- Research
- Open Access

# Robust interval quadratic programming and its application to waste management under uncertainty

- Yongping Li
^{1}Email author and - Guohe Huang
^{2}

**1**:7

https://doi.org/10.1186/2193-2697-1-7

© Li and Huang; licensee Springer. 2012

**Received:**30 March 2012**Accepted:**27 June 2012**Published:**1 September 2012

## Abstract

### Background

No country has ever experienced as large or as fast an increase in municipal solid waste (MSW) quantities that China is now facing. The MSW generation rate in the City of Changchun continues to increase since it has been encountered swift urbanization, industrialization and economic development during the past decades.

### Results

In this study, a robust interval quadratic programming method is developed for the planning of MSW management in the City of Changchun, China. The developed method can not only tackle uncertainties expressed as interval values, fuzzy sets, and their combinations, but also reflect economies-of-scale effects on waste disposal of cost.

### Conclusions

The results are valuable for helping governmental officials more intuitive to know some basic situation, such as optimal waste-flow allocation, waste-flow routing, facility-capacity expansion, and system cost over the planning horizon. Results can also be used to generate decisions for supporting the city’s long-term MSW management and planning, and thus help managers to identify desired MSW management policies in association with cost minimization under uncertainty.

## Keywords

- Environment
- Management
- Fuzzy sets
- Policy analysis
- Quadratic programming
- Solid waste
- Uncertainty

## Background

For decades, massive urbanization and rapid development of global urban economy have increased municipal solid waste (MSW) generation rate. MSW management is crucial for environmental protection and public health and has become a major challenge confronted by the world, particularly for many urban regions of developing countries. For example, global waste generation rate has nearly doubled since 1960, from 2.7 to 4.4 pounds per capital per day, while more than 70% of MSW generated is disposed of at landfills (USEPA, 2007). Due to the waste management hierarchy, one of the greatest challenges that decision makers face is to figure out how to diversify the treatment options, increase the reliability of infrastructure systems, and leverage the redistribution of waste streams among landfilling, incineration, compost, recycling and other facilities (Chang and Davila, 2007). Consequently, many urban regions and countries have established various kinds of laws and regulations to enhance MSW management and planning. A large number of optimization techniques have been proposed for supporting decisions of MSW management and evaluating relevant operation and investment policies; they involve linear, dynamic, integer and multiobjective programming methods (Baetz, 1990; Lund et al. 1994; Masui et al. 2000; Kollikkathara et al. 2010; Cao and Huang, 2011).

The complexity of planning MSW management can be significantly compounded by the fact that many system components cannot be known with certainty beforehand. Hence, in many real-world applications, the quality of information produced by deterministic optimization techniques can be rendered highly questionable when the input data cannot be expressed with precision (Li and Huang, 2006; Li et al. 2011). The complexities could be further amplified not only by interactions among the uncertain parameters but also through additional economic implications. Such complexities have placed many MSW management problems beyond deterministic programming methods. Data imprecision can be addressed through optimization approaches such as fuzzy, stochastic and interval mathematical programming (Wilson and Baetz, 2001ab; Huang et al. 2005ab; El Hanandeh and El-Zein, 2010; Fan and Huang, 2012). Fuzzy mathematical programming considers uncertainties as fuzzy sets, and is effective in reflecting ambiguity and vagueness in resource availabilities. Robust programming based on the concept of fuzzy interval was able to deal with ambiguous coefficients in the optimization model and reflect the vague information of decision makers’ implicit knowledge (Inuiguchi and Sakawa, 1998; Dubois et al. 2001; Ben-Tal and Nemirovski, 2002; Li et al. 2008). Moreover, this method delimits an uncertain decision space by specifying uncertainties through dimensional enlargement of the original fuzzy constraints, and thus enhances the robustness of the optimization process.

However, economies of scale (EOS) may affect the cost coefficients in a mathematical programming problem and make the relevant objective function nonlinear; difficulties arise due to system nonlinearities when these methods are applied to MSW management problems. Even if a nonlinear model was formulated, the modelers would prefer to convert it into equivalent linear forms and use linear methods for generating solutions (Ko and Chang, 2008; Li et al. 2009). Nonlinear mathematical models for real-world applications are hampered by a general lack of appropriate modeling solutions for effectively addressing uncertainties and nonlinearities simultaneously. Quadratic programming is useful for reflecting nonlinearity in cost objectives and has global optimum under a number of system conditions (Hillier and Lieberman 1986). Previously, a few studies incorporating uncertainty within MSW planning were reported, through introducing fuzzy and/or interval optimization methods into the quadratic programming framework to reflect uncertainty and non-linearity (Chen and Huang, 2001). Fuzzy-based quadratic programming is incapable of dealing with ambiguous coefficients or decision makers’ vague preferences; interval-based quadratic programming has difficulties in addressing uncertainties presented in terms of probabilistic or possibilistic distributions.

China has experienced a very rapid increase in its economy during the last two decades. However, in the absence of a comprehensively sustainable development scheme, this increase has brought severe environmental issues, such as water resources depletion and pollution, soil erosion, desertification, acid rain, sandstorms, forest depletion, and solid waste pollution. Among them, solid waste is becoming a critical issue, not only in terms of the impacts being created but also in terms of resources being consumed. No country has ever experienced as large or as fast an increase in solid waste quantities that China is now facing. China produces around 29% of the world’s solid waste each year, and with the economy continuing to grow rapidly, it is clear that China bears what may be the heaviest solid waste management burden in the world. It has been estimated that the amounts of industrial waste increased by 10% while at the same time municipal waste increased by 15% per year in China. In 2004, China surpassed the United States becoming as the world’s largest waste generator, and by 2030 China’s annual solid waste quantities will increase by another 150% - growing from about 190 million tons in 2004 to over 480 million tons in 2030 (Su et al. 2009). Therefore, development of systems analysis method for effective managing MSW and thus providing scientific bases for decision makers is desired.

The objective of this study is to develop a robust interval quadratic programming method for the planning of MSW management in the City of Changchun, China. Robust programming method will be incorporated within an interval quadratic programming framework for better accounting for uncertainties and nonlinearities. The developed method will then be applied to a case of long-term waste management planning. It can not only handle uncertainties expressed as fuzzy sets and interval values, but also deal with nonlinearities in the objective function to reflect the effect of EOS on waste management cost. The results can be used for generating a range of decision alternatives under various system conditions, and thus helping managers to identify desired waste-management policies.

## Methods

*x*

_{ j }

*x*

_{ k }(

*j*≠

*k*) terms as follows (Chen and Huang, 2001):

where ${a}_{\mathit{ij}}^{\pm}$, ${b}_{i}^{\pm}$, ${c}_{j}^{\pm}$, ${d}_{j}^{\pm}$and ${x}_{j}^{\pm}$ are interval parameters/variables; the ‘-’ and ‘+’ superscripts represent lower and upper bounds of an interval parameter/variable, respectively. If the quadratic programming problem satisfies the Kuhn-Tucker conditions (Kuhn and Tucker, 1951) or has a concave objective function, it will then have a global optimum. Such a problem can then be transformed into two deterministic submodels that correspond to lower and upper bounds of the objective-function value, based on an interactive algorithm (Chen and Huang, 2001). However, model (1) has difficulties in reflecting uncertainties presented as fuzzy sets; moreover, it may lead to infeasibility when the model’s right-hand-side parameters have large intervals.

**ℜ**}

^{m × n}, B ∈ {

**ℜ**}

^{m x 1}, C ∈ {R}

^{1 × n}, X ∈ {R}

^{n × 1},

**ℜ**denotes a set of fuzzy parameters and variables, R denotes a set of deterministic numbers, and $\u02dc\le $ means fuzzy inequality. Let constraints in (2b) take the following specific form:

*A*

_{ j }(

*j*= 1, 2, …,

*n*) and

*B*are fuzzy subsets, and symbol ⊕ denotes addition of fuzzy subsets. Fuzziness of the decision space is due to uncertainties in coefficients

*A*

_{ j }and

*B*. Letting ${\underset{\u02dc}{U}}_{j}$ and $\underset{\u02dc}{V}$ be base variables imposed by fuzzy subsets

*A*

_{ j }and

*B*, we have:

*μ*

_{ Aj }denotes the possibility of consuming a specific amount of resource by activity

*j*, and

*μ*

_{ B }indicates the possible availability of resource

*B*. Fuzzy subset

*N*can be expressed as the following L-R fuzzy numbers (Dubois and Prade, 1978):

*u*is the mean value of

*N*;

*β*and

*δ*are the left and right spreads, respectively;

*F*

_{ L }and

*F*

_{ R }are the shape functions. For a linear case, fuzzy subset

*N*can be defined as follows:

*N*. According to the concept of level set (fuzzy

*α*-cut) and the representation theorem, constraints in (3) can be represented as follows:

*α*

_{1}≤

*α*

_{2}≤ … ≤

*α*

_{k}≤ 1. Then, for each

*α*

_{ s }(

*s*= 1, 2, …,

*k*), constraints in (6a) become:

*j*= 1, 2, …,

*n*;

*s*= 1, 2, …,

*k*) and ${\underset{\u02dc}{B}}_{{\alpha}_{s}}$constitute convex and non-empty fuzzy sets. Then, fuzzy constraints in (8) can be replaced by the following 2

*k*precise inequalities, where

*k*denotes the number of

*α*-cut levels (Luhandjula and Gupta, 1996):

*t*) and inf(

*t*) denote the superior and inferior limits among set

*t*, respectively. Therefore, for a fuzzy robust linear program with

*m*fuzzy constraints, the decision space for problem (2b) can be delimited by the following deterministic constraints based on the relations as defined in (6a) to (9f):

*m*imprecise constraints into

*2 km*precise inclusive ones that correspond to

*k*α-cut levels, such that robustness of the optimization process could be enhanced (Li et al. 2008). However, the robust programming method may become inapplicable when uncertainties and nonlinearities exist in the objective function; besides, it has difficulties in dealing with uncertainties that cannot be presented as membership functions. Therefore, robust programming will be incorporated within an interval quadratic programming framework in response to the above challenges. This leads to a robust interval quadratic programming model as follows:

where ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{\pm}$ and ${\underset{\u02dc}{b}}_{\phantom{\rule{0.12em}{0ex}}t}^{\pm}$ denote a set of intervals with vague lower and upper bounds. Assume that no intersection exists between the fuzzy sets at the two bounds. Letting ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{-}$ and ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{+}$ be lower and upper bounds of ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{\pm}$, we have ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{\pm}=\left[{\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{-}\text{,}\phantom{\rule{0.75em}{0ex}}{\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{+}\right]$.

*f*

^{ – }and

*f*

^{ + }. Consequently, two situations need to be considered. When cost coefficients ${c}_{j}^{\pm}$ and ${d}_{j}^{\pm}$ have different signs (i.e. when ${c}_{j}^{\pm}\ge 0$, then ${d}_{j}^{\pm}$ would be ≤ 0, and vice versa), the optimal bound distribution for ${x}_{j}^{\pm}$ can be identified based on a derivative algorithm proposed by Chen and Huang (2001). Firstly, let all left- and/or right-hand-side coefficients be equal to their mid-values. Then, model (11) can be converted into a robust deterministic quadratic programming problem as follows:

where ${\left({c}_{j}\right)}_{\text{mv}}$, ${\left({d}_{j}\right)}_{\text{mv}}$, ${\left({a}_{\mathit{rj}}\right)}_{\text{mv}}$and ${\left({b}_{r}\right)}_{\text{mv}}$are mid-values of ${c}_{j}^{\pm}$, ${d}_{j}^{\pm}$, ${a}_{\mathit{rj}}^{\pm}$ and ${b}_{r}^{\pm}$ [e.g. ${\left({c}_{j}\right)}_{\text{mv}}=({c}_{j}^{-}+{c}_{j}^{+})/2$; ${\left({\underset{\xaf}{a}}_{\phantom{\rule{0.24em}{0ex}}tj}^{s}\right)}_{\text{mv}}$, ${\left({\overline{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{s}\right)}_{\text{mv}}$, ${\left({\underset{\xaf}{b}}_{\phantom{\rule{0.62em}{0ex}}t}^{s}\right)}_{\text{mv}}$and ${\left({\overline{b}}_{\phantom{\rule{0.12em}{0ex}}t}^{s}\right)}_{\text{mv}}$are mid-values of ${\underset{\u02dc}{a}}_{\phantom{\rule{0.12em}{0ex}}tj}^{\pm}$ and ${\underset{\u02dc}{b}}_{\phantom{\rule{0.12em}{0ex}}i}^{\pm}$ [e.g. ${\left({\underset{\xaf}{a}}_{\phantom{\rule{0.24em}{0ex}}tj}^{s}\right)}_{\text{mv}}=({\underset{\xaf}{a}}_{\phantom{\rule{0.24em}{0ex}}tj}^{-}s+{\underset{\xaf}{a}}_{\phantom{\rule{0.24em}{0ex}}tj}^{+}s)/2$. The solutions for model (12) are${X}_{\text{mv opt}}=\left\{{\left({x}_{j}\right)}_{\text{mv opt}}\right|\forall j\}$, where${\left({x}_{j}\right)}_{\text{mv opt}}\in \left[{x}_{j\phantom{\rule{0.12em}{0ex}}\text{opt}}^{-}\text{,}\phantom{\rule{0.25em}{0ex}}{x}_{j\phantom{\rule{0.12em}{0ex}}\text{opt}}^{+}\right]\text{,}\forall j$.

When criterion (13a) is satisfied, then${x}_{j}^{+}$corresponds to ${f}^{+}$; when criterion (13b) holds, then ${x}_{j}^{-}$corresponds to${f}^{+}$. Assume criterion (13a) is satisfied, we can convert model (11) into two submodels (corresponding to ${f}^{-}$and${f}^{+}$) as follows:

### Submodel (1)

### Submodel (2)

Solving the above two submodels, we can obtain the solutions for model (11): ${x}_{j\phantom{\rule{0.12em}{0ex}}\text{opt}}^{\pm}=\left[{x}_{j\phantom{\rule{0.12em}{0ex}}\text{opt}}^{-}\text{,}\phantom{\rule{0.25em}{0ex}}{x}_{j\phantom{\rule{0.12em}{0ex}}\text{opt}}^{+}\right]\text{,}\forall j$ and${f}_{\text{opt}}^{\pm}=\left[{f}_{\text{opt}}^{-}\text{,}\phantom{\rule{0.25em}{0ex}}{f}_{\text{opt}}^{+}\right]$. When cost coefficients ${c}_{j}^{\pm}$ and ${d}_{j}^{\pm}$ have the same sign (i.e., when ${c}_{j}^{\pm}\ge 0$, then ${d}_{j}^{\pm}$≥ 0, and vice versa), model (11) can be directly transformed into two submodels, which correspond to the lower and upper bounds of the objective function value, respectively. Then, each submodel can be converted into a conventional linear program, provided that condition (7) is satisfied for each fuzzy constraint.

## Case study

The MSW generated typically include paper, yard waste, food waste, plastics, metals, glass, wood and other items. Similar to many other cities in China, the study city mainly relies on the use of landfill for handling its solid waste. The landfill is used directly to satisfy waste disposal demand or alternatively to provide capacity for the other facilities’ residue disposals. Before 2008, approximately 2800 to 3000 tonnes per day of the waste (i.e. over 90% of total residential wastes) were buried at the landfill with simple pretreatment, especially in the areas of rural–urban fringe and countryside. In addition, the amount of residential waste diverted from landfill is still low (i.e. less than 17% of total wastes generated by households); increasing the waste diversion, separation and recovery rate and thus reducing the wastes to landfill are becoming an important goal. In 2009, the city built a waste-to-energy facility with a daily capacity of about 500 tonnes (i.e. the Xinxiang incinerator) to help reduce the amount of wastes that ends up at the landfill. The incinerator can generate approximately 51 million kwh per year and gain profits through the sale of electricity and environmentally-friendly building materials. Over the last ten years, the city has achieved significant improvements in waste management. Regional policies and guidelines for collection, transport, treatment and disposal of municipal and industrial wastes in environmentally safe ways have been established. However, the city has been unable to keep up with the growing demand for waste service coverage, environmental requirements for safe disposal systems, and rationalization of cost-effectiveness in service delivery. The capacities of waste management at regional and central levels, the abilities of the local scientists and technicians, and the environmental protection consciousness of the local communities are required to improve immediately to meet the requirements of environmental protection and regional sustainability.

where *X*_{
t
} is decision variable for waste flow (tonne/day), *X*_{
re
} is a reference waste flow (tonne/day), *C*_{
t
} is the transportation/treatment cost for waste flow *X*_{
t
} ($/tonne), *C*_{
re
} is a known cost for reference waste flow *X*_{
re
} ($/tonne), and *m* is an EOS exponent (0 < *m* < 1).

**Waste-generation rate**

District | k = 1 | k = 2 | k = 3 |
---|---|---|---|

D1 | [590, 670] | [610, 700] | [638, 722] |

D2 | [765, 870] | [795, 907] | [825, 941] |

D3 | [630, 719] | [656, 747] | [683, 776] |

D4 | [486, 552] | [511, 574] | [526, 597] |

D5 | [403, 457] | [419, 476] | [436, 494] |

**Shipping cost for waste to facilities**

k = 1 | k = 2 | k = 3 | |
---|---|---|---|

Waste transport cost (to landfill) ($/t) | |||

D1 | [8.44-0.0073x, 10.34-0.0090x] | [8.87-0.0076x, 10.86-0.0093x] | [9.57-0.0082x, 11.73-0.0100x] |

D2 | [6.74-0.0059x, 9.41-0.0082x] | [7.07-0.0061x, 9.88-0.0085x] | [7.64-0.0065x, 10.67-0.0091x] |

D3 | [5.60-0.0049x, 8.40-0.0073x] | [5.88-0.0051x, 8.82-0.0076x] | [6.35-0.0054x, 9.52-0.0081x] |

D4 | [7.30-0.0063x, 9.59-0.0083x] | [7.67-0.0066x, 10.07-0.0087x] | [8.28-0.0071x, 10.88-0.0093x] |

D5 | [0.98-0.0009x, 6.74-0.0058x] | [1.03-0.0009x, 7.07-0.0061x] | [1.11-0.0009x, 7.64-0.0065x] |

BJ | [3.26-0.0028x, 3.75-0.0032x] | [3.42-0.0029x, 3.93-0.0034x] | [3.69-0.0031x, 4.25-0.0036x] |

NG | [2.88-0.0025x, 3.31-0.0029x] | [3.02-0.0026x, 3.47-0.0029x] | [3.26-0.0028x, 3.75-0.0032x] |

CJ | [3.45-0.0030x, 3.98-0.0034x] | [3.62-0.0031x, 4.18-0.0036x] | [3.91-0.0033x, 4.51-0.0038x] |

XM | [2.51-0.0022x, 2.88-0.0025x] | [2.64-0.0022x, 3.03-0.0026x] | [2.85-0.0024x, 3.27-0.0028x] |

HG | [2.49-0.0022x, 2.87-0.0025x] | [2.62-0.0030x, 3.01-0.0026x] | [2.82-0.0024x, 3.25-0.0028x] |

NH | [3.37-0.0029x, 3.88-0.0034x] | [3.54-0.0030x, 4.08-0.0035x] | [3.83-0.0033x, 4.40-0.0037x] |

Waste transport cost (to incinerator) ($/t) | |||

D1 | [3.27-0.0028x, 5.77-0.0050x] | [3.43-0.0029x, 6.06-0.0052x] | [3.71-0.0032x, 6.54-0.0056x] |

D2 | [4.17-0.0036x, 6.26-0.0054x] | [4.38-0.0038x, 6.57-0.0056x] | [4.73-0.0040x, 7.10-0.0060x] |

D3 | [3.45-0.0030x, 6.66-0.0058x] | [3.62-0.0031x, 6.99-0.0061x] | [3.91-0.0033x, 7.55-0.0064x] |

D4 | [1.04-0.0009x, 2.30-0.0020x] | [1.09-0.0009x, 2.42-0.0021x] | [1.17-0.0010x, 2.61-0.0022x] |

D5 | [2.83-0.0025x, 7.15-0.0062x] | [2.97-0.0025x, 7.51-0.0065x] | [3.21-0.0027x, 8.11-0.0069x] |

BJ | [0.48-0.0004x, 0.55-0.0005x] | [0.50-0.0004x, 0.57-0.0005x] | [0.54-0.0005x, 0.62-0.0005x] |

NG | [0.78-0.0007x, 0.89-0.0008x] | [0.82-0.0007x, 0.94-0.0008x] | [0.88-0.0007x, 1.02-0.0009x] |

CJ | [1.92-0.0017x, 2.21-0.0019x] | [2.01-0.0017x, 2.31-0.0020x] | [2.17-0.0018x, 2.50-0.0021x] |

XM | [1.74-0.0015x, 2.01-0.0017x] | [1.83-0.0016x, 2.10-0.0018x] | [1.98-0.0017x, 2.27-0.0019x] |

HG | [1.35-0.0012x, 1.55-0.0013x] | [1.42-0.0012x, 1.63-0.0014x] | [1.53-0.0013x, 1.76-0.0015x] |

NH | [3.37-0.0029x, 3.88-0.0034x] | [3.54-0.0030x, 4.08-0.0035x] | [3.82-0.0033x, 4.40-0.0037x] |

**Transportation cost for waste from district to transfer station ($/t)**

k = 1 | k = 2 | k = 3 | |
---|---|---|---|

D1-BJ | [3.49-0.0030x, 5.38-0.0047x] | [3.66-0.0032x, 5.65-0.0049x] | [3.96-0.0034x, 6.10-0.0052x] |

D2-BJ | [4.73-0.0041x, 6.44-0.0056x] | [4.97-0.0043x, 6.76-0.0058x] | [5.37-0.0046x, 7.30-0.0062x] |

D3-BJ | [3.39-0.0029x, 5.67-0.0049x] | [3.56-0.0031x, 5.96-0.0051x] | [3.84-0.0033x, 6.43-0.0055x] |

D4-BJ | [1.25-0.0019x, 5.36-0.0047x] | [1.31-0.0011x, 3.06-0.0026x] | [1.41-0.0012x, 3.30-0.0028x] |

D5-BJ | [2.24-0.0019x, 5.36-0.0047x] | [2.35-0.0020x, 5.62-0.0048x] | [2.54-0.0022x, 6.08-0.0052x] |

D1-NG | [1.74-0.0015x, 4.42-0.0038x] | [1.83-0.0016x, 4.64-0.0040x] | [2.00-0.0017x, 5.01-0.0043x] |

D2-NG | [1.50-0.0013x, 4.29-0.0037x] | [1.57-0.0013x, 4.51-0.0039x] | [1.70-0.0014x, 4.87-0.0041x] |

D3-NG | [1.25-0.0011x, 4.24-0.0037x] | [1.31-0.0011x, 4.45-0.0038x] | [1.41-0.0012x, 4.80-0.0041x] |

D4-NG | [0.50-0.0004x, 2.05-0.0018x] | [0.52-0.0004x, 2.15-0.0018x] | [0.57-0.0005x, 2.32-0.0020x] |

D5-NG | [1.49-0.0013x, 5.35-0.0046x] | [1.57-0.0013x, 5.61-0.0048x] | [1.70-0.0014x, 6.06-0.0052x] |

D1-CJ | [0.25-0.0002x, 2.91-0.0025x] | [0.26-0.0002x, 3.06-0.0026x] | [0.28-0.0002x, 3.30-0.0028x] |

D2-CJ | [0.50-0.0004x, 3.22-0.0028x] | [0.52-0.0004x, 3.38-0.0029x] | [0.57-0.0005x, 3.65-0.0031x] |

D3-CJ | [2.24-0.0019x, 3.37-0.0031x] | [2.35-0.0020x, 3.54-0.0030x] | [2.54-0.0022x, 3.83-0.0032x] |

D4-CJ | [1.74-0.0015x, 3.52-0.0031x] | [1.83-0.0016x, 3.69-0.0032x] | [1.98-0.0017x, 4.00-0.0034x] |

D5-CJ | [2.99-0.0026x, 6.14-0.0053x] | [3.14-0.0027x, 6.45-0.0055x] | [3.39-0.0029x, 6.97-0.0059x] |

D1-XM | [2.99-0.0026x, 3.22-0.0028x] | [3.14-0.0027x, 3.38-0.0029x] | [3.39-0.0029x, 3.65-0.0031x] |

D2-XM | [1.74-0.0015x, 2.91-0.0025x] | [1.83-0.0016x, 3.06-0.0026x] | [1.98-0.0017x, 3.30-0.0028x] |

D3-XM | [0.54-0.0004x, 2.45-0.0021x] | [0.56-0.0005x, 2.58-0.0022x] | [0.61-0.0005x, 2.78-0.0024x] |

D4-XM | [1.50-0.0015x, 5.65-0.0049x] | [1.57-0.0013x, 2.74-0.0024x] | [1.70-0.0014x, 2.96-0.0025x] |

D5-XM | [1.74-0.0015x, 5.65-0.0049x] | [1.83-0.0016x, 5.94-0.0051x] | [1.98-0.0017x, 6.41-0.0055x] |

D1-HG | [4.99-0.0043x, 5.98-0.0052x] | [5.24-0.0045x, 6.28-0.0054x] | [5.66-0.0048x, 6.78-0.0058x] |

D2-HG | [2.51-0.0022x, 3.36-0.0029x] | [2.64-0.0023x, 3.53-0.0030x] | [2.85-0.0024x, 3.81-0.0032x] |

D3-HG | [1.01-0.0009x, 3.02-0.0026x] | [1.06-0.0009x, 3.17-0.0027x] | [1.14-0.0010x, 3.42-0.0029x] |

D4-HG | [1.26-0.0011x, 2.91-0.0025x] | [1.33-0.0011x, 3.06-0.0026x] | [1.44-0.0012x, 3.30-0.0028x] |

D5-HG | [0.58-0.0005x, 2.15-0.0019x] | [0.61-0.0005x, 2.25-0.0019x] | [0.66-0.0006x, 2.43-0.0021x] |

D1-NH | [4.73-0.0041x, 5.37-0.0047x] | [4.97-0.0043x, 5.64-0.0049x] | [5.37-0.0048x, 6.09-0.0052x] |

D2-NH | [1.75-0.0015x, 3.35-0.0029x] | [1.84-0.0016x, 3.52-0.0030x] | [1.99-0.0017x, 3.80-0.0032x] |

D3-NH | [0.53-0.0004x, 3.22-0.0029x] | [0.55-0.0005x, 3.38-0.0029x] | [0.60-0.0005x, 3.65-0.0031x] |

D4-NH | [5.31-0.0046x, 6.46-0.0056x] | [5.57-0.0048x, 6.78-0.0058x] | [6.02-0.0051x, 7.33-0.0062x] |

D5-NH | [1.74-0.0015x, 5.79-0.0050x] | [1.83-0.0016x, 6.08-0.0052x] | [1.98-0.0017x, 6.56-0.0056x] |

**Technical modeling inputs**

Residue generation rate when waste is shipped from district to incinerator | [0.25, 0.30] |

Residue generation rate when waste is shipped from TR to incinerator | [0.12, 0.15] |

Density factor for waste disposed of at landfill (m | [1.47, 1.82] |

Volume of the existing landfill (10 | [76, 80] |

Daily landfill capacity (t/day) | [1950, 2020] |

Daily incinerator capacity (t/day) | [[440, 470], [490, 520]] |

Safety coefficient for incinerator | [[0.09, 0.13], [0.17, 0.21]] |

Electricity-generation rate when waste is shipped from district to incinerator (KWh/t) | [280, 300] |

Electricity-generation rate when waste is shipped from TR to incinerator (KWh/t) | [370, 400] |

Material recycling rate at TR | [0.16, 0.20] |

Mass loss rate at TR | [0.22, 0.25] |

Maximum expansion capacity for landfill (10 | 18 |

Maximum expansion capacity once time for incinerator (t/day) | 250 |

Maximum expansion capacity once time for transfer station (t/day) | 600 |

The detailed nomenclatures for the variables and parameters are provided in the Appendix. The objective is to minimize the sum of the expenses for collecting, shipping and disposing wastes as well as costs for expanding transfer stations, landfill and incinerator. The constraints define the interrelationships among the decision variables and the waste generation/management conditions. In detail, constraints (17 b) to (17 e) denote that the wastes disposed by each treatment facility (i.e. landfill, incinerator and transfer stations) must not exceed their existing and expanded capacities; constraint (17 f) means that the mass balance of waste flows at transfer stations, where the volume of wastes can be reduced and various useful wastes can be recycled; constraint (17 g) denotes that the waste flows disposed by the waste-management facilities must be over the total waste generation amounts; constraints (17 h) to (17 k) regulate the expansion scales for waste-management facilities; constraint (17 l) stipulates that the decision variables are non-negative.

## Result analysis

## Discussion and conclusion

A robust interval quadratic programming method has been developed through incorporating techniques of robust programming and interval quadratic programming within a general optimization framework. The developed method can not only tackle uncertainties expressed as interval values, fuzzy sets, and their combinations, but also deal with nonlinearities in the objective function such that economies-of-scale effects can be reflected. Furthermore, the uncertain decision space has been delimited through dimensional enlargement of the original fuzzy constraints, leading to enhanced robustness for the optimization process. Compared with the conventional quadratic programming methods, the developed method can address more uncertainties without unrealistic simplifications or information losses, such that the robustnesses of the optimization processes and solutions can be enhanced. The developed method has been applied to planning long-term municipal solid waste (MSW) management in the City of Changchun, China. The results have been generated, which are valuable for helping governmental officials more intuitive to know some basic situation under complex uncertainties, such as optimal waste-flow allocation, waste-flow routing, facility-capacity expansion, and system cost. They can be used to further generate decisions for supporting long-term MSW management and planning activities in the city, and thus help managers to identify desired MSW policies in association with cost minimization under uncertainty.

The results indicate that, in the future 15 years, the city’s majority of wastes would be disposed of at the landfill due to its relatively low operation cost and low capital for facility development/expansion. However, pollutant emissions from landfill site can take a number of forms: gaseous emissions of volatile organic compounds (VOCs), airborne particulate matter and leachate. For example, surface water could be polluted by rainwater flowing through solid waste piles. Groundwater could be contaminated by leachate from landfill sites where solid wastes are disposed of. The polluted surface water and groundwater can further affect the drinking water safety; leachate containing hazardous materials can enter soil and further reside in the agricultural products that make our foods poisonous. Secondly, landfills are the first and/or second largest contribution of methane (CH_{4}) source (e.g. in 2006, the amount of CH_{4} released from landfills was 5985 Gg, occupying 23% of total US anthropogenic methane emissions) (USEPA, 2007). Recently, there is an increasing concern for CH_{4}, as a major greenhouse gas, while its global warming potential is about 23 on a 100-year time horizon (Mor et al. 2006; Chen et al. 2010). Thirdly, conflicts exist in the urban land resources due to the rapid population growth and swift economy development. Particularly, for the City of Changchun, the serious scarcity of land near urban centers leads to waste disposed of at landfill more and more noneconomic. Therefore, issues of land resource consumption, surface water/groundwater contamination, and greenhouse gas effect may imply higher environmental penalties than the savings obtained from waste buried.

The results also indicate that the city has to expand the incinerator to treat its more and more MSW over the planning horizon. Waste incineration can also generate considerable pollutant emissions (e.g. acid gases, metals and various organic compounds) that can present potential human health hazards. Incinerator emissions are complex and depend on the type of waste, the design of the incinerator, combustion conditions, and pollution control equipment. A number of pollutants (e.g. acid gases, metals and various organic compounds, including dioxins) are associated with health hazards and have thus raised serious concerns in the city. Therefore, research efforts focused on vulnerability analysis and risk assessment of human health and ambient environment for the city’s incinerator are desired. Evaluation of the risk effects of air pollution on human health requires a series of assessment activities such as emission and dispersion of pollutants in the atmosphere, exposure of humans to pollutants, and adverse effects of the pollutants on human health; this will be of challenge for many waste managers. In addition, utilization of source-separated collection is one of the key steps in the city’s future MSW management. Source-separated collection begins at the sources of MSW and involves the whole process of collection, transportation, disposal and recycling, which enables waste minimization, resource utilization, and hazardous waste disposal of. This also requires the local government to establish standards and regulations for the source separated collection.

## Appendix

### Nomenclatures

I: type of waste disposal facility, where i = 1 for landfill, and i = 2 for INCINERATOR; *j*: name of district, and *j* = 1, 2, 3, 4, 5; *r*: name of district, and *r* = 1, 2, 3, 4, 5, 6; *k*: planning period, and *k* = 1, 2, 3; ${L}_{k}$: length of period k (day); ${\alpha}_{\mathit{ijk}}^{\pm}$: slope of transportation cost curve for waste from district j to facility i during period k; ${\alpha}_{\mathit{rjk}}^{\pm}$: slope of transportation cost curve for waste from district j to transfer station r in period k; ${\beta}_{\mathit{rjk}}^{\pm}$: Y-intersect of transportation cost curve for waste from district j to transfer station r in period k; ${\alpha}_{\mathit{rjk}}^{\pm}$: slope of transportation cost curve for waste from transfer station r to facility i in period k; ${\beta}_{\mathit{rjk}}^{\pm}$: Y-intersect of transportation cost curve for waste from transfer station r to facility i in period k; ${\alpha}_{2k}^{\pm}$: slope of transportation cost curve for residue from INCINERATOR to landfill in period k; ${\beta}_{2k}^{\pm}$: Y-intersect of transportation cost curve for residue from INCINERATOR to landfill in period k; $\tilde{{\theta}^{\pm}}$: Safety coefficient of INCINERATOR; $D{F}^{\pm}$: Density factor for waste from district to landfill (m^{3}/t); $F{E}_{1}^{\pm}$: Residue rate when waste is shipped from district to INCINERATOR (% of incoming waste); $F{E}_{2}^{\pm}$: Residue rate when waste is shipped from transfer station to INCINERATOR (% of incoming waste); $DL{C}^{\pm}$: Daily capacity of disposing waste at the landfills (t/day); $\tilde{I{C}^{\pm}}$: Capacity of the INCINERATOR (tonne/day); $O{P}_{\mathit{ik}}^{\pm}$: Operating cost of waste disposal facility i during period k ($/t); $O{P}_{\mathit{rk}}^{\pm}$: Operating cost of transfer station *r* during period k ($/t); $R{E}_{2k}^{\pm}$: Revenue from sale electricity generated at INCINERATOR in period k ($/10^{3} KWh); $R{E}_{\mathit{rk}}^{\pm}$: Revenue from recycling waste at transfer station r in period k ($/10^{3} KWh); $TL{C}^{\pm}$: Volume of existing landfill (m^{3}); $T{R}_{\mathit{irk}}^{\pm}$: Transportation cost for waste flow from transfer station r to facility i in period k ($/t); ${\eta}_{1}^{\pm}$: Electricity-generation rate when waste is shipped from district to INCINERATOR (KWh/t); ${\eta}_{2}^{\pm}$: Electricity-generation rate when waste is shipped from TR to INCINERATOR (KWh/t); ${\eta}_{3}^{\pm}$: Material recycling rate at the transfer station (%); ${\eta}_{4}^{\pm}$: Mass loss ratio at the transfer station (%); $VL{C}_{k}^{\pm}$: Variable cost for expanding landfill in period k ($/m^{3}); $VT{C}_{\mathit{rk}}^{\pm}$: Variable cost for expanding transfer station r in period k ($/t); $VI{C}_{k}^{\pm}$: Variable cost for expanding INCINERATOR in period k ($/t); $W{G}_{\mathit{jk}}^{\pm}$: Amount of waste generated in district j in period k (t/day); ${X}_{\mathit{ijk}}^{\pm}$: Waste flow from district j to landfill or INCINERATOR in period k (t/day); ${X}_{\mathit{jrk}}^{\pm}$: Waste flow from district j to transfer station r in period k (t/day); ${X}_{\mathit{irk}}^{\pm}$: Waste flow from transfer station r to landfill or INCINERATOR in period k (t/day); $EL{C}^{\pm}$: Volume of landfill to be expanded (m^{3}); $EI{C}_{k}^{\pm}$: Capacity expanded for INCINERATOR in period k (t/day); $ET{C}_{\mathit{rk}}^{\pm}$: Capacity expanded for transfer station r in period k (t/day); $\mathit{MELC}$: Maximum volume of landfill is allowed to be expanded (m^{3}); $MEI{C}_{k}$: Maximum allowance of capacity expansion for INCINERATOR in period k (t/day); $MET{C}_{k}$: Maximum allowance of capacity expansion for each transfer station in period k (t/day); $TET{C}_{k}$: Maximum allowance of capacity expansion for total transfer stations in period k (t/day).

## Declarations

### Acknowledgements

This Research was supported by the Natural Sciences Foundation of Beijing (8122038), the Program for Changjiang Scholars and Innovative Research Team in University (IRT1127), and the Program for New Century Excellent Talents in University (NCET-10-0376). The authors are grateful to the editors and the anonymous reviewers for their insightful comments and suggestions.

## Authors’ Affiliations

## References

- Baetz BW:
**Optimization/simulation modeling for waste management capacity planning.***Journal of Urban Planning and Development (ASCE)*1990,**116**(2):59–79.View ArticleGoogle Scholar - Ben-Tal A, Nemirovski A:
**Robust optimization – methodology and applications.***Mathematic Program Search B*2002,**92:**453–480.View ArticleGoogle Scholar - Cao MF, Huang GH:
**Scenario-based methods for interval linear programming problems.***Journal of Environmental Informatics*2011,**17**(2):65–74.View ArticleGoogle Scholar - Chang NB, Davila E:
**Minimax regret optimization analysis for a regional solid waste management system.***Waste Manag*2007,**27:**820–832.View ArticleGoogle Scholar - Changchun Urban Planning Council:
*Master plan of Changchun City (2005–2020)*. Changchun, China (in Chinese); 2004.Google Scholar - Chen MJ, Huang GH:
**A derivative algorithm for inexact quadratic program-application to environmental decision-making under uncertainty.***European Journal of Operational Research*2001,**128:**570–586.View ArticleGoogle Scholar - Chen WT, Li YP, Huang GH, Chen X, Li YF:
**A two-stage inexact-stochastic programming model for planning carbon dioxide emission trading under uncertainty.***Applied Energy*2010,**87:**1033–1047.View ArticleGoogle Scholar - Dubois D, Prade H:
**Operations on fuzzy number.***International Journal of Systems Science*1978,**9:**613–626.View ArticleGoogle Scholar - Dubois D, Prade H, Sabbadin R:
**Decision-theoretic foundations of qualitative possibility theory.***European Journal of Operational Research*2001,**128:**459–478.View ArticleGoogle Scholar - El Hanandeh A, El-Zein A:
**Life-cycle assessment of municipal solid waste management alternatives with consideration of uncertainty: SIWMS development and application.***Waste Manag*2010,**30:**902–911.View ArticleGoogle Scholar - Fan YR, Huang GH:
**A robust two-step method for solving interval linear programming problems within an environmental management context.***Journal of Environmental Informatics*2012,**19**(1):1–12.View ArticleGoogle Scholar - Hillier FS, Lieberman GJ:
*Introduction to operations research*. 4th edition. Holden-Day, Oakland, CA; 1986.Google Scholar - Huang GH, Chi GF, Li YP:
**Long-term planning of an integrated solid waste management system under uncertainty – I.***Model development. Environmental Engineering Science*2005,**22**(6):823–834.View ArticleGoogle Scholar - Huang GH, Chi GF, Li YP:
**Long-term planning of an integrated solid waste management system under uncertainty – II.***A north American case study. Environmental Engineering Science*2005,**22**(6):835–853.Google Scholar - Inuiguchi M, Sakawa M:
**Robust optimization under softness in a fuzzy linear programming problem.***International Journal of Approximate Reasoning*1998,**18:**21–34.View ArticleGoogle Scholar - Kollikkathara N, Feng H, Yu DL:
**A system dynamic modeling approach for evaluating municipal solid waste generation, landfill capacity and related cost management issues.***Waste Manag*2010,**30:**2194–2203.View ArticleGoogle Scholar - Ko AS, Chang NB:
**Optimal planning of co-firing alternative fuels with coal in a power plant by grey nonlinear mixed integer programming model.***J Environ Manage*2008,**88:**11–27.View ArticleGoogle Scholar - Kuhn HW, Tucker AW:
**Nonlinear programming. University of California Press.**In*Proceedings of the Second Berkeley Symposium on Mathematical Statistics and Probability*. Edited by: Neyman J. Berkeley, CA; 1951:481–492.Google Scholar - Li YP, Huang GH:
**Minimax regret analysis for municipal solid waste management: An interval-stochastic programming approach.***J Air Waste Manag Assoc*2006,**56**(7):931–944.View ArticleGoogle Scholar - Li YP, Huang GH, Nie XH, Nie SL:
**A two-stage fuzzy robust integer programming approach for capacity planning of environmental management systems.***European Journal of Operational Research*2008,**189:**399–420.View ArticleGoogle Scholar - Li YP, Huang GH, Yang ZF, Nie SL:
**0–1 Piecewise linearization approach for inexact nonlinear programming – application to environmental management under uncertainty.***Canadian Journal of Civil Engineering*2009,**36:**1071–1084.View ArticleGoogle Scholar - Li YP, Huang GH, Sun W:
**Management of uncertain information for environmental systems using a multistage fuzzy-stochastic programming model with soft constraints.***Journal of Environmental Informatics*2011,**18**(1):28–37.View ArticleGoogle Scholar - Luhandjula MK, Gupta MM:
**On fuzzy stochastic optimization.***Fuzzy Set Syst*1996,**81:**47–55.View ArticleGoogle Scholar - Lund JR, Tchobanoglous G, Anex R, Lawver R:
**Linear programming for analysis of material recovery facilities.***Journal of Environmental Engineering (ASCE)*1994,**120**(5):1082–1094.View ArticleGoogle Scholar - Masui T, Morita T, Kyogoku J:
**Analysis of recycling activities using a multi-sectoral economic model with material flow.***European Journal of Operational Research*2000,**122**(2):405–415.View ArticleGoogle Scholar - Mor S, Ravindra K, De Visscher A, Dahiya RP, Chandra A:
**Municipal solid waste characterization and its assessment for potential methane generation: A case study.***Sci Total Environ*2006,**371:**1–10.View ArticleGoogle Scholar - Su J, Huang GH, Xi BD, Li YP, Qin XS, Huo SL, Jiang YH:
**A hybrid inexact optimization approach for solid waste management in the city of Foshan, China.***J Environ Manage*2009,**91**(2):389–402.View ArticleGoogle Scholar - Tai J, Zhang WQ, Che Y, Feng D:
**Municipal solid waste source-separated collection in China: A comparative analysis.***Waste Manag*2011,**31:**1673–1682.View ArticleGoogle Scholar - Thuesen HG, Fabrycky WJ, Thuesen GJ:
*Engineering Economy*. Prentice-Hall, Englewood Cliffs, New Jersey; 1977.Google Scholar - USEPA:
*Municipal Solid Waste in the United States: 2007 Facts and Figures*. 2007.Google Scholar - Wilson BG, Baetz BW:
**Modeling municipal solid waste collection systems using derived probability distributions I: Model development.***Journal of Environmental Engineering (ASCE)*2001,**127**(11):1031–1038.View ArticleGoogle Scholar - Wilson BG, Baetz BW:
**Modeling municipal solid waste collection systems using derived probability distributions II: Extensions and applications.***Journal of Environmental Engineering (ASCE)*2001,**127**(11):1039–1047.View ArticleGoogle Scholar

## Copyright

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.