Review on Energy Resilience

Hamid Afshari Mem. ASME

Department of Mechanical Engineering,

University of Manitoba,

EITC, 75A Chancellors Circle,

Winnipeg, MB R3T 5V6, Canada

e-mail: [email protected]

Romain Farel PS2E Research and Education Institute,

Les Loges-en-Josas 78354, France

e-mail: [email protected]

Qingjin Peng1 Mem. ASME

Department of Mechanical Engineering,

University of Manitoba,

EITC, 75A Chancellors Circle,

Winnipeg, MB R3T 5V6, Canada

e-mail: [email protected]

Improving the Resilience of Energy Flow Exchanges in Eco-Industrial Parks: Optimization Under Uncertainty Eco-Industrial parks (EIPs) and industrial symbioses (IS) provide cost-effective and envi- ronmental friendly solutions for industries. They bring benefits from industrial plants to industrial parks and neighborhood areas. The exchange of materials, water, and energy is the goal of IS to reduce wastes, by-products, and energy consumption among a cluster of industries. However, although the IS design looks for the best set of flow exchanges among industries at a network level, the lack of access to accurate data challenges the optimal design of a new EIP. IS solutions face uncertainties. Considering the huge cost and long establishment time of IS, the existing studies cannot provide a robust model to investigate effects of uncertainty on the optimal symbioses design. This paper introduces a framework to investigate uncertainties in the EIP design. A multi-objective model is proposed to decide the optimal network of symbiotic exchanges among firms. The model minimizes the costs of multiple product exchanges and environmental impacts of flow exchanges. Moreover, this paper integrates the analysis of uncertainties effects on syner- gies into the modeling process. The presented models are depicted through optimizing energy synergies of an industrial zone in France. The efficiency of single and multiple objective models is analyzed for the effects of the identified uncertainties. In addition, the presented deterministic and robust models are compared to investigate how the uncer- tainties affect the performance and configuration of an optimal network. It is believed that the models could improve an EIP’s resilience under uncertainties. [DOI: 10.1115/1.4035729]

Introduction

Industrial symbioses (IS) are defined as a collective approach to achieve a competitive advantage in the exchange of by-products, wastes, water, and energy among a set of neighboring industries [1]. In EIPs, an industrial symbioses network is formed for eco- nomic and environmental friendly purposes [2]. The formation process of IS has been repeatedly discussed in the literature [3]. In a simplified form, EIPs can be classified into three categories: (a) planned symbioses [4], (b) self-organized [5], and (c) mixed meth- ods [6]. It is observed that the self-organized symbioses are mainly developed through individual negotiations between two or more industrial actors to make use of an accessible extra source (energy or material), strongly in line with their business goals. On the contrary, the planned symbioses are mostly designed by a third party (e.g., EIP managers, government, etc.) based on common strategies and interests. It appears that the number and size of planned symbioses are increasing, probably thanks to the success- ful experience in China and South Korea. This requires a deliber- ate attention to the optimal design and operation of the planned IS.

The optimal design of industrial symbioses is, in its simple form, finding the best pairs of plants for flow exchanges. For the versatile spectrum of flow exchanges (including water, chemical products, by-products of a production process, an excessive steam of a process, etc.), a decision maker should involve the cost of dif- ferent facilities and related requirements in the optimal design process. Besides traditional objectives such as the cost minimiza- tion or the profit maximization, a designer may comprise versatile

goals in the symbioses design. A supplementary goal would emphasize on the minimization of environmental impacts by increasing the flow exchanges in an EIP. An ultimate extension would be the design and optimization of more than two plant synergies, i.e., a distribution network (utility water or steam) with multi suppliers and multi consumers. A schematic view of exchanges between industries is presented in Fig. 1.

For designing an industrial symbiosis network, a further set of technical requirements (e.g., temperature, pressure, mass flow rate, etc.) is required compared to the traditional location optimization problems. For example, studies showed that ignoring temperature and distance in a mathematical model could not pro- vide a feasible solution for industries involving in energy syner- gies [7]. For no doubt, the data collection is a time-consuming task, with non-negligible costs, and confidentially limits. Studies have addressed the lack of access to data as a common challenge to model symbioses [8]. In addition, inadequacy of input data leads to error, uncertainty, and risk [9] of presented solutions for an EIP.

Uncertainty is defined as any lack of data or the lack of trust in available data [10–12]. The uncertainty of data in the symbioses network design generates considerable errors in cost estimation, reliability, and efficiency of the solution. There are different classifications of uncertainties in literature. One common classifi- cation divides uncertainties into Aleatoric and Epistemic [13,14]. The aleatoric uncertainty is from stochastic effects such as the random noise and measurement errors. The aleatoric uncertainty is quantifiable using stochastic terms and the probability theory [15]. As a specification, it is not possible to mitigate the aleatoric uncertainty by additional data or analysis. On the other hand, the epistemic uncertainty refers to a lack of knowledge or informa- tion. Errors in simulating processes, data collection, or human errors are the reasons for this type of uncertainty. In fact,

1 Corresponding author.

Manuscript received September 14, 2016; final manuscript received December 20, 2016; published online February 27, 2017. Assoc. Editor: Konstantin Zuev.

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-1 Copyright VC 2017 by ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

increasing the precision of information can help reducing the epistemic uncertainty [16].

A review of symbioses in the existing EIPs showed that focusing on the optimization of flow exchanges by increasing the number of symbioses could erode the resilience of an IS [17]. Resilience is defined as the ability of a system to sustain in a fluc- tuating market environment [18,19]. Therefore, besides efforts to encourage IS, it is needed to improve the resilience by increasing an EIP adaptability to various disruptions.

The goal of this research is to develop an optimization method for industrial symbioses in an EIP, considering both economic and environmental objectives. To do this, a mixed integer linear pro- gramming model is developed to optimize the location and capacity of flow exchanges between industries under uncertain- ties. This method embeds uncertainties in the modeling as sto- chastic scenarios and a range for estimated values of a parameter to measure the worst-case scenarios. A solution approach is then applied to deal with uncertain data during optimization. The research contributions include: (i) proposing a multi-objective model for minimization of the total annual cost and pollutions and (ii) considering uncertainties in the modeling and solution approach to bridge the gap of the literature.

As follows, literature on the symbioses optimization and uncer- tainty modeling is reviewed. The methodology for the industrial network optimization under uncertainty is then presented. The modeling approach is depicted using a dataset from an EIP.

Literature Review

Several studies address the optimization of decisions in an EIP [8]. A review by Kastner et al. classified the existing models for cultivating symbioses in EIPs [20]. The study shows that modeling methods are typically based on tools developed to opti- mize processes including the pinch analysis [21,22] and mixed integer linear programming (MILP) [23]. As industrial park net- works have traditionally been broken into categories of water, power (including heat), and materials, modeling and optimization approaches have been categorized similarly.

In an energy symbiosis through the heat exchange network, the optimization is used for different objectives, known as the heat exchanger networks (HENs) design. Methods of modeling and optimization in this context are mostly based on the pinch analy- sis, total site analysis (TSA) [24–27], and mathematical program- ming [28–30]. Alternatively, for water exchange networks, the pinch analysis [31–32] and MILP [33] are commonly used in sym- bioses to find the optimum trades in inter- or intrafirm to minimize the fresh water consumption and maximize wastewater exchanges. In material exchange networks, the focus is on unwanted by- products, not the product created by plants [20]. This is important because it distinguishes the optimization of by-products recovery/ distribution networks from business as usual processes, which is out of the EIP scope. Although there are several material sharing examples, the optimization is not often included in the research [8]. Mathematical modeling, specifically MILP, is the common method applied for by-product symbioses networks [34–36].

Since in most industrial cases, a combination of wasted water, heat, and material is available, the majority of research focuses on multiple exchange networks. Life cycle assessment (LCA) [37], mathematical modeling [38,39], and supply chain optimization [40–42] are common approaches in the literature.

Resilience in industrial symbiosis is defined based on a system’s ability to keep eco-efficient flow exchanges under dis- ruptions [17]. While the uncertainty is inevitable in engineering system design, designers have developed methods for the design under uncertainties to reduce effects of the disruption [43,44]. Methods to investigate effects of uncertainties are classified into two main branches [45]. Using analytical methods, a mathe- matical model of the problem is formulated under uncertainties. Stochastic optimization [46], linear programming with scenario generation [47], and robust optimization [48,49] are the common mathematical approaches for uncertainty studies. The second method is simulation-based approaches. Depending on the case and state of a system, different simulations are available such as static, dynamic, discrete, continuous, deterministic, and stochas- tic. For static simulations, Monte Carlo (MC) simulations are used [50], and for dynamic systems, the discrete event simulation (DES) is in a common usage [51]. The simulation-based method is limited because of its inherently limitation to return the optimal solution of problems. Therefore, simulation-based optimization methods are developed by combining analytical and simulation models for complex problems [52]. Thus, the integration of these two methods improves the solution approach for either existing uncertainties or nonlinear terms in the optimization problem [53].

As presented in Table 1, limited studies have addressed uncer- tainties for the optimization of synergies in an EIP. Therefore, a framework to study uncertainty in EIPs is adopted. The uncer- tainty is widely discussed in areas such as the supply chain opti- mization [46,54]. The goal of location decisions in the supply chain optimization is to locate the best set of facilities in a net- work to achieve desired goals. Similarly, in the optimization of flow exchanges in EIPs, the optimum set of symbioses’ partners is intended. The framework is adopted using reviewed uncertainties in a supply chain [55]. Attributes [56] of the framework are also presented in Table 1.

Table 1 categorizes uncertainties for the business and interac- tions in a supply chain as well as equivalent synergies in an EIP. Synergic uncertainties are then assessed using their nature, rank, and type. Despite the existing research in the uncertainty evalua- tion in supply chains, limited studies have addressed uncertainties in an EIP.

Qiu and Huang proposed to adopt the supply hub in an indus- trial park (SHIP) as a public logistics and warehousing services to industries inside an industrial park [57]. They proposed two math- ematical models (with and without SHIP) and studied the effect of demand uncertainties in the models. Using a simulation approach, the performance of the models was analyzed in terms of cost (the total cost of a manufacturer, the total cost for the SHIP, and the total cost for supply chain). They concluded strategies such as the application of SHIP could be beneficial for an industrial park under the demand uncertainty to save the total cost for all

Fig. 1 Schematic view of industrial symbioses

021002-2 / Vol. 3, JUNE 2017 Transactions of the ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

involved stakeholders. Moreover, each manufacturer can improve its performance under all demand patterns.

Maes et al. explored the literature to apply an appropriate energy strategy within the Flanders industrial park [58]. They claimed that the energy management in industrial parks can be integrated into the entire development process and park manage- ment. To intensify local synergies, buildings and processes should be clustered for energy exchanges, collective production, and joint contracting of energy services. However, they highlighted that uncertainty and variation of the energy consumption can keep developers from tailoring industrial park design and utilities. The uncertainty of the future tax on carbon or other wastes has been deliberated as well.

P�erez-Vald�es et al. presented a decision making model for a natural-gas powered industrial park [59]. The model maximizes the net present value in the industrial park to determine the type of plants and connections between them. A stochastic mixed- integer programming model was employed to handle the uncer- tainty of future prices and costs of raw materials and finished products. It has been discussed that costs of emissions of CO2 and nitrogen oxides (NOx) can be considered as stochastic parameters in the model. The application of the model in a Norwegian indus- trial park showed the model well suited for analyzing small to moderately sized scenarios considering variations in the most important stochastic parameters.

The existing research for optimization of decisions in EIPs is summarized as follows:

� Although the need for uncertainty analysis is discussed in several studies, the majority of optimization models are for- mulated for deterministic parameters.

� A single objective (mostly for the cost minimization) has been used in the existing studies for energy exchange net- works optimization. Such modeling approach does not directly address economic and environmental concerns in designing industrial symbioses.

� Compared to the uncertainty problems issued in the supply chain literature (as shown in Table 1), there are several uncertainty types for optimizing decisions of EIPs.

Thus, there are opportunities for research in the domain of EIPs to include the uncertainty in the problem formulation as well as in problem solving, as highlighted earlier.

Methodology

The synergy optimization problem in an EIP initiates with gen- erating potentially feasible flow exchanges between plants, and then selecting the best flow exchanges based on predefined eco- nomic and environmental criteria. The decision model uses the

flow data for the matching algorithm. Since uncertainty can affect the optimal decision of symbioses and the further selection of future partners, it is necessary to assess effects early in the method. In addition, there is a need to select partners under the formulated uncertainty. Figure 2 depicts a framework of the proposed method in a step by step manner.

To identify uncertainties, internal and external sources are investigated. As presented in Table 2, the demand uncertainty and supply uncertainty are selected as internal uncertainties. If a supplier fails to fulfill its customers’ demand, such failure impacts customers’ activities. To minimize the impacts, the supply uncer- tainty as well as the demand uncertainty should be considered in the symbioses network design.

The similar consideration should be taken into account for uncertainty in the supply price and tax on carbon. If other supply sources (other than industrial symbioses) provide lower prices, industries would revise their ties with current industry partners. For the tax on carbon, it is expected to increase for a short term, but any reduction in the tax rate would impact established symbioses.

Figure 3 presents the framework to optimize the energy sym- bioses under uncertainty. Since uncertainties affect the synergies from different internal and external sources, two models including deterministic and stochastic models are developed. Optimal design parameters (flow exchanges between the firms) are decided after the performance of models are compared. An optimal opti- mization strategy is then selected under studied uncertainties.

The first developed model is a deterministic MILP model. To formulate symbioses, the total annual cost of symbioses in an EIP and the total pollutions the EIP are minimized using a bi-objective model.

The first objective function minimizes the total annual cost to establish symbioses networks in industries. Different terms are included in the cost objective function to represent various symbi- osis costs as shown in Eq. (1)

min Z1 ¼ XK k¼1

XI i¼1

XJ j¼1

CDkj b k ijy

k ij þ

XK k¼1

XJ j¼1

CIkj D k j 1 �

XI i¼1

xkij

!

þ XK k¼1

XJ j¼1

XI i¼1

UkijL k ija

k ij þ CC

k ijb

k ij

� � ykij

þ XK k¼1

XI i¼1

XJ j¼1

RCkijD k j x

k ij �

XK k¼1

XJ j¼1

TCkj D k j

XI i¼1

xkij

!

� XK k¼1

XI i¼1

TSki XJ j¼1

Dkj x k ij

0 @

1 A (1)

Table 1 Framework to study uncertainties in the optimization of industrial symbioses

Uncertainity in supply chains Uncertainity in EIPs

Level Business as usual Level Industrial symbiosis Nature Rank Type Existing research

Firms Product characteristics Internal Input specifications (e.g., price) A 2 O Process/manufacturing Process change; Technology improvement;

Operation Failure E 2 T

Network End customer demand Change of end customer demand A 1 O Supplier performance Uncertainity of demand supply A 1 O [57] Configuration, infrastructure and facilities

Uncertainity in matching proper symbioses partners

E 3 S

Energy consumption Uncertainity of energy, water, etc., consumption

A 1 O [58]

External Economic environment External Price of other energy sources A 1 S [59] Political environment Tax on CO2 or other wastes A 2 S [59] Natural environment Disasters; weather conditions A 3 S [58,59]

Note: Nature—aleatoric (A) and epistemic (E). Rank—knowing the probabilities (1), knowing the outcomes (2), knowing a little (3). Type—operational (O), tactical (T), and strategical (S).

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-3

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

In Eq. (1), the annual fixed and variable costs of preparing product k from sources other than symbioses partners are added to the annual cost of a conditioning center embedded inside firms. If the symbioses partner does not supply the demand, unsupplied products will be prepared from other external sources. In this case, the buyer may purchase conditioning centers for any product k. Because a part of the demand is supplied by external sources, its tax saving is reduced from the total cost. Other costs include the cost for piping network (if required), by-product/waste/energy recovery facilities at suppliers, and tax saving of suppliers for recovered product k.

The second objective function minimizes the environmental impacts of symbioses in industries as presented in Eq. (2). Because the pollution of recovered by-product/waste/energy is less than the supplying raw material/energy (PEkij < PI

k j ), Eq. (2)

motivates flow exchanges in industries

minZ2 ¼ XK k¼1

XI i¼1

PEkij XJ j¼1

DkJ x k ij þ

XK k¼1

XI j¼1

PIkj D k j 1 �

XI i¼1

xkij

!

(2)

Constraints of the model are as follows:

XI i¼1

xkij � 1 8k; j (3)

Dkj x k ij � S

k i y

k ij 8k; i; j (4)

XJ j¼1

Dkj x k ij � S

k i 8k; i (5)

CIkj D k j

XI i¼1

xkij þ TC k j D

k j

XI i¼1

xkij � CD k j

XI i¼1

bkijy k ij

> XI i¼1

CEkijD k j x

k ij 8k; j (6)

CEkijD k j x

k ij > ðU

k ijL

k ija

k ij þ CC

k ijb

k ijÞy

k ij þ RC

k ijD

k j x

k ij

� TSki D k J x

k ij 8k; i; j (7)

ðLkij � cÞ x k ij � 0 (8)

xkij � y k ij (9)

0 � xkij � 1; y k ij 2f0; 1g (10)

Table 2 Uncertainties identified to model industrial symbioses for optimization

Uncertainty Type Definition

Tax on carbon (A) External Reduction/increase in Tax on carbon due to national/international regulations Supply price (B) Internal/external Any change either internal or external that makes other supply sources more interesting Supply/demand (C and D) internal Any drastic variation in predefined values of demand/supply

Fig. 3 Framework for optimization of energy symbioses under uncertainty

Fig. 2 Steps for the optimal design of industrial symbioses under uncertainty

021002-4 / Vol. 3, JUNE 2017 Transactions of the ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

Constraint (3) sets the satisfaction up to 100% of demand. Con- straints ((4) and (5)) balance demand and supply equations in industries. Constraint (4) decides that only suppliers with the enough capacity can establish symbioses with demanding indus- tries. Constraint (5) highlights that the total supplied demand should not exceed the supply capacity of an industry. Constraint (6) ensures that each demanding industry pays the less money when it contributes to symbioses than that it supplies all demands from other methods. In other words, only financially feasible sym- bioses are selected in the model. Constraint (7) checks if an indi- vidual investment for each symbioses (piping and recovery cost) is economical for the network. Investments should be reimbursed by selling the recovered energy in a defined period. Constraint (8) enforces establishing symbiosis within a specified distance limit. Constraint (9) limits exchange flows to related industries. Con- straint (10) defines variable types in the model.

If the symbioses are established to exchange recovered energy, constraints ((4) and (5)) are revised to constraints ((11) and (12)). Constraint (11) demonstrates that in terms of temperature, the energy capacity of a supply firm should be enough to be selected for supplying the demanded energy. Constraint (12) defines that

the total energy supply should not exceed the supply capacity of an industry regarding temperatures of demand and supply. For further information, refer to Refs. [60] and [61]

Dkj x k ij � S

k i 1 �

TMPkj

TMPki

! ! ykij 8k; i; j (11)

XJ j¼1

Dkj x k ij

1 � TMPj=TMPi � �� � � Ski 8k; i (12)

After modeling symbioses, uncertainties should be embedded in the model. Several methods exist to model uncertainties in a model. In this research, stochastic and robust models are devel- oped for the uncertain parameters. In stochastic optimization, the values of uncertain parameters are defined by decision makers as probability distributions. The goal is to optimize the expected value of some objective functions. In robust optimization prob- lems, the goal is to optimize the worst-case performance of a model [62].

Fig. 4 Proposed solution approach

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-5

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

In the second model, deterministic values of parameters are substituted by statistics of decided uncertainties using the stochas- tic optimization. Therefore, the new objective functions are for- mulated as follows:

min Z1 ¼ Eðf1ðnðDkj Þ; nðCI k j Þ; nðTC

k j ÞÞÞ (13)

min Z2 ¼ Eðf2ðnðDkj ÞÞÞ (14)

For example, Eq. (2) is revised as Eq. (15)

min Z2 ¼ XS s¼1

Ps � �XK

k¼1

XI i¼1

PEkij XJ j¼1

nðDkj Þx k ij

þ XK k¼1

XI j¼1

PIkj nðD k j Þ �

1 � XI i¼1

xkij

�� (15)

Therefore, Eqs. (3)–(12) should be revised as well to include uncertain parameters in the model.

As the solution approach, a sample average method (SAM) is applied to handle the stochastic nature of the model by estimating objective functions [53]. Because the most popular way to deal with randomness in a model is to optimize the expected value of an arbitrary function of parameters, the objective functions are rewritten using average values, as shown in Eq. (16) [63]. Such arbitrary function of parameters should be defined over an appro- priate probability space

1

N

XN v¼1

f# x; xvð Þ� E f# x; xð Þ � �

(16)

In this equation, N random and independent scenarios are used to approximate objective function values. Each scenario reflects randomness by xv, where v ¼ð1; 2; …Þ represents stochastic parameters. Using an estimation of the expected value of functions, the stochastic optimization model is changed to a deter- ministic one which is relatively easier to be solved. Thus, the multi-objective model is simplified as presented in the below equation:

min 1

N

XN v¼1

f1 x; xvð Þ; 1

N

XN v¼1

f2 x; xvð Þ; … !

(17)

An algorithm is developed to solve the model as a weighted lin- ear combination of objective functions as presented in Fig. 4. First, the model is solved using a single objective. The values are used to normalize the weighted multi-objective function. The multi-objective model is then solved, and results are stored. The process continues to test all desired weights. Because of using the sample average method, all described steps are followed for a new set of uncertainty scenarios. Finally, solutions are evaluated to decide the desired network topography.

For the uncertain parameters that are defined in a range (using upper/lower boundaries), a nonprobabilistic robust optimization model is developed. In the robust optimization, a deterministic equivalent of a model, called robust counterpart, is generated. If the robust counterpart of a robust model is computationally tracta- ble, the model can be solved in a reasonable time for the optimal solution [64]. To model a problem using the robust optimization, the uncertain variables and constraint uncertainty are defined. The robust counterpart of the model is generated to search solutions.

The first contribution of this research is opening the discussion on the vulnerability of EIPs to disruptions and uncertainties for our expectations. Thus, it is needed to evaluate the resilience of proposed optimal energy exchange networks using novel methods. The second contribution is that solutions are proposed to improve the resilience of established networks when facing uncertainties,

as presented in Fig. 4. Such solutions includes: (1) identify the symbioses that are resilient under all uncertainties, (2) evaluate the effect of uncertainties on the performance of an optimal net- work of symbioses, and (3) help decision makers applying proper improvement scenarios to maintain the technical feasibility and adaptability of EIPs when key parameters of a system are changed. In the next paragraph, a case study is presented to elabo- rate the proposed method.

Case Study

The proposed model has been applied to optimize energy sym- bioses using anonymous data inspired by a set of industries in France. The objective is to find the optimal energy exchange flows between industries considering technical and economic con- straints and uncertainty of the demand fulfillment. It is expected that the optimum network motivates energy exchanges to reduce the extra fuel consumption for economic and environmental pur- poses. Tables 3–5 present data used for mathematical modeling.

In addition, Table 6 presents uncertain parameters. The uncer- tain data in the model are defined using a range or statistical distri- butions. In each scenario, a sample from the bounded space/ distribution function is obtained for modeling. In this table, the demand and supply declared as the normal distribution refers to the current value presented in Table 4.

Table 3 Distances (km) between plants

Buyers

Suppliers 1 2 3 4 5 6

S1 8 17 2 7 4 6 S2 14 2 11 13 7 4 S3 10 8 5 3 3 4

Table 5 Major parameters included in the models

Parameter Unit Value

Depreciation period of pipelines Year 20 Depreciation of facilities (e.g., HEN) Year 10 Interest rate % 5 Heat recovery cost e/kWh 0.028 Heat price generated from gas e/kWh 0.063 Gas pollution kg CO2/ kWh 0.063 Tax on carbon e/kWh 0.0042

Table 6 Uncertain parameters included in the models

Parameter Type Data

Demand Distribution Normal (demand, dj ) Supply Distribution Normal (Supply, di ) Supply price Range [0.054, 0.063] Tax on carbon Range [0.0042, 0.01]

Table 4 Energy specification of plants in the studied area

Suppliers Buyers

Energy specs. S1 S2 S3 1 2 3 4 5 6

Heat (Ton/yr.) 90 70 100 35 25 10 70 60 30 Temp. (K) 673 523 473 403 373 443 403 423 383

021002-6 / Vol. 3, JUNE 2017 Transactions of the ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

The proposed multi-objective model is used to minimize total environmental impacts of synergies while minimizing the total annual cost of the symbioses network. The model is formulated using AIMMS optimization package software [65]. Using a 4 GB RAM, 2.0 GHz personal computer, the AIMMS could solve the problem in less than one second. To analyze the effect of objec- tive functions on the solution, each objective function is applied separately. Using a weighting method, the multi-objective model is solved and the solutions are compared. Figure 3 presents the optimized synergy network using single and multiple objectives. The multi-objective model and cost minimization model provide a similar network structure. Arrows in Fig. 5 show the direction of the flow from suppliers (i¼3) to buyers (j¼6).

The models are compared using indices presented in Table 7. Because both the multi-objective model and total cost minimiza- tion provide a similar network structure, the indices are presented together. Moreover, the length of a pipeline is used to compare the models on required pipeline networks.

The indices show that the model for the environmental impacts minimization gives the lower pollution compared to other models. On the contrary, the other models could save more cost compared to the model for the environmental impacts minimization.

The effects of the uncertainties are evaluated using the weighted multi-objective model. Equal weights are assigned to both cost minimization and environmental impacts models. The sampling average method is used to solve the stochastic multi- objective model. The objective function values are estimated based on average values of scenarios generated using Table 6. Each uncertain parameter is assessed separately to clarify its effects on the model. In addition, “supply price” and “tax on car- bon” are introduced in a range of values using the AIMMS for robust solutions.

Demand Uncertainty. This uncertainty originates from the expansion of industries that need more raw material/energy con- sumption. It is assumed that each industry can increase its demand up to 50%. After generating N¼10 scenarios, the multi-objective model is solved and the results are evaluated. The liability of flow exchanges between suppliers and users are decided based on the number of confirmed connections and average amount of the sup- ply. To compare effects of the demand uncertainty, results of the stochastic multi-objective model is presented as well in Fig. 5. A connection is considered for unreliable, when few flow exchanges are observed using the connection in ten simulated scenarios.

As shown in Fig. 6, no supplier can satisfy the demand of user 3 under the demand uncertainty. In addition, supplier 3 is consid- ered to supply the demand of user 2, but the connection will be established for a sever increase in demand. Other flow exchanges remain unchanged.

Supply Uncertainty. If a supplier cannot support the promised supply, the user will miss a required demand. In this case, a user

Fig. 5 Optimized symbioses networks using (a) the minimiza- tion of environmental impacts and (b) the minimization of the total cost, and the multi-objective model

Table 7 Comparing optimized symbioses networks using sin- gle and multiple objective functions

Indices Min environmental

impacts Min total cost and multi obj.

Total pollution (kg CO2) 100,255,706 100,380,207 Total cost (e) 69,892,020 69,274,618 Pipeline (km) 27 20

Fig. 6 Comparing effects of the demand uncertainty on optimized flow exchanges in the multi-objective model

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-7

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

may prepare its material/energy from other sources. It is assumed that each supplier may decrease its supply up to 50%. The multi- objective model is solved using ten generated scenarios. Similar to the demand uncertainty, the number of confirmed connections and the average amount of supplies are used to decide liability of flow exchanges. The effects of the supply uncertainty are eval- uated using the stochastic multi-objective model as presented in Fig. 7.

The results show that under the supply uncertainty, user 2 needs to shift its supplier from 2 to 3. Suppliers can partially sup- port the demand of user 3. The supply of user 5 is weakened as well. Under the supply uncertainty, user 6 is always supplied completely.

Uncertainty in supply prices: In a real condition of the supply market, there are several competing sources of supply. Improve- ments in technology could lead to the cheaper price for a raw material. In the energy market, the evolution of energy genera- tions using renewable sources in small and medium size enter- prises could reduce supply prices. In this case, a user prefers to shift its material/energy supply into other sources. It is assumed that the current cost of supplying energy will be reduced to 0.054 (e/kWh) depending on energy sources of users. Effects of the supply uncertainty are evaluated using a robust multi- objective model as presented in Fig. 8.

Figure 8 shows that the uncertainty in supply prices would motivate user 4 to increase its purchase from supplier 3. For user 6, it is preferred to supply more than 50% of the demand from supplier 1 as the result of price uncertainty. Other flow exchanges remain almost the same as the deterministic model.

Tax on Carbon. This uncertainty arises from the national or international regulation to limit produced carbon in industries’ processes. A review of trends in the carbon tax shows that in a short term, countries plan to increase the carbon tax [58,59]. How- ever, in the long term, another scenario could be regulated. It is assumed that the tax on carbon can increase up to 2.5 times of its

current rate. For this uncertain parameter, the rate is directly increased in the robust model to evaluate the effects. The result shows no change in the optimized network. The amount of the demand satisfaction is changed for the optimal cost.

Analysis and Discussion

Besides the network structure, objective function values are analyzed to investigate the effects of uncertain parameters. The indices are for the cost and environmental impacts of the synergic networks as summarized in Table 8.

Table 8 shows the percentage of changes (D) in each index compared to the deterministic multi-objective model. This is use- ful to evaluate effects of uncertainty parameters on the model.

The initial demand uncertainty seems to have the worst effect on indices. But the most effect is for the increase of flow exchanges, not the uncertainty. Thus, the supply uncertainty affects the network more comprehensively than other uncertain parameters. In addition, Fig. 7 demonstrates that the uncertainty in the supply uncertainty has provided three uncertain connections in the network. A surprising result in Table 8 is the limited effect of increasing the carbon tax on reducing pollutions.

A deeper analysis is conducted to evaluate effects of uncertain- ties on performance of the optimal energy symbioses network. The indices are defined for economic and environmental perform- ances of networks as presented in Table 9.

The effects of uncertainties are measured using the performance indices, as presented in Table 10. In this table, the percentage of changes (D) is compared to the values of indices for the determin- istic model.

Comparing the demand satisfaction (DS) and weighted demand satisfaction (WDS) indices shows that the uncertain demand and supply could reduce the total demand satisfaction. The reduction of demand satisfaction is on the contrary of demand increase. Other indices including supply capacity (SC), carbon tax saving

Fig. 7 Comparing effects of the supply uncertainty on optimized flow exchanges in the multi-objective model

Fig. 8 Comparing effects of the uncertainty in supply prices and tax on carbon on optimized flow exchanges in the multi-objective model

021002-8 / Vol. 3, JUNE 2017 Transactions of the ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

(CTS), and network density (ND) present a negative effect of demand and supply uncertainties on optimal energy symbioses. The suppliers’ eco efficiency (SEE) and customers’ eco efficiency (CEE) record different performances for uncertain demand and supply parameters. Thus, the demand uncertainty reduces stake- holders’ financial performances, and the uncertain supply has a positive effect of suppliers and users’ savings.

Because the price and tax on carbon uncertainties present no effect of performance indices, it is useful to investigate a bound- ary for the parameters values. The boundary defines the lower bound and the upper bound where the model is feasible. For the supply price, a feasible boundary is found as [0.026, 0.067]. It means that the energy symbiosis is still feasible if the price is

reduced to 0.026, or increased up to 0.067. For the tax on carbon, it is observed that the symbioses are possible in all ranges of tax. Thus, the policy to reduce/rise tax on carbon is not a barrier for the optimal IS. However, the increase of tax on carbon helps redirecting costs into more symbioses instead of paying the tax for carbon emissions.

In summary, it is believed that some connections are resil- ient to be affected by uncertainties. The flow exchanges between supplier 2 and user 2 (S2-U2), (S3-U4), and (S1-U6) exemplify such resilience in the reviewed industrial case. Thus, the EIP managers could trust such flow exchanges for the opti- mized symbioses network under studied uncertainties. Finding resilient flow exchanges is essential for a long-term operation

Table 8 Evaluation of objectives under the uncertainties

Uncertain parameters

Index Demand Supply Price and carbon tax

Total cost of suppliers 25,236,970 18,214,138 25,535,686 D (%) �1.17% �28.67% 0% Total cost of users (minus tax) 67,915,783 58,862,931 43,738,932 D (%) 55.28% 34.58% 0%

Total cost (with tax income) 86,202,535 72,229,720 62,409,902 D (%) 38.12% 15.73% 0%

Total cost (minus tax) 93,152,754 77,077,069 69,274,618 D (%) 34.47% 11.26%

Total pollution 129,396,198 103,742,485 100,380,207 D (%) 28.91% 3.35% 0%

Table 9 Defined performance indices to evaluate effects of uncertainties

Type

Index Definition Economic Environmental Efficiency

Demand satisfaction (DS) Percentage of covered demand � Weighted demand satisfaction (WDS) Total covered demand � Supply capacity (SC) Percentage of used capacity of suppliers � Carbon tax saving (CTS) Carbon emissions reduction in terms of cost � Network density (ND) [66] Density of cooperative ties over firms � Suppliers’ economic efficiency (SEE) Suppliers’ savings over costs Customers’ economic efficiency (CEE) Customers’ costs after over before symbioses �

Table 10 Measured performance indices to evaluate effects of uncertainties

Uncertain parameters

Index Demand Supply Price and carbon tax

DS (0.01, 0.92, 0, 0.7, 0, 0.97) (0.01, 0.85, 0, 0.66, 0, 0.83) (0.08, 1, 0, 0.95, 0, 1) D (%) N/A N/A N/A

WDS 0.443 0.465 0.541 D (%) �18% �14% 0% SC 0.226 0.207 0.276 D (%) �18% �25% 0% CTS 42,158,557 38,585,880 51,485,368 D (%) �18% �25% 0% ND 0.44 0.48 0.56 D (%) �21% �14% 0% SEE 0.854 1.051 1.03 D (%) �17% 2% 0% CEE 0.22% 0.16% 0.17% D (%) 29% �6% 0%

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-9

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

of the energy exchange system to motivate and encourage investments.

Conclusions

The creation of EIPs means the successful convergence of several mutual objectives in strategy, local policy, and business interests of stakeholders. One of the main tasks, if not the basic of the evaluation, is the structure of synergies and their eco- nomic and environmental gains. This paper brings the attention to the importance of considering uncertainties in the optimiza- tion problem for the flow exchange optimization in designing EIPs, and proposes a multi-objective model for the symbioses creation. The model minimizes the total annual cost of synergic exchanges while minimizing the total environmental impacts of flow exchanges between industries. This is important because the sustainability of EIPs and its symbiosis not only relies on economic bases, such as the oil price decreased recently, but also on the environmental motivation such as the CO2 emission regulated by the state. As a result, the resilience of industrial symbioses is improved by increasing the adaptability of an EIP to disruptions.

This paper reviews briefly the uncertainty factors in EIPs, and studies the effect of formulating uncertainty parameters in the optimization problem using a case application of anonymous industrial data. The deterministic multi-objective model is com- pared to models with uncertain parameters to highlight the effect of uncertainties on the industrial symbioses decisions. Contribu- tions in this paper are as follows:

� A multiple objectives model is proposed to optimize symbio- ses networks for multiple types of symbioses to address mini- mizing the environmental impacts directly.

� Formulation of technical and economic measures is dis- cussed in the flow exchange.

� Uncertain parameters are integrated in the optimal structure of energy symbioses networks.

� Effects of uncertain parameters are compared to identify the most resilient flow exchanges under uncertainties.

As confirmed by other studied in optimizing industrial symbio- ses networks [8], the lack of access to more detailed technical data is an important limitation observed in this research. However, the proposed model has included available technical features in the formulation.

This research also faces some limitations in the approach and application. The match making of flows is simplified to tempera- ture and flow rate compatibility, which is essential but not sufficient for exploring all the thermodynamically feasible synergies. Although integrating the flow details implies important efforts for the data col- lection, it would be interesting to do the exercise with those details.

In the further work, we plan to extend the model for studying other objectives such as maximizing efficiency indices. Moreover, we plan to measure values for uncertain parameters instead of assuming the values. Agent-based modeling can be used for uncertainty studies. Because the uncertainty is inevitable in engi- neering systems, suggesting models to overcome multiple uncer- tainties will be examined in the future research.

Acknowledgment

The authors wish to acknowledge that this research has been supported by UMGF Grant from the University of Manitoba, Discovery Grant No. (RGPIN-2015-04173) from the Natural Sci- ences and Engineering Research Council (NSERC) of Canada, and GLOBALINK research award from the Mitacs Canada and Campus France.

Nomenclature

I ¼ set of supplier industries J ¼ set of demand industries

K ¼ set of by-product/waste/energy types S ¼ set of scenarios for the stochastic model

Parameters

CDkj ¼ fixed cost of preparing k from sources other than i CCkij ¼ fixed cost of conditioning k if supplied from i to j CEkij ¼ selling price of k from industries i to j CIkj ¼ variable cost of preparing k from sources other than i Dkj ¼ demand of industry j from by-product type k Lkij ¼ distance of industries i and j for k Ps ¼ probability of each scenario in the stochastic model

PEkij ¼ environmental impacts of produced k by i PIkj ¼ environmental impacts of prepared k by j

RCkij ¼ cost of recovering k for industry j in i Ski ¼ supply of industry i from k

TCkj ¼ tax on carbon for k imposed to industry j TSki ¼ tax saving of industry i by exporting k

TMPki ¼ temperature of energy k supplied by i TMPkj ¼ temperature of energy k demanded by j

Ukij ¼ unit cost of network between i and j for k (if any) akij ¼ depreciation rate of pipeline between i and j bkij ¼ depreciation rate of facilities between i and j c ¼ distance limit for industries to build synergies

Variables

xkij ¼ percentage of demand supply from i to j for k ykij ¼ binary variable if symbioses exist between i and j

References [1] Chertow, M., 2000, “Industrial Symbiosis: Literature and Taxonomy,” Annu.

Rev. Energy Environ., 25(1), pp. 313–337. [2] Chertow, M., 2004, “Industrial Symbiosis,” Encycl. Energy, 3(19), Elsevier,

Amsterdam, The Netherlands, pp. 407–415. [3] Farel, R., Charrière, B., Thevenet, C., and Yune, J. H., 2016, “Sustainable Man-

ufacturing Through Creation and Governance of Eco-Industrial Parks,” ASME J. Manuf. Sci. Eng., 138(10), p. 101003.

[4] Behera, S. K., Kim, J.-H., Lee, S.-Y., Suh, S., and Park, H.-S., 2012, “Evolution of ‘Designed’ Industrial Symbiosis Networks in the Ulsan Eco-Industrial Park: ‘Research and Development Into Business’ as the Enabling Framework,” J. Cleaner Prod., 29–30, pp. 103–112.

[5] Domenech, T., and Davies, M., 2011, “Structure and Morphology of Industrial Symbiosis Networks: The Case of Kalundborg,” Procedia-Soc. Behav. Sci., 10, pp. 79–89.

[6] Van Beers, D., and Biswas, W. K., 2008, “A Regional Synergy Approach to Energy Recovery: The Case of the Kwinana Industrial Area, Western Australia,” Energy Convers. Manage., 49(11), pp. 3051–3062.

[7] Hui, C. W., and Ahmad, S., 1994, “Total Site Heat Integration Using the Utility System,” Comput. Chem. Eng., 18(8), pp. 729–742.

[8] Boix, M., Montastruc, L., Azzaro-Pantel, C., and Domenech, S., 2015, “Optimization Methods Applied to the Design of Eco-Industrial Parks: A Liter- ature Review,” J. Cleaner Prod., 87, pp. 303–317.

[9] Wang, L., Ma, Y., Zhang, J., Zhang, X., and Liu, Y., 2015, “Uncertainty Quan- tification and Structural Reliability Estimation Considering Inspection Data Scarcity,” ASCE-ASME J. Risk Uncertainty Eng. Syst, Part A: Civil Eng., 1(2), p. 04015004.

[10] Afshari, H., Peng, Q., and Gu, P., 2016, “Design Optimization for Sustainable Products Under Users’ Preference Changes,” ASME J. Comput. Inf. Sci. Eng., 16(4), p. 041001.

[11] Afshari, H., and Peng, Q., 2015, “Modeling and Quantifying Uncertainty in the Product Design Phase for Effects of User Preference Changes,” Ind. Manage. Data Syst., 115(9), pp. 1637–1665.

[12] Wynn, D. C., Grebici, K., and Clarkson, P. J., 2011, “Modelling the Evolution of Uncertainty Levels During Design,” Int. J. Interact. Des. Manuf., 5(3), pp. 187–202.

[13] Engelhardt, R., Eifler, T., Mathias, J., Kloberdanz, H., Birkhofer, H., and Bohn, A., 2011, “An Approach to Analyse Uncertainties in the Product Development Process,” International Conference on Engineering Design (ICED11), Copenha- gen, Denmark, pp. 433–442.

021002-10 / Vol. 3, JUNE 2017 Transactions of the ASME

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019

[14] Saravi, M., Goh, Y. M., Newnes, L., and Mileham, A., 2011, “Modeling Uncer- tainty in Through-Life Costing at the Early Design Stages-I,” ASME Paper No. DETC2011-47863.

[15] Eifler, T., Engelhardt, R., Mathias, J., Kloberdanz, H., and Birkhofer, H., 2010, “An Assessment of Methods to Analyse Uncertainty in Different Stages of the Development Process,” ASME Paper No. IMECE2010-39126.

[16] Anderi, R., Mecke, K., Sprenger, A., and Weitzmann, O., 2010, “Ontology- Based System for the Control of Uncertainty in the Product Lifecycle,” TMCE Symposium, Ancona, Italy, April 12–16, pp. 1259–72.

[17] Zhu, J., and Ruth, M., 2013, “Exploring the Resilience of Industrial Ecosystems,” J. Environ. Manage., 122, pp. 65–75.

[18] Korhonen, J., and Seager, T. P., 2008, “Beyond Eco-Efficiency: A Resilience Perspective,” Bus. Strategy Environ., 17(7), pp. 411–419.

[19] Ayyub, B. M., 2015, “Practical Resilience Metrics for Planning, Design, and Decision Making,” ASCE-ASME J. Risk Uncertainty Eng. Syst. Part A: Civil Eng., 1(3), p. 04015008.

[20] Kastner, C., Lau, R., and Kraft, M., 2015, “Quantitative Tools for Cultivating Symbiosis in Industrial Parks; A Literature Review,” Appl. Energy, 155, pp. 599–612.

[21] Dimian, A. C., Bildea, C. S., and Kiss, A. A., 2014, “Pinch Point Analysis,” Comp. Aided Chem. Eng., 35, pp. 525–564.

[22] Flower, J. R., and Linnhoff, B., 1979, “Thermodynamic Analysis in the Design of Process Networks,” Comput. Chem. Eng., 3(1), pp. 283–291.

[23] Hasani, A., and Khosrojerdi, A., 2016, “Robust Global Supply Chain Network Design Under Disruption and Uncertainty Considering Resilience Strategies: A Parallel Memetic Algorithm for a Real-Life Case Study,” Transp. Res. Part E: Logist. Transp. Rev., 87, pp. 20–52.

[24] Dhole, V. R., and Linnhoff, B., 1993, “Total Site Targets for Fuel, Co-Generation, Emissions, and Cooling,” Comput. Chem. Eng., 17(Supplement 1), pp. 101–109.

[25] Mar�echal, F., and Kalitventzeff, B., 1998, “Energy Integration of Industrial Sites: Tools, Methodology and Application,” Appl. Therm. Eng., 18(11), pp. 921–933.

[26] Kenney, W. F., 1984, Energy Conservation in the Process Industries, Academic Press, San Diego, CA.

[27] Hackl, R., Harvey, S., and Andersson, E., 2010, “Total Site Analysis (TSA),” Chalmers University of Technology, Stenungsund, Sweden.

[28] Shenoy, U. V., 2010, “Targeting and Design of Energy Allocation Networks for Carbon Emission Reduction,” Chem. Eng. Sci., 65(23), pp. 6155–6168.

[29] Chae, S. H., Kim, S. H., Yoon, S.-G., Chae, S. H., and Park, S., 2010, “Optimization of a Waste Heat Utilization Network in an Eco-Industrial Park,” Appl. Energy, 87(6), pp. 1978–1988.

[30] Taskhiri, M. S., Behera, S. K., Tan, R. R., and Park, H.-S., 2015, “Fuzzy Opti- mization of a Waste-to-Energy Network System in an Eco-Industrial Park,” J. Mater.Cycles Waste Manage., 17(3), pp. 476–489.

[31] Olesen, S., and Polley, G., 1996, “Dealing With Plant Geography and Piping Constraints in Water Network Design,” Process Saf. Environ. Prot., 74(4), pp. 273–276.

[32] Foo, D. C. Y., 2009, “State-of-the-Art Review of Pinch Analysis Techniques for Water Network Synthesis,” Ind. Eng. Chem. Res., 48(11), pp. 5125–5159.

[33] Boix, M., Montastruc, L., Pibouleau, L., Azzaro-Pantel, C., and Domenech, S., 2012, “Industrial Water Management by Multiobjective Optimization: From Individual to Collective Solution Through Eco-Industrial Parks,” J. Cleaner Prod., 22(1), pp. 85–96.

[34] Cimren, E., Fiksel, J., Posner, M. E., and Sikdar, K., 2012, “Material Flow Opti- mization in by-Product Synergy Networks,” J. Ind. Ecol., 15(2), pp. 315–332.

[35] Haslenda, H., and Jamaludin, M. Z., 2011, “Industry to Industry By-Products Exchange Network Towards Zero Waste in Palm oil Refining Processes,” Resour., Conserv. Recycl., 55(7), pp. 713–718.

[36] Khosrojerdi, A., Zegordi, S. H., Allen, J. K., and Mistree, F., 2016, “A Method for Designing Power Supply Chain Networks Accounting for Failure Scenarios and Preventive Maintenance,” Eng. Optim., 48(1), pp. 154–172.

[37] Sokka, L., Pakarinen, S., and Melanen, M., 2011, “Industrial Symbiosis Con- tributing to More Sustainable Energy Use—An Example From the Forest Indus- try in Kymenlaakso, Finland,” J. Cleaner Prod., 19(4), pp. 285–293.

[38] Gu, C., Yassine, A., Estel, L., and Leveneur, S., 2013, “A Multiobjective Opti- mization Model for Designing and Optimizing an Ecological Industrial Park,” 9th International Conference on Natural Computation (ICNC), pp. 595–600.

[39] Kim, S. H., Yoon, S.-J., Chae, S. H., and Park, S., 2010, “Economic and Environmental Optimization of a Multi-Site Utility Network for an Industrial Complex,” J. Environ. Manage., 91(3), pp. 690–705.

[40] Côt�ea, R. P., and Cohen-Rosenthal, E., 1998, “Designing Eco-Industrial Parks: A Synthesis of Some Experiences,” J. Cleaner Prod., 6(3–4), pp. 181–188.

[41] Brownbridge, G., Azadi, P., Smallbone, A. J., Bhave, A., Taylor, B. J., and Kraft, M., 2014, “The Future Viability of Algae-Derived Biodiesel Under

Economic and Technical Uncertainties,” Bioresour. Technol., 151, pp. 166–173.

[42] Liu, S., Gikas, P., and Papageorgiou, L. G., 2010, “An Optimisation-Based Approach for Integrated Water Resources Management,” Comput. Aided Chem. Eng., 28, pp. 1075–1080.

[43] Ross, A. M., and Hastings, D. E., 2006, “Assessing Changeability in Aerospace Systems Architecting and Design Using Dynamic Multi-Attribute Trade Space Exploration,” Collection of Technical Papers—Space 2006 Conference 1, pp. 551–568.

[44] Eckert, C., Clarkson, J. P., and Zanker, W., 2004, “Change and Customization in Complex Engineering Domains,” Res. Eng. Des., 15(1), pp. 1–21.

[45] Awudu, I., and Zhang, J., 2012, “Uncertainties and Sustainability Concepts in Biofuel Supply Chain Management: A Review,” Renewable Sustainable Energy Rev., 16(2), pp. 1359–1368.

[46] Afshari, H., Sharafi, M., Elmekkawy, T. Y., and Peng, Q., 2016, “Multi-Objective Optimization of Facility Location Decisions Within Inte- grated Forward/Reverse Logistics Under Uncertainty,” Int. J. Bus. Perform. Supply Chain Modell., 8(3), pp. 250–276.

[47] Lababidi, H. M. S., El-Wakeel, M. A., Alatiqi, I. M., and Al-Enzi, A. F., 2004, “Optimizing the Supply Chain of a Petrochemical Company Under Uncertain Operational and Economical Conditions,” Ind. Eng. Chem. Res., 43(1), pp. 63–73.

[48] Kang, J.-S., Chang, C.-C., Lee, D.-Y., and Lee, T.-Y., 2012, “Robust Optimiza- tion of Microgrids-An Application to Taichung Industrial Park,” 11th Interna- tional Symposium on Process Systems Engineering, Parts A and B, Vol. 31, pp. 1020–1024.

[49] Nagel, J. B., and Sudret, B., 2016, “Hamiltonian Monte Carlo and Borrowing Strength in Hierarchical Inverse Problems,” ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civil Eng., 2(3), p. B4015008.

[50] Jung, J. Y., Blau, G., Pekny, J. F., Reklaitis, G. V., and Eversdyk, D., 2004, “A Simulation Based Optimization Approach to Supply Chain Management Under Demand Uncertainty,” Comput. Chem. Eng., 28(10), pp. 2087–2106.

[51] Miranda, P. A., and Garrido, R. A., 2004, “Incorporating Inventory Control Deci- sions Into a Strategic Distribution Network Design Model With Stochastic Demand,” Transp. Res. Part E: Logist. Transp. Rev., 40(3), pp. 183–207.

[52] Khosrojerdi, A., Xiao, M., Sarikprueck, P., Allen, J. K., and Mistree, F., 2014, “Designing a System of Plug-in Hybrid Electric Vehicle Charging Stations,” ASME Paper No. DETC2013-12764.

[53] Sharafi, M., and ElMekkawy, T. Y., 2015, “Stochastic Optimization of Hybrid Renewable Energy Systems Using Sampling Average Method,” Renewable Sustainable Energy Rev., 52, pp. 1668–1679.

[54] Afshari, H., Sharafi, M., Elmekkawy, T. Y., and Peng, Q., 2014, “Facility Loca- tion Decisions Within Integrated Forward/Reverse Logistics Under Uncertainty,” Procedia CIRP, 17, pp. 606–610.

[55] Simangunsong, E., Hendry, L. C., and Stevenson, M., 2012, “Supply-Chain Uncertainty: A Review and Theoretical Foundation for Future Research,” Int. J. Prod. Res., 50(16), pp. 4493–4523.

[56] Skinner, D. J. C., Rocks, S. A., Simon, and Pollard, J. T., 2014, “A Review of Uncertainty in Environmental Risk: Characterising Potential Natures, Locations and Levels,” J. Risk Res., 17(2), pp. 195–219.

[57] Qiu, X., and Huang, G. Q., 2011, “On Storage Capacity Pooling Through the Supply Hub in Industrial Park (SHIP): The Impact of Demand Uncertainty,” IEEE International Conference on Industrial Engineering and Engineering Man- agement, pp. 1745–1749.

[58] Maes, T., Van Eetvelde, G., De Ras, E., Block, C., Pisman, A., Verhofstede, B., Vandendriessche, F., and Vandevelde, L., 2011, “Energy Management on Industrial Parks in Flanders,” Renewable Sustainable Energy Rev., 15(4), pp. 1988–2005.

[59] P�erez-Vald�es, G., Kaut, M., Nørstebø, V., and Midthun, K., 2012, “Stochastic MIP Modeling of a Natural Gas-Powered Industrial Park,” Energy Procedia, 26, pp. 74–81.

[60] Bergman, T. L., and Incropera, F. P., 2011, Introduction to Heat Transfer, 6th ed., Wiley, Hoboken, NJ.

[61] Giedt, W., 1971, Thermophysics, Van Nostrand Reinhold Company, New York. [62] Snyder, L. V., 2006, “Facility Location Under Uncertainty: A Review,” IIE

Trans., 38(7), pp. 547–564. [63] Gutjahr, W. J., and Reiter, P., 2010, “Bi-Objective Project Portfolio Selection and

Staff Assignment Under Uncertainty,” Optimization, 59(3), pp. 417–445. [64] Ben-Tal, A., El Ghaoui, L., and Nemirovski, A., 2009, Robust Optimization,

Princeton University Press, Princeton, NJ. [65] Paragon Decision Technology B.V., 2016, “AIMMS Optimization Package Software,”

AAIMS, Bellevue, WA, last accessed November 2016, www.aimms.com [66] Romero, E., and Ruiz, M. C., 2014, “Proposal of an Agent-Based Analytical

Model to Convert Industrial Areas in Industrial Eco-Systems,” Sci. Total Envi- ron., 468–469, pp. 394–405.

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering

JUNE 2017, Vol. 3 / 021002-11

D ow

nloaded from https://asm

edigitalcollection.asm e.org/risk/article-pdf/3/2/021002/6071826/risk_003_02_021002.pdf by The P

ennsylvania S tate U

niversity user on 17 O ctober 2019