capacity limits, plant discharge limits, reservoir end condi-
tions (104 m3) and plant generation limits (MW)
achieved by the proposed PSO-based method is superior to that
of EP-based and SA-based methods in terms of the cost and also Plant Vmin Vmax Vini Vend Qmin Qmax Pmin
h
Pmax
h

less CPU time is required in case of PSO method. 1 80 150 100 120 5 15 0 500
2 60 120 80 70 6 15 0 500
3 100 240 170 170 10 30 0 500
4 70 160 120 140 6 20 0 500

Fig. 2. Cost convergence from EP method. Fig. 3. Cost convergence from SA method.
 K.K. Mandal et al. / Applied Soft Computing 8 (2008) 1392–1399 1399

Table 12
Cost curve coefficients and operating limits of thermal generators
Unit as ($/h) bs ($/MWh) cs ($/(MW)2 h) ds ($/h) es (1 MW1) Pmin
s (MW) Pmax
s (MW)

1 100 2.45 0.0012 160 0.038 20 175

2 120 2.32 0.0010 180 0.037 40 300
3 150 2.10 0.0015 200 0.035 50 500

[2] S.A. Soliman, G.S. Christensen, Application of functional analysis to

optimization of variable head multi reservoir power system for long term
regulation, Water Resour. Res. 22 (6) (1986) 852–858.
[3] Q. Xia, N. Xiang, S. Wang, B. Zhang, M. Huang, Optimal daily scheduling
of Cascaded plants using a new algorithm of non-linear minimum cost
network flow, IEEE Trans. PWRS 3 (3) (1988) 929–935.
[4] M.V.F. Pereira, L.M.V.G. Pinto, A decomposition approach to the eco-
nomic dispatch of the hydrothermal systems, IEEE Trans. PAS 101 (10)
(1982) 3851–3860.
[5] K.P. Wong, Y.W. Wong, Short-term hydrothermal scheduling part 1:
simulated annealing approach, IEE Proc. Gen. Transm. Distribution
141 (5) (1994) 497–501.
[6] P.C. Yang, H.T. Yang, C.L. Huang, Scheduling short-term hydrothermal
generation using evolutionary Programming techniques, IEE Proc. Gen.
Transm. Distribution 143 (4) (1996) 371–376.
[7] S.O. Orero, M.R. Irving, A genetic algorithm modeling framework and
solution technique for short term optimal hydrothermal scheduling, IEEE
Trans. PWRS 13 (2) (1998).
[8] E. Gil, J. Bustos, H. Rudnick, Short-term hydrothermal generation sche-
duling model using a genetic algorithm, IEEE Trans. Power Syst. 18 (4)
(2003) 1256–1264.
[9] J.M. Ramirez, P.E. Ontae, The short-term hydrothermal coordination via
genetic algorithms, Electr. Power Components Syst. 34 (2006) 1–19.
[10] N. Sinha, R. Chakrabarty, P.K. Chattopadhyay, Fast evolutionary pro-
gramming techniques for short-term hydrothermal scheduling, Electr.
Power Syst. Res. 66 (2003) 97–103.
[11] X. Yuan, Y. Yuan, Application of cultural algorithm to generation sche-
duling of hydrothermal systems, Energy Convers. Manage. 47 (2006)
Fig. 4. Hydraulic system network. 2192–2201.
[12] Z. Gaing, Particle swarm optimization to solving the economic dispatch
considering the generator constraints, IEEE Trans. PWRS 18 (3) (2003)
1187–1195.
6. Conclusion
[13] H. Yoshida, K. Kawata, Y. Fukuyama, S. Takayama, Y. Nakanishi, A
particle swarm optimization for reactive power and voltage control
In this paper, a particle swarm optimization based approach considering voltage security assessment, IEEE Trans. PWRS 15 (4)
has been proposed and demonstrated to solve the short-term (2000) 1232–1239.
hydrothermal scheduling problem. The effect of valve point [14] M.A. Abido, Optimal design of power-system stabilizers using particle
loading is taken into consideration. Numerical results show that swarm optimization, IEEE Trans. Energy Convers. 17 (3) (2002) 406–413.
[15] J.B. Park, K.S. Lee, J.R. Shin, K.Y. Lee, A particle swarm optimization for
highly near-optimal solutions can be obtained by particle solving the economic dispatch with nonsmooth cost functions, IEEE
swarm optimization method, when compared with the Trans. Power Syst. 20 (1) (2005) 34–42.
evolutionary programming and simulated annealing technique. [16] B. Yu, X. Yuan, J. Wang, Short-term hydro-thermal scheduling using
particle swarm optimization method, Energy Convers. Manage. 48 (2007)
1902–1908.
References [17] Z. Michalewicz, M. Schoenauer, Evolutionary algorithms for constrained
optimization problems, Evol. Comput. 4 (1) (1996) 1–32.
[1] S. Chang, C. Chen, I. Fong, P.B. Luh, Hydroelectric generation scheduling [18] M. Basu, An Interactive fuzzy satisfying method based on evolutionary
with an effective differential programming, IEEE Trans. PWRS 5 (3) programming technique for multiobjective short-term hydrothermal sche-
(1990) 737–743. duling, J. Electr. Power Syst. Res. 69 (2004) 277–285.