Professional Documents
Culture Documents
Abstract
This paper presents an interactive fuzzy satisfying method based on evolutionary programming technique for short-term multiobjective
hydrothermal scheduling. The multiobjective problem is formulated considering two objectives: (i) cost and (ii) emission. Assuming that the
decision maker (DM) has fuzzy goals for each of the objective functions, evolutionary programming technique based fuzzy satisfying method
is applied for generating a corresponding optimal noninferior solution for the DM’s goals. Then, by considering the current solution, the DM
acts on this solution by updating the reference membership values until the satisfying solution for the DM can be obtained. 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. Thermal plants with nonsmooth fuel cost and emission level function
are also taken into consideration. Results of the application of the proposed method are presented.
© 2003 Elsevier B.V. All rights reserved.
Keywords: Hydrothermal scheduling; Cascaded reservoirs; Interactive fuzzy satisfying method; Membership function; Evolutionary programming technique
0378-7796/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2003.10.003
278 M. Basu / Electric Power Systems Research 69 (2004) 277–285
functions so as to minimize the fuel cost and emission level time m, Psimin and Psimax , the lower and upper generation limits
of thermal plants simultaneously while satisfying the vari- for ith thermal unit, Ns the number of thermal generating
ous constraints on the hydraulic and power system network. units, m and M the time index and scheduling period, fq the
The main constraints include: the time coupling effect of qth objective function.
the hydro sub problem, where the water flow in an earlier
time interval affects the discharge capability at a later period 2.1.2. Emission
of time, the varying system load demand, the cascade na- Thermal power stations are major causes of atmospheric
ture of the hydraulic network, the varying hourly reservoir pollution, because of high concentration of pollutants they
inflows, the physical limitations on the reservoir storage cause. In this study, nitrogen oxides (NOx ) emission is taken
and turbine flow rate and the loading limits of both thermal as the selected index from the viewpoint of environment
and hydro plants. Here the objective functions (i.e. cost and conservation. The amount of emission from each generator
emission) are modeled with fuzzy sets. Fuzzy sets were first is given as a function of its output [16], which is the sum of
introduced by Zadeh [13] as an effective means of solving a quadratic and an exponential function. The total emission
nonprobabilistic problems. The different objectives are eas- in the system can be expressed as
ily integrated because all the membership function values of
these objectives are in the same range [0, 1]. It is assumed Ns
M
that the decision maker (DM) has imprecise or fuzzy goals f2 = [αsi + βsi Psim + γsi Psim
2
+ ηsi exp(δsi Psim )]
for each of objective functions. The fuzzy goals are quanti- m=1 i=1
fied by defining their corresponding membership functions. (2)
The DM then specifies the reference membership values
for each of the objective functions and the corresponding where αsi , βsi , γ si , ηsi , δsi are emission curve coefficients of
optimal noninferior solution can be obtained. Through the ith thermal unit subject to
interaction the DM’s reference membership values are up- (i) Power balance constraints
dated by considering the current values of the membership The total active power generation must balance the
functions as well as the objectives until a satisfying solution predicted power demand plus losses, at each time in-
for the DM is obtained. The proposed method has been terval over the scheduling horizon
validated by applying it to a test system [12,17].
Ns
Nh
Psim + Phjm − PDm − PLm = 0 m∈M (3)
2. Problem formulation i=1 j=1
2.1.1. Economy where Cij , C2j , C3j , C4j , C5j , C6j are the power gen-
The fuel cost function of each thermal generating unit eration coefficients of jth hydro unit, Qhjm the water
considering valve-point effects [15] is expressed as the sum discharge rate of jth reservoir at time m, Vhjm the stor-
of a quadratic and a sinusoidal function. The total fuel cost age volume of jth reservoir at time m.
in terms of real power output can be expressed as (ii) Generation limits
Ns
M
Phj
min
≤ Phjm ≤ Phj
max
, j ∈ Nh , m ∈ M (5)
f1 = [asi + bsi Psim + csi Psim
2
m=1 i=1
and
+|dsi sin{esi (Psimin − Psim )}|] (1) Psimin ≤ Psim ≤ Psimax , i ∈ Ns , m ∈ M (6)
where asi , bsi , csi , dsi , esi are the cost curve coefficients of where Phjmin , P max are the lower and upper generation
hj
ith thermal unit, Psim the output power of ith thermal unit at limits for jth hydro unit.
M. Basu / Electric Power Systems Research 69 (2004) 277–285 279
where Qmin
hj , Qhj are the minimum and maximum wa-
max
where fqmin is the minimum value of qth objective, fqmax the
ter discharge rate of jth reservoir; Vhj
min , V max : minimum
hj
maximum value of qth objective, n the number of objective
and maximum storage volume of jth reservoir. functions.
(b) The continuity equation for the hydro reservoir network After defining the membership functions, the DM is asked
to specify the reference (desirable) levels of achievement of
Vhj(m+1) = Vhjm + Ihjm − Qhjm − Shjm the membership functions, called the reference membership
Ruj values µrq , q = 1–n. Then to obtain the satisfying solution,
+ (Qhl(m−tlj ) + Shl(m−tlj ) ), the following minimax problem is solved [14].
l=1 min {max|µrq − µfq (X̄)|} (11)
j ∈ Nh , m ∈ M (9) X∈Ω q=1–n
where Ihjm is the inflow rate of jth reservoir at time where Ω is the set of noninferior solutions, µfq the qth mem-
m, Ruj the number of upstream units directly above jth bership function; µrq the qth reference membership value.
hydro plant, Shjm the spillage of jth reservoir at time m, In this paper the minimax problem is solved using evolu-
tlj the water transport delay from reservoir l to j. tionary programming technique. If the DM is not satisfied
with the current solution, then through the interaction, the
reference membership values can be updated and the updat-
3. Fuzzy satisfying method ing is done by considering the current values of the mem-
bership functions as well as the objectives. This interactive
Considering the imprecise nature of the decision-maker’s updating process is continued until the satisfying solution
judgment, it is natural to assume that the decision-maker for the DM is obtained.
may have fuzzy or imprecise goals for each objective func-
tion. The fuzzy sets are defined by equations called mem-
bership functions. The higher the value of the membership 4. Evolutionary programming
function implies a greater satisfaction with the solution. The
membership function consists of a lower and upper boundary Evolutionary programming (EP) is a technique in the field
value together with a strictly monotonically decreasing and of evolutionary computation. It seeks the optimal solution
continuous function. Fig. 1 illustrates the graph of the possi- by evolving a population of candidate solutions over a num-
ble shape of a strictly monotonically decreasing membership ber of generations or iterations. During each iteration, a sec-
function. The lower and upper bounds, fqmin (X̄), fqmax (X̄) of ond new population is formed from an existing population
each of the objective functions fq (X̄), q = 1–n under given through the use of a mutation operator. This operator pro-
constraints are established to elicit a membership function duces a new solution by perturbing each component of an
µfq (X̄) for each objective function fq (X̄). The DM is fully existing solution by a random amount. The degree of opti-
satisfied with the objective value fq (X̄) if µfq (X̄) = 1, and mality of each of the candidate solutions or individuals is
not satisfied at all if µfq (X̄) = 0. The qth membership func- measured by their fitness, which can be defined as a function
tion is now defined as: of the objective function of the problem. Through the use
of a competition scheme, the individuals in each population
0, if fq (X̄) ≥ fqmax
f max − fq (X̄) compete with each other. The winning individuals form a re-
q sultant population, which is regarded as the next generation.
µfq (X̄) = , if fqmin (X̄) < fqmax (10)
fqmax − fqmin For optimization to occur, the competition scheme must be
1, if fq (X̄) ≤ fqmin such that the more optimal solutions have a greater chance
280 M. Basu / Electric Power Systems Research 69 (2004) 277–285
of survival than the poorer solutions. Through this the popu- following equation:
lation evolves towards the global optimal point. The Evolu-
Ns
Nh
tionary programming technique is iterative and the process
is terminated by a stopping rule. The rule widely used is ei- Psdg m = PDm + PLm − Psim − Phjm , m∈M
i=1 j=1
ther (a) stop after a specified number of iterations or (b) stop i=d
when there is no appreciable change in the best solution for
(14)
a certain number of generations. Rule (a) is adopted in the
present work. The dependent thermal generation must satisfy the con-
straints in Eq. (6).
economic emission load dispatch problem of hydrothermal respectively, where i = 1–Ns , j = 1–Nh and m = 1–m. U(a,
power system is described below. b) denotes a uniform random variable ranging over [a, b].
Let pk = [Ps1 , Ps2 , . . . , Psi , . . . , PsNs , Qh1 , Qh2 , . . . ,
5.2. Creation of offspring
Qhj , . . . , QhNh ]T be a trial matrix designating the kth
individual of a population to be evolved and Psi =
An offspring vector p
k is created from each parent pk by
[Psi1 , Psi2 , . . . , Psim , . . . , PsiM ], Qhj = [Qhj1 , Qhj2 , . . . ,
adding to each component of pk a Gaussian random variable
Qhjm , . . . , QhjM ]. The elements Psim and Qhjm are the
with zero mean and a standard deviation proportional to the
power output of the ith thermal unit and the discharge rate
scaled cost values of the parent trial solution, i.e.
of the jth hydro plant at time m. The range of the elements
Psim and Qhjm should satisfy the thermal generating ca- p
k = [Ps1
, Ps2 , . . . , Psi
, . . . , PsN
, Q
h1 , Q
h2 , . . . ,
s
pacity and the water discharge rate constraints in Eqs. (6)
and (8), respectively. Assuming the spillage in Eq. (9) to be × Q
hj , . . . , Q
hNh ]T
zero for simplicity, the hydraulic continuity constraints are
Psim = Psim + N(0, σi2 ), m∈M (15)
Ruj
M
M
M
Q
hjm = Qhjm + N(0, σj2 ), m∈M (16)
Vhj0 − VhjM = Qhjm − Qhl(m−tlj ) − Ihjm ,
m=1 m=1 l=1 m=1 The standard deviations σ i and σ j indicate the ranges of
j ∈ Nh (12) the offspring created around the parent trial solution and are
given according to the following equation:
where Vhj0 is the initial storage volume of jth reservoir; λpk
σi = ε (P max − Psimin ) (17)
VhjM the final storage volume of jth reservoir. λmin si
To meet exactly the restrictions on the initial and final
reservoir storage in Eq. (9), the water discharge rate of the λpk
σj = ε (Qmax − Qmin
hj ) (18)
jth hydro plant Qhjd in the dependent interval d is then λmin hj
calculated by where λmin is the minimum value of λ among the Np trial
solutions, ε is a scaling factor, λpk is the value of the function
M
associated with the trial vector pk , i.e.,
Qhjd = Vhj0 − VhjM + Ihjm
m=1
λpk = {max|µrq − µfq (pk )|} (19)
q=1,2,...,n
Ruj
M M
+ Qhl(m−tlj ) − Qhjm , where λpk is to be minimized. The offspring p
k is created
m=1 l=1 m=1 according to the relative value of λpk ; if λpk is relatively
m=d low, the offspring trial solution is created near the current
j ∈ Nh (13) solution pk ; if λpk is relatively high, the next trial solution
will be searched within a wider range.
The dependent water discharge rate must satisfy the con-
straints in Eq. (8). 5.3. Competition and selection
Also to meet exactly the power balance constraints in
Eq. (3), the thermal generation Psdg m of the dependent ther- The Np parent trial vectors pk , k = 1–Np along with their
mal generating unit dg can then be calculated using the corresponding offspring p
k formed by mutation, k = 1–Np
M. Basu / Electric Power Systems Research 69 (2004) 277–285 281
Table 1
Results of the interactive fuzzy satisfying procedure in obtaining solutions
Interactive Reference Output Whether satisfied with
count #I membership value current noninferior solution
Membership Objective function
function {f1 (US$)} and {f2 (lb)}
0 (initial) µr1 = 0.55 µf 1 = 0.5496 f1 = 51470 No
µr2 = 0.65 µf 2 = 0.6490 f2 = 27873
1 µr1 = 0.80 µf 1 = 0.8001 f1 = 47906 Yes
µr2 = 0.70 µf 2 = 0.6998 f2 = 26234
282 M. Basu / Electric Power Systems Research 69 (2004) 277–285
Table 2
Hydrothermal generation (MW) schedule for µr1 = 0.55 and µr2 = 0.65
Hour Ph1 Ph2 Ph3 Ph4 Ps1 Ps2 Ps3
outputs after the initial interaction, the DM is not satisfied generation and emission level are decreased. This gives the
with the current noninferior solution. In interaction 1, the satisfactory noninferior solution for the DM and the inter-
DM sets the reference membership value µr1 = 0.80 and active process is terminated. The determined hydrothermal
µr2 = 0.70 in order to reduce the cost of generation and generation schedules and water discharge rates are given in
emission level. The problem is again solved with these new the Tables 2–5, respectively. The average CPU time using
interactive inputs and the outputs are observed. Both cost of Pentium 3 PC was 1 h, 16 min and 22 s.
Table 3
Hourly plant discharge (×104 m3 ) for µr1 = 0.55 and µr2 = 0.65
Hour Qh1 Qh2 Qh3 Qh4
Table 4
Hydrothermal generation (MW) schedule for µr1 = 0.80 and µr2 = 0.70
Hour Ph1 Ph2 Ph3 Ph4 Ps1 Ps2 Ps3
Table 5
Hourly plant discharge (×104 m3 ) for µr1 = 0.60 and µr2 = 0.70
Hour Qh1 Qh2 Qh3 Qh4
8. Discussion
Table A.3
Weighted sum approach becomes inconvenient to coordi- Reservoir inflows (×104 m3 )
nate objectives as they may possess different units of mea-
Hour Reservoir Hour Reservoir Hour Reservoir
surement making it difficult to justify the weighing factors
to generate proper Pareto optimal solutions. But in case of 1 2 3 4 1 2 3 4 1 2 3 4
fuzzy satisfying method different objectives are easily inte- 1 10 8 8.1 2.8 9 10 8 1 0 17 9 7 2 0
grated because all the membership function values of these 2 9 8 8.2 2.4 10 11 9 1 0 18 8 6 2 0
objectives are in the same range [0, 1]. 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
Appendix A 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
Fig. 2, Table A.1–A.6.
Table A.1
Load demand
Hour PD (MW) Table A.4
Reservoir storage capacity limits, plant discharge limits, reservoir end
1 750
conditions (×104 m3 ) and plant generation limits (MW)
2 780
3 700 Plant Vmin Vmax Vini Vend Qmin Qmax Phmin Phmax
4 650
1 80 150 100 120 5 15 0 500
5 670
2 60 120 80 70 6 15 0 500
6 800
3 100 240 170 170 10 30 0 500
7 950
4 70 160 120 140 6 20 0 500
8 1010
M. Basu / Electric Power Systems Research 69 (2004) 277–285 285
Table A.5
Cost curve coefficients and operating limits of thermal generators
Unit as (US$/h) bs (US$/MWh) cs (US$/(MW)2 h) ds (US$/h) es (1/MW) Psmin (MW) Psmax (MW)
Table A.6
Emission curve coefficients of thermal generators
Unit αi (lb/h) βi (lb/MWh) γ i (lb/(MW)2 h) ηi (lb/h) δi (1/MW)
References minimum cost network flow, IEEE Trans. PWRS 3 (3) (1988) 929–
935.
[9] M.V.F. Pereira, L.M.V.G. Pinto, A decomposition approach to the
[1] IEEE Current Operating Problems Working Group, Potential impacts
economic dispatch of the hydrothermal systems, IEEE Trans. PAS
of clean air regulations on system operations, IEEE Trans. PWRS
101 (10) (1982) 3851–3860.
10 (1995) 647–653.
[10] K.P. Wong, Y.W. Wong, Short-term hydrothermal scheduling. Part 1.
[2] J. Nanda, D.P. Kothari, K.S. Lingamurthy, Economic emission dis-
Simulated annealing approach, IEE Proc. Generation Transm. Distrib.
patch through goal programming technique, IEEE Trans. Energy
141 (5) (1994) 497–501.
Conversion 3 (1988) 26–32.
[11] P.C. Yang, H.T. Yang, C.L. Huang, Scheduling short-term hydrother-
[3] T.D. King, M.E. El-Hawary, F. El-Hawary, Optimal environmental
mal generation using evolutionary programming techniques, IEE
dispatching of electric power systems via an improved Hopfield
Proc. Generation Transm. Distrib. 143 (4) (1996) 371–376.
neural network model, IEEE Trans. PWRS 10 (3) (1995) 1559–
[12] S.O. Orero, M.R. Irving, A genetic algorithm modeling framework
1565.
and solution technique for short term optimal hydrothermal schedul-
[4] D. Srinivasan, A. Tettamanzi, Heuristic-guided evolutionary ap-
ing, IEEE Trans. PWRS 13 (2) (1998).
proach to multiobjective generation scheduling, IEE Proc. Generation
[13] L.A. Zadeh, Fuzzy sets, Inf. Control 8 (1965) 338–353.
Transm. Distrib. 143 (6) (1996) 553–559.
[14] M. Sakawa, H. Yano, An interactive fuzzy satisfying method using
[5] Y.H. Song, G.S. Wang, P.Y. Wang, A.T. Johns, Environmen-
augmented minimax problems and its application to environmental
tal/economic dispatch using fuzzy logic controlled genetic algorithm,
systems, IEEE Trans. SMC 17 (6) (1985) 720–729.
IEE Proc. Generation Transm. Distrib. 144 (4) (1997) 377–381.
[15] H. Yang, P. Yang, C. Huang, Evolutionary programming based eco-
[6] S. Chang, C. Chen, I. Fong, P.B. Luh, Hydroelectric generation
nomic dispatch for units with non-smooth fuel cost functions, IEEE
scheduling with an effective differential programming, IEEE Trans.
Trans. PWRS 11 (1) (1996) 112–118.
PWRS 5 (3) (1990) 737–743.
[16] M.R. Gent, J.W. Lamont, Minimum emission dispatch, IEEE Trans.
[7] S.A. Soliman, G.S. Christensen, Application of functional analysis
PAS 90, 2650–2660.
to optimization of variable head multi reservoir power system for
[17] P.K. Hota, R. Chakrabarti, P.K. Chattopadhyay, A simulated
long term regulation, Water Resources Res. 22 (6) (1986) 852–
annealing-based goal-attainment method for economic emission load
858.
dispatch with nonsmooth fuel cost and emission level functions,
[8] Q. Xia, N. Xiang, S. Wang, B. Zhang, M. Huang, Optimal daily
Electr. Machines Power Syst. 28, 1037–1049.
scheduling of cascaded plants using a new algorithm of non-linear