Content uploaded by Elizaveta Kuznetsova
Author content
All content in this area was uploaded by Elizaveta Kuznetsova on May 26, 2019
Content may be subject to copyright.
Integrated Decision-Support Methodology for Combined Centralized-
Decentralized Waste-to-Energy Management Systems Design
Elizaveta Kuznetsovaa,c, Michel-Alexandre Cardinb, Mingzhen Diaoa, Sizhe Zhanga
a Department of Industrial Systems Engineering and Management, National University of
Singapore, Singapore
b Dyson School of Design Engineering, Imperial College London, United Kingdom
c GERAD at Department of Mathematics and Industrial Engineering, École Polytechnique
Montréal, Canada
Abstract
The rapid expansion of urban populations and concomitant increase in the generation of
municipal solid waste (MSW) exert considerable pressure on the conventional centralized
MSW management system and are beginning to exceed disposal capacities. To tackle this issue,
the conventional centralized MSW management system is more likely to evolve toward a more
decentralized system with smaller capacity waste treatment facilities that are integrated at
different levels of the urban environment, e.g., buildings, districts, and municipalities. In
addition, MSW can become an important urban resource to address the rising energy
consumption through waste-to-energy (WTE) technologies capable of generating electricity,
heat, and biogas. This shift toward the combined centralized-decentralized waste-to-energy
management system (WtEMS) requires an adapted decision-support methodology (DSM) that
can assist decision-makers in analyzing MSW generation across large urban territories and
designing optimal long-term WtEMS.
The proposed integrated DSM for WtEMS planning relies on: i) an MSW segregation and
prediction methodology, ii) an optimization methodology for the deployment of multi-level
urban waste infrastructure combining centralized and decentralized facilities, and iii) a multi-
criterion sustainability framework for WtEMS assessment. The proposed DSM was tested on
a case study that was located in Singapore. The proposed WtEMS not only reduced the total
operational expenses by about 50%, but also increased revenues from electricity recovery by
two times in comparison with the conventional MSW management system. It also allowed
more optimal land use (capacity-land fragmentation was reduced by 74.8%) and reduced the
size of the required transportation fleet by 15.3% in comparison with the conventional MSW
system. The Global Warming Potential (GWP) was improved by about 18.7%.
Keywords: mega-cities, municipal solid waste, waste-to-energy, combined centralized-
decentralized management systems, waste sources segregation, long-term planning
Nomenclature
Abbreviations
AD – Anaerobic Digestion
CF – Centralized Facility
DF – Decentralized Facility
DSM – Decision Support Methodology
GWP – Global Warming Potential
KPI – Key Performance Indicator
LCA – Life Cycle Analysis
MSW – Municipal Solid Waste
WGS – Waste Generation Source
WtE – Waste-to-Energy
WtEMS – Waste-to-Energy Management System
WTF – Waste Treatment Facility
Sets
||, – Life span period of a WTF [year]
||, – Number of waste generators [unit]
||, – Number of candidate sites where decentralized (on-site) and centralized (off-site)
treatment facilities can be installed [unit]
||, – Set of technologies available for the deployment
||, – Possible number of units of each technology that can be deployed at each candidate
site [unit]
Parameters and variables
, – Amount of waste generated at each time step, t, by each waste generator, i [tons of
waste/year]
– Unit transformation capacity of treatment facility [tons of processed waste/day]
, – Limitation of land space represented by the maximum number of units of technology, a,
to be installed at candidate site, j [units]
,
– Amount of recovered resource per ton of treated waste of technology, a [Amount of
recovered energy/material/ton of processed waste]
,
– Amount of additional resources, i.e., water and electricity, required for one ton of waste
treatment by technology, a [amount of consumed resource/ton of treated waste]
– Amount of waste which can be transported by one transportation unit (e.g., a truck) [tons
of waste/transportation unit]
, – Distance between a waste generator, , and candidate site, [km]
,
– Waste treatment facility deployment cost for the technology, a, and unit, l, per one-ton
capacity [$/unit capacity]
– Cost associated with waste transportation per kilometer [$/km]
– Operations cost of technology, a, including facilities maintenance and other possible
expenditures, e.g., lease costs for treatment machines [$/unit capacity]
,,
– Cost associated with land use required for waste treatment unit, l, of technology, a, at
each candidate site, [$/ unit capacity]
,
– Manpower cost depending on type of technology, a, and number of installed units, l
[$/unit capacity]
– CO2 penalties or taxes associated with waste transportation and treatment [$/ton of CO2
equivalent]
– Resources price for resource consumption or recovery during waste treatment process
[$/amount of energy/material]
– Emissions per ton of treated waste [gCO2 equivalent/ton of processed waste]
– Emissions per km of waste transportation [gCO2 equivalent/km]
– Discount rate
, - The coefficient for cost reduction for every subsequent unit, l, of technology, a
– Economy of Scale factor
– Arbitrarily large number
,,, – Decision variable indicating the number of capacity units, l, of technology type, a, to
be deployed at candidate site, , at time step,
,,, – Decision variable indicating the amount of waste generated by waste generator, ,
assigned (i.e., transported) to the technology of type, a, installed at candidate site, , at time
step,
,
- Additional continuous variables to determine the smaller value between the quantity
of waste, ,, generated at WGS i at time step is greater than or equal to the system capacity,
,,, installed at candidate site, j
– Binary decision variable, equal to 1 if the quantity of the waste, ,, generated at WGS i
at time step, , is greater than or equal to the system capacity, ,,, installed at candidate site,
j
1. Introduction
Over the past decades, the global urban population increased significantly, has reached 3.9
billion residents, and is projected to increase 66% by 2050 (UN 2014). The disparity between
urban and rural populations will become drastic for some regions, with about 90% of the
population living in cities and driving the creation of large mega-cities (UN 2014). Continuous
urbanization amplifies challenges related to the adequate delivery of basic services and
infrastructure to ensure a minimum quality of life for the residents (UN Habitat 2016). One of
these challenges concerns the efficient management of municipal solid waste (MSW), the
amount of which is expected to increase from 1.3 billion tons per year in 2012 to almost 2.2
billion tons per year in 2025 (Hoornweg & Bhada-Tata 2012).
Classical strategies for MSW management rely on incineration, sanitary landfills, and open
dumps. For Organisation for Economic Co-operation and Development (OECD) countries,
incineration covered 21% of waste, whereas sanitary landfills hosted more than 42% of MSW
generated in 2012 (Hoornweg & Bhada-Tata 2012). In AFR countries, 78% was sent to open
dumps and 88% was landfilled (Hoornweg & Bhada-Tata 2012).
More recently, several countries started their journeys toward more responsible MSW
management, with an emphasis on increasing resource recovery and decreasing waste disposal.
Countries with modest land territories showed exemplary results in achieving waste
management goals by transforming their disposal space shortage bottleneck into a driver for
the deployment of sustainable MSW management. Japan went under an 11% threshold of
MSW disposal rates by promoting new waste management incentives starting in 1970 (JESC
2014). South Korea decreased its landfill rate from over 90% in the 1980s to under 10%, while
its recycling rate increased to 80% (WMR 2015). Germany reduced the number of landfill sites
from 50,000 in 1950 to 300 in 2016, and is planning to recommission all remaining landfills
by 2022 (Greentumble 2016). Belgium is regarded as one of the top performers in waste
management, with 75% of its waste being reused, recycled, or composted. This resulted from
the implementation of waste management plans that were crafted 25 years ago (Greentumble
2016).
At the same time, the reliance on centralized MSW management architecture has been proven
to be inefficient by many scholars based on the experience of other countries. The increasing
MSW amounts require the expansion of the waste collection fleets and extension of
transportation journeys that contribute to traffic jams (Yukalang et al. 2017), local pollution,
and road deterioration (Ejaz et al. 2010). Increasing MSW amounts also stimulate the
deployment of new large disposal and incineration facilities to replace the existing ones whose
capacities are being rapidly exceeded (UN 2011). These factors make centralized MSW
management expensive and unsustainable in practice, which can also decrease the resilience of
cities and amplify risks related to public health and the environment (UN Habitat 2016). The
option of shrinking large disposal capacities has been recognized to be an important world
concern in the face of rapid urbanization (Yadav et al. 2017) (Figure 1(a)).
Various waste recycling technologies and initiatives have been adopted as alternatives to the
classical centralized strategies for waste disposal. Indeed, by considering MSW as a valuable
resource, new recycling technologies can generate electricity and useful heat (Xiong et al.
2016), syngas and biodiesel (Wen et al. 2016), compost and liquid fertilizer (Wei et al. 2017),
and other by-products. In some cases, waste becomes an important alternative to address
resource scarcity, e.g., waste-to-energy (WtE) technologies that can generate electricity, heat,
and biogas. Such technologies have been considered in different countries to overcome energy
production scarcity issues (Korai et al. 2017; Halder et al. 2015). To enable this multi-resource
recovery, the classical approach to MSW management that consists of large centralized plants
located in the city periphery is more likely to evolve towards a management system with waste
treatment facilities of smaller capacity integrated directly into the urban environment (Xiong
et al. 2016) (Figure 1(b)). On one hand, this decentralization of waste treatment will ensure the
minimization of waste collection areas, transportation distances, and requirements for the
transportation fleet by treating MSW at a site closer to waste generation sources (WGS) and
recovering valuable resources closer to the final consumers. On the other hand, this
decentralization will relieve the pressure on existing centralized landfill infrastructures. This
new combined waste management system will become integral to the city metabolism that is
aimed at eliminating waste and pollution resulting from residents, municipal activities, and
businesses. An urban-integrated MSW management framework will further contribute to
mitigating climate change. A relatively modest effort toward waste source segregation and
recovery can lead to a considerable environmental improvement (Kayakutlu et al. 2017) and
even conversion of the waste management system into a carbon sink (Menikpura et al. 2013).
While a wide range of waste management planning tools (Morrissey & Browne 2004) and,
more specifically, decision-support tools (Vitorino de Souza Melaré et al. 2017) exists, they
are not adapted for current planning conditions and may not always accommodate planners’
concerns. Indeed, a better understanding of the factors related to multi-level centralized and
decentralized waste treatment, resource recovery, and associated economic outcomes are
required in modern planning tools. Additional considerations include extended environmental,
social, and urban planning constraints (e.g., available land and transportation limitations) and
clear solution benchmarking. This is confirmed by the ongoing international initiatives for the
development of digital and data-driven management systems, e.g., in Amsterdam (Fitzgerald
2016) and Singapore (Bhunia 2018).
A decision-support methodology (DSM) is required to establish a more systematic long-term
system planning approach that capitalizes on the examples of the foremost countries in terms
of implementing MSW management strategies. The DSM must allow a successful transfer of
MSW management practices between counties, while considering local peculiarities and
constraints. Extensive work is still required for the development of coherent MSW
management solutions in an urban context in the presence of multiple stakeholders and decision
factors (Kayakutlu et al. 2017). Thus, the focus of this paper is on the development of a novel
DSM for MSW recycling that is consistent with this analysis and the hierarchy of MSW
management measures defined by (DIRECTIVE 2008/98/EC 2008).
A detailed review of these recent advancements and their bottlenecks is provided in Section 2.
In response to these bottlenecks, the paper addresses the issue of establishing a sustainable
WtEMS urban architecture based on MSW source territorial distribution. It does so by
developing a novel integrated DSM with a demonstration application and contributions along
the following three pillars:
(i) It makes an important advancement toward segregation of MSW sources and
modeling of their distribution across large urban territories. The proposed approach
explicitly defines the relationship between MSW generation and explanatory
variables based on different urban activities and their intensity across large urban
territories. This approach requires neither the collection of large data amounts nor
extended surveys. In addition, it provides MSW estimations depending on the
evolution of the urban landscape defined by urban planners.
(ii) It proposes a methodology for WtEMS design optimization that considers multi-
level candidate locations (e.g., at the level of buildings, districts, and global cities)
for facilities combining various treatment technologies of different capacities. In
addition, it takes into account not only specific urban-planning constraints in
transportation flows (when waste can be transported only to one treatment site) but
also limitations in land-space occupation. The proposed methodology provides the
optimization schedule for MSW treatment facilities deployment over a large
planning horizon, alongside optimal waste assignment (i.e., transportation
schedule) for different time periods.
(iii)It provides an extended multi-criteria framework as an additional filter to evaluate
the compliance of the WtEMS design with economic, environmental, and social
indicators. This evaluation method avoids the complexity arising from multi-
objective optimization accounting for these factors.
The proposed integrated methodology provides guidance to decision-makers to identify
WtEMS with an optimal balance between centralized and decentralized facilities by selecting
optimal technologies, their locations, capacities, and waste assignment. A tractable
optimization model provides an optimal solution for decision-makers in a reasonable time and
illustrates the trade-off between economic, environmental, and social factors.
The remaining sections of the paper are organized as follows. Section 2 provides an extensive
literature review related to the three key research pillars for sustainable WtEMS and analyses
existing bottlenecks. Section 3 outlines an integrated DSM for sustainable planning of
combined centralized-decentralized WtEMS. In Section 4, the methodology is applied to the
analysis of a Singapore case study as an illustration. In Section 5, a critical analysis of the
proposed methodology is provided and future research directions are identified. Section 5
concludes with a synthesis and discussion of the main research outputs.
a)
b)
Figure 1. Conceptual illustration of a MSW management system relying on a) centralized and
b) combined centralized-decentralized WtEMS configurations.
2. Expanded literature review
This section summarizes the main research contributions presented in the introduction by
enumerating the major bottlenecks and challenges for each. The analysis relies on the findings
of the previously conducted extensive review of existing DSM, such as (Vitorino de Souza
Melaré et al. 2017). It also integrates recent bibliographical references in the area of MSW
modelling and prediction, management system optimization, and solution assessment.
The major challenges related to MSW modeling concern the prediction of MSW output based
on either the statistics of MSW generation or construction of complex prediction models
relying on an available (although extended) number of input parameters (Table 1). The first
group of these type of models, such as the one proposed in (Abbasi & El Hanandeh 2016),
struggles to capture changes in future MSW trends since their estimations are based on MSW
historical data. They also do not account for the impact of other explanatory variables, such as
taxes. The second group of prediction techniques involves big data analytics and
implementation of extended surveys to perform spatially-distributed predictions (Keser et al.
2012), or reliance on advanced prediction models that integrate a large number of explanatory
variables (Li et al. 2011). However, the relationship between MSW generation and explanatory
variables is not usually explicitly identified. In view of this, additional data collection (Keser
et al. 2012; Li et al. 2011; Lebersorger & Beigl 2011) and model training (Abbasi & El
Hanandeh 2016) may be required to perform MSW estimations for predictions over different
time horizons. In addition, only a limited number of studies, such as (Keser et al. 2012), have
attempted to provide global estimations of MSW outputs or other subcategories, or to model
MSW distribution for urban territories.
Table 1. Detailed review of the recent MSW prediction approaches.
Bibliographical
reference
Waste category
Tested approaches
Modelled
period and
granularity
Explanatory variables (1)
Case study
(Keser et al.
2012)
MSW
Spatial auto regression
(SAR) and
geographically
weighted regression
(GWR) models
Total
generation
for 1-year
period
(2000)
Population density, higher education
graduation ratio, infant mortality rate, number
of facilities in small organized industrial
districts, a
gricultural production value,
asphalt-paved road ra
tio in rural areas,
unemployment rate, a
nnual average
temperature, and annual rainfall
Turkey
divided
into 81
provinces
(Li et al. 2011)
- Kitchen waste
- Recyclable
materials
- Other wastes
Statistical analysis, a
sampling survey and
the Analytic
Hierarchy Process
5 years
(2004 –
2008),
yearly
time step
Activities: maintenance, subsistence and
leisure
Social parameters: floating population, non-
civil servants, retired people, civil servants,
college students (including both
undergraduates and graduates), primary and
secondary students, and preschoolers
Beijing,
China
(Lebersorger &
Beigl 2011)
Commercial
and household
waste
Explorative data
analysis and
a multiple regression
analysis
Total
generation
for 1-year
period
(2001)
23 main variables divided into groups: Private
households and demographic variables,
economic variables, integrated waste
treatment fac
ilities (local solid fuel heating
and composting), general indicators,
describing regional structure
Styria
region,
Austria
(Abbasi & El
Hanandeh
2016)
MSW
Support vector
machine, adaptive
neuro-fuzzy inference
system, artificial
neural network and k-
nearest neighbors
5 years
(2015 –
2020),
monthly
time step
Amount of waste generation
Logan
city,
Australia
(Adamovic et
al. 2017)
MSW
General regression
neural network and
Structural break
general regression
neural network
models
2 years
(2011 –
2012),
yearly
time step
GDP, urban population, average household
size, tourism expenditure, unemployed rates,
household expenditure, domestic material
consumption, population density, industry
value added, population from 20 to 65, alcohol
consumption and co2 emission
44
countries
(OECD
and non-
OECD
countries)
(Chen et al.
2012)
MSW divided
into recycled,
industrial and
domestic waste
System dynamics
prediction model
40 years
(2005 –
2045),
five-year
time step
total population, birth rate, industrial gdp,
industrial growth rate
Singapore
(1) The input variables used in the prediction approaches to estimate waste generation.
The major recent developments involved solving the problems of facility allocation,
technology selection, and capacity expansion (Table 2). By extension, these studies can be
naturally connected to MSW modelling since optimization models require inputs related to the
amounts and location of generated wastes and to assessments of the sustainability of MSW
management strategies. However, current optimization approaches usually rely on statistical
data related to a specific case study.
Table 2. Research topics addressed in recent bibliographical references.
Reference
Research topics as core pillars of DSM
Waste
modeling and
prediction
Optimization
Solution
assessment
Vehicle
routing Facilities
location
Technologies and
connection to the
end user Size/capacity
Realistic
representation
of the
amounts and
types of
generated
waste in a
given
territory
The amount
of waste to be
transported;
vehicle
routing and
fleet
Selection of
optimal
location for a
treatment
facility for a
given number
of candidate
locations
Selection of
optimal treatment
technologies and
energy /material
output
Expansion of
facility for
the long-time
planning
horizon,
assignment to
different
treatment
technologies
Complex
assessment
framework
taking into
account
economic,
environmental
and social
considerations
(Mirdar Harijani et al. 2017)
X
X
X
X
(Yadav et al. 2017)
X
X
(Lee et al. 2016)
X
X(1)
(Rentizelas et al. 2014)
X
X
X
X(2)
(Dai et al. 2011)
X
X
X
(Minciardi et al. 2008)
X
X
(Yu et al. 2012)
X
X
(Santibanez-Aguilar et al. 2015)
X
X
(1) The facility location problem is approximated by a waste assignment problem where CAPEX represents plant
opening cost.
(2) Environmental impact is accounted for in the objective function through CO2 emissions monetization.
Table 3 summarizes the key details of the reviewed optimization approaches for MSW
management system deployment.
In most of the papers reviewed, except (Mirdar Harijani et al. 2017), the optimization models
did not include a selection of waste treatment technology for each candidate site. Each
candidate site was predefined for the deployment of a specific waste treatment technology, e.g.,
incineration or biomass treatment. Although this can be explained by specific urban constraints
and requirements, it restricted exploration of the types of technologies considered for
deployment at each site. In addition, many studies (Mirdar Harijani et al. 2017; Rentizelas et
al. 2014; Dai et al. 2011), did not distinguish between centralized and decentralized waste
treatment facilities. This oversight is of note because the maximum treatment capacity for
decentralized facilities can be considerably different (e.g., up to 106 times smaller) than the
maximum capacity of centralized facilities. Therefore, decentralized deployment implies
different company sizes and business models, which generate different investments and
operation costs per ton of waste treated in comparison with the more sizeable centralized
facilities.
Table 3. Detailed review of selected deployment approaches.
Bibliographical
reference
Aim of the optimization
approach and case study
System specific focus – waste categories and
waste treatment technologies
Objective function and
additional assessment
indicators
Deployment
horizon
(Mirdar
Harijani et al.
2017)
NPV optimization of
recycling and disposal
network in Tehran (Iran)
Waste generator clustering
around collection points or
22 municipalities centers.
Waste categories: plastic, glass, paper, metal,
organic, others. Transformation into recyclable
material (plastic, paper and metal), electricity,
compost fertilizer.
Technologies: material recovery facilities,
anaerobic digester, composting facilities,
landfill with gas recovery system, advanced
thermal treatment (pyrolysis and gasification).
Off-site centralized facilities of important
capacity (about 250 – 500 tons/day)
Revenues (output generation
+ gate fees for waste
processing)
CAPEX
OPEX
Environmental cost
Medium – 5-
year
planning
horizon
(Yadav et al.
2017)
Selection location for
transf
er station for waste
collection
Hypothetical urban center
of 192 km2 and 1.8 million
of habitant in 2035
Waste categories: compostable, recyclable and
landfill.
Accounts for different collection schedules,
transportation capacities of public and private
companies from residential, commercial and
institutional sources.
CAPEX of transfer station
deployment
OPEX for transportation and
operation of existing and
deployed facilities
Lon-term –
20-year
planning
horizon
(Lee et al.
2016)
Optimization of waste
transfer, collection truck
management strategies,
optimal locations for new
waste treatment facilities
Hong Kong (waste transfer
to China is a possible
feasible solution)
Technologies: incineration and landfills
CAPEX (incinerator and
warehouse)
OPEX (operational cost in the
incinerator and landfill,
transportation cost from
each two points, cost of
moving replacement truck to
waste collection point and
incinerator, truck cost)
Revenues (from incinerator)
Short-term –
1-year
planning
horizon
(Rentizelas et
al. 2014)
NPV of WtE facilities and
associated electricity grid
and heating/cooling
infrastructures deployment
Thessaly district,
Greece
Biomass-type waste from MSW and
agricultural sources related to wheat straw,
maize, cotton stalks and prunings from olive
and almond trees.
CAPEX
OPEX (related to the power
plant, the supply chain of
MSW and biomass, the
district heating and cooling
(district energy) network with
the connection to the
customers, as well as the
electricity transmission line
and connection to the grid)
Revenues
Long-term –
20-year
planning
horizon
(Dai et al.
2011)
Waste generation
prediction and expansion of
the existing composting and
incineration facilities in
Beijing, China
Technologies: landfill, composting and
incineration.
CAPEX
OPEX
Revenues
Medium – 5-
year
planning
horizon
(Minciardi et
al. 2008)
Waste assignment to
different waste treatment
facilities in Genova, Italy
Waste categories: paper, plastic, glass, wood,
organic, metals, inert matter, scraps, textiles
Technologies: landfill, incineration plant,
plant for organic materials treatment and
refuse derived fuel plant
OPEX
Revenues
Unrecycled waste
Sanitary landfill disposal
Environmental impact
(incinerator emissions)
Short-term –
1-year
planning
horizon
(Yu et al. 2012)
Waste management for the
abstract case study of three
cities in China
Waste categories: glass and other types of
waste
Technologies: glass recycling plant,
incineration and sanitary landfill
OPEX (collection,
transportation, recycling,
treatment and disposal costs)
Risks associated to waste
management procedures and
technology used
Medium – 5-
year
planning
horizon
(Santibanez-
Aguilar et al.
2015)
Waste management for the
case study of five cities in
Mexico, each one divided
into 10 subzones.
11 waste categories including MSW, brown
glass, paper, aluminum and non-recyclable
waste.
Technologies: material recycling, thermal and
chemical recycling, pyrolysis, incineration,
pyrolysis and gasification, plasma arc
gasification, conventional gasification
Net profit (Revenues, CAPEX
and OPEX)
Amount of processed waste
Total number of fatalities
Short-term –
1-year
planning
horizon
Another bottleneck is related to the way environmental and social impacts of MSW
management systems are integrated as part of the DSM. To avoid compromising computational
tractability, multi-objective optimization may require aggregating several objectives into one
function, either by converting environmental impacts into an economic unit of global
optimization objective (Mirdar Harijani et al. 2017) or by using a weighting approach to
aggregate social risk (Yu et al. 2012). Despite allowing easy aggregation, this method may
underestimate the weights or prices of different objectives and “hide” their effects on
optimization results. Another way to illustrate the optimization trade-off is to perform a
classical multi-objective optimization. This may imply greater complexity for the optimization
model and can drastically increase the computational burden in comparison with single-
objective optimization. This requires a range of assumptions or simplifications to deal with the
computational complexity that arises, e.g., by adopting a reference point (Minciardi et al. 2008)
or obtaining Pareto fronts through optimization of individual objectives (Santibanez-Aguilar et
al. 2015). In this view, the attempt to incorporate LCA, or social oriented criteria, into the
optimization model poses additional challenges to it, such as the introduction of a greater
complexity into the problem, tractability, and the difficulty of aggregating different objectives
into one term. To date, no multi-criteria assessment framework based on economic,
environmental, and social indicators exists to assess the sustainability of an MSW management
system.
3. Integrated Decision Support Methodology (DSM)
The paper proposes an integrated DSM by addressing the aforementioned challenges. In
addition, a promising solution to overcome the limitations of individual approaches lies in the
combination of modeling, optimization, and assessment frameworks for the development of
the extended models. Figure 2(a) depicts the DSM flowchart composed of three main modules:
(1) the waste modelling and prediction, (2) optimization of WtEMS, and (3) a multi-
dimensional assessment. In Step 1, the MSW sources are categorized and their distribution
across the urban territory is modelled. By relying on the projections for explanatory variables,
e.g., demographic and economic conditions, this module provides MSW source predictions and
helps quantify uncertainties in MSW generation for all planning horizon durations (Figure
2(b)). These MSW output scenarios, as well as data related to the abstract models of waste
treatment technologies, are used as input data in Step 2 focusing on WtEMS optimization. The
optimization module encompasses all related technical and cost parameters and, guided by the
optimization objective, aims at finding the optimal configuration for the WtEMS. After the
optimization module yields an optimal deployment plan, Step 3 evaluates it using a multi-
criterion assessment framework. The assessment module uses the base line MSW treatment
strategy typically represented by the current MSW treatment with incineration. The waste
treatment strategy assessment can be done under projections of future operational conditions,
e.g., resource costs or specific urban planning conditions, and can lead to the update of the
specific optimization model constraints. In this view, projections of economic conditions and
urban planning strategies may lead to adjustments of the optimization constraints related to
maximum available local space for waste technology deployment in order to improve the final
sustainability key performance indicator (KPI) of the global WtEMS solution. Eventually, after
multi-criteria assessment of different WtEMS and comparison with benchmark scenario is
done, optimal WtEMS designs with the associated deployment schedule can be selected.
a)
b)
Figure 2. Integrated DSM for WtEMS design: a) methodology flowchart and b) WtEMS
design procedure.
3.1. MSW distribution across urban territories
Several attempts have been made to narrow waste quantification by categories. These attempts
are mainly based on long-term campaigns of waste sampling covering large territories and
interviews at various stages of the existing MSW management systems. The general tendency
of MSW segregation by category is summarized in (Hoornweg & Bhada-Tata 2012) that
discussed similar waste proportions in different regions across the globe.
However, the waste management system is a spatial problem requiring not only knowledge of
the global amounts of MSW generated in a territory, but also an understanding of the
distribution of these sources across this territory, which is also referred as the “geography of
waste”. Indeed, WGS quantification and distribution impact not only the choice of waste
treatment technologies, but also waste collection and transportation, e.g., fixed routines for
regularly produced waste of large amounts, infrequent schedules for seasonal waste, and upon
request collection for irregular and bulky waste types (Nilsson & Christensen 2011). The waste
management system of each territory is defined by its administrative subdivision responsible
for performing, organizing (e.g., hiring private companies) and supervising waste management
profiles. For example, waste management profiles have been found to differ by metropolitan
areas regrouped into regions in Turkey (Goren & Ozdemir 2010), by municipalities in the
Metropolitan Region of São Paulo (Brazil) (Jacobi & Besen 2011), and by urban districts or
communes in the municipality of Bamako (Mali) (Kéita 2001). In this view, the waste
management follows the municipal ordinances for collecting waste management taxes and
prescribing collection routes, frequency, bin systems, etc. (Nilsson & Christensen 2011).
However, WGS are usually non-uniformly distributed across urban territories depending on
residential, commercial, office, industrial and mixed activities subzones, illustrated in Figure
3(a), with different waste distribution proportions for each waste category. The MSW is broken
down into categories and analyzed to determine its sources (i.e., activities subzones) and
associated factors affecting its waste generation and distribution (Figure 3(b)). Subsequently,
the activities subzones are analyzed and linked to specific datasets characterizing those
subzones. Apart from industry data and population census, historical data on waste generation
and distributions will be required for modeling and validation purposes. Another reason to
model the waste distribution by administrative subzones lies in the typical availability of
datasets by these administrative subdivisions.
a)
b)
Figure 3. MSW analysis: a) abstract illustration of urban territory subdivision on the activities
and administrative subzones and b) steps for MSW profiling.
The first step to determine the MSW distribution starts with profiling the subzones and waste
types for a better understanding of associated explanatory variables. This information can then
be used for a weight calculation for each waste type in each subzone. In this view, it is ideal to
identify the breakdown of activities in each subzone for accurate modeling of waste
distributions, e.g., on manufacturing, retail and construction, and their intensity. However,
since such specific information is usually unavailable, more generalized methods and
assumptions can be used to obtain waste distributions.
Figure 4. Framework for modeling of MSW distribution across urban territories.
The procedure for waste distribution modeling across territories is as follows:
1. MSW profiling. Following the example of MSW profiling in Figure 3(c), MSW must
be analyzed and split into categories defined by scenario, s. These categories can be
related to the global MSW categories identified and quantified for the whole urban area.
The possible activities, responsible for the generation of each of the waste categories in
urban territory, along with explanatory variables, must be identified.
2. Activities subzones. The urban territory of interest must be split into the activities
subzones, n. This split can be supported by the data issued from urban development
strategies (Figure 3(a)).
3. Administrative subzones. The division of the same urban territory into administrative
subzones, i, according to urban governance structure, e.g., by districts (Figure 3(b)),
must be performed.
4. Administrative subzones classification. Each administrative subzone must be
classified by the occupation fraction of each urban activity. The framework proposes
the classification into residential, industrial, commercial, natural, mixed, and other
types of activities groups.
5. Weight factors calculation. The subzone layer calculates the weightage of each
relevant land occupation by different urban activities. For example, under domestic
waste, the relevant subzones are residential subzones. In this view, the residential
population in each subzone will be weighed against the total residential subzone
population to obtain a weight or fraction of the domestic waste generated in each
subzone. The sum of all subzones weights is equal to1.
a. Domestic Waste
Domestic waste is generated from only residential areas and the population is chosen
as the weightage factor. The domestic subzone weightage (DSW) for domestic waste
is calculated based on its residential population:
=
, C
(1)
where C is the set of subzones i with residential classification.
Subzones that do not have residential occupations have been assigned with a zero
weightage for domestic waste.
b. Non-Domestic Waste
For non-domestic waste, the intensity of activities has been assumed to be
proportional to the territory occupied by this activity, i.e., more land space leads to
greater intensity of the activity or business transactions. In this view, the land area is
used as the weightage factor. However, commercial spaces are likely to be denser in
terms of their activities and, thus, in specific waste generation per land area than the
mixed land categories. To take this into account, a modifier matrix is introduced to
calculate the effective area matrix (EAM) as a weightage factor appropriately:
,= ,,
(2)
where s is the scenario defining MWS categories under consideration, n is the
number of activities/purpose selected for subzone i classification. The land area
matrix, ,, is calculated by multiplying an occupation fraction by the land area of
the corresponding subzone. In case additional data is available for the explanatory
variables influencing MSW generation, the waste output can be adjusted through the
Modifier matrix, ,. This possibility is discussed further in Section 4.1.
The non-domestic subzone weightage (NDSW) for scenario, s, in subzone, i, is
calculated as:
,=,
,
(3)
6. Waste distribution. By using global records on the amounts of MSW categories and
weight factors for domestic and non-domestic MSW, the actual waste distribution
across urban territories is calculated.
3.2. Optimization of waste management
3.2.1. Overview and assumptions for the optimization model
Figure 5 presents the conceptual superstructure for WtEMS deployment including WGS,
energy, and material flows exchanged in the urban territory and surrogate model of waste
treatment technology.
The assumptions underlying the optimization model are as follows:
- The term “on-site machine” refers to the decentralized facilities (DF) of smaller capacity
located in the proximity of each waste source. Conversely, the term “off-site machine”
refers to the centralized facility (CF) of larger capacity treating waste flows transported
from different WGS.
- Each WGS, as well as the candidate site for deployment, is abstracted as a geographical
(waste generation and treatment) node. The distance between a WGS and a candidate site
(DF or CF) is calculated based on the longitude and latitude coordinates by applying the
triangle location algorithm (Ivis 2006).
- All candidate locations can host various waste treatment technologies targeted by the
decision-maker. The on-site installed capacity has to be such that the on-site installed
machine can process all waste generated locally without transportation of any outstanding
waste to off-site facilities to limit transportation flows.
- The capacities of on-site and off-site Waste Treatment Facility (WTF) cannot be reduced
upon deployment.
- The optimization model has been developed using global planning perspectives and does
not account for economic relationships between WtEMS stakeholders. In this view, no
disposal cost or tipping fees have been considered in the model.
- The optimization model for deployment of MSW treatment infrastructure has been
formulated conceptually and independently from the MSW source type considered for
infrastructure deployment. In this view, the theoretical framework can deal with various
types of MSW, e.g., paper and cardboard, horticultural waste and plastic, and associated
treatment technologies as model inputs.
Figure 5. Conceptual superstructure of WtEMS.
3.2.2. Optimization model formulation
The optimization model has been formulated as a mixed integer linear programming (MILP).
The objective is to minimize the “absolute” expenses over the long-term period, T, represented
as the differences between the total costs (Eq. (6)) and the revenues, , obtained from the
resources recovery (Eq. (5)):
,()
(4)
=1
(1 + )
,
,
(5)
By taking into account the aforementioned installation cost, , and operation cost, ,
the total WtEMS cost, , over a period, , is defined as the summation of all relevant costs
discounted over lifecycle period, , to obtain the net present value of future cash flows:
=1
(1 + )+
(6)
The optimization problem accounts for the installation cost of WTF for which capacity can be
progressively deployed during a long-term planning horizon (from several years up to decades).
For more convenience, the investment cost has been divided into two terms: (i) initial
investment cost at time t =0 and (ii) deployment cost for the remaining future planning horizon
for t = 1,…,T:
= ,
,,,
,= 0
(7)
= ,
, ,,,,,,,
= 1, … ,
(8)
where
,=( 1) .
(9)
The EoS factor has been also integrated in Eqs. 9 – 11 to account for the reduction in variable
costs. It is of note that the EoS factors for different costs can vary for different industries and
types of plants. However, it has been concluded that many plants exhibit substantial savings
due to their increased capacities (Haldi & Whitcomb 1967). For the purpose of this study, the
same formulation of EoS based on the capacity expansion has been assumed for fixed and
variables costs.
The operational cost, , encompasses transportation, , land use, , operation and
maintenance, (O&M) , manpower, , cost of additional resources required for waste
recovery (e.g., water and electricity inputs), , and pollution cost :
=+++++, .
(10)
The transportation cost, , is proportional to the transportation distances and the amount
of waste allocated to each technology installed at candidate site, j:
=365 ,,,,,
,
.
(11)
Expenditures involving O&M, land use, and manpower are calculated by using Eq. (12), Eq.
(13), and Eq. (14), respectively. For simplicity, the EoS factor for these variable costs has been
expressed as a function of installed capacity:
= ,
,,,
,
(12)
= ,,,
,,,
,
(13)
= ,,
,,,
,
(14)
The pollution cost (Eq. (15)) consists of two parts: (i) the amount of waste transported
from the WGS to WTF and (ii) the emissions generated by the waste treatment activities:
=365 ,,,,,
+
,,, ,
,
.
(15)
It is of note that the pollution cost integrated in the can account for the specific economic
measures for the reduction in CO2 emissions adopted in different countries.
Finally, during waste recovery, WTF consumes energy and materials (e.g., electricity and
water). The expense for the consumption of these resources is monetized with and is
proportional to the quantity of treated waste:
=
,
,,,
,
,
(16)
Equations (15) and (16) define the costs of pollution and resource consumption and introduce
non-linearity into the optimization model. To linearize it, additional variables
and
that
are introduced in Eqns. (18) and (19) are used to determine which of the quantity of generated
waste, ,, or the capacity of the system, ,,, is smaller. The constraint in Eq. (17) ensures
that the waste from WGS, i, assigned and transported to technology, a, at candidate site, j, at
period, t, cannot be greater that the capacity of this technology, i.e., , ,, is equal
to zero when ,>,,:
,
,,
=
,
(17)
0
,
(18)
0
(1 ),
(19)
In addition, the amount of waste sent to a given treatment facility cannot be greater than the
installed capacity:
,,,,
,,,
, , ,
(20)
The waste generated at WGS cannot be sent to a candidate site where the WTF has not been
installed yet (Eq. (21)). Also, to limit the number of transportation flows, the waste from one
WGS, i, can be transported to one and only one candidate site (Eq. (22)):
,,, ,,,,= 0, , , , ;
(21)
,,,
= 1, , , ,
(22)
Equation (23) shows that the capacity deployment of technology, a, at each candidate site, j, is
progressive. Equation (24) ensures that the installed capacity cannot be reduced throughout the
time:
,,, ,,,, , , ,
(23)
,,, ,,,, , , ,= 1, … ,
(24)
It should be noted that the land surface available in each candidate site and its price can be
defined by the decision maker in the input parameters as the function of maximum possible
installed capacity. For each technology type, a, and at each time step, t, the total installed
capacity cannot exceed the limitation of land space reflected by the total number of possible
waste treatment units which can be hosted by each candidate site, j:
,,,
,, , ,
(25)
3.3. Multi-dimension assessment of WtEMS
Although the economic performance remains dominant for WtEMS design optimization, other
key performance indicators (KPI) must be taken into consideration in the selection of the
optimal WtEMS architecture. Until now, the amount of CO2 emissions generated by the new
waste treatment strategy played the role of this additional non-economic KPI able to evaluate
WtEMS sustainability. However, with the progressive shift toward the combined centralized-
decentralized strategy, WtEMS becomes an inherent part of the urban environment and
warrants a more extended multi-dimensional assessment framework.
The proposed multi-dimensional framework supports the decision-maker in evaluating the
WtEMS optimality from the point of view of:
- Deployed capacity. The technological specifications of waste treatment technologies allow
their deployment by unit blocks of predefined capacities. In this regard, large centralized
facilities are typically composed of “building blocks” of several dozens or hundreds of
tons per day capacities. Under some conditions, the optimization model can lead to an
important capacity over-deployment to cover the total MSW generation. This over-
deployment can create undesirable “lock-in” effects, when the large capital investments
but relatively low capital and operating costs can push higher-valued mechanisms of waste
recycling (e.g., DF) out of the market (WEF 2016). The effectiveness of capacity usage
can be quantified through the capacity utilization rate (Mahadevan 2007) or over-
deployment rate as follows:
-
= 1 ,
(26)
where is the total deployed waste treatment capacity. The parameter, , takes its value
from the range of [0 1] and tends to zero with increasing capacity usage effectiveness.
It is noteworthy that, at the same time, the over-deployment rate represents a reserve
capacity that can be useful to deal with uncertainties in waste generation.
- Reserved land. Land use required for the waste treatment facilities deployment can become
a critical asset not only in the context of land-constrained territories, such as Singapore
and Hong Kong, but also large mega-cities. Indeed, efficient land allocation between
municipal activities represents a major challenge in the context of rapidly growing cities
(Ichimura 2003). In this regard, the decision-maker must be able to select WtEMS in order
to avoid large land occupation at one candidate location and ensure the optimal dispersion
of waste treatment capacities across urban territories. Indeed, the capacity dispersion can
offer important advantages. The spatial spread of WTF can minimize potential risks due
to facility failures, ensure a presence of WTF in different urban districts, and provide a
uniform waste distribution across all urban territories. To evaluate this capacity dispersion,
the indicator of urban fragmentation index (Demetriou 2014) is converted into a capacity-
land fragmentation factor as follows:
= ,
,,
(27)
where ., is the land surface occupied by unit, l, of technology, a, at candidate site, j.
Variable ,
is the total waste treatment capacity of technology, a, deployed at candidate
site, j, and S is the total surface of case study. Variable takes its value from the range of
[0 1] and tends to zero in case of high capacity-land fragmentation across urban
territories.
Of note, this KPI equips the decision-maker with information regarding land occupation
required for WtEMS deployment. Its values must be analyzed individually for each
particular urban territory, since the decision-maker can aim for low or high capacity-land
fragmentation for different urban situations. For example, for dense or land-constrained
areas, it can be assumed that a big land portion, i.e., for CF, can be inconvenient. The more
feasible alternative is instead to have more dispersed land occupation, i.e., with DF, when
the KPI of land-capacity fragmentation will decrease. However, a small KPI may
nonetheless generate urban planning challenges since it will imply the need to earmark
numerous lots of small land parcels in the city. In this view, the final analysis about the
suitability of the WtEMS design based on this KPI must involve urban planners.
- Pressure on the transportation network. Another important KPI concerns the pressure
exerted by the WtEMS on the existing urban transportation system. Indeed, detailed
evaluation and optimization of the waste transportation routine is typically performed after
the problem of WtEMS capacity allocation. The transportation cost is considered in the
WtEMS capacity allocation problem; however, this does not enable evaluating the pressure
applied by the waste transportation fleet on the urban mobility network. In this view, the
additional KPI on the waste transportation fleet has been included in the multi-dimensional
assessment framework. The pressure on the transportation network is evaluated through
an average number of trucks per day required to transport the MSW from generation
source, i, to candidate site, j. It has been calculated based on the results from the
optimization model regarding the waste assigned from WGS to WtE facility by using Eq.
(11). This indicator gives a first approximation about the fleet size required for waste
transportation; however, the number of trucks can be further optimized with geographical
information system models for waste collection.
- Global Warming Potential. A carbon emission tax is included in the optimization OPEX
for the process and transportation impact. However, pollutant emissions may originate
from other waste treatment related processes, e.g., electricity and material consumption,
which are not considered in the economic model. In addition, carbon value can be defined
based on different economic mechanisms and different criteria across countries. Indeed,
one of the major bottlenecks of a carbon tax is related to the difficulty of estimating the
real costs of carbon emissions for the environment. The amount of equivalent pollutants
from different MSW management strategies was evaluated in terms of their global
warming potential in tons of CO2 equivalent emitted throughout the system lifecycle under
consideration:
=365 ,,
+
,,,
,
(28)
where ,, is the waste treatment capacity of technology, a, deployed at candidate site, j, at
time period, t. Variable ,,, is the amount of waste transported from WGS, i, to the
technology, a, located at candidate site, j, per time period, t. Variables
and
are the amounts of equivalent CO2 emissions associated with operation of waste
treatment technology, a, and waste transportation, respectively.
As mentioned in Section 3.2, these factors are already indirectly accounted for in the
optimization model through cost functions, i.e., the technology unit, land rental, transportation
and emission costs. However, the associated prices cannot fully reflect the importance of these
factors. For example, land rental reflects the actual land value but does not allow the direct
evaluation of land scarcity. The transportation cost provides estimations of transportation fleet
maintenance and fuel cost but does not provide estimations of the pressure exerted by the waste
transportation fleet on urban mobility. Therefore, in assessing the WtEMS optimality, the
decision-maker needs to assess these additional factors independently from the purely
economic-based optimization. To this end, this paper proposes a multi-dimensional assessment
framework for the decision-maker after the optimization model has identified the
economically-optimal solution.
4. WtEMS in Singapore – waste distribution modeling, system deployment
optimization, and design evaluation
This section provides a demonstration of the complex integrated methodology presented in
Section 3. It mirrors each subsection to present the outputs and analysis for Singapore based
on publicly available information for MSW distribution modeling, WtEMS optimization, and
multi-criterion assessment.
4.1. MSW distribution modeling
4.1.1. Urban area profiling and MSW distribution results
By relying on Singapore’s 2011 land-use plan (URA 2013) that shows subzone activities with
the administrative subdivisions of Singapore, the island activities have been classified into five
different categories: Residential (R), Nature (N), Commercial (C), Industrial (I), and Other (O)
(Figure 6). The “Other” category consists of special use, infrastructure and areas reserved for
further development. As shown in Figure 6, certain subzones can fall into more than one
category due to their diverse land use; such subzones are labeled as Mixed (M).
Figure 6. Singapore administrative subzones classification.
The assignment of land subzones to specific categories (e.g., residential, industrial) has been
done based on a detailed review of all subzone activities, in addition to the (URA 2013) map
that gives a broad and simplified view and the particular assumptions of the land space these
activities occupy. In other words, to some extent, all subzones will include residences, parks,
infrastructure installations, and businesses, which makes all subzones mixed by default.
However, the goal of this paper is to make a first step toward a “geography of waste” concept
and to model waste distributions only from the major waste contributors in each subzone. To
identify dominant activities in each subzone, an occupation threshold has been fixed. If land
space occupied by an activity in the specific subzone exceeds this threshold, this activity is
qualified to be one of the major ones in this subzone. If the subzone accounts for two or more
dominant activities, it is qualified as mixed; otherwise, it is qualified according to its major
dominant activity. Figure 6 has been generated with the threshold of 20% of land occupation
by activity by subzone.
Table 4 shows the MSW profiling based on its possible sources.
Table 4. Waste type profiles.
Waste Type
Examples
Possible Sources
Classification
Construction Debris
Unwanted material from constructions
Construction Sites
Non-Domestic
Used Slag
Waste matter from smelting
Steel Mills
Non-Domestic
Ferrous Metals
Steel Cans, Aluminum
Households, Factories
Mixed
Wood/Timber
Pallets, Furniture, Crates
Households, Factories
Mixed
Horticultural Waste
Tree trunks, Branches
Maintenance of trees
Non-Domestic
Paper/Cardboard
Books, Boxes
Households, Offices, Factories
Mixed
Food Waste
Meat, Fish, Vegetables
Households, Restaurants
Mixed
Plastics
Plastic bottles, Plastic bags
Households, Offices
Mixed
Others
-
-
Mixed
Following this waste profiling, Table 5 identifies the type of waste based on its possible
sources, depending on the subzone classification.
Table 5. Waste type subzone classification
Waste Type
Subzone Classification
Residential
Commercial
Industrial
Nature
Others
Construction Debris
X
X
X
X
Used Slag
X
Ferrous Metals
X
X
X
X
Wood/Timber
X
X
X
X
Horticultural Waste
X
X
X
X
X
Paper/Cardboard
X
X
X
X
Food Waste
X
X
X
X
Plastics
X
X
X
X
Others
X
X
X
X
X
The above tables offer insights into the assignment of waste to the different subzones. The
classification of the areas was done using qualitative online research and analysis of the land
use plan. Categories of land activities present in the administrative subzones are given a score
of 1 while those absent are given 0. Based on this analysis, mixed subzones involving more
than one activity have been assigned a score between 0 and 1 based on the land space
percentage occupied by each activity in each administrative subzone. For this purpose,
Singapore’s land development plan has been used. Attention should be drawn to the possibility
that these scores can be adjusted to reflect different activity densities, e.g., the population
density in the respective residential sectors and intensity of commercial activities.
Figure 7 gives an overview of the available statistical data for waste generation in Singapore
from 2003 to 2015. The available data was used to obtain a fixed percentage ratio for waste
types falling in both categories. The percentages of domestic and of non-domestic wastes for
mixed waste types were estimated to be 60.85 and 39.15%, respectively.
The major sources of food waste generation have been estimated to be Bedok, Woodlands, and
Jurong West with 5.9, 5.1, and 4.8% of the total Singapore food waste output, respectively
(Figure 11(c)). The highly dense residential areas involve a high number of markets, food
centers, and restaurants responsible for the considerable food waste generation.
4.1.2. MSW model analysis
While the calculation based on land area and population may be logical for domestic and non-
domestic waste generation, it may present some drawbacks.
Firstly, the MSW distribution model considers only basic geographical and demographic
attributes. However, more accurate MSW distribution modeling requires more data related to
different subzones activities. By considering this, subzones situated remotely from residential
areas could be less frequented and may generate less MSW than remote areas holding attraction
elements. Accordingly, hotspot places such as Tampines with many conveniently located
shopping malls or Geylang with nightlife activities could hold an increased human traffic and,
thus, increased waste generation. In addition, the types of waste generated during day and night
times can vary. Accordingly, a modifier matrix becomes important to encompass such social
and economic parameters in the MSW distribution model. A possible contribution could be in
the MSW modeling distribution and urban development areas in order to quantify the
attractiveness of different urban areas and, thus, model their MSW distribution. The modifier
matrix can be established based on the centrality index quantifying the centrality of a given
location by combining the number of people attracted to locations and the range of their
activities engaged at these locations (Zhong et al. 2017).
Secondly, the MSW distribution uses 12 years of statistical data to obtain the fixed percentage
ratios for waste types falling simultaneously in categories of domestic and non-domestic waste.
However, for accurate waste profiling, a more detailed analysis is required of the waste
categories and activities sources. This work can be done in collaboration with local authorities
by holding survey campaigns for waste generation and collection.
Moreover, MSW distribution modeling is based on fixed 2015 population census data and land
area subzones for weight calculations. In this view, the weights for each subzone will be
constant and not change over time. One possible improvement would be to link the MSW
distribution model to the prediction of explanatory variables to determine the evolution in
domestic and non-domestic MSW. For example, the growth of a subzone population based on
projects for development of residential areas or the extent of industrial development based on
the opening of new industrial sites could be considered.
4.2. WtEMS design optimization
4.2.1. Optimization model input
The optimization model can exploit the output of the MSW distribution modeling (Section 4.1)
or the specific information of MSW generation provided by a decision-maker. For simplicity,
the proposed methodology has been illustrated here through one stream of food waste
representing an important portion of the world MSW (Chainey 2015). The disposal rate of food
waste reaches almost 86% for Singapore (NEA 2016).
Figure 9. Food WGS established on the Google Map representation of Singapore.
For the purpose of this study, the WGS have been represented by the nodes of food waste
generation in 111 food courts, hawker centers, and markets across Singapore. To estimate the
amount of food waste produced daily by each hawker center, waste generation data have been
collected in several targeted sources. By using this data, as well as WGS area estimation from
the GIS software, the average food waste per unit area has been estimated to be 1.409 kg/m2
per day. Under the assumption that the same food waste amount, ,, is generated per unit area,
the waste output for 111 WGS was estimated (Figure 9). The amount of waste generated at
each WGS, i, has been assumed to grow linearly over the considered lifecycle with a constant
increase of 2% per year.
Table A1 in Appendix A summarizes the data related to possible technologies to be deployed
at candidate locations. The first prototype of micro-scale anaerobic digestion (AD) technology,
considered here for the on-site deployment, was built in 2013 in London (UK) to process urban
food waste and continues to operate to date (Walker et al. 2017). Currently, three similar pilot
plants have been established, with two in Central London (Izabelanair 2017). The equivalent
large scale AD plant has been considered for the off-site deployment (Izabelanair 2017).
Although, current waste treatment in Singapore mainly relies on incineration, the technical
specifications of the benchmark technology have been defined based on open source data for
waste-to-energy technology from (Cook 2014).
Two candidate sites have been preselected for the off-site facilities locations: the reserved
construction area in Seletar subzone and the area near the water reclamation plant in Changi.
At each candidate site, five AD units can be deployed, subjected to the limitation of available
land space. The transportation distances have been calculated between the WGS and these
candidate sites, i.e., the average distances from the WGS to candidate sites at Seletar and at
Changi are 11.6 km and 17.6 km, respectively.
The candidate sites for on-site DF coincide with WGS coordinates. In this view, no specific
transportation efforts are needed to ensure the supply of food waste from the WGS to the DF
equipped with micro-scale AD. The land cost is assumed to be $15/m2 per year for the industrial
areas and $25/m2 per year for residential areas (JTC 2016). The CO2 emission tax has been
assumed to be $10 per ton of CO2 equivalent (Lam 2017). The CO2 emissions for food waste
transportation has been defined to represent on average 600 g CO2 per kilometer travelled
(Dunnebeil & Lambrecht 2012). The electricity price has been fixed to $0.15/kWh (EMA
2017). The discount rate, EoS factor, and lifecycle duration have been set up to 0.01, 0.8, and
15 year, respectively. The convergence gap has been fixed to 1%.
4.2.2. WtEMS deployment
The following figures illustrate a progressive deployment of waste treatment infrastructure for
Year 1 (Figure 10(a)), Year 10 (Figure 10(b)), and Year 15 (Figure 10(c)) across Singapore. In
addition, they depict the evolution of food waste assignment from the WGS to different WtE
CF. The optimization starts with the deployment of AD at Year 1 at both off-site candidate
sites of 250 and 200 tons/day capacity, respectively. The capacity at site 2 is expanded at Year
3 up to 250 tons/day. The DF are deployed progressively throughout the lifecycle to treat food
waste exceeding the WtE CF capacity.
a)
b)
c)
Figure 10. Progressive WtEMS deployment over the lifecycle: a) Year 1, b) Year 10, and c)
Year 15.
The deployment strategy is visible from the detailed cash flows distribution for the entire
lifecycle (Figure 11(a)). The main CAPEX investment for CF deployment is done in Year 1
and another additional CF unit is deployed in Year 3. The DF starts its deployment in Year 2
to deal with waste exceeding the capacity of the WtE CF. Following the increase in the food
waste generation, a progressive addition of DF treatment capacities continues to be observed
in the optimization. The installed capacity proportion of DF/CF reaches about 84% for CF and
16% for DF in the final Year 15.
The discounted cash flow distribution of the total lifecycle is illustrated in Figure 11(b). It
shows that the major expenses are shared by CAPEX, O&M, and manpower costs representing
17.5%, 31%, and 31%, respectively. Resource consumption and transportation expenses
occupy around 11.8% and 6.8% of the total lifecycle investment, respectively. At the current
levels of carbon price and rent cost, pollution tax and land cost represent 0.4% and 1.4%,
respectively, and are not significant in the decision-making.
a)
b)
Figure 11. Optimized solution discounted cash flows: a) yearly cash flows and b) total
lifecycle cash flows.
4.2.3. Sensitivity analysis
a. EoS impact
Figure 12(a)-(c) illustrates the WtEMS deployment for different EoS factors. The CF is used
to treat the majority of the waste generation whereas the DF are used to adjust the installed
treatment capacity to tackle the increase in the waste generation.
a) b)
c)
Figure 12. Progressive WtEMS deployment for a) EoS = 0.6, b) EoS = 0.8, and c) EoS = 1.
This deployment is in line with observations done in (Manne 1967) for the addition of new
capacities under the constant growth of demand and a non-zero discount factor. While
decreasing the EoS factor (or reinforcement of EoS) and maintaining the same discount factor,
it is preferable to build a large capacity earlier in the planning period, even though operators
need to pay immediately for capacity that will only be used later. This is confirmed by the
evolution of the over-deployment of the total WtEMS capacity for different EoS factors, as
illustrated in Figure 13. The over-deployment peaks are situated in order from EoS = 0.6 to
EoS = 1 during the planning lifecycle.
Figure 13. Total WtEMS capacity over-deployment for different EoS factors.
b. Initial capacity input
An important consideration concerns the selection of parameters used by a decision-maker as
input to the optimization model. Such inputs will define an optimized WtEMS configuration
regarding the sizes of waste treatment vertexes, such as the WTF capacity and edges length
that involves the transportation distance. Indeed, the decision-maker must specify the input
parameters related to the capacities of waste treatment units, number of units allowed to be
deployed, and candidate sites locations that define the distances between the WGS and WTF.
As shown in Section 4.2.2, transportation expenses have a moderate contribution to the total
NPV for the case study under consideration. As a consequence, the distance does not exert a
significant influence on the deployment results for the case study size similar to Singapore (i.e.,
several dozens of km) at a similar transportation price. In this regard, the decision-maker has
the freedom to predefine candidate sites in this perimeter without considerably affecting the
final WtEMS configuration. The analysis below focuses on the impact of the initial waste
treatment capacities, preselected for deployment by the decision-maker, on the optimized
WtEMS configuration. Indeed, even if the producers of the technologies could offer various
waste treatment capacities, the decision-maker’s input is required to be more specific. A large
range of preselected capacities can lead to waste treatment technologies of various capacity
sizes in the deployment solution, resulting in customized and costly WtEMS. Moreover, some
capacity sizes selected for deployment can lead to the decreased effectiveness in capacity
usage.
The following analysis shows the influence of the initial CF capacities (defined as inputs by
decision-maker) on the proportion of CF- and DF-deployed capacities for EoS = 0.8 (Figure
14).
Figure 14. Proportions of CF and DF capacities deployed under different initial input of CF
unit capacities.
The WtEMS configuration or CF- or DF-deployed capacities proportions can be affected by
the initial input of the CF capacity unit. By minimizing the objective function, the optimization
model identifies the optimal WtEMS configurations under different initial input of CF
capacities. In this context, for the initial input of CF capacities is equal to 50, 100, and 250
tons/day, and the total deployed capacity is composed of about 84% of CF and 16% of DF
facilities. For the initial input of CF capacity equal to 150 tons/day, the deployed DF portion is
reduced to about 1%, whereas for the initial input of CF capacity equal to 200, 300, and 350
tons/day, no DF facilities have been deployed.
One possible explanation is related to the fact that the optimization problem searches for the
best combination of CF and DF units to address the total waste generation at the most optimal
cost. In this view, when the total waste generation exceeds the CF capacity but is insufficient
to activate the deployment of another centralized unit, the optimization model deploys a DF to
cover this outstanding waste generation. Also, with an increase of the initial input to the CF
unit capacity, the EoS influence increases as well. This leads to a situation where the
optimization model attempts to rely entirely on the centralized deployment. However, this
output can change with the modification of costs associated with CF and DF deployment and
operation. In this view, the deployment mechanism for the combined centralized-decentralized
WtEMS must be further analyzed in detail. Moreover, this mechanism must be accounted by
the decision-makers at the early stage prior to the optimization to achieve the optimal balance
between centralized and decentralized capacities suitable for the specific urban territory.
The influence of DF capacities on the deployment results for the combined centralized-
decentralized WtEMS has been found to be minor and are therefore not discussed in this paper.
4.3. Multi-dimensional assessment
To assess the performance of the optimized deployment strategy integrating both the CF and
DF, the combined WtEMS has been compared with the pure centralized WtE and decentralized
MSW management strategies (Table 6). The metrics for the benchmark case, if all generated
waste has been processed in the conventional incineration facility, have also been calculated.
It was assumed that this conventional treatment is done in the existing facility, which does not
require CAPEX investment.
The inputs for the global warming potential (GWP) assessment of different MSW strategies
were retrieved from the LCA study by (Tong & Tong 2016) conducted for Singapore. The
incineration used for the benchmark was associated with 113∙106 g CO2 equivalent/ton of
treated waste and the AD with 83∙106 g CO2 equivalent/ton of treated waste. These values were
calculated by taking into account the avoidance factor of electricity generation during waste
treatment. The GWP input for transportation activities was estimated to be 1014∙g CO2
equivalent/km for a six-ton load truck (Tong & Tong 2016).
Table 6 summarizes the core economic indicators from the optimization (Section 4.2) and the
multi-dimensional assessment for different MSW management strategies based on i) combined
centralized-decentralized WtE facilities, ii) centralized WtE facilities, iii) decentralized WtE
facilities integrated in the urban environment, and iv) conventional waste treatment by
incineration.
Table 6. Comparison of MSW management strategies.
MSW management
strategies / Parameters Combined Centralized Decentralized
Conventional
treatment (existing
incineration)
Best
Total NPV, ,
M$/lifecycle
-20.82 -8.82 -47.57 - Centralized
Total NPV CAPEX,
M$/lifecycle -22.18 -15.66 -55.02 - Centralized
Total NPV OPEX,
M$/lifecycle -100.64 -94.9 -98.55 -202.8 Centralized
Total NPV Revenues,
M$/lifecycle
102 101.74 106 50.87 Decentralized
Average capacity over-
deployment rate 0.022 0.055 0.11 -(1) Combined
Total reserved land, m2
(2)
11,860 (84.32% for CF
and 15.68% for DF) 12,000 12,860 12,000 Combined
Land-capacity
fragmentation 4.09∙10-12 1.62∙10-11 1.98∙10-13 1.62∙10-11
Decentralized /
Combined
Average transportation
fleet, trucks 72 85 - 85 Decentralized
Global Warming
P
otential, Mtons
CO2/lifecycle 238 245.4 254.8 292.75 Combined
(1) Capacity over-deployment for the conventional treatment could appear if currently installed capacity is
insufficient to handle the increase in food waste generation and will depend on the capacity of incineration unit
assumed for plant expansion.
(2) Average land use does not take into account the land use by auxiliary installations and equipment (e.g.,
warehouse, office).
The centralized WtEMS shows good performance in the economic KPI. It requires the lowest
CAPEX and OPEX over the planning horizon. In comparison, the deployment of the combined
WtEMS incurs more expenses and, as a result, the Total NPV more than doubled in comparison
with that of the centralized waste management strategy. The decentralized WtEMS requires
important Total NPV related to the deployment of multiple stand-alone facilities and associated
infrastructures for local waste treatment. However, it shows slightly higher revenues than the
centralized and combined cases due to its slightly better transformation efficiency. All three
strategies based on the AD technology perform better than the conventional incineration. More
specifically, the proposed combined WtEMS reduces OPEX by half and more than doubles
revenues in comparison with conventional MSW treatment.
The combined centralized-decentralized case outperforms other waste management strategies
in terms of over-deployment of the average capacity, total land surface reserved for treatment
facilities, and environmental impact. Indeed, the combined WtEMS allows the optimal
combination of large centralized units with micro-scale decentralized facilities. This desirably
leads to the minimum rate of the total capacity of over-deployment and land occupation. In
addition, the combined WtEMS reduces GWP by about 18.7% in comparison with the
conventional strategy and performs slightly better than the purely centralized and decentralized
strategies due to the minor over-deployment in capacity.
Figure 15. MSW management strategies comparison.
Although a decentralized WtEMS achieves the least capacity-land fragmentation, the
combined WtEMS allows an important decrease in capacity-land fragmentation. The capacity-
land fragmentation was reduced by 74.8% in comparison with centralized and conventional
waste treatment strategies. In addition, the combined WtEMS relieved the need for
transportation by reducing the number of fleet by 15.3% in comparison with centralized and
conventional systems. Figure 15 shows some of these tradeoffs graphically.
Although the results above were obtained for a realistic case study, the authors wish to highlight
that these are not practical recommendations for waste treatment system deployment in
Singapore. Such applied recommendations must be defined with strong implication of
government authorities.
5. Conclusions and further contributions
The paper proposed a novel integrated decision-support methodology (DSM) for waste-to-
energy management system (WtEMS) development in an urban environment. It made an
important advancement toward segregation of MSW sources and modelling of their
distributions across large urban territories. It proposed a WtEMS design optimization
methodology accounting for multi-level candidate locations (e.g., at building, district, and
global city levels) for facilities combining various treatment technologies of different
capacities. The proposed methodology provides the optimization schedule for waste treatment
capacity deployment over a large planning horizon together with optimal waste allocation
schedule for different time periods. It provides a multi-criteria evaluation framework helping
to assess optimal WtEMS design using not only economic criteria, but also environmental and
social aspects important for urban planning.
The proposed DSM was tested using a case study for food waste management in Singapore
using publicly available information and considering the deployment of a combined
centralized-decentralized WtEMS. To identify a sustainable food waste management strategy,
the promising technologies of micro- and large-scale AD were successfully tested under real
urban conditions and have been considered for deployment. The scale effect for different
installations was accounted for in two ways. On the one hand, the optimization methodology
explicitly defines models for the decentralized (on-site) technologies of small capacities and
the equivalent centralized (off-site) technologies of large treatment capacity. The economic and
technical parameters for the technology were established based on the peculiarities of the real
installations of different capacities ranges. On the other hand, the decrease in the installation
cost with the increase of treatment unit capacities was accounted for through the economy of
scale (EoS) for both decentralized and centralized facilities. The capacity utilization was
indirectly accounted for in the objective function through the costs and revenues formulation.
In this view, the optimization model naturally tends to maximize the capacity utilization of
each installation in order to increase their revenues from electricity recovery and to avoid
investments for new facilities deployment. The performance of the proposed combined
WtEMS was compared with purely centralized, decentralized, and conventional MSM
management strategies. The results show that the combined WtEMS reduced total operational
expenses by about 50% and increased revenues from electricity recovery almost two-fold in
comparison with conventional MSW management. It also allowed more optimal land use (i.e.,
capacity-land fragmentation was reduced by 74.8%) and reduced the required transportation
fleet by 15.3% in comparison with conventional MSW systems. The global warming potential
(GWP) was improved by about 18.7%.
Future developments were discussed around major topics related to MSW modelling,
optimization of WtEMS deployment, and assessment. The design of cost-efficient and
sustainable waste treatment infrastructure requires clear segregation of MSW into categories
(e.g., food waste, paper/cardboard, horticultural waste, etc.), estimation of their generation
amounts, and distribution across large urban territories. The MSW generation depends on
various demographic and economic variables. Moreover, MSW amounts, categories, and
generation schedules in different urban subzones can be influenced by various factors, e.g.,
social attractiveness of urban subzones. A possible axis for future research will be devoted to
the development of an explicit MSW generation meta-model by connecting MSW categories
with specific demographic and economic variables. Following the analysis in Section 4.1.2,
different urban development indicators, e.g., centrality index, can be explored for MSW
distribution modelling.
The preselection of MSW treatment technologies to be considered for possible deployment is
crucial. This choice is highly dependent on the properties and composition of waste, economic
parameters, treatment and resource recovery efficiency, and environmental factors. As per
usual practice, the decisions regarding deployed technologies are guided not only by the
facilities’ CAPEX and OPEX, but also by the amounts of consumed and generated resources
and their respective costs. In addition, the importance of other performance criteria, analyzed
in Section 4.3, could significantly rise in future decades and require careful consideration for
WtEMS design. To provide an adequate decision-support tool, an optimization methodology
must enable realistic modelling of environmental and operational factors, but also of related
uncertainties. The integration of uncertainties in the operational and environmental conditions
will require further development and integration with the proposed optimization methodology.
This can be considered by integrating the current deterministic optimization model with data-
driven real options analysis, supporting flexible and adaptable deployment strategies in the face
of uncertainty (Cardin et al. 2017; Caunhye & Cardin 2017). Moreover, different decision rules
can be tied with different uncertainty conditions to trigger different deployment options, i.e.,
system expansion with amounts of waste generation, technological shift with advancements in
technology development, and system reconfiguration with changes in resources prices. This
will allow integrating emerging technologies with more attractive economic, technical, and
environmental performance into the MSW treatment deployment schedule, as well as
accounting for treatment plant deterioration and decommissioning.
The paper provides deployment results for a combined centralized-decentralized WtEMS and
analyzes the main advantages of this strategy. Optimal results show that the proportion of CF
and DF deployment under EoS = 0.8 represents 84.32% and 15.68%, respectively. However,
possible conversion of non-economic criteria, evaluated in Section 4.3, into additional
objective functions for the multi-objective optimization can change the CF/DF proportion. The
WtEMS deployment optimization based not only on economic, but also on environmental (e.g.,
GWP) and operational (e.g., fleet size and land use) objectives, could lead to an increase in
installed capacity at decentralized candidate sites. In this view, future work can be related to
the exploration of optimal equilibrium between CF and DF under different optimization
objectives. Under current incentives for high recycling rates, the major factor selecting
appropriate type of waste treatment technology remains the type of MSW sources. Indeed, to
ensure coverage of the major recycling pathways, the appropriate technology must be identified
for each MSW source. Furthermore, to improve the recycling outcomes, different symbiotic
cross-relations (i.e., closed material loops) can be considered between major pathways when
the resource recovered from treatment of one WMS source can be used in another MSW source
technology or urban activity (Geng et al. 2010). In this view, DF enables treating MSW locally
and closer to final consumers of the recovered resources. However, urban symbiosis brings not
only opportunities but also bottlenecks. One of them is that it requires involvement from
various stakeholders possibly holding conflicting objectives, which makes the overall decision-
making process less straightforward and more complex (Kayakutlu et al. 2017). In this view,
other techniques like graph theory (Melese et al. 2017) or agent-based modelling (He et al.
2017) may be required to model complex decisions in a multi-stakeholder setting.
However, under current incentives for high recycling rates the major factor for this selection
remains number and type of MSW sources identified in the territory. Indeed, to ensure the
major recycling pathways, the appropriate technology must be identified for each MSW source.
Further, to improve recycling outcomes, different symbiotic cross-relations can be considered
between major pathways when the resource recovered from treatment of one WMS source can
be used in another MSW source technology or urban activity (Geng et al. 2010). This joins the
above conclusion on the need of deeper exploration of urban symbiotic relations applied to
MSW treatment.
The integration of decentralized technologies into an urban environment leads to
reconsideration of urban planning strategies and the increased importance of social cohesion
and acceptance (Adil & Ko 2016). To ensure the sustainability of WtEMS, a proper design
approach integrate different perspectives, including environmental and social considerations
(Chong et al. 2016). In this view, one of the future developments relates to the expansion of
the multi-criteria assessment framework in order to include indicators related to social safety.
Despite the attempt of (Yu et al. 2012) to address social risk issues, the proposed methodology
still contains some bottlenecks (e.g., equal weighting factors to economic and risk objectives
that made the cost dominant and social risk effect approaching zero in the decision framework).
In this view, the consideration of social factors requires an appropriate analysis before a more
adequate evaluation framework can be proposed.
A more detailed analysis of the empirical deployment strategy and on field diagnostic test can
be done through the application of the developed DSM for an MSW management pilot project.
Such a pilot project, selected to play the role of the representative urban context on the
restricted territory perimeter, enables simultaneous development, test, and implementation of
the technological developments for waste treatment and data-driven decision-support
methodologies. This work is foreseen for the next phase of the current research project.
6. Acknowledgements
This research is supported by the National Research Foundation and the Prime Minister’s
Office of Singapore under its Campus for Research Excellence and Technological Enterprise
(CREATE) programme. The authors acknowledge the valuable inputs from Yeo Teck Kian on
MSW modelling and Tong Huanhuan for sharing her expertise on waste treatment processes.
The authors also wish to thank anonymous referees for their scrupulous review of the work and
for the comments provided, which have helped improving the paper significantly.
Appendix A. Optimization model input
Table A1. Overview of food waste treatment technologies.
Technology
Category
Resource output
Resource
requirements
Advantages
Capacity(1)
CAPEX
O&M
Manpower
Carbon emissions(2)
Waste-to-
Energy
Incineration
(Benchmark)
(Cook 2014)
Thermal
processing
Energy positive
Electricity ~ 130 kWh/ton
of processed MSW;
Ashes 15 – 20% of MSW
by weight
Electricity ~
70 kWh/ton
of processed
MSW (Khoo
et al. 2010)
Suitable for energy
generation for urban
usage, close location
to municipalities
facilitating MSW
transportatio
n and
recovered electricity
supply
Large
capacity: ~
500 – 3,000
ton of MSW
per day
150 tons/day
~
$650/ton
of annual
capacity
35,6
M$/unit
Regular on-going
costs associated
with operation,
cleaning and ash
removal
$900,000/year per
unit
Required
trained
operating staff
– requires
around 50
staff to operate
$900,000/year
per unit
5.93∙105 g CO
2
equivalent/ton of
treated waste (Khoo
et al. 2010)
Wet/Dry
Anaerobic
Digestion
(AD) (Cook
2014;
Sanscartier
et al. 2012;
Leffertstra
2003)
Biological
digestion +
thermal
processing
Energy positive
70% of CH4 converted into
electricity ~ 260 kWh/ton
of processed waste (Khoo
et al. 2010);
Organic solid sludge (can
be further processed into
compost);
Waste water (may require
further treatment)
Electricity ~
32 kWh/ton
of processed
MSW (Khoo
et al. 2010)
Qualified as a
versatile and
adaptable to
different
applicati
ons and
sizes; viable and
economically
feasible technology
for large application
Small and
medium
capacity >
50 ton of
MSW per
day
~ $110–
150 /ton of
annual
capacity
1.08
M$/unit
Regular
maintenance
required on all
machines, costs
associated with
removal of
digested material
$360,000/year per
unit
Large
industrial
units
requires >20
staff to operate
$360,000/year
per unit
0.2∙105 g CO
2
equivalent/ton of
treated waste (Khoo
et al. 2010)
Micro-scale
Anaerobic
Digestion
(Walker et
al. 2017;
Yaman et al.
2013)
Biological
digestion +
thermal
processing
Energy positive
Net electricity output for
the whole site - 271
kWh/ton
Electricity ~
117 kWh/ton
(without
considering
logging
system)
Perfect for small and
medium size urban
applications (e.g.,
restaurants)
Small
capacity: >
20 m3
digester
equivalent
to about 1
ton/day
~ $345/ton
of annual
capacity
0.125
M$/unit
Regular technical
servicing, removal
of residues
$4300
/year per
unit
Almost
automated
charging
$4300/year
per unit
0.2∙105 g CO
2
equivalent/ton of
treated waste (Khoo
et al. 2010)
(1) CF units capacities have been downscaled to fit with total food waste generation in WGS of Figure 9.
(2) Process emissions from (Khoo et al. 2010) have been converted into CO2 emissions equivalent to estimate their Global Warming Potential (IPCC 2007).
References
Abbasi, M. & El Hanandeh, A., 2016. Forecasting municipal solid waste generation using
artificial intelligence modelling approaches. Waste management (New York, N.Y.), 56, pp.13–
22.
Adamovic, V.M., Antanasijevic, D.Z., Ristic, M., Peric-Grujic, A.A. & Pocajt, V. V., 2017.
Prediction of municipal solid waste generation using artificial neural network approach
enhanced by structural break analysis. Environmental Science and Pollution Research, 24,
pp.299 – 311.
Adil, A.M. & Ko, Y., 2016. Socio-technical evolution of Decentralized Energy Systems: A
critical review and implications for urban planning and policy. Renewable and Sustainable
Energy Reviews, 57, pp.1025–1037. Available at:
http://dx.doi.org/10.1016/j.rser.2015.12.079.
Bhunia, P., 2018. S$10.8 million Environmental Robotics Programme and Closing the Waste
Loop research funding initiative from NEA Singapore. OpenGovAsia, pp.4–5. Available at:
https://www.opengovasia.com/articles/s-10-8-million-environmental-robotics-programme-
and-closing-the-waste-loop-research-funding-initiative-from-nea-singapore [Accessed
February 27, 2018].
Cardin, M., Xie, Q., Ng, T.S., Wang, S. & Hu, J., 2017. An approach for analyzing and
managing flexibility in engineering systems design based on decision rules and multistage
stochastic programming. IISE, 49(1), pp.1–12.
Caunhye, A.M. & Cardin, M., 2017. An approach based on robust optimization and decision
rules for analyzing real options in engineering systems design. IISE Transactions, 49(8),
pp.753–767.
Chainey, R., 2015. Which countries waste the most food? World Economic Forum, pp.1–3.
Available at: https://www.weforum.org/agenda/2015/08/which-countries-waste-the-most-
food/ [Accessed August 2, 2017].
Chen, M., Giannis, A. & Wang, J.-Y., 2012. Application of system dynamics model for
municipal solid waste generation and lanfill capacity evaluaiton in Singapore. The Macrotheme
Review, 1, pp.101–114.
Chong, Y.T., Teo, K.M. & Tang, L.C., 2016. A lifecycle-based sustainability indicator
framework for waste-to-energy systems and a proposed metric of sustainability. Renewable
and Sustainable Energy Reviews, 56, pp.797–809. Available at:
http://dx.doi.org/10.1016/j.rser.2015.11.036.
Cook, P.R.M., 2014. An Analysis of New and Emerging Food Waste Recycling Technologies
and Opportunities for Application, New York, USA.
Dai, C., Li, Y.P. & Huang, G.H., 2011. A two-stage support-vector-regression optimization
model for municipal solid waste management - A case study of Beijing, China. Journal of
Environmental Management, 92(12), pp.3023–3037.
Demetriou, D., 2014. Chapter 2. Land Fragmentation. In The Development of an Integrated
Planning and Decision Support System (IPDSS) for Land Consolidation. pp. 1–340.
DIRECTIVE 2008/98/EC, 2008. Waste and repealing certain directives., Brussels: The
European Parliament and the Council. Available at: http://eur-lex.europa.eu/legal-
content/EN/TXT/PDF/?uri=CELEX:32008L0098&from=EN.
Dunnebeil, F. & Lambrecht, U., 2012. Fuel efficiency and emissions of trucks in Germany. An
overview,
Ejaz, N., Akhtar, N., Nisar, H. & Naeem, U.A., 2010. Environmental impacts of improper solid
waste management in developing countries : a case study of Rawalpindi City. Transactions on
Ecology and the Environment, 142, pp.379–387.
EMA, 2017. Electricity Tarifs Singapore. Available at:
https://www.ema.gov.sg/Residential_Electricity_Tariffs.aspx [Accessed June 27, 2017].
Fitzgerald, M., 2016. Data-driven City Management. A Close Look at Amsterdam’s Smart City
Initiative. MITSLoan Management Review, pp.1–13. Available at:
https://sloanreview.mit.edu/case-study/data-driven-city-management/ [Accessed February 28,
2018].
Goren, S. & Ozdemir, F., 2010. Regulation of waste and waste management in Turkey. Waste
Management & Research, 29(4), pp.433–441.
Greentumble, 2016. Countries with the Most Sofisticated Waste Management. Greentumble.
Available at: https://greentumble.com/countries-with-the-most-sophisticated-waste-
management/ [Accessed February 26, 2018].
Halder, P.K., Paul, N., Joardder, M.U.H. & Sarker, M., 2015. Energy scarcity and potential of
renewable energy in Bangladesh. Renewable and Sustainable Energy Reviews, 51, pp.1636–
1649. Available at: http://dx.doi.org/10.1016/j.rser.2015.07.069.
Haldi, J. & Whitcomb, D., 1967. Economies of Scale in Industrial Plants. Journal of Political
Economy, 75(4), pp.373–385.
He, Z., Xiong, J., Sheng, T., Fan, B. & Shoemaker, C.A., 2017. Managing competitive
municipal solid waste treatment systems : An agent-based approach. European Journal of
Operational Research, 263(3), pp.1063–1077. Available at:
http://dx.doi.org/10.1016/j.ejor.2017.05.028.
Hoornweg, D. & Bhada-Tata, P., 2012. What a waste: A Global Review of Solid Waste
Management. The World Bank.
Ichimura, M., 2003. Urbanization , Urban Environment and Land Use : Challenges and
Opportunities An Issue Paper. , (January), pp.1–14.
IPCC, 2007. Direct Global Warming Potentials. Intergovernmental Panel on Climate Change.
Available at: https://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch2s2-10-2.html
[Accessed August 7, 2017].
Ivis, F., 2006. Calculating Geographic Distance : Concepts and Methods. In NEGUS. pp. 1–
10.
Izabelanair, 2017. Giant LEAP for micro-scale AD. Food Science and Technology, pp.1–5.
Available at: http://www.fstjournal.org/features/31-3/micro-scale-ad [Accessed March 1,
2018].
Jacobi, P.R. & Besen, G.R., 2011. Solid Waste Management in Sao Paulo: The challenges of
sustainability. Estudos Avancados, 25(71), pp.135–158.
JESC, 2014. History and Current State of Waste Management in Japan. , pp.1–32. Available
at: https://www.env.go.jp/en/recycle/smcs/attach/hcswm.pdf.
JTC, 2016. Land for General Industries. , pp.1–6. Available at:
http://www.jtc.gov.sg/industrial-land-and-space/pages/land-for-general-industries.aspx
[Accessed November 18, 2016].
Kayakutlu, G., Daim, T., Kunt, M., Altay, A. & Suharto, Y., 2017. Scenarios for regional waste
management. Renewable and Sustainable Energy Reviews, 74(April 2016), pp.1323–1335.
Available at: http://dx.doi.org/10.1016/j.rser.2016.11.147.
Kéita, M., 2001. Building partnerships for urban waste management in Bamako. In “Making
Decentralisation Work” programme (MDW), Drylands Programme, IIED. pp. 1–21.
Keser, S., Duzgun, S. & Aksoy, A., 2012. Application of spatial and non-spatial data analysis
in determination of the factors that impact municipal solid waste generation rates in Turkey.
Waste Management, 32(3), pp.359–371.
Khoo, H.H., Lim, T.Z. & Tan, R.B.H., 2010. Food waste conversion options in Singapore:
Environmental impacts based on an LCA perspective. Science of the Total Environment,
408(6), pp.1367–1373. Available at: http://dx.doi.org/10.1016/j.scitotenv.2009.10.072.
Korai, M.S., Mahar, R.B. & Uqaili, M.A., 2017. The feasibility of municipal solid waste for
energy generation and its existing management practices in Pakistan. Renewable and
Sustainable Energy Reviews, 72(December 2016), pp.338–353. Available at:
http://dx.doi.org/10.1016/j.rser.2017.01.051.
Lam, L., 2017. Singapore Budget 2017: Singapore to implement carbon tax from 2019; diesel
taxes restructured. The Straits Times, pp.1–5. Available at:
http://www.straitstimes.com/singapore/environment/singapore-budget-2017-singapore-to-
implement-carbon-tax-from-2019-2-vehicle [Accessed August 7, 1BC].
Lebersorger, S. & Beigl, P., 2011. Municipal solid waste generation in municipalities:
Quantifying impacts of household structure, commercial waste and domestic fuel. Waste
Management, 31(9–10), pp.1907–1915.
Lee, C.K.M., Yeung, C.L., Xiong, Z.R. & Chung, S.H., 2016. A mathematical model for
municipal solid waste management - A case study in Hong Kong. Waste Management, 58,
pp.430–441.
Leffertstra, H., 2003. Anaerobic Digestion - turning organic waste into energy and fertilizer,
Zaragoza, Spain.
Li, Z.S., Fu, H.Z. & Qu, X.Y., 2011. Estimating municipal solid waste generation by different
activities and various resident groups: A case study of Beijing. Science of the Total
Environment, 409(20), pp.4406–4414.
Mahadevan, B., 2007. Operations Management : Theory and Practice Pearson Education, ed.,
Manne, A.S., 1967. Investments for Capacity Expansion: Size, Location, and Time-Phasing,
MIT Press, Cambridge, MA (USA).
Melese, Y., Heijnen, P., Stikkelman, R. & Herder, P., 2017. International Journal of
Greenhouse Gas Control An approach for fl exible design of infrastructure networks via a risk
sharing contract : The case of CO2 transport infrastructure. International Journal of Greenhouse
Gas Control, 63(March), pp.401–411. Available at:
http://dx.doi.org/10.1016/j.ijggc.2017.06.006.
Menikpura, S.N.M., Sang-Arun, J. & Bengtsson, M., 2013. Integrated Solid Waste
Management: An approach for enhancing climate co-benefits through resource recovery.
Journal of Cleaner Production, 58, pp.34–42.
Minciardi, R., Paolucci, M., Robba, M. & Sacile, R., 2008. Multi-objective optimization of
solid waste flows: Environmentally sustainable strategies for municipalities. Waste
Management, 28(11), pp.2202–2212.
Mirdar Harijani, A., Mansour, S., Karimi, B. & Lee, C.-G., 2017. Multi-period sustainable and
integrated recycling network for municipal solid waste – A case study in Tehran. Journal of
Cleaner Production, 151, pp.96–108.
Morrissey, A.J. & Browne, J., 2004. Waste management models and their application to
sustainable waste management. Waste Management, 24, pp.297–308.
NEA, 2016. Waste statistics and overall recycling. Available at:
http://www.nea.gov.sg/energy-waste/waste-management/waste-statistics-and-overall-
recycling [Accessed February 14, 2017].
Nichols, W., 2018. UK’s “largest” on-site anaerobic digestion plant comes online. Business
Green, pp.2–4. Available at: https://www.businessgreen.com/bg/news/2335484/uks-largest-
on-site-anaerobic-digestion-plant-comes-online [Accessed March 2, 2018].
Nilsson, P. & Christensen, T.H., 2011. Waste Collection: Systems and Organization. In T. H.
Christensen, ed. Solid Waste Technology & Management. Blackwell Publishing Ltd., pp. 277–
295.
Rentizelas, A.A., Tolis, A.I. & Tatsiopoulos, I.P., 2014. Combined Municipal Solid Waste and
biomass system optimization for district energy applications. Waste Management, 34(1),
pp.36–48.
Sanscartier, D., Maclean, H.L. & Saville, B., 2012. Electricity Production from Anaerobic
Digestion of Household Organic Waste in Ontario: Techno-Economic and GHG Emission
Analyses. Environmental Science and Technology, 46, pp.1233–1242.
Santibanez-Aguilar, J.E., Martinez-Gomez, J., Ponce-Ortega, J.M., Napoles-Revera, F., Serna-
Gonzalez, M., Gonzalez-Campos, J.B. & El-Halwagi, M.M., 2015. Optimal planning for the
reuse of municipal solid waste considering economic, environmental, and safety objectives.
Americam Institute of Chemical Engineers Journal, 61(6), pp.1881–1899.
Tong, H. & Tong, Y.-W., 2016. Life cycle assessment of Food Waste Management Options in
Singapore. In 15th International Conference on Sustainable Energy Technologies 19 - 22 of
July. Singapore.
UN, 2011. Shanghai Manual - A Guide for Sustainable Urban Development of the 21st
Century, Shanghai, China.
UN, 2014. World Urbanization Prospectives. United Nations. Available at:
https://esa.un.org/unpd/wup/Publications/Files/WUP2014-Report.pdf.
UN Habitat, 2016. Urbanization and development. Emerging Futures. UN Habitat.
URA, 2013. Concept Palan 2011 and MND’s Land Use Plan. Urban Development Auhtority,
pp.1–2. Available at: https://www.ura.gov.sg/uol/concept-plan.aspx?p1=View-Concept-
Plan&p2=Land-Use-Plan-2013 [Accessed March 10, 2017].
Vitorino de Souza Melaré, A., Montenegro González, S., Faceli, K. & Casadei, V., 2017.
Technologies and decision support systems to aid solid-waste management: a systematic
review. Waste Management, 59, pp.567–584.
Walker, M., Theaker, H., Yaman, R., Poggio, D., Nimmo, W., Bywater, A., Blanch, G. &
Pourkashanian, M., 2017. Assessment of micro-scale anaerobic digestion for management of
urban organic waste : A case study in London , UK, Elsevier Ltd. Available at:
http://dx.doi.org/10.1016/j.wasman.2017.01.036.
WEF, 2016. The New Plastics Economy Rethinking the future of plastics, Cologny/Geneva
Switzerland.
Wei, Y., Li, J., Shi, D., Liu, G., Zhao, Y. & Shimaoka, T., 2017. Environmental challenges
impeding the composting of biodegradable municipal solid waste: A critical review. Resources,
Conservation and Recycling, 122, pp.51–65.
Wen, Z., Wang, Y. & De Clercq, D., 2016. What is the true value of food waste? A case study
of technology integration in urban food waste treatment in Suzhou City, China. Journal of
Cleaner Production, 118, pp.88–96.
WMR, 2015. South Korea Legislates Towards a Zero Waste Society. Waste Management
Review, pp.1–5. Available at: http://wastemanagementreview.com.au/south-korea-legislates-
towards-a-zero-waste-society/ [Accessed February 26, 2018].
Xiong, J., Ng, T.S.A. & Wang, S., 2016. An optimization model for economic feasibility
analysis and design of decentralized waste-to-energy systems. Energy, 101, pp.239–251.
Yadav, V., Bhurjee, A.K., Karmakar, S. & Dikshit, A.K., 2017. A facility location model for
municipal solid waste management system under uncertain environment. Science of The Total
Environment, (October 2016), pp.1–12.
Yaman, R., Bywater, A., Murcott, J., Blanch, G., Neylan, D., Poggio, D., Walker, M., Andrews,
C. & Kibber, C., 2013. Optimising Urban Micro AD Networks, Banbury, UK.
Yu, H., Solvang, W.D. & Yuan, S., 2012. A Multi-objective Decision Support System for
Simulation and Optimization of Municipal Solid Waste Management System. In 3rd IEEE
International Conference on Cognitive Infocommunications. Kosice, Slovakia, pp. 193–199.
Yukalang, N., Clarke, B. & Ross, K., 2017. Barriers to Effective Municipal Solid Waste
Management in a Rapidly Urbanizing Area in Thailand. Environmental Research and Public
Health, 14.
Zhong, C., Schla, M., Mu, S., Batty, M. & Schmitt, G., 2017. Revealing centrality in the spatial
structure of cities from human activity patterns. Urban Studies, 54(2), pp.437–455.
