Sunteți pe pagina 1din 32

EUGENE: an optimization model for solid waste management planning 2

EUGENE: an optimization model for integrated regional solid waste


management planning

C. Berger
GERAD, École des Hautes Études Commerciales, 3000, chemin de la Côte-Sainte-Catherine, Montréal,
Quebec, Canada, H3T 2A7
(e-mail: claudeb@crt.umontreal.ca)

G. Savard
GERAD and Département de Mathématiques et de Génie Industriel, École Polytechnique, C.P. 6079,
Succ. Centre-Ville, Montréal, Quebec, Canada, H3B 3A7
(e-mail: gilles@crt.umontreal.ca)

A. Wizere
Département de Mathématiques et de Génie Industriel, École Polytechnique, C.P. 6079, Succ. Centre-
Ville, Montréal, Quebec Canada, H3B 3A7
(e-mail: agnes@crt.umontreal.ca)

Abstract: This paper presents a sophisticated mixed-integer linear programming model developed to help regional
decision-makers in the long-term planning of solid waste management activities. The model removes practically all
the limitations of earlier integrated waste management models. All the features and capabilities of the tool are
described in technical and non-technical language. The paper includes statistics from a real-world application, and
some directions for future developments.

Keywords: integrated solid waste management, mixed-integer programming, modelling, optimization, regional
systems, systems analysis.

Reference to this paper should be made as follows: Berger, C., Savard, G. and Wizere, A. (1999) ‘EUGENE: an
optimization model for integrated regional solid waste management planning’, Int. J. Environment and Pollution, Vol.
12, Nos. 2/3, pp. 280–307.

1 Introduction
The aim of this paper is to present in a detailed way a new optimization model to deal with the integrated solid waste
EUGENE: an optimization model for solid waste management planning 3

management planning problem at a regional level. This model is named EUGENE, in honour of Eugène René Poubelle, the
prefect of Seine who in 1884 forced the landlords to provide trash cans (in French, poubelle) to their tenants. It is the first
result of a long-term project aimed at producing a decision-support system for regional decision-makers. The first objective
pursued was to develop a model which allows the user to capture practically all aspects of this planning problem. The heart
of this paper is a mathematical description of the model.
At all times, human activities have generated solid wastes that were discarded because they were considered as negative
value goods. Traditionally, the most prevalent method of disposing of those wastes has been to first collect them from the
source and then bury them in a landfill. However, the steady increase of landfilled wastes and the emergence and
development of a new public environmental consciousness have created a strong negative attitude toward landfilling.
Regional and national regulations to protect the environment have increased the cost of developing new landfills and siting
them has become increasingly difficult because the public opposes having such a facility nearby. Solid waste management
has thus become a major concern in industrialized countries. The ideal way to improve the situation would be to reduce
consumption and, thus, the generation of waste. But this goes against the people’s will to preserve their lifestyle and thus to
consume more goods. Consequently, society is searching for improved methods of waste management and ways to reduce
the amount of waste to be landfilled.
Fortunately, many of these waste materials can be reused or transformed, into useful materials or energy, if properly
managed. Many waste management options have been proposed and implemented during recent years: waste source
reduction, reusing, recycling, composting, incineration with or without energy recovery, fuel production, and landfilling.
However, none of these options can solve the solid waste problem by itself. They must be combined intelligently in such a
way as to reduce the environmental and social impacts at an acceptable cost for the community. This is what is usually
referred to as integrated solid waste management.
Operations research models are particularly well suited for the description of complex systems which involve many
variables and constraints. For example, these models have been successfully used in the energy sector since the advent of
digital computers. They were first used for very specific problems with a limited scope, but large and integrated models
were developed during the last 25 years to help in understanding the complexity of the systems as well as assessing the
long-term role and impacts of new technologies (see Fishbone and Abilock, 1981; Jadot et al., 1979; and Schrattenholzer,
1981).
The use of optimization models in solid waste management has followed a similar pattern with some delay. The first
models dealt with specific aspects of the problem: vehicle routing, assignment of generating sources to landfills, transfer
stations siting, site selection for landfills, etc. The first generation of integrated models incorporated simplified descriptions
of the system and were subject to many limiting assumptions: weak disaggregation of material flows, one processing option
of each type, sites dedicated to one particular processing or landfilling technology, only one time period,
recyclables/organics collections rarely taken into account, poor (or no) description of markets for recyclables, a single waste
generating source, insufficient user’s control on the accuracy of the investment cost functions, etc.
More recent models, which appeared in the 1990s, remove some of the above limitations but not all. Gottinger (1991)
proposed a dynamic network flow model with non-linear costs for waste management and facility-siting decisions. Shekdar
et al. (1992) described a dynamic goal-programming model for the management of existing facilities of the waste system. A
multi-objective mixed-integer programming approach was proposed by Caruso et al. (1993) for the study of a regional
EUGENE: an optimization model for solid waste management planning 4

system over a single time period. An interesting dynamic mixed-integer programming model incorporating a large set of
technologies and dealing with pollution costs was presented by Chang et al. (1993). The preceding model has been
transformed into a multi-objective one by Chang and Wang (1996). It takes four different criteria into account, three of them
being environmental functions. Revenues from sales to markets are taken into account in the dynamic mixed-integer
programming model of Baetz and Neebe (1994). The model has a limited choice of technologies and only one new
landfilling site may be developed. A multiperiod and multi-regional model developed by Everett and Modak (1996) has
some interesting distinguishing features. Amongst them, there is the consideration of (1) aggregated and disaggregated
flows of materials and (2) a number of collection options for the components of the waste stream. The model does not deal
with capacity addition. A very detailed static non-linear programming model, MIMES/WASTE, has been proposed by
Sundberg et al. (1994) to address municipal and regional problems. The main objective for the model is cost minimization,
but emissions control is integrated into the model via explicit restrictions and fees. Recycling and energy production goals
may also be imposed. The MWS model of Ljunggren (1997) is an extension of MIMES/WASTE to national problems. In
the Swedish application, the national waste management system is represented by ten generalized municipal waste
management subsystems. The model also enables the analysis of regional transports of waste.
The EUGENE model, described below, removes all of the above limitations encountered in the first generation of
integrated models. It can describe very general real-life systems with many different options and variants of them. The next
section presents in non-mathematical language the addressed problem and the major features of the model. This is followed
by two sections dealing respectively with the implementation of the model and its future developments. The paper ends with
a short conclusion and two Appendices containing the mathematical description of the model.

2 An overview of the model


EUGENE is a process-oriented model based on the mixed-integer programming paradigm. As stated by Wilson (1977), the
merit of such an approach is that ‘large numbers of strategies may be evaluated against several criteria on a routine and
consistent basis, giving the planner better information on which to base his recommendations’. The problem addressed by
EUGENE can be stated as follows:

Given
1 the locations of sources,
2 the generation of each type of waste from each source,
3 the various types of collection of waste, recyclables and organics,
4 the allowed combinations of collections at each source,
5 a set of sites where facilities can be located,
6 a set of existing or potential processing (like composting facilities, incinerators, material recovery facilities, etc.) and
landfilling technologies,
7 a set of markets for energy, materials and recyclables/organics collections,
EUGENE: an optimization model for solid waste management planning 5

8 the distances between the sources, locations and markets,


9 a number of time periods and their length.

Determine
1 the combination of collections at each source for each time period,
2 a schedule of capacity additions for each landfilling/processing facility at each location,
3 the assignment at each time period of each collection of each source to processing/landfilling locations and markets,
4 the activity of each technology for each operating mode at each location and time period,
5 the assignment of the materials generated at each site to other sites and markets.
Figure 1 contains a simple case illustration of the schematic network of the municipal solid waste (MSW) system handled
by EUGENE. It will help the reader to understand some of the model features.
As can be seen from Figure 1, the model works with three types of node and two types of flow. The three types of node
are associated with the waste-generating sources, the processing and landfilling locations, and the markets. The number of
nodes of each category is under the user’s control. The region under study is usually segmented into a number of relatively
homogeneous sources relative to their waste-generation rates. Note that a user may distinguish processing nodes from the
landfilling nodes, as Figure 1 may suggest, but EUGENE does not make any distinction between them. This statement will
be clarified later in this section.
EUGENE: an optimization model for solid waste management planning 6

Figure 1 The schematic network of the system.


These three types of node are linked together using aggregated and disaggregated flows. Aggregated flows (i.e. those
associated with the bold arcs of Figure 1) are those leaving the source nodes. They are associated with waste, recyclables
and organics collections and they are sent to the various processing/landfilling and market locations. Recyclables/organics
collections are not allowed to be sent to landfilling locations. Disaggregated flows are those leaving the
processing/landfilling locations for markets or other processing/landfilling locations. They correspond to the individual
materials that are contained in the collections or produced by a processing/landfilling technology. Energy flows may also
exit the technologies. This is the case, for instance, of landfill gas or electricity produced from this gas at a landfilling
location. Processing/landfilling technologies may have as input one or several aggregated and disaggregated flows.
However, their output can only be associated with disaggregated flows.
Here are some examples illustrating the use of aggregated/disaggregated flows at the processing/landfilling nodes. They
are in agreement with Figure 1. An incinerator may burn waste collections (aggregated flows) or individual products
(disaggregated flows) with high heating values (like the paper and plastic rejects of a material recovery facility) and obtain
energy and ashes (disaggregated flows). Ash must be landfilled and, in some cases, treated before being landfilled. It is also
possible to define some collections for recyclables with high heating values and send them directly to the incinerator, thus
avoiding the material recovery facility (MRF). A MRF receives collections of recyclables (aggregated flows) and separates
them into flows of individual marketable products (like newspapers, magazines, cardboards, and different plastics) and
EUGENE: an optimization model for solid waste management planning 7

rejects (disaggregated flows) that leave the site for market locations or other processing/landfilling locations. A composting
plant may process organics collections (aggregated flows) and some of the materials (disaggregated flows), like paper and
cardboard, obtained at the gate of the MRF. MSW composting facilities may process municipal solid waste alone or mixed
in some proportion with sewage sludge (a recipe consisting of an aggregated flow and a disaggregated flow) and produce
compost, rejects and some marketable recyclable products (disaggregated flows).
A high flexibility is thus available for the description of the individual technologies. Collections of recyclables and/or
organics may also be sent directly from the sources to the markets: one may define a collection of recyclables corresponding
to a group of specialized drop-off centres consisting of containers that are frequently replaced and hauled to some market
locations. Collections of recyclables and/or organics are not allowed in landfills; waste collections and disaggregated
materials are permitted.
Each collection of recyclables/organics is defined by a list of allowed source generated materials it contains. The
quantity of materials in such a collection in a given region depends on the participation rate of the households and the
capture rates of each of the collected materials.
These recyclables/organics collections are grouped together by the user into various valid combinations. For instance, a
possible combination of collections may consist of a collection for recyclable paper once every two weeks, a collection for
food waste once every two weeks and a weekly collection of (remaining) waste. The same collections, but with a different
set of collection frequencies, may define another combination. For each source a list of allowed combinations must be
supplied. The model does not choose individual collections but instead selects exactly one combination of collections at
each source and time period. There are thus no flow variables associated with the combinations. However, there are binary
variables indexed by combination names, generating sources and time periods. Once a particular combination is selected for
a given source and time period (i.e. when its binary variable equals unity), then the flow variables of its recyclables/organics
collections may take positive values. Note also that a waste collection is implicitly associated with each triplet (combination,
source, period). It is the collection of remaining wastes after the recyclables/organics collections have been done. The empty
combination is allowed: in this case there is no recyclables/organics collection and all the waste is directed into the waste
collection.
Some parameters may be used to specify that a particular recyclables/organics collection of a given combination is
available for only a fraction of the households of a given source. It is possible this way to have two or more non-disjoint
collections (i.e. collections with at least one common collected material) in a given combination for a particular source. For
instance, a combination may consist of two collections defined with the same list of materials. The first one is a curbside
collection available for 50% of the households while the second one is a group of drop-off centres serving 35% of the
households. In this case the model assumes that the two collections serve disjoint parts of the source’s territory and that 15%
of the households are not served by any collection of recyclables/organics (but are still served by a waste collection). When
collections are disjoint, they may serve the same part of the territory.
Once the wastes, recyclables and organics are collected, they may be transported to one or more sites where they can be
processed or landfilled (recyclables/organics collections may also be sent to markets). Each site is characterised by its
location, its area, its unit cost, and a list of technologies that are suitable for it. For EUGENE, landfilling is a technology just
like composting, incineration, and material recovery. Unlike many other existing models, EUGENE does not force the user
to a priori decide on the use of the sites. Thus the cardinality of the list of allowed technologies for a given site is not
EUGENE: an optimization model for solid waste management planning 8

necessarily equal to unity. It is very common with the models encountered in the literature to have lists of dedicated sites for
landfilling, incineration, composting, material recovery facilities, etc. In those models, only one technology is available for
a particular processing/landfilling location. Those models can decide only if they will install some capacity of the dedicated
technology on a given site. With EUGENE, a given undeveloped site may eventually be developed for any of the uses
implied by its list of technologies, but for no more than one. EUGENE thus decides if the site needs to be developed, the
type of technology that will be installed on it, and the schedule of capacity additions.
Each technology may be characterized by one or more operating modes. For instance, an incinerator may burn waste
collections, or different categories of paper, or some proportions of specified products (like 45% of papers, 25% of plastics,
and 30% of organics). In this case, operating modes may be specified for the waste collections of each combination, for the
individual types of paper, and for each specified mixture of products. A composting facility may be operated using one or
many valid recipes, like paper (67%) with sewage sludge (33%), or food waste (78%) with sawdust (22%). Each valid
recipe may be associated with a given operating mode with its specific unit variable operating cost.
All the waste processing or disposal facilities that appear in this model involve a capital cost to set them up, a fixed cost
for each year that they are kept open, and a variable operating cost that depends on the operating mode and the waste
throughput. The general form assumed for the total investment cost is a concave function of capacity, equivalent to an
economy of scale. Instead of assuming a single investment cost function for each technology, the model allows for two such
functions: one for investment in a new facility and the other for capacity expansion of an already existing plant. These
functions are approximated with concave piecewise linear functions. The number of linear pieces for each of the two cost
functions is under the user’s control.
Any existing or potential landfill site may be characterized by a list of materials that are allowed to be landfilled at the
site. It is thus possible to describe specialized landfills (such as those for fly ashes) or to use the model to study the impacts
of some regional or national policies, such as the banning of organics in all or some landfills. It is also possible to specify
landfilling taxes, varying with time, for each material that can be landfilled. This is another way to study the impacts of
regulatory policies.
Markets may be defined for some or all of the materials, energy, and collections of recyclables and/or organics. This
feature may be used to investigate ways to reduce the landfilling activities. Each market is characterized by a concave
revenue–quantity curve represented by a piecewise linear approximation where the number of pieces is again under the
user’s control.
Finally, because the model is a dynamic one with the number and the length of the time periods defined by the user,
almost all data are allowed to vary with time.
These are the main features of the model that was proposed to attack the problem defined at the beginning of this section.
They were described within a 0-1 mixed-integer programming model. Its general structure is thus the following:
minimize ∑ di Xi + ∑ e j Z j
i j

subject to
EUGENE: an optimization model for solid waste management planning 9

≤ 
∑ aki Xi + ∑ bkj Z j = c k , ∀k
i j ≥ 
Xi ≥ 0, ∀i and Z j ∈ {0,1}, ∀j

The coefficients di and ej for the objective function and aki, bkj and ck for the constraints are the known parameters. The
continuous variables Xi and binary variables Zj are the decision variables to be determined.
A slightly simplified version of the mathematical structure of EUGENE may be found in the Appendices. It will make the
above statements more precise. Other capabilities of the model are also described.

3 Implementation
EUGENE was implemented using the AMPL algebraic modelling language (Fourer et al., 1993) in a Windows NT/95
environment via the AMPL Plus application (Compass Modeling Solutions, Inc., 1995) and the CPLEX 4.0 optimizer
(CPLEX Optimization, Inc., 1995). The powerful capabilities of AMPL allows easy modification of the model and
experimentation of many alternate formulations in a short time. AMPL Plus facilitates the access of the model data in
ASCII files as well as many popular spreadsheet and database application formats. Results may also be saved in any of
these forms.
A first database was constructed for the City of Montréal, Canada. The city was segmented into 10 generating areas, and
15 processing/landfilling sites were provided for the installation of technologies. Three of these 15 sites are already
developed. Fifty-seven different materials, including 24 waste components, two types of energy, 11 recyclables/organics
collections and 18 combinations of collections were considered. The database contains descriptions of 20 technologies
(three for incineration, three for recycling, six for composting, one for MSW composting and seven for landfilling). Forty
operating modes were considered for the incinerators, 18 for the composting technologies, four for the MSW composting
technologies, two to six for the MRFs and one for each landfilling technology. A detailed description of the data (and even
more processing technologies) and the results for many scenarios can be found in Live (1997), Soucy (1997), Ratel (1998)
and Ben Abdallah (1997).
The base case scenario for two time periods (of five years each) produces, after a presolve phase, a model comprising
19 918 variables (including 770 binary variables), 5216 constraints and 149 706 non-zero coefficients. These size statistics
may seem impressive but the database is very detailed. Most of the models documented in the literature have much simpler
databases and produce significantly smaller problems. We believe that EUGENE would produce problems with sizes
comparable to those of the original models if we transform their databases into EUGENE databases.
The two time period problems associated with the various tested scenarios generated with the above data are usually
solved in two to three hours on a Pentium based microcomputer equipped with a 150 MHz processor. Some larger problems,
with four time periods, were also solved in approximately six hours but three sites, two collections and three combinations
were deactivated for these runs. Reductions in solution times can still be achieved with the use of integrality cuts and the
strengthening of some constraints (in particular, constraints 7–10 of the Appendix). Some work is in progress in this area
EUGENE: an optimization model for solid waste management planning 10

and will be reported later. Note the enormous progress accomplished during the last 15 years: all the waste management
models reported by Jenkins (1982) were computationally intractable when they had more than 20 binary variables.

4 Future developments
EUGENE is the result of an intense team effort that produced several improved and more sophisticated versions of it over a
two-year period. These improvements were dictated by the needs of its first users and some other improvements may occur
in the future. The most important among these are:
The addition of various social and environmental indicators, actually under study, for an eventual use in multicriteria
analyses.
The description of the transfer station options for transportation of collections and materials to remote locations (this
means that we should also allow the use of aggregated flows as possible outputs of processing/landfilling technologies:
a transfer station not only receives but also generates aggregated flows).
The blending of the various organic materials that can be used by the composting units. Actually, the model works with
recipes provided by the user. The introduction of blending constraints, very similar to those encountered in the oil
refinery models for the production by blending of the refined products (see Haverly Systems, Inc., 1982), allows the
implicit consideration of an infinite number of different recipes. This will increase the data needs of the model since the
characteristics of each organic material and the standards of the various grades of marketable compost will have to be
taken into account by the model.
An improved description of the MSW composting processes where some separation of the recyclable constituents of
the waste is done prior to, or after, the composting stage. It would be desirable to have this separation done
endogenously by the model. Actually, the user must define an operating mode for the waste collection associated to a
combination of recyclables/organics collections with explicit outputs for compost and each recyclable material. The
model should be able to determine itself what the outputs and their proportions are.
A better description of the incinerators where the energy content of each collection and the ash production of these
collections are endogenously computed by the model from the data of the individual components of the collections (as
in Carlson, 1986).
The possibility of importing aggregated and disaggregated flows from outside the system boundary (for instance
sawdust may be imported to improve the composting yields, waste collections from outside may be forced into the
system because of existing agreements between two different regions for waste treatment).
Some mechanisms to deal with uncertainty.
EUGENE: an optimization model for solid waste management planning 11

5 Conclusion
In this paper, a detailed mixed-integer programming model for integrated planning of regional solid waste management
systems has been described. It contains many innovative features and removes many limitations frequently encountered in
other existing optimization models for waste management. The model has been applied for the City of Montréal. Several
interesting studies were done using it. For instance, EUGENE has proven very useful to determine the impacts of various
recycling targets imposed on the system. It was also used to study the effects of various policies designed to reduce (through
landfilling taxes) or eliminate organics landfilling. By comparing the base case with a scenario based on higher participation
rates of households in the various recyclables/organics collections, it is possible to determine maximum annual sensitizing
budgets (in monetary units per household) for improvement of the participation rates.
Many future improvements have been proposed and most of them will be implemented. The model will eventually be
the heart of a decision-support system for the regional decision-makers. It was developed with the objective of allowing the
user to capture practically all aspects of the planning problem.

6 Appendix A: model constraints


This appendix contains a mathematical description of the model constraints. It allows the reader to gain a deeper
understanding of the model. It also gives enough details for someone who would like to implement a similar model. The
implemented objective functions are described in Appendix B.
The following is a list of some important sets that are used in the model.
Mat : The list of all materials included in the model
Energy : The set of energy flows included in the model
Src : The list of all waste generating sources
RO_col : The list of all recyclables and/or organics waste collections
Combination : The list of all non-empty valid combinations of recyclables and/or organics collections
Process : The list of all processing and landfilling technologies
Site : The list of all locations where a processing or landfilling technology may be installed
Markets : The list of all markets for materials, energy and recyclables/organics collections
Links : The list of all valid origin-destination pairs. It is a subset of
(Src ∞ Site) ≈ (Src ∞ Markets) ≈ (Site ∞ Site) ≈ (Site ∞ Markets).
Other sets will also be introduced and defined as they are needed. Whenever possible, the following convention has been
followed: all variables names are in capital letters, sets names in small letters with the first letter capitalized and parameters
names in small letters.
EUGENE: an optimization model for solid waste management planning 12

Finally, note that it is a slightly simplified description of the model. In particular, many sets and coefficients mentioned
below are defined or computed by the model from the information contained in the user’s database. Most of the constraints
involve summations over valid pairs of origins and destinations, i.e. over a subset of the set Links defined above. In order to
simplify the notations, it has been decided to omit this restriction. For ease of description, the remaining part of this
Appendix is separated into subsections dealing with homogeneous blocks of constraints, and objective functions.

Wastes, recyclables and organics collections


The following four types of constraint ensure that all generated wastes are collected from the available combinations of
collections and that exactly one combination is selected for each source at each time period.
∑ COMB[src,comb, t] + COMB[src,"empty",t ] = 1, ∀src ∈Src, ∀t ∈{1, ..., T } (1)
comb ∈Comb[src]

Q_ RC[src, comb,col,t ] = q[src,t ]× frac[src, comb, col,t ] × COMB[src, comb,t ]


∀src ∈Src, ∀comb ∈Comb[src], ∀col ∈Col[comb], ∀t ∈ {1,. .., T }

(2)
Q_ WC[src,comb, t] = q[src,t ] × COMB[src, comb,t ] – ∑ Q_ RC[src, comb,col, t]
col ∈Col[comb ]
(3)
∀src ∈Src, ∀comb ∈Comb[src], ∀t ∈ {1,... ,T }

Q_ WC[src,"empty",t ] = q[src, t ]× COMB[src,"empty",t ], ∀src ∈Src,∀t ∈{1, ..., T } (4)


EUGENE: an optimization model for solid waste management planning 13

where
T is the number of time periods;
Comb(src) ∏ Combination is the list of all valid combinations of collections at source
src Src;
COMB[src,comb,t] is a binary variable taking the value 1 if combination of collections comb is selected at source src during
time period t, and 0 otherwise;
COMB[src,"empty",t] is a binary variable taking the value 1 if the "empty” combination (i.e. if no recyclables and organics
collection) is selected at source src during time period t, and 0 otherwise;
Q_RC[src,comb,col,t] ε 0 is annual quantity (in tonnes) of recyclables/organics materials collected by collection col of
combination comb at source src during time period t;
q[src,t] is the total annual quantity (in tonnes) of wastes generated by source src in time period t;
frac[src,comb,col,t] is the weight fraction of the wastes generated by source src that is collected by collection mode col of
combination comb at time period t (it is a computed parameter);
Col[comb] is the list of all recyclables and/or organics collections associated to combination comb;
Q_WC[src,comb,t] ε 0 is the annual quantity (in tonnes) of wastes collected at source src in time period t after the
recyclables and/or organics of combination comb have been collected;
Q_WC[src,"empty",t] ε 0 is the annual quantity (in tonnes) of wastes collected at source src in time period t when there is no
recyclables and/or organics collection.
Constraints (1) ensure that exactly one valid combination of collections is chosen for each waste generating source in each
time period. Relations (2) compute the amount (in tonnes) of recyclables and/or organics that are collected by each
collection of the selected combination at each location and time period. Relations (3) and (4) compute the amount of wastes
to be collected after the recyclables and/or organics were collected.
Note that the total waste q[src,t] generated at each source and time period is exogenously defined. It is given to the
model. An interesting option to consider would be to introduce price sensitivity of total waste generation in order to test
waste reduction policies. This would however significantly increase the complexity of the problem because many additional
binary variables would be required.

Allocation of wastes, recyclables and organics collections to sites


The following constraints ensure that the wastes, recyclables and organics collected at each source are transported to sites
for processing or landfilling, or to the markets.
∑ Q_ RC[src,comb, col,t ] = ∑ RC_ SF[src, col, dest, t] +
comb ∈Comb _sc[src,col ] dest ∈Sfl
EUGENE: an optimization model for solid waste management planning 14

∑ RC _ BULK[src, col, dest,t ] + ∑ RC _ MRK[col, src, mar, t] (5)


dest ∈Site mar ∈Market[col ]

∀src ∈Src, ∀col ∈Roc[src], ∀t ∈ {1,... ,T }

Q_ WC[src,comb, t] =
∑ WC_ PRC[ src, comb , dest, t ] + ∑ WC_ LF [src , comb, dest , t ] (6)
dest ∈Site dest ∈Lf _ c [ src , comb, t]

∀src ∈Src, ∀comb ∈Comb[src] ∪ {"empty"}, ∀t ∈ {1,. .., T }

∑ ∑ RC_ SF[src,col, site,t ] ≤ M × ∑ EXIST_ P[site, sf , t ]


src∈Src col∈Roc[src] sf ∈Sf [site]
(7)
∀site ∈Sfl, ∀t ∈{1, ..., T }

∑ ∑ WC_ LF[src,comb, site, t] ≤ M × ∑ EXIST _ P[site, lf , t]


src∈Src comb ∈ lf ∈Lf [site]
Comb[src]∪ {"empty "}:
site∈Lf _ c[src,comb ,t ]

(8)
∀site ∈ Lfl, ∀t ∈ {1,. .., T }

∑ ∑ RC_ BULK[src,col, dest, t ]


src∈Src col∈Roc[src]

 WC_ PRC[src, comb,site, t] + 


+ ∑ ∑ 
δ ×

src∈Src comb∈  LF [src,comb, site,t ] WC_ LF[src, comb, site, t ]
Comb[src]∪{"empty"}

 
≤ M × 1 – ∑ EXIST_ P[site, sf , t]
 sf ∈Sf [ site] 
(9)
∀site ∈Sfl, ∀t ∈{1, ..., T }

∑ ∑ WC _ PRC[src, comb, site, t]


src∈Src comb ∈
Comb[src]∪ {"empty "}

 RC _ BULK[src, col, site, t] + 


+ ∑ ∑  
src ∈Src col ∈Roc[ src ]  δ SF [site] × RC _ SF[src, col, site, t]
EUGENE: an optimization model for solid waste management planning 15

 
≤ M × 1 – ∑ EXIST_ P[site,lf ,t ]
 lf ∈Lf [ site ] 
(10)
∀site ∈ Lfl, ∀t ∈ {1,. .., T }

where
Comb_sc[src,col] ∏ Combination is the list of all valid combinations at source src Src defined using collection col
RO_col;
Sfl ∏ Site is the list of locations where a separation facility can be installed;
RC_SF[src,col,des,t] ε 0 is the annual quantity (in tonnes) of recyclables and/or organics collected at source src by
collection col and transported to location dest in time period t for separation;
RC_BULK[src,col,dest,t] ε 0 is the annual quantity (in tonnes) of recyclables and/or organics collected at source src by
collection col and transported for bulk processing to location dest in time period t;
RC_MRK[col,src,mar,t] ε 0 is the annual quantity (in tonnes) of recyclables and/or organics collected at source src by
collection col and transported to market mar in time period t;
Market[col] ∏ Markets is the list of available markets for the recyclables and/organics collection col;
Roc[src] ∏ RO_col is the list of recyclables and/or organics collections available at source src;
WC_PRC[src,comb,dest,t] ε 0 is the annual quantity (in tonnes) of waste collected by combination comb at source src and
transported for processing to location dest in time period t;
WC_LF[src,comb,dest,t] ε 0 is the annual quantity (in tonnes) of waste collected by combination comb at source src and
transported for landfilling at location dest in time period t;
Lf_c[src,comb,t] ∏ Site is the list of locations where the collected waste of source src by combination comb may be
landfilled at time period t;
δLF[src,comb,site,t] equals 1 if site Lf_c[src,comb,t], and 0 otherwise;
EXIST_P[site,prc,t] is a binary variable taking the value 1 if processing or landfilling technology prc exists at location site
at time period t, and 0 otherwise;
Sf[site] is the set of the allowed separation technologies at location site;
δSF[site] equals 1 if site Sfl, and 0 otherwise;
M is a large constant;
EUGENE: an optimization model for solid waste management planning 16

Lf[site] ∏ Process is the set of valid landfilling technologies allowed for location site;
Lfl ∏ Site is the set of locations suitable for landfilling.
Constraints (5) assign recyclables and/or organics collections of each source to locations and markets. Those collections
may be separated by a material recovery facility or bulk-processed (composted, incinerated, etc.). Since separation facilities
have a particular representation in the model, and since a site may a priori contain any processing or landfilling technology,
it is necessary to generate two sets of flow variables for collections from sources to sites (the RC_SF and the RC_BULK
variables). Any of the processing technologies other than the separation facilities is allowed to receive any bulk process
recyclables/organics collections.
Constraints (6) assign waste collections to sites. They may be bulk-processed or landfilled. Again, since landfilling
technologies have a particular representation in the model, it is necessary to consider two types of MSW flow variable.
We will see later (constraints (30)) that only one type of processing or landfilling technology may be installed at a given
location. This means that it is not possible to simultaneously have positive values for the RC_SF and RC_BULK variables of
a given location or positive values for the WC_PRC and WC_LF variables. Constraints (7)-(10) enforce this.
Relation (7) forces the RC_SF variables of a site suitable for a separation facility to take the zero value if no separation
facility exists at that site. Relation (8) ensures that the landfilling variables WC_LF of a given location take the zero value
when the location is not used for landfilling. Relation (9) forces the nullity of the RC_BULK, WC_PRC and WC_LF
variables when a separation facility exists at a given site. Finally, relation (10) says that the existence of a landfilling
technology at a given location forbids the positiveness of the WC_PRC, RC_BULK, and RC_SF variables associated with
this location.

Processing of collections and materials


The following constraints relate the activity variables of the processing technologies to the incoming flows of materials and
collections.
∑ ∑ input[mat , prc, mode, t ]× ACT[prc, mode, site, t]
prc ∈Prc [site] mode∈Mode[ prc]:
mat∈Inp[ prc, mode]

= ∑ ∑ FLOW _ II[trn, mat, orig, site,t ]


trn∈Trn _ Int: orig∈Site
mat∈Transp _Mat [trn ]

(11)
∀mat ∈Mat, ∀site ∈Site,∀t ∈{1,..., T }

∑ ∑ input[comb, prc, mode,t ] × ACT[prc, mode, site, t]


prc ∈Prc [site] mode∈Mode[ prc]:
comb ∈Inp[ prc,mode ]
EUGENE: an optimization model for solid waste management planning 17

= ∑ WC _ PRC[src,comb, site,t ] (12)


src∈Src:
comb∈Comb[src ]

{
∀comb ∈Combination ∪ "empty"},∀site ∈Site, ∀t ∈{1, ..., T }

∑ ∑ input[col, prc, mode, t] × ACT[ prc, mode, site,t ]


prc ∈Prc [site] mode∈Mode[ prc]:
c ol∈Inp[prc,mode ]

= ∑ ( RC_ BULK[src, col, site, t ]+ δ SF [site] × RC _ SF[src, col, site,t ]) (13)


src∈Src:
col∈Col [src ]

∀col ∈RO_ col,∀site ∈Site, ∀t ∈{1, ..., T }

where
Prc[site] ∏ Process is the set of technologies that can be installed at location site;
Mode[prc] is the list of operating modes of technology prc;
Inp[prc,mode] is the set of inputs to technology prc under operating mode mode;
input[inp,prc,mode,t] is the quantity of input inp per unit of activity of technology prc under operating mode mode in time
period t;
ACT[prc,mode,site,t] ε 0 is the activity level of technology prc of location site operating under operating mode mode during
time period t;
Trn_Int is the list of transportation means that can be used to transport the outputs of the technologies located on each
location of set Site;
Transp_Mat[trn] is the list of materials of set Mat that can be transported by transportation mean trn;
FLOW_II[trn,mat,orig,site,t] ε 0 is the annual flow (in tonnes) of materials mat from location orig to location site
transported by transportation mean trn during time period t.
The model also allows the user to define groups of materials and/or collections that can be used as inputs to the processing
technologies. When two or more materials or collections belong to a same group, they are considered as identical by the
technologies using it.

Outputs of processing technologies


The following constraints compute the material and energy outputs of the processing technologies.
EUGENE: an optimization model for solid waste management planning 18

 
 ∑ FLOW_ II[trn, mat, site, dest, t] 
 dest∈Site 
∑ + ∑ FLOW_ IL[trn, mat, site,lf ,t ] 
trn ∈Trn _ Int:  lf ∈Lf _ m[mat,t ] 
mat ∈Transp _Mat[trn]  
+ ∑ FLOW _ IM[trn, mat , site, mar,t ]
 mar ∈Market [mat] 
= ∑ ∑ output[mat , prc, mode, t ]× ACT[ prc,mode, site,t ]
prc∈Prc[site] mode ∈Mode[ prc]:
mat∈Out[ prc,mode ]

 fr_ col[src,col, garb,t ] × fr _sep[garb, mat ,t ]


+ δ SF [site] × ∑ ∑  
garb∈Garb[mat] src∈Src:  × RC_ SF[src,col, site,t ] 
col∈Roc [src]

∀mat ∈Mat, ∀site ∈Site,∀t ∈{1,..., T } (14)

where
Lf_m[mat,t] is the set of locations where material mat may be landfilled at time period t if they are developed as landfills;
FLOW_IL[trn,mat,site,lf,t] ε 0 is the annual flow (in tonnes) of material mat transported by transportation mode trn from
location site to landfilling location lf during time period t;
FLOW_IM[trn,mat,site,mar,t] ε 0 is the annual flow (in tonnes) of material mat transported by transportation mode trn from
location site to market mar during time period t;
Out[prc,mode] is the set of outputs of technology prc under operating mode mode;
output[out,prc,mode,t] is the quantity of output out per unit of activity of technology prc under operating mode mode in
time period t;
Garb[mat] is the set of (collected) garbages that generate material mat in a separation facility;
fr_col[src,col,garb,t] is the weight fraction of garbage garb in the recyclables/organics collection col of generating source
src in time period t (it is a computed value obtained from many basic data supplied by the user);
fr_sep[garb,mat,t] is the weight fraction of waste component garb that becomes material mat at the gate of a separation
facility in time period t.
The right-hand side of balance constraint (14) is the total quantity of each material obtained at the gate of processing
technologies located at each site. It is equated to the flow variables leaving the site for further processing at other sites, for
landfilling, or for sales on the markets. Usually, the separation technologies (i.e. those receiving and processing the flows of
recyclables/organics collections associated to the RC_SF variables) do not have inputs and outputs because the model
assumes that the components of the collections referred to by these variables must be separated.
EUGENE: an optimization model for solid waste management planning 19

Note also that constraint (14) is a simplified version of the real constraint. It is possible for each material of a
recyclables/organics collection to be ‘separated’ in other materials. For instance, the user may inform the model that the
paper part of a given collection will, after the separation, be replaced by various types of paper and some contaminated
paper. This is a way to describe losses and a finer separation process.
The following constraint is a simplification of constraint (14). It is related to energy production and it is assumed that
energy can only be sold on energy markets.
∑ FLOW _ NRG_ M[ener ,site, mar,t ]
mar ∈Market [ener]

≤ ∑ ∑ output[ener , prc, mode, t ]× ACT[ prc, mode, site, t] (15)


prc∈Prc [site] mode∈Mode[prc]:
ener ∈Out[prc,mode ]

∀ener ∈Energy, ∀site ∈Site,∀t ∈{1,..., T }

where
FLOW_NRG_M[ener,site,mar,t] ε 0 is the annual flow (in the appropriate energy units) of energy carrier ener
generated at location site and sold at market mar during time period t.

Landfilling
The landfilling activities at each landfilling location are computed by the following constraints.
Q_ LF[lf , mat, t] = ∑ ∑ fr[src, comb,mat , t] × WC _ LF[src, comb,lf ,t ]
src∈Src comb ∈
Comb[src]∪ {"empty"}

+ ∑ ∑ FLOW _ IL[trn, mat, site, lf ,t ] (16)


trn ∈Trn _ Int: site∈Site
mat∈Transp _Mat[trn ]

∀mat ∈Mat, ∀t ∈ {1,... ,T }, ∀lf ∈ Lf _ m[mat,t ]

∑ ∑ ACT[ prc, mode, lf ,t ] =


prc ∈Lf [lf ] mode ∈Mode[prc]

∑ weight_ to_ vol[mat ] × Q_ LF[lf , mat, t] (17)


mat∈Mat:
lf ∈Lf _ m[mat,t ]

∀lf ∈Lfl,∀t ∈{1, ..., T }

where
EUGENE: an optimization model for solid waste management planning 20

Q_LF[lf,mat,t] ε 0 is the annual quantity (in tonnes) of material mat landfilled at location lf in time period t;
fr[src,comb,mat,t] is the weight fraction of material mat in the solid waste collection associated to combination comb of
generating source src in time period t (it is a computed value obtained from many basic data supplied by the user);
weight_to_vol[mat] is the conversion factor in volume units of material mat. It is expressed in cubic metres per tonne.
Constraints (16) compute the landfilled quantities of each material at each location during each time period. Constraints (17)
compute the total landfilling activity of each location at each time period. Note that the landfilling technologies have, unlike
the other technologies, activity variables expressed in volume units (cubic metres).

Capacity constraints
The activity variables of the collection and transportation modes, the processing technologies and the landfilling
technologies are linked to their capacity variables through the following constraints.
∑ ACT[prc, mode, site, t] ≤ max_ af [ prc,t ] × CAP _ P[prc, site,t ]
mode ∈Mode[prc]
(18)
∀site ∈Site, ∀prc ∈Prc[site] \ Lf [site], ∀t ∈ {1,.. .,T }
ACT[ prc, op_ mode, site,t ] ≤
max _ uf _ mode[ prc, op_ mode, t ]× ∑ ACT[ prc, mode, site,t ] (19)
mode ∈Mode[ prc]

∀site ∈Site, ∀prc ∈Prc[site], ∀op_mode ∈ Mode[ prc], ∀t ∈{1, ..., T }:


max _ uf _ mode[ prc, op_ mode,t ] < 1
t
nypp × ∑ ∑ ACT[lft,mode, lf , s] ≤ CAP _ P[lft, lf , t]
s =1 mode ∈Mode[lft ] (20)
∀lf ∈Lfl,∀lft ∈ Lf[lf ], ∀t ∈{1, ..., T }

 ∑ WC _ PRC[src, comb, dest, t] 


 dest ∈Site 
∑ ∑ + ∑ WC _ LF[src, comb,dest, t]
src∈Src comb ∈  dest∈ 
Comb[src]∪ {"empty "} Lf _ c[src,comb ,t ] 
≤ CAP _T ["gb",t ] (21)
∀t ∈{1, ..., T }
EUGENE: an optimization model for solid waste management planning 21

∑ ∑ Q_ RC[src, comb,col,t ] ≤ CAP _T [col,t ] (22)


src∈Src comb∈
Comb_ sc[src,col]

∀col ∈RO_ col,∀t ∈{1, ..., T }

 
 ∑ FLOW_ II[trn, mat, orig, dest, t] 
 dest∈Site 
∑ ∑ +  ∑ FLOW_ IL[trn, mat, orig,lf , t ]  ≤ CAP_ T[trn,t ] (23)
mat ∈Transp _Mat[trn] orig ∈Site  lf ∈Lf _ m[mat,t ] 
 
+ ∑ FLOW _ IM[trn, mat, orig, mar,t ]
 mar∈Market [mat] 
∀trn ∈Trn_ Int, ∀t ∈ {1,... ,T }

where
CAP_P[prc,site,t] ε 0 is the annual capacity (in tonnes) of processing technology or the cumulative capacity (in cubic
metres) of landfilling technology prc at location site in time period t;
max_af[prc,t] is the maximum annual availability factor of processing technology prc during time period t;
max_uf_mode[prc,op_mode,t] is the maximum annual utilization factor of technology prc under operating mode op_mode
during time period t;
nypp is the number of years per time period;
CAP_T["gb",t] ε 0 is the annual capacity of waste collections ("gb" stands for plastic green bags) in time period t;
CAP_T[xxx,t] ε 0 is the annual capacity of collection and transportation (from the processing or landfilling locations) xxx
during time period t.
Constraints (18) ensure that total activity of each processing technology at each location does not exceed its existing
capacity. The capacity of a given processing technology is affected by its annual availability factor. Similar constraints also
exist for minimum utilization factors. Constraints (19) control the activity variables for specific operating modes through
upper bounds on their shares. Note that similar constraints also exist for minimum operation of technologies for some
operating modes. Activities of landfilling technologies are controlled by constraints (20). Note the cumulative nature of
these constraints and of the capacity variable. Constraints (21) control the waste collection capacities. The summation of the
quantities of wastes left after the recyclables/organics collections of each combination, including the "empty" combination
should not exceed the annual waste collection capacity at each time period. Constraints (22) ensure that the aggregated
flows of recyclables and organics collections are bounded by their collection capacities. Finally, constraints (23) relate the
flows of materials produced at each processing/landfilling location to the capacity variables of their transportation modes.
EUGENE: an optimization model for solid waste management planning 22

Investment in processing or landfilling capacity


The following constraints relate the investment variables of processing/landfilling technologies to their capacity variables
and to their binary decision variables. These binary variables are used to determine if an investment is for a new plant or for
capacity expansion of an existing facility. Remember that these two types of investment may have different concave
investment cost functions. Finally, there is a constraint that bounds to unity the number of different technologies at each
location.
CAP _ P[ prc, site,t ] =
 n _seg_ new[ prc] 
t
 ∑ INV_ P_ N[prc, site,seg, ∆] 
seg=1
resid _ prc[ prc,site, t] + ∑   (24)
s =u[prc,t ] n_seg_ add[prc] 
+ ∑ INV_ P_ A[ prc,site, seg, ∆]
 seg=1 

∀site ∈Site, ∀prc ∈Prc[site],∀t ∈{1,..., T }

where
1 if prc ∈Lf [site]
u[ prc, t] = 
max {t – life[ prc]+ 1,1}otherwise
INV_ P_ N[prc, site,seg,t ] ≤ M × D_ NEW_ P[ prc, site, seg, t]
(25)
∀site ∈Site, ∀prc ∈Prc[site], ∀seg ∈ {1,. .., n_ seg_ new[ prc]},∀t ∈ {1,... ,T }

INV_ P_ A[ prc,site, seg,t ] ≤ M × D_ ADD_ P[ prc,site, seg,t ]


(26)
∀site ∈Site, ∀prc ∈Prc[site], ∀seg ∈ {1,. .., n_ seg_ add[ prc]}, ∀t ∈ {1,. .., T }

cap_ min[ prc] × EXIST _ P[site, prc,t ] ≤ CAP_ P[prc, site, t ]


(27)
≤ cap_ max[ prc] × EXIST_ P[site, prc, t]

∀site ∈Site, ∀prc ∈Prc[site], ∀t ∈ {1,.. .,T }


n _seg_new [ prc ]
∑ D_ NEW_ P[ prc, site, seg, t ]+ EXIST _ P[site, prc, t –1] ≤ 1 (28)
seg=1

∀site ∈Site, ∀prc ∈Prc[site], ∀t ∈ {1,.. ., T }


n _seg_ add[prc]
∑ D_ ADD_ P[prc, site,seg,t ] – EXIST_ P[site, prc, t – 1] ≤ 0 (29)
seg=1
EUGENE: an optimization model for solid waste management planning 23

∀site ∈Site, ∀prc ∈Prc[site], ∀t ∈ {1,.. ., T }

∑ EXIST_ P[site, prc, t] ≤ 1 ∀site ∈Site, ∀t ∈{1, ..., T } (30)


prc ∈Prc [site]

where
resid_prc[prc,site,t] is the residual installed capacity of technology prc at location site built prior to the time horizon and
available in time period t. The investment cost of this capacity is considered to be a sunk cost expended before the first time
period;
n_seg_new[prc] is the number of straight lines used to describe the concave investment cost function for a processing or
landfilling technology prc at an undeveloped location;
n_seg_add[prc] is the number of straight lines used to describe the concave investment cost function for a capacity addition
to an already existing processing or landfilling technology prc;
INV_P_N[prc,site,seg,t] ε 0 is a variable whose value is the capacity of a new plant or landfilling technology prc installed at
location site at the beginning of time period t using linear investment cost function seg;
INV_P_A[prc,site,seg,t] ε 0 is a variable whose value is the capacity added to an already existing plant or landfilling
technology prc at location site at the beginning of time period t using linear investment cost function seg;
life[prc] is the lifetime of technology prc, i.e. the number of time periods for which new capacity may be utilized;
D_NEW_P[prc,site,seg,t] is a binary decision variable taking the value 1 if and only if a new plant or landfill of type prc is
installed at location site using the linear investment cost function seg at the beginning of time period t;
D_ADD_P[prc,site,seg,t] is a binary decision variable taking the value 1 if and only if there is a capacity expansion to a
processing plant or to a landfill of type prc at location site using the linear investment cost function seg at the beginning of
time period t;
cap_min[prc] > 0 is the minimum capacity of processing or landfilling technology prc when it exists;
cap_max[prc] is the maximum capacity of processing or landfilling technology prc when it exists.
Constraints (24) say that the installed capacity of a technology at a location in a given time period is equal to its residual
capacity plus the investment made in the current time period and previous time periods within its lifetime. Landfilling
technologies are assumed to have an infinite or, at least, a very long (of T or more time periods) lifetime. Investments are
either for new capacity (variables INV_P_N) or expansion (variables INV_P_A) of existing capacity.
Constraints (25) and (26) relate the investment variables to the binary decision variables D_NEW_P and D_ADD_P
because new capacity and capacity expansion have different investment cost functions. Constraints (27) ensure that the
existing capacity of an installed technology is within a valid range determined by technical feasibility considerations.
Constraints (28) force the decision variables D_NEW_P for new capacity of a technology to take the zero value when
there is already some capacity of this technology at a location. Otherwise these decision variables are free. Constraints (29)
EUGENE: an optimization model for solid waste management planning 24

force the decision variables D_ADD_P associated to capacity expansion of a technology to be fixed at zero when there is no
existing capacity of this technology at a location. Otherwise they are free.
Finally, constraints (30) imply that no more than one processing or landfilling technology may exist at each location
during each time period. This does not forbid, as the preceding constraints show, future investment at a given location if
there already exists some capacity of a given technology at this location. Capacity expansions are allowed for this
technology.

Land-use constraints
Each processing or landfilling technology has some land-use requirement, and each location suitable for such a technology
is characterized by its area (and by a unit land cost). The following constraints compute the land requirements at each
location. These land requirements are then bounded by the total area of the sites. The land requirement for the installation of
a new plant is described by the following linear relationship
Area = a + b × Capacity
where constants a and b vary with each technology. In the case of a capacity addition to an already existing plant, the model
omits constant a.
USED_ AREA[site, t] = USED_ AREA[site, t – 1] +
 n_seg_ new[ prc]  fix_ area[ prc] × D_ NEW _ P[prc, site,seg,t ]  
 ∑  
 seg=1  +var _ area[prc] × INV _ P_ N[ prc, site, seg, t] 
∑ (31)
prc ∈Prc [site] 
n_seg _add[ prc]
+ ∑ var _ area[ prc] × INV _ P_ A[ prc, site, seg, t] 
 seg=1 
∀site ∈Site, ∀t ∈{1, ..., T }

USED_ AREA[site, t] ≤ area[site], ∀site ∈Site, ∀t ∈{1, ..., T } (32)

where
USED_AREA[site,t] ε 0 is a variable indicating the area of location site occupied by a processing or landfilling technology
during time period t. If t = 0, this is the used area at the end of time period 0, i.e. at the beginning of time period 1, before
the model can decide if it will add some capacity or not;
fix_area[prc] is the fixed area for a new plant of processing/landfilling technology prc. This is the minimum area occupied
by such a technology;
var_area[prc] is the land requirement per unit of capacity of technology prc;
area[site] is the total area of location site.
EUGENE: an optimization model for solid waste management planning 25

Landfilling diversion
It is possible to set some minimum landfill diversion targets that will generate constraints limiting the quantities of materials
landfilled during each time period. They are generated when the user makes use of some optional parameters to study the
impacts of diversion policies.
∑ ∑ Q_ LF[lf , mat ,t ] ≤ ∑ q[src,t ] – min _ div[t ] (33)
mat ∈Mat lf ∈Lf _m[mat, t] src∈Src

∀t ∈{1, ..., T }

∑ ∑ Q_ LF[lf , mat ,t ] ≤ (1 – div_rate[t] ) × ∑ q[src,t ] (34)


mat ∈Mat lf ∈Lf _m[mat, t] src∈Src

∀t ∈{1, ..., T }

where
min_div[t] ε 0 is the minimum annual quantity (in tonnes) of materials diverted from landfills in time period t;
div_rate[t] [0,1] is the minimum diversion rate of materials from the landfills in time period t.
Constraints (33) are generated only when min_div[t] is strictly positive. They constrain the annual landfilled tonnage during
time period t to be lower than or equal to the total tonnage of wastes generated minus min_div[t].
Constraints (34) are only generated for those t for which div_rate[t] is non-zero. These relations constrain the annual
landfill tonnage to be lower than or equal to total waste generation times a constant smaller than 1.

Market constraints
Materials, recyclables/organics collections and energy may be sold on markets. Markets are described using piecewise
revenue functions with as many straight lines as the user needs. Each product may have one or more markets. Each valid
pair product-market is described using one or several intervals characterized by a specific price and an upper bound on the
quantity of product that can be sold on that market segment. The last interval has no finite upper bound.
∑ ∑ FLOW_ IM[trn, mat, site, mar, t] = DEMAND[mat, mar, t ]
trn ∈Trn _ Int: site∈Site
mat ∈Transp _Mat[trn]

(35)
∀mat ∈Mat, ∀mar ∈ Market[mat], ∀t ∈ {1,. .., T }

∑ RC _ MRK[col, src, mar, t ] = DEMAND[col, mar, t] (36)


src∈Src:
col∈Roc[src]
EUGENE: an optimization model for solid waste management planning 26

∀col ∈RO_ col,∀mar ∈Market[col],∀t ∈{1,... ,T }

∑ FLOW _ NRG_ M[ener, site, mar,t ] = DEMAND[ener, mar, t] (37)


site ∈Site

∀ener ∈Energy, ∀mar ∈ Market[ener ],∀t ∈{1,..., T }


n _seg _mrk [prod , mar ]
DEMAND[ prod, mar,t ] = ∑ SALES[ prod, mar,seg,t ] (38)
seg=1

∀prod ∈Mat ∪ RO_col ∪ Energy, ∀mar ∈Market[ prod ],∀t ∈{1,..., T }


SALES[ prod, mar, seg, t] ≤ max_ sales[ prod, mar,seg,t ]
∀prod ∈Mat ∪ RO_col ∪ Energy, ∀mar ∈Market[ prod ], (39)
{
∀seg ∈ 1,..., n_ seg_ mrk[ prod , mar] – 1}, ∀t ∈{1, ..., T }

min _ demand[ prod, mar, t ] ≤ DEMAND[prod , mar, t] ≤ max_ demand[ prod, mar,t ]
(40)
∀prod ∈Mat ∪ RO_col ∪ Energy, ∀mar ∈Market[ prod ],∀t ∈{1,..., T }

where
DEMAND[prod,mar,t] ε 0 is the annual demand of product prod by market mar during time period t;
SALES[prod,mar,seg,t] ε 0 is the annual quantity of product prod sold on segment seg of market mar during time period t;
max_sales[prod,mar,seg,t] is the maximum annual quantity of product prod that can be sold on segment seg of market mar
during time period t;
min_demand[prod,mar,t] is the minimum annual demand for product prod by market mar during time period t;
max_demand[prod,mar,t] is the maximum annual demand for product prod by market mar during time period t.
Relations (35)-(37) are market equilibrium constraints for materials, collections and energy. They ensure that supply is
equal to demand. Constraints (38) relate the demand variables to the market structure defined in terms of market segments,
as discussed above. They allow to have price sensitive demands since each market segment is characterized by a specific
price and an upper bound. Relations (39) are upper bounds on the SALES variables., i.e. the absorption capacity of each
market segment. Finally, relations (40) are lower and upper bounds on the demand variables. They can be used to test some
market constraints (like signed agreements and scenarios on the maximum size of the markets).

Miscellaneous constraints
Some parameters allow the user to define minimum and maximum installed capacity of a technology or a group of
EUGENE: an optimization model for solid waste management planning 27

technologies on a specific location or on all locations at some or all time periods. Other parameters may also be used to
control the minimum and maximum number of installed units of a group of technologies at each time period. They will not
be described here.

Appendix B: objective functions


The EUGENE model actually has two possible objective functions that can be chosen by the user. The first one is the total
discounted system cost, and the second one is the cumulative landfilling.

Objective function I: total discounted system cost (TDSC)


This objective function represents the sum of all costs in the system minus the sum of all revenues from the markets sales
discounted for all time periods. It can be summarized as follows:
T  df _ ac[t ] × ( ANNUAL_COST [t ] − ANNUAL_ SALES[t ]
TDSC = ∑   (41)
t =1  + df _ ic[t ] × (INVEST [t ] − SALVAGE [t ]) 
with
nypp
df _ac[t ] = ∑ (1+ disc )1–y (1 + disc )nypp ×( t –1)
y =1

df _ic[t] = 1 (1 + disc )nypp×( t –1)


and where
disc is the annual discount rate;
df_ac[t] is the discounting factor for the annual flows of time period t (the numerator discounts the annual flows of the
period back to the beginning of the time period and the denominator discounts the cost from the beginning of the time
period to the beginning of the first time period);
ANNUAL_COST[t] is a variable equal to the sum of all annual costs occurring during time period t;
ANNUAL_SALES[t] is a variable equal to the sum of all annual revenues from sales to the markets costs occurring during
time period t;
df_ic[t] is the discounting factor for the investment costs of time period t. Those costs are assumed to be made at the
beginning of the time period;
INVEST[t] is a variable equal to the sum of investment costs of time period t;
EUGENE: an optimization model for solid waste management planning 28

SALVAGE[t] is a variable whose value is the correction to the end-effect problem which occurs in time-phased
mathematical programming models. The correction reduces the cost for investment in a technology when part of its
technical lifetime will extend beyond the last time period. The relation defining these variables will not be described in this
paper. Its description can be found in Fishbone and Abilock (1981).
EUGENE: an optimization model for solid waste management planning 29

The relations defining the precedent variables, except the salvage variables, are given below.
ANNUAL_COST[t]=
∑ ∑ q[src, t] × unit_ col_ cost[comb, t ] × COMB[src, comb, t ]
src∈Src comb ∈
Comb[src]∪ {"empty "}

 
 ∑ dist[src, dest]× RC_ BULK[src,col, dest, t] 
 dest∈Site 
+ ∑ ∑ varom_ transp[col, t] ×  + ∑ dist[src, dest]× RC _ SF[src, col,dest, t]  + varom_ transp["gb", t]
src∈Src col∈Roc[src]  dest∈Sfl 
 
+ ∑ dist[src, mar] × RC _ MRK[col,src, mar, t]
 mar∈Market[col ] 
 ∑ dist[src, dest] × WC_ PRC[src, comb, dest, t] 
 dest∈Site 
× ∑ ∑ 
src∈Src comb∈ + ∑ dist[src, dest] × WC _ LF[src, comb, dest, t ]
Comb[src]∪{"empty"}  dest ∈Lf _c [src,comb,t ] 
 unit _ trn_ cost[trn, t] × CAP_ T[trn, t ]+ varom_ transp[trn, t ]× 
 
  
  
  ∑ dist[orig, dest] × ∑ FLOW_ II[trn, mat, orig,dest, t] 
dest ∈Site mat∈Transp _Mat[trn ]
+ ∑   
trn∈Trn _ Int ∑  + ∑ dist[orig,lf ] × ∑ FLOW _ IL[trn, mat,orig, lf ,t ] 
 orig∈Site  lf ∈Lfl mat∈Transp_ Mat [trn]:
lf ∈Lf _m [mat,t ]

  
 + ∑ ∑ dist[orig, mar]× FLOW _ IM[trn, mat,orig, mar,t ]
  mat∈Transp _ Mat[trn ] mar ∈Market [mat] 
 fixom[ prc, t ]× CAP_ P[prc, site, t ] 
+ ∑ ∑  + ∑ varom[ prc, op_ mode, t] × ACT[ prc, op_ mode, site, t]
site∈Site prc∈Prc[site]
 op _mode ∈Mode[ prc] 
+ ∑ lf _ tax[mat ,t ] × ∑ Q_ LF[lf , mat, t]
mat∈Mat lf ∈Lf _ m[mat,t ]

where
unit_col_cost[comb,t] is the weighted unit cost (per tonne) of the waste and recyclables/organics collections belonging to
combination comb during time period t;
varom_transp[xxx,t] is the unit cost of transportation (in monetary units per tonne-km) of waste collection (xxx = "gb"),
recyclables/organics collections (xxx RO_col) and materials obtained at the gate of the processing technologies (xxx
EUGENE: an optimization model for solid waste management planning 30

Trn_Int) during time period t;


dist[orig,dest] is the distance between locations orig and dest;
unit_trn_cost[trn,t] is the unit cost (per tonne) for collecting in time period t, using the transportation mode trn, the
materials produced at each location where a processing/landfilling technology may be located;
fixom[prc,t] is the annual fixed operating and maintenance cost associated with the installed capacity of technology prc in
time period t and charged regardless of the plant utilisation (this cost may include charges such as insurance’s and property
taxes);
varom[prc,op_mode,t] is the annual variable operating and maintenance cost associated to operating mode op_mode of
technology prc during time period t;
lf_tax[mat,t] is the landfilling tax (in monetary units per tonne) of material mat during time period t.
ANNUAL_ SALES[t ] =
n_ seg_mrk [prod , mar]
∑ ∑ ∑ ( price[ prod, mar, seg,t ] × SALES[ prod, mar,seg,t ])
prod ∈ mar ∈Market [prod ] seg=1
Mat ∪RO_col ∪Energy

where
price[prod,mar,seg,t] is the unit price of product prod in segment seg of market mar during time period t.
INVEST[t] =
∑ ∑ sens_ cost[src, comb,t ] × COMB[src, comb,t ]
src∈Src comb ∈Comb[src]

  fixinv_ new[ prc, seg, t ]   


   × D_ NEW _ P[prc, site,seg,t ]  
n_ seg_new[ prc]  + fix _ area[prc] × land _ cost[site]  
 ∑   
 seg=1 +  varinv_ new[ prc, seg, t ]   
+ ∑    +var _ area[prc] × land _ cost[site] × INV _ P_ N[ prc, site, seg, t] 

site∈Site prc∈Prc[Site]  
  fixinv_ add[ prc, seg,t ] × D_ ADD_ P[ prc, site, seg, t] 
 n_ seg_add[ prc]  varinv_ add[ prc, seg,t ] 
+ ∑  
+

×

seg =1 
 +var_ area[ prc] × land _ cost[site]  INV_ P_ A[prc, site, seg,t ]
    

where
sens_cost[src,comb,t] is the sensitizing cost associated with the establishment or continuation of the combination of
collections comb at source src during time period t. This cost is charged regardless of the existence of the combination
EUGENE: an optimization model for solid waste management planning 31

during the preceding time period;


fixinv_new[prc,seg,t] is the fixed investment cost for a new plant based on technology prc in time period t using linear
investment cost function seg;
varinv_new[prc,seg,t] is the cost per capacity unit for a new plant based on technology prc using linear investment cost
function seg during time period t;
land_cost[site,t] is the unit cost (in monetary units per area unit) of land at location site during time period t;
fixinv_add[prc,seg,t] is the fixed investment cost for capacity expansion of an existing plant based on technology prc in
time period t using linear investment cost function seg;
varinv_add[prc,seg,t] is the unit cost for capacity expansion of an existing plant based on technology prc using linear
investment cost function seg during time period t.

Objective function II: cumulative landfilling


This is the summation over all time periods of the landfilled volumes of materials in all the landfills.
T
nypp × ∑ ∑ ∑ ∑ ACT[prc, mode, lf , t] (42)
t =1 lf ∈Lfl prc∈ mode∈
Lf [lf ] Mode [ prc ]

The EUGENE model is the following:


minimize Objective function (41) or (42)
subject to constraints (1)–(40)
and the integrity and non-negativity restrictions of the variables

Acknowledgements
The authors acknowledge Nadia Live, France Ratel and Maxime Soucy for their many constructive comments and
suggestions that helped us to gradually improve the quality and the scope of the model. They also acknowledge two
anonymous referees and Amit Kanudia for suggestions that improved the paper.

References
Baetz, B.W. and Neebe, A.W. (1994) ‘A planning model for the development of waste material recycling programmes’, Journal of the
Operational Research Society, Vol. 45, No. 12, pp. 1374–1384.
EUGENE: an optimization model for solid waste management planning 32

Ben Abdallah, M. (1997) Les marchés des matières secondaires au sein d’une gestion intégrée des déchets, Master thesis, École des
Hautes Études Commerciales, Montréal, Canada.
Carlson, R.L. (1986) ‘The impact of materials recycling programs on energy recovery facility economics’, Journal of Resource
Management and Technology, Vol. 15, No. 1, pp. 28–36.
Caruso, C., Colorni, A. and Paruccini, M. (1993) ‘The regional urban solid waste management system: a modelling approach’, European
Journal of Operational Research, Vol. 70, No. 1, pp. 16–30.
Chang, N.-B., Schuler, R.E. and Shoemaker, C.A. (1993) ‘Environmental and economic optimization of an integrated solid waste
management system’, The Journal of Resource Management and Technology, Vol. 21, No. 2, pp. 87–100.
Chang, N.-B. and Wang, S.F. (1996) ‘Solid waste management system analysis by multiobjective mixed integer programming model’,
Journal of Environmental Management, Vol. 48, No. 1, pp. 17–43.
COMPASS Modeling Solutions, Inc. (1995) Using AMPL Plus, Reno, Nevada.
CPLEX Optimization, Inc. (1995) Using the CPLEX Callable Library, CPLEX Optimization, Inc., Incline Village, NV.
Everett, J.W. and Modak, A.R. (1996) ‘Optimal regional scheduling of solid waste systems I: model development’, Journal of
Environmental Engineering, Vol. 122, No. 9, pp. 785–792.
Fishbone, L.G. and Abilock, H. (1981) ‘MARKAL, a linear-programming model for energy systems analysis: technical description of the
BNL version’, Energy Research, Vol. 5, pp. 353–375.
Fourer, R., Gay, D.M. and Kernighan, B.W. (1993) AMPL: A Modeling Language for Mathematical Programming, The Scientific Press.
Gottinger, H.W. (1991) Economic Models and Applications of Solid Waste Management, Gordon and Breach Science Publishers, New
York.
Haverly Systems, Inc. (1982) Advanced Refinery Modeling Course, Denville, New Jersey.
Jadot, P., Heirwegh, T. and Thonet, C. (1979) ‘A multinational energy system: data base, simulation and optimisation models’, in Strub,
A. (Editor) Energy Models for the European Community. An Energy Policy Special, pp. 72–88.
Jenkins, L. (1982) ‘Developing a solid waste management model for Toronto’, INFOR, Vol. 20, No. 3, pp. 237–247.
Live, N. (1997) Modélisation des technologies de traitement et d’élimination dans un système de gestion intégrée des déchets, Master
thesis, École Polytechnique de Montréal, Canada.
Ljunggren, M. (1997) A Systems Engineering Approach to National Solid Waste Management, Thesis for the Degree of Licentiate of
Engineering, Energy Systems Technology Division, Chalmers University of Technology, Göteborg, Sweden.
Ratel, F. (1998) Gestion intégrée des déchets sur le territoire de la Ville de Montréal: analyse de la filière compostage, Master thesis (in
preparation), Université du Québec à Montréal, Canada.
Schrattenholzer, L. (1981) The Energy Supply Model MESSAGE, International Institute for Applied Systems Analysis, Laxenburg,
Austria, Report RR-81-31.
Shekdar, A.V., Krishnaswamy, K.N., Tikekar, V.G. and Bhide, A.D. (1992) ‘Indian urban solid waste management systems - jaded
systems in need of resource augmentation’, Waste Management, Vol. 12, pp. 379–387.
Soucy, M. (1997) Modélisation des sources génératrices et des types de collecte liés à un système de gestion intégrée des déchets, Master
thesis, École Polytechnique de Montréal, Canada.
Sundberg, J., Gipperth, P. and Wene, C.-O. (1994) ‘A systems approach to municipal solid waste management: a pilot study of Göteborg’,
Waste Management and Research, Vol. 12, No. 1, pp. 73–91.
Wilson, D.C. (1977) ‘Strategy evaluation in planning of waste management to land: a critical review of the literature’, Applied
EUGENE: an optimization model for solid waste management planning 33

Mathematical Modelling, Vol. 1, pp. 205–217.

S-ar putea să vă placă și