Professional Documents
Culture Documents
Combined Heat and Power Economic Dispatch by Harmony Search Algorithm PDF
Combined Heat and Power Economic Dispatch by Harmony Search Algorithm PDF
com
a
Power System Research Laboratory, Department of Electrical and Electronic Engineering, K.N.Toosi University of Technology,
322-Mirdamad Avenue West, 19697 Tehran, Iran
b
Department of Mechanical Engineering, Amirkabir University of Technology, 424-Hafez Avenue, Tehran, Iran
Received 19 September 2006; received in revised form 30 May 2007; accepted 9 June 2007
Abstract
The optimal utilization of multiple combined heat and power (CHP) systems is a complicated problem that needs powerful methods
to solve. This paper presents a harmony search (HS) algorithm to solve the combined heat and power economic dispatch (CHPED) prob-
lem. The HS algorithm is a recently developed meta-heuristic algorithm, and has been very successful in a wide variety of optimization
problems. The method is illustrated using a test case taken from the literature as well as a new one proposed by authors. Numerical
results reveal that the proposed algorithm can find better solutions when compared to conventional methods and is an efficient search
algorithm for CHPED problem.
Ó 2007 Elsevier Ltd. All rights reserved.
Keywords: Economic dispatch; Combined heat and power; Harmony search algorithm; Optimization
0142-0615/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2007.06.006
714 A. Vasebi et al. / Electrical Power and Energy Systems 29 (2007) 713–719
cogeneration units. The analysis and interpretation of the increases as the power generation decreases. The CHP eco-
connection have led to the development of the two-layer nomic dispatch problem of a system is to determine the
algorithm. The outer layer uses the Lagrangian relaxation unit heat and power production so that system production
(LR) technique to solve the power dispatch, and the inner cost is minimized while the heat-power demands and other
layer uses the gradient searching method to solve the heat constraints are met. Mathematically, the problem is to min-
dispatch with the unit heat capacities passed by the outer imize the following objective function:
layer. Song and Xuan [3] employed an improved penalty X
np X
nc X
nh
function formulation for the genetic algorithm (GA) to Minimize C¼ ci ðpi Þ þ cj ðhj pj Þ þ ck ðhk Þ ð1Þ
solve the CHPED problem. Chang and Fu [4] used a i¼1 j¼1 k¼1
multi-objective method by using a fuzzy decision index X
np X
nc
and GA to a seven-generator sample system. Song et al. Subject to pi þ pj ¼ P D ð2Þ
[5] introduced an ant colony search algorithm (ACSA) i¼1 j¼1
where f(x) is the objective function and g(x) is the inequal- GA [12]. Here, HMCR and PAR are parameters that are
ity constraint function; h(x) is the equality constraint func- used to improve the solution vector. Both are defined in
tion. x is the set of each decision variable, xi, and X is the Step 3.
set of the possible range of values for each decision vari-
able, that is Lxi 6 Xi 6 Uxi, where Lxi and Uxi are the lower 3.2. Initialize the harmony memory
and upper bounds for each decision variable. The HS algo-
rithm parameters are also specified in this step. These are In Step 2, the HM matrix is filled with as many ran-
the harmony memory size (HMS), or the number of solu- domly generated solution vectors as the HMS
tion vectors in the harmony memory; harmony memory 2 3
x11 x12 . . . x1N 1 x1N
considering rate (HMCR); pitch adjusting rate (PAR); 6 x2
number of decision variables (N) and the number of impro- 6 1 x22 x2N 1 x2N 77
6 7
visations (NI), or stopping criterion. 6 .. . . . .. 7
.
HM ¼ 6 . .. .. .. 7 ð9Þ
The harmony memory (HM) is a memory location 6 7
6 HMS-1 7
where all the solution vectors (sets of decision variables) 4 x1 xHMS-1
2 . . . xHMS-1
N 1 xHMS-1
N 5
are stored. This HM is similar to the genetic pool in the xHMS
1 xHMS
2 ... xHMS
N 1 xHMS
N
716 A. Vasebi et al. / Electrical Power and Energy Systems 29 (2007) 713–719
Table 1
Optimal results for case study I
Methods Optimal results Cost ($)
p1 p2 h2 p3 h3 h4
IGA_ MU [7] 0.00 160.00 39.99 40.00 75.00 0.00 9257.07
GA [3] 0.00 159.23 39.94 40.77 75.06 0.00 9267.20
LR [2] 0.00 160.00 40.00 40.00 75.00 0.00 9257.07
ACSA [5] 0.08 150.93 48.84 49.00 65.79 0.37 9452.20
GT [6] 0.00 157.92 26.00 42.08a 89.00a 0.00 9207.64
Proposed method 0.00 160.00 40.00 40.00 75.00 0.00 9257.07
a
Outside the feasible operating region of cogeneration unit 3.
ð14Þ
c4 ¼ 23:4h4 ð16Þ
X
4
Minimize C¼ ci i ¼ 1; . . . ; 4 ð17Þ
i¼1
4. Case studies
This case study was originally proposed by Gou et al. Fig. 7. Feasible operation region for the cogeneration unit 4 (case study
[2]. The problem consists of a conventional power unit, II).
718 A. Vasebi et al. / Electrical Power and Energy Systems 29 (2007) 713–719
Table 3
Optimal results for case study II
Cases Method Demand Unit 1 Unit 2 Unit 3 Unit 4 Unit 5 C ($)
PD HD P1 P2 h2 P3 h3 P4 h4 h5
I HS 300 150 134.74 48.20 81.09 16.23 23.92 100.85 6.29 38.70 13723.20
GA 135.00 70.81 80.54 10.84 39.81 83.28 0.00 29.64 13779.50
II HS 250 175 134.67 52.99 85.69 10.11 39.73 52.23 4.18 45.40 12284.45
GA 119.22 45.12 78.94 15.82 22.63 69.89 18.40 54.99 12327.37
III HS 160 220 41.41 66.61 97.73 10.59 40.23 41.39 22.83 59.21 11810.88
GA 37.98 76.39 106.0 10.41 38.37 35.03 15.84 59.97 11837.4
After 25,000 function evaluations the best solution is c1 ¼ 254:8863 þ 7:6997p1 þ 0:00172p21 þ 0:000115p31 ð22Þ
obtained at (p1 = 0, p2 = 160, h2 = 40, p3 = 40, h3 = 75, c2 ¼ 1250 þ 36p2 þ 0:0435p22 þ 0:6h2 þ 0:027h22 þ 0:011p2 h2
h4 = 0) with corresponding function value equal to
ð23Þ
$9257.07 as shown in Fig. 5. No constraints are active
for this solution. Table 1 presents the best solution of this c3 ¼ 2650 þ 34:5p3 þ 0:1035p23 þ 2:203h3 þ 0:025h23 þ 0:051p3 h3
problem obtained using the HS algorithm and compares ð24Þ
the HS results with solutions reported by other researchers. c4 ¼ 1565 þ 20p4 þ 0:072p24 þ 2:3h4 þ 0:02h24 þ 0:04p4 h4
It is obvious from the Table 1 that the result obtained using
HS algorithm is the same as the best known solution ð25Þ
reported previously in the literature [2,7]. Although c5 ¼ 950 þ 2:0109h5 þ 0:038h25 ð26Þ
Sudhakaran and Slochanal [6] reported better solution, X
4
however, this solution due to violation of constraints is Minimize C¼ ci i ¼ 1; . . . ; 5 ð27Þ
not feasible. i¼1
To determine the impacts of different parameters of the Subject to 35 6 p1 6 135 MW ð28Þ
HS algorithm on the solution quality and convergence
0 6 h5 6 60 MWth ð29Þ
behavior, an empirical study is performed. To show the
effects of single parameter changes, seven different scenar- P D ¼ p 1 þ p2 þ p 3 þ p 4 ð30Þ
ios are tested as shown in Table 2. Each scenario is tested H D ¼ h2 þ h3 þ h4 þ h5 ð31Þ
over 30 runs and maximum number of iterations is fixed
Boiler and generator output constraints are listed in
to 40,000 for all runs. In Table 2 the average total cost
Eqs. (28) and (29). The feasible operating region of cogen-
for case study(I) is summarized.
eration units 2–4 are shown in Figs. 4, 6 and 7, respectively.
As mentioned earlier, the HMCR determines the rate of
The performance of the proposed algorithm is validated by
choosing one value from the historical values stored in the
means of comparison with an existing method such as
HM. The larger the HMCR is the less exploration is
genetic algorithm. Table 3 shows the results obtained from
achieved; and the algorithm further relies on stored values
the HS algorithm and GA method for three different heat
in HM and this potentially leads to the algorithm getting
and power demand cases. The results indicate that the pro-
stuck in a local optimum. On the other hand, choosing
posed approach can yield to better results in comparison
the HMCR too small decreases the algorithm efficiency
with those obtained using GA.
and the HS algorithm behaves like a pure random search,
with less assistance from the historical memory. As shown
in Table 2 large and small HMCR values lead to a decrease 5. Conclusion
in the solution quality. Large and small HMS values
decreases the efficiency of the harmony memory as seen This paper has introduced a harmony search algorithm
in Table 2. For most problems, a HMS between N and to solve the combined heat and power economic dispatch
2N is reasonable. It is observed that the algorithm has problem. The method illustrated using two test cases; one
small sensitivity to PAR values. is taken from the literature and the other one proposed
by authors. Three demand cases from the new test system
4.2. Case study II have been shown in details. The proposed new test system
can be used as a standard test system for assessing cogen-
A new test system is used to show the effect of the pro- eration economic dispatch algorithms. The results obtained
posed method. This problem consists of a conventional by the proposed method are compared against those gener-
power unit, three cogeneration units and a heat-only unit. ated with other (evolutionary and mathematical program-
The cogeneration unit 2 is taken from the previous case ming) techniques reported in the literature and they
study and the cost functions of the units are shown in clearly demonstrate that the proposed method is practical
Eqs. (22)–(26). and valid for CHPED applications.
A. Vasebi et al. / Electrical Power and Energy Systems 29 (2007) 713–719 719
Acknowledgement [7] Su CT, Chiang CL. An incorporated algorithm for combined heat
and power economic dispatch. Electric Power Syst Res
2004;69:187–95.
The assistance and suggestions of Mrs. M. Partovi are [8] Wong Kit Po, Algie Cameron. Evolutionary programming approach
kindly acknowledged. for combined heat and power Dispatch. Electric Power Syst Res
2002;61:227/232.
References [9] Eriksen Peter Børre. Economic and environmental dispatch of
power/CHP production systems. Electric Power Syst Res
[1] Rooijers FJ, van Amerongen RAM. Static economic dispatch for 2001;57:33–9.
cogeneration systems. IEEE Trans Power Syst 1994;9 (3):1392–8. [10] Geem ZW, Kim JH, Loganathan GV. A new heuristic optimization
[2] Guo T, Henwood MI, van Ooijen M. An algorithm for combined algorithm: harmony search. Simulation 2001;76 (2):60–8.
heat and power economic dispatch. IEEE Trans Power Syst 1996;11 [11] Kim JH, Geem ZW, Kim ES. Parameter estimation of the nonlinear
(4):1778–84. Muskingum model using harmony search. J Am Water Resour Ass
[3] Song YH, Xuan QY. Combined heat and power economic dispatch 2001;37 (5):1131–8.
using genetic algorithm based penalty function method. Electric [12] Geem ZW, Kim JH, Loganathan GV. Harmony search optimization:
Mach Power Syst 1998;26:363–72. application to pipe network design. Int J Model Simulat 2002;22
[4] Chang CS, Fu W. Stochastic multi-objective generation dispatch of (2):125–33.
combined heat and power systems. IEE Proc-Gen Trans Distrib [13] Lee KS, Geem ZW. A new structural optimization method based on
1998;145 (5):583–91. the harmony search algorithm. Comput Struct 2004;82 (9-10):781–98.
[5] Song YH, Chou CS, Stonham TJ. Combined heat and power [14] Geem ZW, Tseng C, Park Y. Harmony search for generalized
economic dispatch by improved ant colony search algorithm. Electric orienteering problem: best touring in China. Springer Lect Notes
Power Syst Res 1999;52:115–21. Comput Sci 2005;3412:741–50.
[6] Sudhakaran M, Slochanal SMR. Integrating genetic algorithms and [15] Lee KS, Geem ZW. A new meta-heuristic algorithm for continues
tabu economic dispatch search for combined heat and power. Power engineering optimization: harmony search theory and practice.
Syst Stabil Control 2003:67–71. Comput Method Appl Mech Eng 2004;194:3902–33.