Вы находитесь на странице: 1из 13

A grey hop, skip, and jump approach: generating

alternatives for expansion planning of waste


management facilities
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

G.H. Huang, B.W. Baetz, and G.G. Patry

Abstract: A grey hop, skip, and jump (GHSJ) approach is developed and applied to the area of municipal solid waste
management planning. The method improves upon existing modelling to generate alternative approaches by allowing
uncertain information to be effectively communicated into the optimization process and resulting solutions. Feasible
decision alternatives can be generated through interpretation of the GHSJ solutions, which are capable of reflecting
potential system condition variations caused by the existence of input uncertainties. Results from a hypothetical case
study indicate that useful solutions for the expansion planning of waste management facilities can be generated. The
decision alternatives obtained from the GHSJ solutions may be interpreted and analyzed to internalize
environmental-economic tradeoffs, which may be of interest to solid waste decision makers faced with difficult and
controversial choices.
Key words: grey programming, modelling to generate alternatives, hop-skip-jump approach, waste management
planning, uncertainty, public sector decision making.

Resume : Une approche dite de triple saut de zones incertaines (GHSJ) est dCveloppte et appliquCe au domaine de
planification de gestion de dCchets municipaux. La mtthode amtliore les rnodkles existant pour gtntrer des approches
alternatives en permettant i des informations incertaines d'Ctre communiquCes efficacement dans le processus
d'optimisation et les solutions rtsultantes. Les alternatives de dtcisions faisables peuvent Ctre gCnCrCes B travers
l'interprttation des solutions de la GHSJ, qui sont capables de refltter les conditions de variations potentielles du
systkme causCes par l'existence des incertitudes introduites. Des risultats d'un cas hypothttique d'Ctude indiquent que
des solutions utiles pour la planification d'expansion des amtnagements de gestion de dtchets peuvent Ctre gCnCrCes. Les
alternatives de dCcision obtenues des solutions de la GHSJ peuvent &tre interprCtCes et analysies pour assimiler les
compromis environnementaux et Cconorniques, qui peuvent Ctre inttressants pour les dtcideurs de gestion de dCchets
solides confrontCs avec des choix difficiles et controversCs.
Mots elks : programmation d'incertitudes, modtlisation pour gCnCrer des alternatives, approche de triple saut,

planification de gestion de dCchets, incertitudes, prise de dtcision en secteur publique.


[Traduit par la rCdaction]

1. Introduction
The planning of municipal solid waste management systems
to satisfy increasing waste disposal and treatment demands
is often subject to a variety of impact factors. Therefore,
optimization may b e useful for reflecting the effects of these
factors and generating optimal solutions. However, due to
the presence of uncertainty and many nonquantifiable factors
relating to environmental and economic objectives, and the
possibility that public opposition may eliminate the optimal
Received October 12, 1995.
Revised manuscript accepted June 6, 1996.
G.H. Huang. Faculty of Engineering, University of Regina,
Regina, SK S4S OA2, Canada.
B.W. Baetz. Department of Civil Engineering, McMaster
University, Hamilton, ON L8S 4L7, Canada.
G.G. Patry. Faculty of Engineering, University of Ottawa,
ON KIN 6N5, Canada.

Written discussion of this paper is welcomed and will be


received by the Editor until April 30, 1997 (address inside
front cover).
Can. I. Civ. Eng. 23:

1207 - 1219 (1996).

solution from further consideration, solid waste decision


makers faced with difficult and controversial choices may
prefer a set of alternatives s o that they can bring implicit
knowledge (i.e., knowledge that cannot b e incorporated within
an optimization model) to bear o n the problem (Gidley and
Bari 1986).
Methods f o r modelling to generate alternatives (MGA)
have been proposed in response to the above situation. T h e
M G A approaches provide an optimal solution and several
near-optimal alternatives for a planning problem. Preferably,
the alternatives are close to the optimal solution with respect
to the objective function value, but vary considerably from
the optimal solution in terms of system variables. A decision
maker can then review the generated alternatives and internalize the tradeoffs between the differences i n the objective
function value and the differing system characteristics.
Previously, a number of M G A approaches have been
proposed and applied. Brill (1979) developed a technique
named hop, skip, and jump (HSJ) for linear and mixed integer
programming problems to generate alternatives that are good
with respect to the model objective and different from one
another with respect to the specified decisions. Church and

Printed in Canada / Imprimt au Canada

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

Can. J. Civ. Eng. Vol. 23, 1996

Huber (1979) used a reverse heuristic to find close-tooptimal solutions for maximal covering location problems.
Falkenhausen (1979) employed a heuristic evolution strategy
to generate alternative solutions for a regional wastewater
treatment system planning problem. Chang et al. (1980) developed a random technique for generating decision alternatives
by maximizing the sum of several randomly selected decision
variables. Brill et al. (1981) applied the HSJ method to a multiobjective linear programming problem to generate alternatives
for a hypothetical land use planning problem. Chang et al.
(1982) discussed an approach named branch and boundlscreen
(BBS) for obtaining good and different alternatives by first
generating many solutions efficiently and then applying a
screening process to determine the alternatives. More recently,
Baetz et al. (1990) developed an MGA approach for dynamic
programming-based planning problems and applied it to solid
waste management planning.
The major deficiency with the existing MGA approaches
is that they are based on deterministic mathematical programming models, which may not be able to effectively communicate uncertain information into the optimization framework.
Therefore, a grey hop, skip, and jump (GHSJ) approach is
developed in this paper to mitigate this problem. The GHSJ
approach can directly communicate uncertainty into the optimization process and the resulting solutions, such that optimal
and close-to-optimal solutions for the decision variables and
the objective function value can be obtained (Huang et al.
1995). Thus, decision alternatives can be generated by adjusting different combinations of the decision variable values
within their solution intervals according to projected applicable
conditions, which will reflect potential system condition variations caused by the existence of input uncertainties.
The purpose of this paper is to develop the GHSJ approach
and apply it to a hypothetical case study of municipal solid
waste management planning. The results will be interpreted
and analyzed to show the potential applicability of the developed methodology to waste management planning and other
types of public sector decision making problems.

2. The grey hop, skip, and jump approach


In solid waste management systems, uncertainties may exist
in many system components related to environmental, socioeconomic, and resources concerns, and the associated information may not be known with certainty but instead as follows:
"the capital cost for expanding the composting facility will
be in the range of $1 000000 to $1 200000," "the waste
generation rate is approximately 90 to 100 tonnes per week,"
"the incinerator has a capacity of 2000 to 2500 tonnes per
week", and so on (Inuiguchi et al. 1990). Difficulties may
arise when modelling such systems with deterministic mathematical programming methods. Therefore, a GHSJ method
is now developed, where concepts of grey systems and grey
decisions are introduced into an HSJ modelling framework to
reflect the effects of uncertainties, and interactive solution
algorithms are used for solving the related grey programming problems (Huang 1994).
First, we introduce some definitions. Let x denote a closed
and bounded set of real numbers. A grey number x* is
defined as an interval with known upper and lower bounds
but unknown distribution information for x:

where x- and x + are the lower and upper bounds of x i ,


respectively. When x- = x + , x + becomes a deterministic
number, i.e., x i = x- = x + .
Let '$3' denote a set of grey numbers. A grey vector (or
matrix) is defined as
[2.2]

X+ = {x:

[x;, x;]

[2.3]

xi = {x:

[x-,
x$]
rJ
JJ

Vi}
'v'i, j }

xi E

('$3')

XF E {(SZ~}IJ'XI~

A grey system is a system containing information presented


as grey numbers, and a grey decision is a decision made
within a grey system. Thus a grey mathematical programming (GMP) model can be defined as follows (Huang 1994):
Minimize

subject to

where X i is a grey vector of decision variables, f*(X*) is


a grey objective function, and g: (Xi) Ib:, Vi, are grey
constraints.
When model [2.4] is linear, integer linear, or quadratic,
it has interval solutions as follows:

[2.6]

x;~pt

[ ~ j , xTOpt],
~~'

xTOpt> x j O p tand V j

where xf can be discrete or continuous variables. The detailed


solution algorithms for the G M P models have been provided
by Huang (1994).
A simple example for the G M P model can be presented
as follows:
Minimize

subject to

where f * is a grey objective function, x: a r e grey decision


variables, and the coefficient [2, 31 represents a grey number
with its lower and upper bounds being 2 and 3, respectively
(and so on for the other coefficients).
According to the GMP solution algorithm proposed by
Huang (1994), the above model can be converted to two submodels as follows:
(i) Minimize

Huang et al.
Fig. 1. Graphical depiction of a grey mathematical programming problem and its solution.

-------

constraint [2 8b]

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

constrarnt [2 8c]

0
I

A = feaslble for [2 8b] and [2 8c]


B = feaslble for [2 8b] + softly-feaslble for
C = feasrble for [2 8c] + softly-feasible for
D = feasible for [2.8b]+ infeasible for [2.8;]
E = softly-feasible for [2.8b]and [2.8c]
F = feasible for [2.8c]+ infeasible for [2.8b]
G = softly-feas~blefor 12 8b] + Infeasible for [2.8c]
H = softly-feasible for [2 8c] + Infeasible for [2 8b]
I = infeas~blefor [2 8b] and [2 8c]
U4 0
= solution set

x2

subject to

(ii) Minimize

subject to

the upper bounds of their solutions, we get a lower f* value


but a higher possibility of violating the constraints. Conversely,
when xf and x $ approach the lower bounds, a higher f *
value but lower possibility of violation can b e obtained.
In an application to solid waste management, we can assume
that x$ and x $ represent waste flows, and f * is system cost.
Thus, there exists a tradeoff between system cost and solution feasibility. A conservative strategy will correspond to
estimation of higher waste flows and higher operation and
transportation costs. In comparison, an optimistic strategy
will relate to lower waste flows and lower operation and
transportation costs. Consequently, decision alternatives can
be generated through interpretation of the grey solutions
according to projected applicable system conditions.
The GMP solution algorithms are significantly different
from ordinary best or worst case analysis. In the GMP, the
solution corresponding to f - (lower bound of the objective
function value) can be first solved (when the objective is to
be minimized), and the relevant solution corresponding to
f (upper bound of the objective function value) was proven
to be feasible as one of the two bounds of the desired grey
solution (Huang 1994). Thus, the results corresponding to
f and f - lead to a set of optimal and stable grey solutions
(the grey solutions are stable if the objective function value
varies between f; and f ip,
as the decision variables change
between xyoptan$ xTOpJ.In a bestlworst case analysis, in
comparison, the major concern is the solution of the objective function value, while decision variable solutions for the
best and worst cases may not necessarily construct a set of
feasible and stable grey solutions (i.e., when the best case
(corresponding to f-) is first solved, the relevant worst case
solutions for decision variables may be infeasible as one of
the two bounds of the grey solution; conversely, when the
worst case (corresponding to f +) is first calculated, poor
grey solutions may be generated).
To obtain a second grey solution which is different from
the optimal solution, we introduce a GHSJ approach, where
the sum of nonzero variables in the initial solution is minimized
subject to a target constraint on the cost objective as follows:
+

where x j and xf represent the lower and upper bounds of


x:, respectively. x ; , and x ; , are the solutions obtained
from solving submodel [2.9].
Thus, the solution of model [2.8] is

Figure 1 shows a graphical depiction of this problem, where


"softly feasible" means that the feasibility is dependent upon
the location of the decision variables {xf,x $ ) and the conditions of the grey constraints. The solution set for x: and
x 3 is presented as a rectangle. When x: and x $ are close to

Can. J. Civ. Eng. Vol. 23, 1996

Fig. 2. Modelling process for the grey hop, skip, and jump
(GHSJ) approach.

I uncertain ~arametersI

Fig. 3. Hypothetical study municipalities and waste management


facilities.
Munici~alitv
. .2
Municipality 3

e
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

interactive solution algorithm

and objective function value

Facility 2

implicit knowledge

examination of different solutions and different combinations


of decision variable values within their solution intervals

,* * ,U W E Facility 1

potential decision scheme

interest groups:

I justifcation of the generated alternat~vesI

Minimize

subject to

- - - -,

Municipality 1
municipal solid waste
residue from waste-to-energy (VVTE) facility

represent uncertain inputs and outputs. This is particularly


meaningful for practical applications because (i) it is typically more difficult to specify distributions than to define
fluctuation intervals; and (ii) the existing optimization methods
that deal with distribution uncertainties have difficulties in
solution algorithms, computational requirements, and results
interpretation (Inuiguchi et al. 1990; Marti 1990).

3. Application to solid waste decision


making

where xf E X * , and a is the increment for the target


constraint. Normally, even though unmodelled issues are
considered, good alternative solutions are unlikely to be more
than 10% worse than the initial optimal solution (Chang and
Brill 1982). Additional alternative solutions can be obtained
by minimizing the sum of different combinations of nonzero
variables that appear in one or more of the previous solutions.
Figure 2 shows the general modelling process for the GHSJ
approach. The GHSJ provides alternative solutions represented
as grey numbers for the decision variables and objective
function value, which can be further used to generate several
deterministic alternatives by adjusting different combinations
of decision variable values within their solution intervals.
For each GHSJ solution, when the decision variable values
vary within their solution intervals, the objective function
value will change within its solution interval correspondingly.
The GHSJ has an advantage of low computational requirements, since it does not lead to more complicated intermediate
models due to the characteristics of the GMP solution algorithm (Huang 1994). Moreover, the method does not require
distribution information, since grey numbers are used to

3.1. Overview of the hypothetical problem


A hypothetical problem has been developed to illustrate the
GHSJ modelling approach based upon representative cost and
technical data from the solid waste management literature.
The study region is assumed to include three municipalities,
as shown in Fig. 3. Three time periods are considered with
each having an interval of 5 years. At the beginning of the
time horizon, an existing landfill and two waste-to-energy
(WTE) facilities are available to serve the region's solid
waste disposal needs. The landfill has an existing capacity of
[0.625, 0.7751 x lo6 t, and WTE facilities 1 and 2 have
capacities of [loo. 1251 and [200, 2501 t/d, respectively. The
WTE facilities generate residues of approximately 30% (on a
mass basis) of the incoming waste streams, and their revenues
from energy sales are approximately [15, 251 $/t combusted.
Over the 15-year planning horizon, the landfill capacity
can be expanded once by an increment of [ 1 .55, 1.701 X
lo6 t, and the WTE facilities can be expanded by any of
four options in each of the three time periods (see Table 1
for detailed information), with a maximum expansion limit
of 250 t/d. Table 1 also shows the capital costs for capacity
expansions for the three facilities, which are expressed in
terms of present value dollars, with the costs being escalated
to reflect anticipated conditions and then discounted to gener-

Huang et al

Table 1. Capacity expansion options and their costs for the


landfill and WTE facilities.

Table 2. Waste generation, transportation costs, and facility


operating costs.
Time period

Time period
Symbol

k = l

k = 2

Capacity expansion option for WTE facility i, i = 2, 3 (tld)


ATC, (option 1)
100
100
100
ATC,,, (option 2)
150
150
150
ATCi,, (option 3)
200
200
200
ATCi4, (option 4)
250
250
250

,,

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

Symbol

k = 3

Capacity expansion option for the landfill (lo6 t)


ALC
[1.55, 1.701 [1.55, 1.701
[1.55, 1.701

Capital cost of WTE facility expansion, i = 2, 3


($lo6 present value)
FTC,,, (option 1)
10.5
8.3
6.5
FTC,,, (option 2)
15.2
11.9
9.3
FTC,,, (option 3)
19.8
15.5
12.2
FTC,, (option 4)
24.4
19.1
15.0
Capital cost of landfill expansion ($lo6 present value)
FLC:
t13, 151
[13, 151
[13, 151

ate present value cost terms for the objective function.


Table 2 contains waste generation values for the three
municipalities, operating costs of the three facilities, and transportation costs for the waste flows between municipalities and
facilities in the three time periods. It is indicated that the
municipal solid waste generation rates and the costs for waste
transportation and treatment vary temporally and spatially.
Therefore, the problem under consideration is how to obtain
preferred facility expansion alternatives during different periods
and how to effectively allocate the relevant waste flows, in
order to minimize total system cost. Since the majority of
data for the system have uncertain features and are known
only as intervals without distribution information, the GHSJ
approach is considered to be appropriate for this problem.

k = l

k = 2

Waste generation (tld)


WG:, (Municipality 1) [200, 2501
[225, 2751
WGk (Municipality 2) [375, 4251
[425, 4751
WG:, (Municipality 3) [300, 3501
[325, 3751

[250, 3001
[475, 5251
[375, 4251

Cost of waste transportation


TR:,, (Municipality 1) [12.1, 16.11
TRf,, (Municipality 2) [10.5, 14.01
TR:,,(Municipality 3) [12.7, 17.01

to the landfill
[13.3, 17.71
t11.6, 15.41
[14.0, 18.71

($It)
[14.6, 19.51
[12.8, 16.91
[15.4,20.6]

Cost of waste transportation to


TR:,, (Municipality 1) [9.6, 12.81
TR&, (Municipality 2) [10.1, 13.41
TR:,, (Municipality 3) [8.8, 11.71

WTE facility
t10.6, 14.11
[11.1, 14.71
[9.7, 12.81

1 ($It)
[11.7, 15.51
[12.2, 16.21
[10.6, 14.01

Cost of waste transportation to


TR:,, (Municipality 1) [12.1, 16.11
TR?,, (Municipality 2) [12.8, 17.11
TR;,, (Municipality 3) [4.2, 5.61

WTE facility
[13.3, 17.71
[14.1, 18.81
[4.6, 6.21

2 ($It)
[14.6, 19.51
[15.5, 20.71
[5.1, 6.81

Cost of residue transportation from the WTE Facilities to the


landfill ($It)
FT& (WTE facility 1) [4.7, 6.31
[5.2, 6.91
[5.7, 7.61
FT& (WTE facility 2) [13.4, 17.91 L14.7, 19.71 i16.2, 21.71
Operational cost ($It)
OP:, (Landfill)
[30, 451
[40, 601
OP: (WTE facility 1) [55, 751
[60, 851
OP$ (WTE facility 2) [50, 701
[60, 801

[50, 801
[65, 951
[65, 851

subject to
3

[31b]

3.2. Model formulation

k'

C C L,[X',~+
j=l k = ~

C x&FE]
i=2

k'

In the municipal solid waste management system under consideration, grey decision variables include two categories:
continuous and binary. The continuous variables represent
"municipality
facility" waste flows over the time horizon,
and the binary variables represent facility expansion decisions.
The objective is to achieve optimal planning for facility
expansion and relevant municipal solid waste flow allocation
with minimum system cost. The constraints include all relationships between the decision variables and the waste generation and management conditions. A grey integer programming
(GIP) model can be formulated as follows:
Minimize

<

C ALC*y: + LC*;

k t = 1, 2, 3

k= l

(landfill capacity constraints)

i = 2 , 3 and k'

1, 2, 3

(WTE facility capacity constraints)

(waste disposal demand constraints)

[3.le]

C
Ill=

z:,,~ 5 1;

= 2,

3 and V k

Can. J. Civ. Eng. Vol. 23, 1996

(only one WTE facility expansion may occur in any given


time period)

(landfill expansion may only be considered once)

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

[3. lg]

xSk r 0;

V i, j , k

(nonnegativity constraints)

= integer;

= integer;

Vk

= 2,

3 and V m , k

(nonnegativity and binary constraints)


where i is the type of waste management facility (i = 1, 2,
3, with i = 1 for the landfill, and 2 and 3 for WTE facilities
1, and 2, respectively); j is the municipality, j = 1, 2, 3
(Fig. 2); k is the time period, k = 1, 2, 3; m is the expansion
option for the WTE facilities, m = 1, 2, 3, 4; xik is the waste
flow from municipality j to facility i during period k (tld);
y: is the binary decision variable for landfill expansion at
the start of period k ; z:f,k is the binary variable for WTE
facility i with expansion option m at the start of period k ,
i = 2, 3; C t k is the total cost of waste management for
waste flow from municipality j to facility i in period k ($It):
Ci:ilk

= TRL
ilk

C$

= TR$

+ OP$; for i = 1 and V j , k


+ OP$ + FE(FT$ + OP&) - RE:;
for i

= 2,

3 and V j , k

FE is the residue flow rate from the WTE facility to the landfill (% of incoming mass to the WTE facility); FLC: is the
capital cost of landfill expansion in period k ($); FT: is the
landfill" waste
transportation cost for "WTE facility i
flow during period k , i = 2 , 3 ($It); FTC,,,,k is the capital
cost of expanding WTE facility i by option m in period k ,
i = 2, 3 ($); Lk is the length of time period k (d); LC* is
the existing landfill capacity (t); OP$ is the operating cost of
facility i during period k ($It); RE: is the revenue from the
WTE facilities during period k ($It); TC! is the existing
capacity of WTE facility i, i = 2, 3 (tld); TR& is the transportation cost for waste flow from municipality j to facility i
during period k ($It); WG$ is the waste generation rate in
municipality j during period k (tld); ALC is the amount of
capacity expansion for the landfill (t); and ATC,,,,,, is the
amount of capacity expansion (option m) for WTE facility i
at the start of period k , i = 2, 3 (tld).
Equation [3. la] implies that the objective is to minimize
system cost which is related to the benefits and costs of
different waste management activities and capital costs for
related facility expansions. Constraints [3. lb] and [3. lc] stipu-

late that the upper limit for waste treatment and disposal in
any time stage is determined by both the existing and expanded
capacities for the landfill and WTE facilities. This dynamic
nature is related to economic development, population increase,
and environmental management activities. Constraint [3. Id]
states waste disposal demand for the three municipalities.
Constraint [3. le] requires that only one WTE facility expansion may occur for any given time period, and constraint
[3.lf ] stipulates that the landfill can only be expanded once
for the entire planning time horizon. Constraints [3.lg] to
[3. li] define technical relationships for the decision variables.
The detailed solution algorithm for the above GIP model
is provided by Huang et al. (1995). Generally, interactive
relationships between objective and constraints, between decision variables and parameters, and between different decision
variables are analyzed and quantitatively presented. Submodels
corresponding to the upper and lower bounds of the objective
function value are formulated based on the interactive relationships. Grey solutions are then generated through interpretation of solutions from the two submodels.
The solutions for the grey binary variables have four possible representations ([0, 01, [ I , 11, [O, 11, and [ l , 01). These
variables represent the related grey decisions that reflect
potential system condition variations caused by the input
uncertainties. For example, if an incinerator has two options
for capacity expansion in a given time period: 500 or 600 t/d
capacity (assuming that only one expansion is allowed in the
period, and the option of 600 tld has a higher capital cost
than that of 500 t/d), let the 500 t/d option correspond to a
binary variable x,, and the 600 t/d option correspond to
another binary variable x2. When both x , and x2 are deterministic variables, we will have one of the following three
possible solutions: (i) no expansion (xl = 0 and 12 = 0);
(ii) expanding by 500 t/d (xl = 1 andx2 = 0); o r (iii) expanding by 600 tld (x, = 0 and x2 = 1). Thus, the solutions
are deterministic and cannot effectively reflect the effects of
uncertainties.
In comparison, when x, and x2 are grey binary variables,
we have x l = x: and x2 = x$. Assume that the objective is
to minimize f * , and that x; and x; correspond to f - . Thus
we will have one of the following six possible solutions
(through formulating the related constraints):
(i) no expansion (x; = 0 , x; = 0; and x: = 0, x: = 0);
j = 1,
(ii) expanding by 500 t/d (x; = 1, x; = 0 ; and x
x: = 0);
(iii) expanding by 600 t/d (x; = 0, x; = 1; and x: = 0,
x; = 1);
(iv) expanding by [O, 5001 t/d (x; = 0, x; = 0; and x: =
1, x; = 0);
(v) expanding by [O, 6001 t/d (x; = 0, x; = 0; and x: =
0 , x; = 1);
(vi) expanding by [500, 6001 tld (x; = 1, x; = 0; and
x: = 0, x; = 1).
The solutions in (i) to (iii) are the same as those when x ,
and x2 are deterministic. However, the grey solutions in (iv)
to (vi) reflect potential system condition variations caused by
the existence of input uncertainties. The lower bound expansion values (and thus lower capital costs) correspond to advantageous system conditions (e.g., conditions when recycling,
reduction, and reuse (3R) initiatives are effective in controll-

Huang et al

Table 3. Optimal solutions obtained through the grey integer


programming (GIP) model.

Table 3 (concluded).
--

Symbol

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

Symbol

Facility

Expansion

Period

Binary decision variable


1
1
1
2
1
3

Solution

yf opl
y: opt
Y: opt

zI:

WTE facility 1
WTE facility 1
WTE facility 1

1
1
1

1
2
3

0
0
0

WTE facility 1
WTE facility 1
WTE facility 1

2
2
2

1
2
3

0
0
0

WTE facility 1
WTE facility 1
WTE facility 1

3
3
3

1
2
3

1
1
0

WTE facility 1
WTE facility 1
WTE facility 1

4
4
4

1
2
3

0
0
0

WTE facility 2
WTE facility 2
WTE facility 2

1
1
1

1
2
3

0
0
0

WTE facility 2
WTE facility 2
WTE facility 2

2
2
2

1
2
3

11, 01
1
0

WTE facility 2
WTE facility 2
WTE facility 2

3
3
3

1
2
3

10, 11
0
0

WTE facility 2
WTE facility 2
WTE facility 2

4
4

1
2
3

0
0
0

z:12

op1

I:3'

opt

z:21

op1

~ $ 2 2opt
6 2 3 opt
3:'I

opl

z:z

opt

3:z

opt

'$41

opt

~ $ 4 2opt

~ f l opt
l
G I 2 opt
'$13

opt

z:2,

apt

z;22

opt

'f23

opt

'$31

opt

z:3?

opt

~ $ 3 3opt

,z:

opt

2:z

opt

z:43

opt

*,,,

x:,?

Landfill
Landfill
Landfill

I opt

1
0
0

x
.r f321
~ $ 2

x$,
x:~,
x:~?

x;~,

Facility

Expansion

WTE facility 2
WTE facility 2
WTE facility 2

1
1
1

WTE facility 2
WTE facility 2
WTE facility 2

2
2
2

WTE facility 2
WTE facility 2
WTE facility 2

3
3
3

Period

Solution

System cost, f* ($lo6):

Fig. 4. Expansion scheme for waste-to-energy (WTE) facilities


(optimal solution).
(a)

Continuous decision variable (tld)


Landfill
1
1
0
Landfill
1
2
0
Landfill
1
3
0
Landfill
Landfill
Landfill

2
2
2

1
2
3

[263, 2711
[51, 721
[125, 1371

Landfill
Landfill
Landfill

3
3
3

1
2
3

0
0
0

800 -1

10

15

Time (year)

(b)
800 -

S 700-

10

15

Time (year)

WTE facility 1
WTE facility 1
WTE facility 1

1
1
1

1
2
3

[200, 2381
122
150

WTE facility 1
WTE facility 1
WTE facility 1

2
2
2

1
2
3

87
[374, 4031
[350, 3631

WTE facility 1
WTE facility 1
WTE facility 1

3
3
3

2
3

0
0

ing waste generation), and the upper bound expansion values


(and thus higher capital costs) correspond to m o r e demanding system conditions. For example, [500, 6001 tld in ( v i )
means that the expansion level is quite flexible and can b e
either 500 o r 600 t/d, which correspond to different system
conditions.
Thus, through the GHSJ approach as applied to model
[3.1], decision alternatives (close-to-optimal solutions) based
o n the GIP model can be obtained by minimizing the sum of

1214

Can. J. Civ. Eng. Vol. 23, 1996

Table 4. Close-to-optimal solutions under different target constraints for the cost objective (tld).
- --

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

Symbol

1I:'

opt

'$12

opt

'$13

opt

'$21

opt

~ $ 2 2opt
z:23

op1

';4I

opt

'i42

opt

'$3

op1

I: '

l opt

'fl2

opt

:I'3

opt

'$21

opt

2:'2

opt

'i23

opl

z:31

opt

3:'2

opt

z:33

.pi

~ f 4 opt
I
'$42

apt

4:'3

opt

x T I l op1
x T 1 2 opt
'?I3

opt

x T 2 1 opt
x T 2 2 op1
~ T 2 o3 p ~
'$31

opl

~ T 3 opt
2
x T 3 3 op1
~ $ l1 opt
x:12

op1

~ 1 1 opt
3
~ $ 2 1opt
x:22

apt

x;23

opt

~ : 3l opt
x:3?

opt

~ $ 3 3opt
~ f l opt
l
x:12

opt

:I3'

opt

Facility

Expansion

Period

Alternative 1.1

Binary decision variable


1
1
2
0
3
0

Alternative 1.2 Alternative 1.3

Landfill
Landfill
Landfill

1
1
1

WTE facility 1
WTE facility 1
WTE facility 1

1
1
1

1
2
3

0
0
0

0
0
0

WTE facility 1
WTE facility 1
WTE facility 1

2
2
2

1
2
3

0
0
0

0
0
0

WTE facility 1
WTE facility 1
WTE facility 1

3
3
3

1
2
3

[1,01
1
1

[ I , 01
1
1

WTE facility 1
WTE facility 1
WTE facility 1

4
4
4

1
2
3

10, 11
0
0

10, 11
0
0

WTE facility 2
WTE facility 2
WTE facility 2

1
1
1

1
2
3

0
0
0

0
0
0

WTE facility 2
WTE facility 2
WTE facility 2

2
2
2

1
2
3

0
0
0

0
0
0

WTE facility 2
WTE facility 2
WTE facility 2

3
3
3

1
2
3

1
1
1

1
1
0

WTE facility 2
WTE facility 2
WTE facility 2

4
4
4

1
2
3

0
0
0

0
0
[O, 11

Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill

Continuous decision variable (tld)


1
1
0
0
1
2
[O, 421
LO,421
1
3
0
0
2
1
262
262
2
2
59
59
2
3
116
116
3
1
0
0
3
2
0
0
3
3
0
0

WTE
WTE
WTE
WTE
WTE
WTE
WTE
WTE
WTE

1
0
0

1
1
1
1
1
1
1
1
1

1
1
1
2
2
2
3
3
3

1
2
3
1
2
3
1
2
3

[200, 2501
122
125
[63, 751
366
359
10, 501
10, 501
10, 501

[200, 2381
[122, 1301
150
88
366
359
[O, 501
[O, 501
[O, 501

WTE facility 2
WTE facility 2
WTE facility 2

1
1
1

1
2
3

0
[103, 1111
[125, 1751

[O, 131
103
[loo, 1501

facility
facility
facility
facility
facility
facility
facility
facility
facility

Huang et at.

Table 4 (concluded).

Symbol

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

x:,,
x:,~
x:,, opt
x : ~ ,opt
x : ~ , opt
x:,, opt

Facility
WTE
WTE
WTE
WTE
WTE
WTE

facility 2
facility 2
facility 2
facility 2
facility 2
facility 2

Expansion Period Alternative 1.1 Alternative 1


2
2

2
3
3
3

Alternative 1.3

[25, 751
10, 501
[O, 501
300
325
375

System cost, f* ($lo6)

different combinations of nonzero variables that appeared in


the previous solutions subject to different target constraints
on the cost objective.

3.3. Modelling results


Table 3 and Fig. 4 contain and illustrate the results for the
initial optimal solution obtained through the GIP model.
Table 4 presents close-to-optimal solutions obtained through
the GHSJ approach, where alternatives 1.1, 1.2, and 1.3 were
generated under the conditions where the sum of all nonzero
continuous variables in the initial solution was minimized
subject to different target constraints on the cost objective
( a = 0.02, 0.05, and 0.08). The solutions for alternative 1.3
( a = 0.08) are illustrated in Fig. 5. Table 5 contains the
close-to-optimal solutions for the conditions when the sum of
different combinations of continuous variables (waste flows
to landfill, WTE facility 1, and WTE facility 2, respectively)
in the initial solution was minimized for an a value of 0.05.
It is indicated that solutions for the objective function value
and many decision variables are grey numbers.
The results indicate that the landfill should be expanded
at the start of period 1 ( y+ = [ l , I]), but not expanded further in periods 2 and 3 ( y$ and y$ are both equal to [O, 01).
The amount of expansion is the [1.55, 1.701 x lo6 t level
input into the model (Table 1). Figure 4 shows the optimal
expansion schemes for WTE facilities 1 and 2, respectively.
It is indicated that WTE facility 1 should be expanded by
200 tld in both periods 1 and 2 (z:~, = [I, 11 and z:~, =
[ l , 11) and WTE facility 2 should be expanded by 1150, 2001
tld in period 1 and 150 tld in period 2 ,z(:
= [ I , 01, z,:
=
11, 11, and zz3, = [O, 11). The expansion of 1150, 2001 tld
in period 1 means that there are two alternatives for the
expansion, where 150 tld corresponds to f - , and 200 tld
corresponds to f +. Thus, when the decision scheme tends
toward f - under advantageous conditions, it may be applicable to expand WTE facility 2 by 150 tld in both periods 1
and 2; and when the scheme tends toward f + under more
demanding conditions, it may be suitable to expand WTE
facility 2 by 200 tld in period 1 and 150 tld in period 2. No
expansion should be carried out in period 3 for either of the
facilities, since sufficient capacity has been developed in the
previous periods (see Fig. 4 for details).
For the grey continuous variable solutions, it is indicated
in the optimal solution that the landfill accepts wastes only
from municipality 2 because of its close proximity to the
municipality and landfill capacity limitations, in addition to
residues from the WTE facilities. All waste flows from
municipality 3 and delivered to WTE facility 2 due to its

Fig. 5. Expansion scheme for waste-to-energy (WTE) facilities

(alternative 1.3).
(a)

800 -1

Time (year)

% 300

g 200
I00
0

10
Time (year)

close proximity to the facility. WTE facility 2 also accepts


a portion of the flows from municipalities 1 and 2. The
remaining waste flows from municipalities 1 and 2 are determined to be hauled to WTE facility 1. As shown in Table 6 ,
[30.1, 26.4]%, 132.8, 31.7]%, and 137.1, 41.91% of the
total waste are determined to be routed to landfill, WTE
facility 1, and WTE facility 2, respectively, in period 1.
Similar distribution patterns can also be found in the solutions for periods 2 and 3, although direct flows to the landfill
are reduced owing to capacity limitations.
Alternative 1.1 was generated by minimizing the sum of
all nonzero continuous variables in the initial solution subject
to a target constraint off* 1 1.02 f&,. It is indicated that

1216

Can. J. Civ. Eng. Vol. 23, 1996

Table 5. Close-to-optimal solutions when different variable combinations are optimized.


-

Symbol

Y:
Y:
Y'

opt
opt

3 opt

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

I:1'

opt

6 1 2 opt
:I3'
I*: '

opt
opt

z:22

opt

2:'3

opr

I:'

opr

;'32

opr

z:33

opt

'&I

opr

4:'2

opt

G 4 3 opt
'$1

l opt

2:12

opt

:I3'

opt

'$21

opt

2:22

opt

'$23

opt

3:'l

opt

'$32

opr

'$33

opr

'$41

opr

'$42

opt

'$43

opt

x:l

I opt

~ f l opt
2
x i 1 3 opt
x i 2 1 opt
x:22

opt

'?23

opt

'731

opt

~ ? 3 2opt
x i 3 3 opr

Facility

Expansion

1
1
1

1
2
3

0
0
0

WTE facility 1
WTE facility 1
WTE facility 1

2
2
2

1
2
3

0
0
1

WTE facility 1
WTE facility 1
WTE facility 1

3
3
3

1
2
3

1
[I, 01
0

WTE facility 1
WTE facility 1
WTE facility 1

4
4
4

1
2
3

0
[o, 11
0

WTE facility 2
WTE facility 2
WTE facility 2

1
1
1

1
2
3

0
[I, 01
0

WTE facility 2
WTE facility 2
WTE facility 2

2
2
2

1
2
3

0
[o, 11
0

WTE facility 2
WTE facility 2
WTE facility 2

3
3
3

1
2
3

[1,01
0
1

WTE facility 2
WTE facility 2
WTE facility 2

4
4
4

1
2
3

[o, 11
0
0

Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill
Landfill

Continuous decision variable (tld)


1
1
0
0
2
0
0
1
1
3
0
0
2
1
237
262
2
2
34
[59, 621
2
3
112
116
3
1
0
0
2
0
0
3
o
[O, 391
3
3

opt
opt

x;33

opt

opr

WTE facility 2
WTE facility 2
WTE facility 2

~ $ 2 1opr
~ $ 2 2opr
x:23

'$13

opt

Binary decision variable


1
1
2
0
3
0

WTE facility 1
WTE facility 1
WTE facility 1

lx:

~ ; 1 3 opt

Alternative 3

1
1
1

2:x

l opr

x i 1 2 opt

Alternative 2

Landfill
Landfill
Landfill

WTE facility
WTE facility
WTE facility
WTE facility
WTE facility
WTE facility
WTE facility
WTE facility
WTE facility

x:l

Period

1
1
1
1
1

1
1
1
1

Alternative 4

Huang et al.

Table 5 (concluded).

Symbol

Facility

WTE facility 2
WTE facility 2
WTE facility 2
WTE facility 2
WTE facility 2
433
WTE facility 2
System cost, f+ ($lo6)

xi2,opt
xi2?opt
xi2,
xi,, op,
xt3,

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

O ~ I

Expansion

Period

2
2
2
3
3
3

1
2
3
1
2
3

Alternative 2

Alternative 3

[50,100]
0
[o, 501
[300, 3501
[325, 3751

[50, 1001
to, 471
[o, 501
[300, 3501
[325, 3751

25
0
0
[300, 3271
325

[375, 4251

[375, 3861

375

[405.1, 725.41

[405.1, 725.41

[405.1, 725.41

both WTE facilities 1 and 2 should have more expansion


capacity in this alternative than in the optimal solution, which
leads to an increase in system cost. The detailed allocation
facility" waste flows in this alternative
for "municipality
is also different from the optimal solution. However, the
general percentage-distribution pattern for waste flows to
the landfill, WTE facility 1, and WTE facility 2 is similar to
that of the optimal solution (see Table 6). This is an option
that provides alternative waste flow allocation patterns and
increased waste management capacity relative to the optimal
solution based on a 2 % increase in the system cost. Alternative 1.2 was subject to a target constraint off * 2 1.05f:pt.
Generally, the facility-expansion and waste-flow-allocation
solutions in alternative 1.2 are similar to those in alternative 1.1. Alternative 1.3 was subject to a target constraint of
f * r 1.08f Zpt. It is indicated that, compared to the optimal
solution and alternatives 1.1 and 1.2 solutions, more waste
flows to WTE facilities (i.e., sum of waste flows to WTE
facilities 1 and 2) and more expansions of the two WTE facilities are determined, which corresponds to the target 8%
increase in the system cost. Since more flow to WTE facilities and less flow to the landfill are provided, this alternative
is an option that emphasizes land resource conservation under
an assumption that increased utilization of the WTE facilities
(and thus increased system cost) is allowed.
When the a value is increased, a higher system cost will
be incurred for the waste management activities. The increased
cost can be devoted to increased expansion of waste management facilities to provide more capacity for waste disposal
and treatment in the future, and (or) allocation of increased
waste flows to the WTE facilities (and thus less flows to the
landfill) for the purpose of land resource conservation.
Alternatives 2, 3, and 4 in Table 5 were generated by
minimizing the sum of different combinations of decision variables in the initial solution (min Cj Ck x:,~, min Cj Ck x % ~ ,
and min C, Ckx $ ~for alternatives 2, 3, and 4, respectively),
for an increase in system cost of 5 % for all three cases. When
the sum of flows to a facility is minimized, the relevant solutions for these flows will be lower than those in other alternatives. For example, in alternative 2, waste flows to the landfill
are generally lower than those in other alternatives, since
Cj Ck x S . ~is minimized (see Table 6 for detailed compari,
on increased system costs, alternative 2
sons). ~ h u s based
is an option that emphasizes land resource conservation under
an assumption that increased utilization of the WTE facilities
is allowed; alternative 3 relates to the situation when environmental regulations or community perception may influence

Alternative 4

'

the utilization of WTE facility 1 (where more waste flows are


determined to be routed to WTE facility 2); and alternative 4
relates to the situation when the utilization of W T E facility 2
is affected, where more waste flows are routed to WTE
facility 1.
Generally, the results indicate that, through the proposed
modelling approach, uncertain information can be effectively
communicated into the GHSJ optimization processes and
resulting solutions. Thus, decision alternatives can be generated from the GHSJ solutions according to projected planning
situations. For each alternative individually, lower decision
variable values (i.e., lower flows to waste management facilities and lower expansion for the WTE facilities) within their
solution intervals should be used under advantageous system
conditions (e.g., conditions when recycling, reuse, and reduction initiatives are effective in reducing waste disposal and
treatment demands), and higher decision variable values within
their solution intervals should be used under more demanding system conditions. This GHSJ solution feature may be
favored by decision makers because of the increased flexibility
and applicability for determining final decision schemes.
Thus the GHSJ solutions provide various alternatives that
can be evaluated from the stand point of environmental and
economic tradeoffs. Decision makers can compare a range of
systems with respect to cost and relative utilization of various
facilities, and can obtain optimal system operation characteristics for a range of policy alternatives (e.g., minimize
landfill usage) that will in turn assist in the development of
a preferred operational strategy under certain conditions.

4. Concluding remarks
A GHSJ approach has been applied to a hypothetical case
study of municipal solid waste management planning. The
method improves upon the existing HSJ approach by allowing
uncertain information, presented as interval numbers, to be
effectively communicated into the optimization process and
resulting solutions. Feasible decision alternatives can be generated through interpretation of the optimal and close-to-optimal
grey solutions (presented as stable intervals) according to
projected applicable system conditions. Results from the hypothetical case study indicate that potentially useful solutions
for the expansion planning of solid waste management facilities
have been generated. The decision alternatives obtained may
be interpreted and analyzed to internalize environmental and
economic tradeoffs, which may be of interest to public sector
decision makers faced with difficult and controversial choices.

Can. J. Civ. Eng. Vol. 23, 1996


Table 6 . Distributions of waste flows to different waste management facilities in time
period k.
k = l
Lower
bound

Upper
bound

Lower
bound

Optimal solution
To landfill (tld)
To WTE facility 1 (tld)
To WTE facility 2 (tld)
To landfill (%)
To WTE facility 1 (%)
To WTE facility 2 ( % )

263
287
325
30.1
32.8
37.1

27 1
325
429
26.4
31.7
41.9

51
496
428
5.2
50.9
43.9

Alternative 1.1 ( a = 0.02)


To landfill (tld)
To WTE facility 1 (tld)
To WTE facility 2 (tld)
To landfill (%)
To WTE facility 1 (%)
To WTE facility 2 (%)

262
263
350
29.9
30.1
40.0

262
375
388
25.5
36.6
37.9

59
488
428
6.1
50.0
43.9

Alternative 1.2 ( a = 0.05)


To landfill (tld)
To WTE facility 1 (tld)
To WTE facility 2 (tld)
To landfill (%)
To WTE facility 1 (%)
To WTE facility 2 (%)

262
288
325
29.9
33.0
37.1

262
376
388
25.5
36.6
37.9

59
488
428
6.1
50.0
43.9

Alternative 1.3 ( a = 0.08)


To landfill (tld)
To WTE facility 1 (tld)
To WTE facility 2 (tld)
To landfill ( % )
To WTE facility 1 ( % )
To WTE facility 2 (%)

247
278
350
28.2
31.8
40.0

247
328
450
24.1
32.0
43.9

34
488
453
3.5
50.0
46.5

Alternative 2 (minimize flows to landfill)


To landfill (tld)
237
237
To WTE facility 1 (tld) 288
325
350
463
To WTE facility 2 (tld)
27.1
To landfill (%)
23.1
T o W T E f a c i l i t y l ( % ) 32.9
31.7
To WTE facility 2 (%)
40.0
45.2

34
500
441
3.5
51.3
45.2

Alternative

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

Alternative 3 (minimize flows to WTE


To landfill (tld)
262
To WTE facility 1 (tld) 263
350
To WTE facility 2 (tld)
29.9
To landfill (%)
To WTE facility 1 (%)
30.1
To WTE facility 2 (%)
40.0

facility 1)
262
59
263
463
500
453
25.5
6.1
25.7
47.5
48.8
46.5

Alternative 4 (minimize flows to WTE


To landfill (tld)
262
To WTE facility 1 (tld) 288
To WTE facility 2 (tld) 325
29.9
To landfill (%)
To WTE facility 1 (%)
32.9
To WTE facility 2 (%)
37.0

facility 2)
298
59
375
488
352
428
29.1
6.1
36.6
50.0
34.3
43.9

2
Upper
bound

k
Lower
bound

3
Upper
bound

Huang et al.

Acknowledgments
This research has been supported by the Natural Sciences and
Engineering Research Council of Canada. We are also grateful
to the anonymous reviewers for their helpful comments.

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 06/04/13
For personal use only.

References
Baetz, B.W., Pas, E.I., and Neebe, A.W. 1990. Generating alternative solutions for dynamic programming-based planning problems.
Socio-Economic Planning Sciences, 24: 27 -34.
Brill, E.D. 1979. The use of optimization models in public sector
planning. Management Science, 25: 413 -422.
Brill, E.D., Chang, S.Y., and Hopkins, L.D. 1981. Modelling to
generate alternatives: the HSJ approach and an illustration using
a problem in land use planning. Management Science, 27:
314-325.
Chang, S.Y., Brill, E.D., and Hopkins, L.D. 1980. Efficient random
generation of feasible alternatives: a land use planning example.
Staff Paper 12, Institute for Environmental Studies, University
of Illinois at Urbana-Champain, Urbana, Ill.
Chang, S.Y., Brill, E.D., and Hopkins, L.D. 1982. Use of mathematical models to generate alternative solutions to water resources
planning problems. Water Resources Research, 18: 58-64.
Church, R.L., and Huber, D.L. 1979. On finding close to optimal
solutions as well as determining a noninferior set of tradeoff
envelopes for multi-objective location models. Research Series 34,
Department of Civil Engineering, University of Tennessee,
Knoxville, Tenn.

Falkenhausen, K. 1979. Facility location planning for regional


waste treatment systems by branch and bound vs. evolution
strategy concepts. The Joint Meeting and Operations Research
Society of America and the Institute of Management Science,
Milwaukee, Wis.
Gidley, J.S., and Bari, M.F. 1986. Modelling to generate alternatives. ASCE Water Forum '86, American Society of Civil
Engineers, New York, pp. 1366- 1373.
Huang, G.H. 1994. Grey mathematical programming and its application to municipal solid waste management planning. Ph.D.
dissertation, McMaster University, Hamilton, Ont.
Huang, G.H., Baetz, B.W., ,and Patry, G.G. 1995. Grey fuzzy
integer programming: an application to regional waste management planning under uncertainty, Socio-Economic Planning
Science. 29: 17-38.
Inuiguchi, M., Ichihashi, H., and Tanaka, H. 1990. Fuzzy programming: a survey of recent developments. In Stochastic versus
fuzzy approaches to multiobjective mathematical programming
under uncertainty. Edited by R. Slowinski and J. Teghem. Kluwer
Academic Publishers, Dordrecht, The Netherlands. pp. 45-70.
Marti, K. 1990. Stochastic programming: numerical solution techniques by semi-stochastic approximation methods. In Stochastic
versus fuzzy approaches to multiobjective mathematical programming under uncertainty. Edited by R. Slowinski and J. Teghem.
Kluwer Academic Publishers, Dordrecht, The Netherlands.
pp. 23-44.

Вам также может понравиться