Particle Swarm Optimization Using Economic Dispatch
Particle Swarm Optimization Using Economic Dispatch
on MATHEMATICAL METHODS and COMPUTATIONAL TECHNIQUES IN ELECTRICAL ENGINEERING, Sofia, 27-29/10/05 (pp372-377)
Abstract: - A solution to unit commitment using binary particle swarm optimization (BPSO) is presented. The
minimum up and down time constraints, start-up and shutdown cost, spinning reserve, and generation limit are
taken into account. The minimum up and down time constraints are considered in generating the particles to
narrow the search space. Penalty factors are introduced to calculate the fitness of particles, which tend to avoid
infeasible combinations. Problem formulation, representation and the simulation results are presented. The
results show that the proposed method is effective.
Key-Words: - unit commitment, power system, binary particle swarm optimization, economic dispatch, penalty
factors
∑Pu
i =1
ij ij − PDj = 0 (2)
are that PSO is easy to implement and there are few
parameters to adjust. Therefore, PSO has been
(b) Spinning reserve requirement successfully applied in many areas.
N Each individual in PSO flies in the search space
∑P i =1
i
max
uij ≥ PDj + PRj (3) with a velocity which is dynamically adjusted
according to its own flying experience and its
(c) Generation limit companions’ flying experience. Each individual
Pi min ≤ Pij ≤ Pi max (4) keeps track of its coordinates in the problem space,
which are associated with the best solution (fitness) it
(d) Minimum up/down time has achieved so far. This value is called pbest.
TijON > MUTi (5) Another best value that is tracked by the global
version of the particle swarm optimizer is the overall
TijOFF > MDTi (6)
best value, and its location, obtained so far by any
where particle in the population. This location is called
N number of units, gbest. At each time step, the particle swarm
T scheduling period in hours, optimization concept consists of velocity changes of
Pij generation of unit i for hour j, each particle toward its pbest and gbest locations.
Acceleration is weighted by a random term, with
ai , bi , ci fuel cost coefficients of unit i,
separate random numbers being generated for
uij on(1)/off(0) status of unit i at time j, acceleration toward pbest and gbest locations.
Sij start-up cost of unit i at time j, where If xi = ( xi1 , xi 2 ,L , xiD ) represent the ith particle
−TijOFF / τ i
in the D-dimensional space, the binary version of
Sij = σ i + δ i (1 − e ) , and σ i , δ i , τ i PSO can be formulated as follows [16].
are start-up cost coefficients of unit i, vidk +1 = w ⋅ vidk + c1 ⋅ rand () ⋅ ( pbestid − xidk )
Dij shutdown cost of unit i at time j, (7)
+c2 ⋅ rand () ⋅ ( gbestd − xidk )
PDj system load demand at time j,
⎧1 rand () < S (vidk +1 )
PRj system spinning reserve required at time j, xidk +1 = ⎨ (8)
⎩0 otherwise
Pi min minimum generation limit of unit i, where
Pi max
maximum generation limit of unit i, vidk velocity of individual i at iteration k,
T ON
ij ON period of unit i at time j, v min ≤ vik ≤ v max ,
w inertia weight factor, often decrease linearly
TijOFF OFF period of unit i at time j,
from about 0.9 to 0.4 during a run [17].
MUTi minimum up time of unit i, wmax − wmin
w = wmax − × iter .
MDTi minimum down time of unit i, itermax
c1 , c2acceleration constant, often set to be 2,
rand() uniform random number between 0 and 1,
3 Binary Particle Swarm xidk current position of individual i at iteration k,
Optimization (BPSO)
Kennedy and Eberhart first introduced the particle
pbesti pbest of individual i,
swarm optimization (PSO) method, which is an gbest gbest of the group,
7th WSEAS Int. Conf. on MATHEMATICAL METHODS and COMPUTATIONAL TECHNIQUES IN ELECTRICAL ENGINEERING, Sofia, 27-29/10/05 (pp372-377)
S(v) a sigmoid limiting transformation function, ⎧1 tij < MUTi & ui ( j −1) = 1
S (v) = 1/(1 + e − v ) . ⎪
uij = ⎨0 tij < MDTi & ui ( j −1) = 0 (9)
⎪
⎩0 or 1 otherwise
4 Solution Methodology This means that the status of the units will be
A proposed binary particle swarm optimization determined by the minimum up/down time
(BPSO) method is proposed in the paper to solve the constraints of the units at first, and then determined
UC problem. Two modifications are made to the by BPSO. Hence, minimum up and down time
solution. One is using a new method to generate the constraints can be considered when initializing or
particles, which insures the particles satisfy modifying the individuals. The individual i in the
minimum up and down time constraints. The other is BPSO would be presented as
introducing penalty factors to avoid infeasible xi = ( xi1 , xi 2 ,L , xi ( N *T ) )
individuals. These modifications will prevent random (10)
generation and test feasibility step in [15]. = (u11 , u21 ,L u N 1 , u12 , u22 ,L u N 2 ,L u NT )
5 Simulation Results
Spinning
Operation Startup Unit
Hour Reserve Generation Schedule (MW)
Cost ($) Cost ($) Schedule
(MW)
1 13683.13 0 210 1100000000 455 245 0 0 0 0 0 0 0 0
2 14554.50 0 160 1100000000 455 295 0 0 0 0 0 0 0 0
3 16301.89 0 60 1100000000 455 395 0 0 0 0 0 0 0 0
4 18637.68 1109.74 90 1101000000 455 365 0 130 0 0 0 0 0 0
5 20020.02 1793.94 202 1101100000 455 390 0 130 25 0 0 0 0 0
6 22387.04 1096.29 232 1111100000 455 360 130 130 25 0 0 0 0 0
7 23261.98 0 182 1111100000 455 410 130 130 25 0 0 0 0 0
8 24150.34 0 132 1111100000 455 455 130 130 30 0 0 0 0 0
9 26588.96 339.31 112 1111110000 455 455 130 130 110 20 0 0 0 0
10 29365.95 519.36 97 1111111000 455 455 130 130 162 43 25 0 0 0
11 31916.06 120.00 167 1111111110 455 455 130 130 162 73 25 10 10 0
12 33205.25 0 117 1111111110 455 455 130 130 162 80 25 53 10 0
13 29365.95 0 97 1111111000 455 455 130 130 162 43 25 0 0 0
14 26588.96 0 112 1111110000 455 455 130 130 110 20 0 0 0 0
15 24150.34 0 132 1111100000 455 455 130 130 30 0 0 0 0 0
16 20895.88 0 152 1101100000 455 440 0 130 25 0 0 0 0 0
17 19608.54 0 72 1100100000 455 455 0 0 90 0 0 0 0 0
18 21891.43 897.67 102 1110100000 455 455 130 0 60 0 0 0 0 0
19 24150.34 913.99 132 1111100000 455 455 130 130 30 0 0 0 0 0
20 29365.95 833.10 97 1111111000 455 455 130 130 162 43 25 0 0 0
21 26588.96 0 112 1111110000 455 455 130 130 110 20 0 0 0 0
22 21891.43 0 102 1110100000 455 455 130 0 60 0 0 0 0 0
23 17684.69 0 172 1100100000 455 420 0 0 25 0 0 0 0 0
24 15427.42 0 110 1100000000 455 345 0 0 0 0 0 0 0 0
Total 551682.71 7623.39 3153 559306.10
The BPSO program was implemented in MATLAB solution. The results are shown in Table 4. As can be
and executed on a Pentium III 800 personal computer seen, the proposed BPSO method has good quality
with 256MB RAM. The program is tested on a and convergence characteristic.
10-unit system, which data is given in Table 1 and 2
[15]. The 10-unit system simulation results of [15]
were incorrect because the startup costs of the units Table 4 The quality of the
which start up at the first scheduling time were not solution
calculated. Moreover, the worst generation cost in best generation cost ($) 559306.10
Table (8) of [15] was also unbelievable. worst generation cost ($) 562383.57
The spinning reserve is assumed to be 5% of the average generation cost ($) 560894.43
load demand. The population size is set to be 20, and standard deviation ($) 751.21
the iteration is set to be 100. The convergence
tendency of the best evaluation value in the
population during BPSO processing is shown in 6 Conclusion
Fig.2. Fig.3 shows the scheduling generation and In this paper, a modified BPSO method is proposed
load demand. Table 3 illustrates the solution obtained to solve the UC problems. A new strategy is
by the BPSO. Operation, startup costs, spinning for employed for representing chromosomes and
the 24h period, unit on/off schedule and generation encoding the problem search space, of which the
supplying the load is also provided in Table 3. 50 minimum up and down time constraints are taken
trials are performed to examine the quality of the into account in initializing and modifying the
7th WSEAS Int. Conf. on MATHEMATICAL METHODS and COMPUTATIONAL TECHNIQUES IN ELECTRICAL ENGINEERING, Sofia, 27-29/10/05 (pp372-377)
particles. Thus, the individuals of the BPSO method [12] K.S. Swarup, S. Yamashiro, Unit commitment
are all satisfy the minimum up and down time solution methodology using genetic algorithm,
constraints. The penalty coefficients are used to IEEE Transactions on Power Systems, Vol.17,
calculate the evaluation value of the individuals, No.1, 2002, pp. 87-91.
which discourage the infeasible combinations. The [13] A.H. Mantawy, Youssef L. Abdel-Magid,
feasibility of the proposed method is demonstrated by Shokri Z. Selim, A simulated annealing algorithm
simulation. for unit commitment, IEEE Transactions on
Power Systems, Vol.13, No.1, 1998, pp. 197-204.
[14] J. Kennedy, R. Eberhart, Particle swarm
References: optimization, Proceddings of IEEE International
[1] A.J. Wood, B.F. Wollenberg, Power Generation Conference on Neural Networks, Vol. IV, Perth,
Operation and Control, John Wiley, New York, Australia, 1995, pp. 1942-1948.
1984. [15] Zwe-Lee Gaing, Discrete particle swarm
[2] Happ H.H., R.C. Johnson, W.J. Wright, Large optimization algorithm for unit commitment,
scale hydro-thermal unit commitment-method IEEE Power Engineering Society General
and results, IEEE Trans. on PAS, Vol.PAS-90, Meeting, Vol.1, 2003, pp. 13-17.
1971, pp. 1373-1383. [16] J. Kennedy, R. Eberhart, A discrete binary
[3] Baldwin, C.J., K.M. Dale, R.F. Dittrich, A study version of the particle swarm algorithm,
of economic shutdown of generating units in Proceeding of IEEE International Conference on
daily dispatch, AIEE Tr. on PAS, Vol.78, 1960, Evolutionary Computation, Anchorage, 1998, pp.
pp. 1272-1284. 84-89.
[4] W.L. Snyder, H.D. Powell, Jr., J.C. Rayburn, [17] Y. Shi, R.C. Eberhart, Empirical study of
Dynamic programming approach to unit particle swarm optimization, Proceedings of the
commitment, IEEE Transactions on Power 1999 Congress on Evolutionary Computation,
Systems, Vol.2, No.2, 1987, pp. 339-350. Piscataway, 1999, pp. 1945-1950.
[5] W.J. Hobbs, G. Hermon, S. Warner, G.B. Sheble,
An enhanced dynamic programming approach for
unit commitment, IEEE Transactions on Power
Systems, Vol.3, No.3, 1988, pp. 1201-1205.
[6] Z. Ouyang, S.M. Shahidehpour, An intelligent
dynamic programming for unit commitment
application, IEEE Transactions on Power
Systems, Vol.6, No.3, 1991, pp. 1203-1209.
[7] F. Zhuang, F.D. Galiana, Towards a more
rigorous and practical unit commitment by
Lagrangian relaxation, IEEE Transactions on
Power Systems, Vol.3, No.2, 1988, pp. 763-773.
[8] S.J. Wang, S.M. Shahidehpour, D.S. Kirschen, S.
Mokhtari, G.D. Irisarri, Short-term generation
scheduling with transmission and environmental
constraints using augmented Lagrangian
relaxation, IEEE Transactions on Power Systems,
Vol.10, No.3, 1994, pp. 1294-1301.
[9] Chuan-Ping Cheng, Chih-Wen Liu, Chun-Chang
Liu, Unit commitment by Lagrangian relaxation
and genetic algorithms, IEEE Transactions on
Power Systems, Vol.15, No.2, 2000, pp. 707-714.
[10] Time T. Maifeld, Gerald B. Sheble,
Genetic-based unit commitment algorithm, IEEE
Transactions on Power Systems, Vol.11, No.3,
1996, pp. 1359-1370.
[11] S.A. Kazarlis, A.G. Bakirtzis, V. Petridis, A
genetic algorithm solution to the unit
commitment problem, IEEE Transactions on
Power Systems, Vol.11, No.1, 1996, pp. 83-92.