Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico

Journal of Cleaner Production xxx (2015) 1e20

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico
n-Gosa
lbez b, c, Pascual Eduardo Murillo-Alvarado a, Gonzalo Guille
lez a,
María Ponce-Ortega a, *, Agustín Jaime Castro-Montoya a, Medardo Serna-Gonza Jose b
nez Laureano Jime a b c


s de Hidalgo, Morelia 58060, Michoaca
n, Mexico Chemical Engineering Department, Universidad Michoacana de San Nicola Departament d'Enginyeria Química (EQ), Universitat Rovira i Virgili (URV), Campus Sescelades, Avinguda Països Catalans 26, Tarragona 43007, Spain Centre for Process Integration, School of Chemical Engineering and Analytical Science, The University of Manchester, Manchester M13 9PL, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Received 7 August 2014 Received in revised form 12 August 2015 Accepted 13 August 2015 Available online xxx

The tequila industry in Mexico generates large amounts of lignocellulosic residues from the cultivation fields as well as from the tequila processing industries. Nowadays, these residues are disposed, but they could be used as feedstock for a biorefinery. Before implementing a biorefinery system to treat the residues from the tequila industry in Mexico, it is necessary to assess the economic and environmental performance of the entire supply chain for biofuel production. Developing a proper optimization framework for supply chain management in the tequila industry in Mexico that will consider all the activities involved along with the conflicting objectives of its daily operation represents a scientific challenge. Therefore, this paper presents a multi-objective optimization approach for designing such a supply chain that accounts for the simultaneous maximization of the net present value and environmental performance of the network. The environmental objective function accounts for impacts in ecosystem quality, human health and damage to resources. These are quantified through the ecoindicator 99 method. Numerical results show that the implementation of a biorefinery system in Mexico based on the residues from the tequila industry can provide significant economic and environmental benefits. Particularly, results indicate that the best economic solution shows a profit of 7.9
108 USD/year. Furthermore, the distributed system involving several central and distributed processing plants allows obtaining significant economic improvements. Finally, the results reported through Pareto curves allow identifying several solutions that are appealing for decision makers. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Biofuels Bioethanol Supply chain Multi-objective optimization Environmental impact Biogasoline

1. Introduction Biofuels have been the subject of intensive research due to their great potential to replace fossil fuels and decrease greenhouse gas emissions (Maes et al., 2015). Nowadays, there are several techniques for biofuels production. One of the most promising alternatives is to use residues containing lignocellulosic materials as feedstock (Liew et al., 2014). Several works have analyzed the benefits of this technology (Patrizi et al., 2013). Kazi et al. (2010) presented a techno-economic analysis comparing several routes for bioethanol production. Cardona et al. (2010) studied the production of bioethanol from sugar cane bagasse considering several

* Corresponding author. Tel.: þ52 443 3223500x1277; fax: þ52 443 3273584. E-mail address: [email protected] (J.M. Ponce-Ortega).

technologies and biological transformations. Albernas-Carvajal et al. (2014) proposed a model for the optimal design of the fermentation process for bioethanol production from molasses and hydrolyzed sugar cane bagasse. Alex-Marvin et al. (2012) presented an optimization study focusing on the net present value of five types of lignocellulosic biomass for ethanol production. ChouinardDussault et al. (2011) incorporated process integration into life cycle analysis for the production of biofuels and Patrizi et al. (2015) evaluated the emergy of bioethanol production. Most of the works mentioned above focused on assessing the economic performance of the technologies that produce biofuels as unique criterion. In practice, however, economic, environmental and social aspects need to be considered for a proper optimization of the network. Unfortunately, quantifying social aspects is challenging, and for this reason, they are typically left out of the analysis.

http://dx.doi.org/10.1016/j.jclepro.2015.08.052 0959-6526/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

2

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

The design of the supply chain associated to a biorefinery system aims to determine the optimal lignocellulosic material to be used as feedstock, the location and capacity of the biorefineries, the technologies used, the distribution of products as well as all the transportation tasks involved. As already mentioned, several objectives must be considered for a proper design of the network. Some research in the field of supply chains are the works proposed by Mele et al. (2009) that addressed the optimal planning of supply chains for bioethanol and sugar cane production considering economic and environmental concerns. In addition, Mele et al. (2011) reported a systematic optimization approach for designing and planning supply chains for the sugar cane industry accounting for economic and environmental aspects. Akgul et al. (2011) presented mixed-integer linear programming (MILP) models for designing bioethanol supply chains. Moreover, Karlsson et al. (2014) incorporated environmental issues in the assessment of bioethanol production. Previous analyses have considered only one product, ~ ez-Aguilar et al. (2011) proposed a multithis way Santiban objective optimization formulation for a supply chain associated to a biorefinery involving economic and environmental aspects, including multiple products. Previous approaches considered
n-Gosa
lbez deterministic information, but other authors like Guille and Grossmann (2010) proposed an approach for the optimal design of supply chains considering the environmental damage in the presence of uncertainty in the life cycle inventory of emissions, and Vujanovic et al. (2015) and Kostin et al. (2012) incorporated uncertainty in the analysis of biorefineries. It should be noted that targeting is an important issue that allows specifying the desired values of the objectives before the design is completed. In this way, Tay and Ng (2012) reported a targeting approach for designing integrated biorefineries. Integrated biorefineries incorporate the production of biofuels and energy, like it was done in the work by Lim et al. (2014), who designed integrated biorefineries based on rice. Identifying the proper objectives in designing biorefineries has become an important task. In this context, El-Halwagi et al. (2013) incorporated safety issues in the design of supply chains for biorefineries, and Martinez-Hernandez et al. (2014) reported an economic and environmental analysis for biorefineries. Process systems engineering is a research area that allows reducing the mass and energy consumption in an industrial process. In this field, Kelloway and Daoutidis (2014) presented a process systems approach for designing biorefineries, and Ng et al. (2015) presented an optimal planning for a bioenergy park. Furthermore, ~ ez-Aguilar et al. (2014) reported a study for the supply Santiban chains associated to biorefineries in Mexico, where the main bioresources required to yield biofuels were identified. Gong and You (2014) developed a global optimization approach for designing biorefineries. Their work provides alternative effective formulations to manipulate the nonconvex terms in the optimization approach. To improve the acceptability of biofuel, recently Yue and You (2014) included transfer price and revenue sharing in the production planning of biorefineries. Agostinho and Ortega (2013) presented an energetic and environmental assessment of bioethanol production in Brazil, and Moncada et al. (2014) reported the evolution from biofuels to integrated biorefineries in Colombia. Mota et al. (2014) analyzed supply chains in Portugal accounting for economic, environmental and social issues. Identifying the proper products from a biorefinery is an important issue. Kajaste (2014) reviewed the production of chemicals from biomass. It should be noted that the above-mentioned investigations have identified that there is a clear need to consider multiple objectives (such as economic, environmental and social aspects) in the optimal planning of supply chains associated to biofuels. Furthermore, previous works have highlighted the importance to consider the specific

available bioresources for each place to determine the optimal planning of supply chains associated to biofuels. Therefore, in the present study, and for the case of Mexico, the biomass obtained as waste from the tequila industry is considered as bioresource in the optimal design of a supply chain of biofuels in Mexico that is carried out considering several objectives simultaneously. The tequila industry, which is quite important in Mexico, generates many lignocellulosic materials that can be used as feedstock for biofuels production. Particularly, several tonnes of leaves (made of lignocellulosic material) are left in the fields, while many tonnes of plant heads (also lignocellulosic material) are discharged from the tequila factories. This lignocellulosic residue is called agave bagasse. For the specific case of Mexico, several approaches were reported for designing biorefinery systems. Saucedo-Luna et al. (2011) studied the optimal conditions of the saccharification of ~ ez et al. (2011) the agave bagasse for bioethanol production. Nún studied the economic viability of biofuels production from agave bagasse in Mexico. Murillo-Alvarado et al. (2014) reported an optimization approach for designing a biorefinery system based on residues from the tequila industry in Mexico. This study considered only the economic performance as main target. It should be noted that the above-mentioned approach did not assess the environmental impact of the associated supply chain. However, considering the environmental performance in the optimal design of a biorefinery system is of paramount importance for taking full advantage of the potential environmental benefits of biofuels production. This paper addresses the design of a biorefinery supply chain based on residues from the tequila industry in Mexico. The proposed approach considers simultaneously economic and environmental aspects. The later ones are quantified following the Econindicator 99, which covers the damage to resources, human health and ecosystem quality. The motivation for selecting this approach is that it considers several impacts in the aforementioned categories. Note, however, that the proposed optimization method is general enough to incorporate any other environmental assessment method. The design task is formulated as a multi-objective MILP model that includes different sets of equations reflecting capacity limitations, mass and energy balances and economic and environmental performance calculations. Numerical results show that it is possible to identify appealing designs achieving good economic and environmental performance. 2. Problem statement The problem under study is summarized in Fig. 1. A threeechelon supply chain is considered (production, storage, transportation to final consumers). Several sources of agave (i.e., the plant used to make tequila) are available, including the stalks from the harvested places and the residues generated by the tequila industry during the production of tequila. Potential places to install the central and distributed bioethanol processing facilities are available. The optimization model must determine the optimal location and size of such facilities. The problem then consists of determining the distribution for the agave residues, the required number, location and capacity of the biorefineries (along with the technology used in each of them), and the distribution of products and feedstocks. Fig. 2 depicts the superstructure of alternatives associated with the problem described above. This superstructure is a mathematical representation that includes all the potential places for feedstocks, biorefineries and markets. Different transportation options are also considered (i.e., truck, train and pipe, depicted in Fig. 2 by colors red, black and blue, respectively, in the web version). The

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

3

Fig. 1. Case study for the supply chain of bioethanol production in Mexico from agave bagasse. Distributed and central plants only differ in the size, but implement the same technology.

biorefineries implement the technologies shown in Fig. 3, which were already tested in a pilot plant, thereby generating the necessary experimental data for the calculations. The production process is as follows. The first process is the pre-treatment step, in

which the raw material is milled to obtain a desired size. The next stage is the hydrolysis process that is followed by a fermentation of the sugars obtained in the two previous stages. Finally, there is a separation step, which considers two stages of separation: pre-

Fig. 2. Superstructure for the biofuels supply chain based on residues from the tequila industry in Mexico. Capacities of the tequila factories are fixed. Input flows to the tequila factories are given, only output flows are optimized.

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

4

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

Fig. 3. Biorefinery processing from agave bagasse.

concentrated and dehydrated. Two types of biorefineries are considered: central and distributed plants, both implementing the same technology as shown in Fig. 3. These plants differ in the volume that can process (i.e., central plants show greater capacity). The time horizon is one year, and the problem is divided in different periods of one month each to account for the seasonality of the system. 3. Model formulation This section describes the proposed multi-objective optimization model of the biofuel supply chain based on residues of the tequila industry in Mexico. The model takes into account the following sets: i stands for harvesting areas, j is an index used to define distributed processing facilities, l is used to represent central processing facilities. Indexes j and l that represent the distributed and central process facilities are grouped into the index Plant for reducing the number equations. The potential sites of the industries for tequila production are associated to index k, while index m represents locations where bioethanol is consumed (a mixing center for bioethanol and gasoline in each location is considered). Index N1 represents places with solid fuel demand. The time periods are identified by index t, while the alternatives for the transportation of raw materials and products are defined by index w. The equations of the model are described in detail as follows. Balance in agave growing areas. For producing tequila, only the plant heads are considered. Therefore, in the growing areas the agave plants are cut to obtain the plant heads and discharge the stalks. Then, the mass of agave plant heads ðFPlantHeadsi;t Þ is determined from the total available agave ðFAgavei;t Þ and the corresponding separation fraction ðzAHi Þ:

FAgavei;t $zAHi ¼ FPlantHeadsi;t ;

c i2I; t2T

(1)

The waste discharged by the plant corresponding to the stalks ðFTotLeavBagi;t Þ is composed of lignocellulosic material that can be used to yield bioethanol. Hence, the total stalk obtained from the agave plant is determined considering the efficiency of separation zALi as follows:

FAgavei;t $zALi ¼ FTotLeavBagi;t ;

c i2I; t2T

(2)

Notice that the agave stalks are obtained from different fields from where it must be transported to the potential biorefineries. Two types of biorefineries are considered. The first one corresponds to central processing facilities, which are located in industrialized zones that correspond to locations close to tequila factories. The second one corresponds to distributed processing facilities that are located near to the fields or growing areas, but far from the industrialized zones. It should be noticed that the central processing facilities are located far from the fields. This structure increases the transportation costs but decreases the operating cost. On the other hand, the distributed facilities decrease the transportation cost for agave stalks, but the processing cost is increased. Then, the agave bagasse from stalks obtained in the different fields is distributed to the potential central and distributed biorefineries as follows:

FTotLeavBagi;t ¼

X

FLeavBagDistributed i;j;t

j2J

þ

X

FLeavBag Central ; i;l;t

c i2I; t2T

(3)

l2L

Maximum available agave. There is an upper limit on the agave available in each field ðFAgaveMAX i;t Þ, which depends mainly on the available land and water. The amount of agave used is therefore constrained to be lower than this limit, as stated in the following relationship:

FAgavei;t  FAgaveMAX i;t ;

c i2I; t2T

(4)

Balances in tequila industry. The plant heads from the fields ðFPlantHeadsi;t Þ can be distributed to the different tequila factories to produce tequila in the different time periods TequilaIndustry

ðFPlantHeadsi;k;t

FPlantHeadsi;t ¼

X

Þ as follows: TequilaIndustry

FPlantHeadsi;k;t

;

c i2I; t2T

k2K

(5) The total mass of plant heads in each tequila factory TequilaIndustry

ðFTotalPlantHeadsk;t

Þ is determined from the sum of plant

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

heads sent from the different fields i over each time period t as follows:

FTotalHeadsTequilaIndustry ¼ k;t

X FPlantHeadsTequilaIndustry ; i;k;t i2I

(6)

c k2K; t2T Residues of agave bagasse from the tequila industry. The tequila processing plants discharges significant amounts of lignocellulosic residues after processing the plant heads. This residue can be used in the biorefineries. Hence, the agave bagasse obtained from the tequila factories ðFTequilaBagassek;t Þ depends on an efficiency factor ðzCookedk Þ as follows:

FTequilaBagk;t ¼ FTotalHeadsTequilaIndustry $zCookedk ; c k2K; t2T k;t

5

Plant FMillPlant ¼ FReact Plant p;t $aReactor p p;t ;

c p2P; t2T

(12)

After the first reactor, a mix of hydrolyzed and no hydrolyzed compounds is separated through a filter. The hydrolyzed part is sent to a fermentation tank, while the no hydrolyzed part is directed towards a second hydrolysis reactor to obtain additional fermentable materials. The separation after the first reactor is modeled through Equations (13) and (14), where the sieve unit is used as filter: Plant FReact Plant ¼ Ffilt1Plant p;t $aFiltred1p p;t ;

c p2P; t2T

(13)

Plant ¼ Ffilt2Plant FReact Plant p;t $aFiltred2p p;t ;

c p2P; t2T

(14)

(7)

There is an additional filtering section to treat the effluent from the second hydrolysis reactor. Hence, the fermentable material

This agave bagasse that is obtained as residue from the tequila factories is sent to the biorefineries:

ðFFilteredSecondPlant p;t Þ obtained from Equation (16) is conducted to

FTequilaBagk;t ¼

X

solid fuel ðFFuelSolidPlant p;t Þ. The latter is a by-product in the global bioethanol production process, and is given by Equation (17):

FBagTequilaDistributed k;j;t

j2J

þ

X

FBagTequilaCentral ; k;l;t

one fermentation process, while the other part is considered as

c k2K; t2T

(8)

l2L

Balances in biorefineries. The equations for central and distributed processing facilities are grouped using the index Plant for both facilities. Hence, index p is the union of j and l. The following balance takes into account the total amount of bagasse that is pro-

Plant Ffilt2Plant ¼ FReact2Plant p;t $aReactor2p p;t ;

c p2P; t2T

Plant ¼ FFilteredSecondPlant FReact2Plant p;t $aFilteredsecond1p p;t ;

c p2P; t2T (16)

cessed in the biorefineries ðFTotalBag Plant p;t Þ, which is equal to the stalk bagasse from cultivation areas ðFLeavBagp;t Þ and the bagasse obtained from the tequila industry ðFBagTequilaPlant k;p;t Þ:

FTotalBagPlant p;t

¼

X

X

FLeavBag Plant i;p;t ;

Fermentable materials

c p2P; t2T

(9)

p2P

Fig. 3 shows the flowsheet of the technology implemented in the biorefineries, which is based on a pilot plant already under operation in the Universidad Michoacana de San Nicol
as de Hidalgo. In this pilot plant, both residues (i.e., the agro-residues and tequila process residues) have been used as feedstock. Furthermore, the optimal operating conditions for both residues have been identified. In addition, several pretreatment options have been assessed. Experimental results show that the best option is to combine first an acid hydrolysis and then an enzymatic hydrolysis for these residues. Both routes have been accounted for in the LCA. It should be noted that currently there is no single bioethanol processing facility under operation in Mexico, but the Mexican Government is indeed interested in building them, which motiaPlant p

vates the present study. Hence, realistic performance factors for each processing stage are defined based on the aforementioned experimental data. The steps involved in the bioethanol production are described in Equations (12)e(23). First, the agave bagasse is preprocessed to decrease the particle size through a milled process: Plant FTotalBagPlant ¼ FMillPlant p;t $aMillp p;t ;

c p2P; t2T

(10)

The milling process produces a juice rich in sugars that can be processed to produce bioethanol: Plant FTotalBagPlant p;t $aJuicep

¼

FJuicePlant p;t ;

Plant FReact2Plant ¼ FFuelSolidPlant p;t $aFilteredSecond2p p;t ; c p2P; t2T

(17)

FBagTequilaPlant k;p;t

k2K

þ

c p2P; t2T

(15)

(11)

The flowrate from milling is treated in a first hydrolysis reactor.

agave bagasse

ðFJuicePlant p;t Þ,

ðFFerment Plant p;t Þ

are juices from primary

and effluents from the first ðFfilt1Plant p;t Þ

and second ðFFilteredSecondPlant p;t Þ hydrolysis reactors (only the hydrolyzed part is considered). These fermentable materials are directed to a fermentation process to obtain bioethanol from sugar. The produced bioethanol is sent to a distillation column that is known as concentrator column:

FFerment Plant ¼ FJuicePlant þ Ffilt1Plant p;t p;t p;t þ FFilteredSecondPlant p;t ;

c p2P; t2T

Plant ¼ FCol1Plant FFerment Plant p;t ; p;t $aFermentationp

(18)

c p2P; t2T (19)

Materials from the concentrator column are sent to a second column that is used to dehydrate the bioethanol (note that one major limitation of the use of bioethanol as fuel is its humidity content when it is blended with hydrocarbons and used in engines without changes). After the dehydrator column, the product is stored in a tank that distributes the product to the consumers according to the associated demand: Plant FCol1Plant ¼ FCol2Plant p;t aDistillp p;t ;

c p2P; t2T

Plant ¼ FStockPlant FCol2Plant p;t aDehydratedp p;t ;

c p2P; t2T

(20) (21)

Distribution of products from processing plants to markets. The total bioethanol obtained from the distributed plants ðFStockPlant p;t Þ is

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

6

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

then collected and sent to the markets ðgprodBioet Distributed Þ as j;m;t follows:

X

FStockPlant ¼ p;t

gprodBioet Plant p;m;t ;

c j2J; t2T

(22)

GSolidFuelMIN n2;t  GSolidFueln2;t  GSolidFuelMAX n2;t ;

m2M

Additional equations enforce that the total bioethanol ðGBioethanolm;t Þ is equal to the bioethanol obtained in all the plants ðgprodBioet Plant p;m;t Þ in each time period t:

X

Upper and lower limits for the required solid fuel in each market are also imposed:

gprodBioet Plant p;m;t ¼ GBioethanolm;t ;

c m2M; t2T

(23)

j2J

c n22N2; t2T

(31)

Cost of the processing plants. The following equation is used to calculate the operating cost for the biorefineries considering all the processing steps involved:

CostPlantOP ¼

X CoperPlantMill þ CoperPlantReactor p p p2P

The total demand of fuels in each market can be satisfied by either pure gasoline or a mixture of gasoline and bioethanol. Hence, the total fuel demand ðTotalFuelm;t Þ in each market m takes into account the flow of gasoline ðGasoPurem;t Þ and the flow of a gasoline mix ðFuelMixm;t Þ. The gasoline mix contains 20% of bioethanol and 80% of gasoline, which is needed for engines without modifications. This is modeled via the following constraint:

TotalFuelm;t ¼ GasoPurem;t þ FuelMixm;t ;

þ CoperPlantTank2 þ CoperPlantColumn p p þ CoperPlantColumn2 þ CoperPlantTankA p p

c m2M; t2T

(25)

The following constraint is used to define the amount of gasoline in the mix:

c m2M; t2T

(26)

The total amount of gasoline consumed is determined as follows:

TotalGasolinem;t ¼ GasoPurem;t þ GasoMixm;t ;

 (32)

Note that the model must sell all the ethanol produced in the plants, as there is no option to store the biofuel. Hence, the model needs to decide whether it is more convenient to cover the fuel demand using pure gasoline or a mixture that contains gasoline blended with bioethanol. As will be shown later in more detail, gasoline is better from an economic perspective, while the mixture gasoline/bioethanol shows lower environmental impact. The following equation defines the percentage of bioethanol in the mixture, which is covered with the amount of bioethanol obtained in the distributed and central biorefineries ðGBioethanolm;t Þ:

FuelMixm;t $0:8 ¼ GasoMixm;t ;

þ CoperPlantSieve2 þ CoperPlantTank p p

c m2M; t2T (24)

FuelMixm;t $0:2 ¼ GBioethanolm;t ;

þ CoperPlantReactor2 þ CoperPlantSieve p p

c m2M; t2T (27)

Where, the operating costs for the different processing steps are obtained multiplying the unitary cost by the amount of material processed:

CoperPlant Mill ¼ p

X

UCostPlant OpMill $FTotalBagPlant p p;t ;

c p2P

t2T

(33) ¼ CoperPlant Reactor p

X

UCostPlant OpReactor $FMillPlant p p;t ;

c p2P

t2T

(34) ¼ CoperPlant Reactor2 p

X

UCostPlant OpReactor2 $Ffilt2Plant p p;t ; c p2P

t2T

(35) ¼ CoperPlant Sieve2 p

X

UCostPlant OpSieve2 $FReact2Plant p;t ; p

c p2P

t2T

(36) ¼ CoperPlant Sieve p

X

UCostPlant OpSieve $FReact Plant p p;t ;

c p2P

t2T MIN Maximum ðTotalFuelMAX m;t Þ and minimum ðTotalFuelm;t Þ demand limits are imposed in each market and time period to ensure a minimum demand satisfaction level.

MAX TotalFuelMIN m;t  TotalFuelm;t  TotalFuelm;t ;

(37) ¼ CoperPlant Tank p

X t2T

c m2M; t2T

c p2P

(28) Distribution of solid fuel from biorefineries to markets. The solid fuel produced in central and distributed plants can be sent to markets n2 (the set m represents the markets for gasoline and the set n2 represents the markets for solid fuel) as follows:

FFuelSolidPlant p;t

¼

X

gSolidFuelPlant p;n2;t ;

c l2L; t2T

(38) CoperPlant Tank2 ¼ p

X

UCostPlant OpTank2 $FFerment Plant p;t ; c p2P p

t2T

(39)

(29)

n22N2

The total solid fuel in each market is determined as follows:

X

  ; UCostPlant OpTank $ Ffilt1Plant þ FJuicePlant p p;t p;t

¼ CoperPlant Column p

X

UCostPlant OpColumn $FCol1Plant p p;t ;

c p2P

t2T

gSolidFuelPlant p;n2;t ¼ GSolidFueln2;t ;

c n22N2; t2T

(30)

(40)

p2P

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

CoperPlant Column2 ¼ p

X

UCostPlant OpColumn2 $FCol2Plant p p;t ; c p2P

7

FReact2CAP Plant  FReact2Plant p p;t ;

c p2P; t2T

(48)

t2T

(41) CoperPlant TankA ¼ p

X

UCostPlant OpTankA $FStockPlant p p;t ; c p2P

 FFerment Plant FFermentCAP Plant p p;t ;

c p2P; t2T

(49)

FCol1CAP Plant  FCol1Plant p p;t ;

c p2P; t2T

(50)

FCol2CAP Plant  FCol2Plant p p;t ;

c p2P; t2T

(51)

t2T

(42) In the same way, the capital cost accounts for the fixed and variable costs for each processing unit as follows:

2

 FStockPlant FStockCAP Plant p p;t ;

2 3 :YpPlant 6 7 Plant MIN Plant 6 7 FTotBagCAP  FTotBagCAP 7 p p 6 7 6 FTotBagCAP Plant ¼ 07 MAX 6 7 6 p 7 6 7 6 FTotBagCAP Plant  FTotBagCAP Plant p p 6 7 Mill 6 7 6 CPlant p ¼ 0 7 Mill Mill Mill Plant 6 7 6 7 CPlant ¼ CFixPlant þ CVarPlant $FFTotBagCAP 6 7 6 Reactor p p p p 7 6 7 ¼ 0 CPlant p 6 7 6 CPlant Reactor ¼ CFixPlant Reactor þ CVarPlant Reactor $FMillCAP Plant 7 6 p p p p 6 7 6 CPlant Reactor2 ¼ 0 7 7 6 7 6 p 7 6 CPlant Reactor2 7 6 ¼ CFixPlant Reactor2 þ CVarPlant Reactor2 $Ffilt2CAP Plant 7 p p p p 6 7∨6 CPlant Sieve ¼ 0 7; p 6 7 6 Sieve Sieve Sieve Plant 7 CPlant ¼ CFixPlant þ CVarPlant $FReactCAP 6 7 6 p p p p 6 7 ¼0 7 CPlant Sieve2 7 p 6 CPlant Sieve2 ¼ CFixPlant Sieve2 þ CVarPlant Sieve2 $FReact2CAP Plant 7 6 7 Tank 6 7 6 p p p p 6 7 CPlant ¼ 0 6 7 6 p Tank Tank Plant 7 6 CPlant Tank 7 ¼ CFixPlant þ CVarPlant $FFermentCAP 6 p p p p 6 7 6 CPlant Column ¼ 0 7 7 p 6 7 Column 7 ¼ CFixPlant Column þ CVarPlant Column $FCol1CAP Plant 6 CPlant p 7 6 Column2 p p p 7 6 7 6 CPlant ¼ 0 4 5 p 6 CPlant Column2 ¼ CFixPlant Column2 þ CVarPlant Column2 $FCol2CAP Plant 7 4 5 TankA p p p p CPlant ¼ 0 p CPlant TankA ¼ CFixPlant TankA þ CVarPlant TankA $FStockCAP Plant p p p p

X CPlantMill þ CPlantReactor p p

þ þ

CPlantTank p

þ

CPlantSieve p

CPlantColumn p

þ 

þ

þ

CPlantColumn2 p

(43) The capital cost for each unit is modeled using a disjunction. To yPlant p

is used to determine the optimal this end, the binary variable location for the distributed or central plants. In addition to the disjunctive term, it is necessary to define the constraints to calculate the capacity of the processing units. The capacity for each unit is given by the maximum flowrate passing through it during all the time periods.

c p2P; t2T

(53)

variable YpPlant inside the disjunction) is one when the distributed or

CPlantSieve2 p

þ CPlantTankA p

FTotBagCAP Plant  FTotBagPlant p p;t ;

c p2P

The binary variable yPlant (which corresponds to the Boolean p

p2P

CPlantReactor2 p

(52)

3

YpPlant

CostPlantCAP ¼ KF

c p2P; t2T

central plant p is installed, and zero otherwise. If the distributed or central plant is installed, then the amount of processed agave bagasse must lie within upper and lower limits:

FTotBagPlant p

MIN

$yPlant  FTotBagCAP Plant ; p p

 FTotBagCAP Plant FTotBagCAP Plant p p

MAX

c p2P

$yPlant ; p

c p2P

(54) (55)

Binary variable yPlant is also used to calculate the capital cost for p each processing unit, where the fixed part of the capital cost is multiplied by the binary variable and the variable part is equal to a unitary cost multiplied by the greatest flowrate processed during the time horizon.

(44) 3.1. Transportation costs

 FMillPlant FMillCAP Plant p p;t ;

c p2P; t2T

(45)

 Ffilt2Plant Ffilt2CAP Plant p p;t ;

c p2P; t2T

(46)

FReactCAP Plant  FReact Plant p p;t ;

c p2P; t2T

(47)

This section describes how the transportation costs are obtained for every transportation mode w (i.e., tank, train, and duct). Transportation cost from cultivation areas i to distributed plants j. TDIS is defined for modeling the transportation Boolean variable Yi;j;w costs associated to bagasse from cultivation areas i to distributed plants j. Hence, when the corresponding binary variable ðyTDIS i;j;w Þ

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

8

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

takes a value of one, the transportation costs are activated considering only one of the transportation options:

3 TDIS Yi;j;w 6 7 X ∨ 4 5 CostUnitTransi;j;w $FLeavBag Distributed w2W CostTransi;j ¼ i;j;t

X

FLeavBagCentral ¼ i;l;t CostTransi;l ¼

X

DCostTransi;l;w ;

c i2I; l2L

DCostTransi;l;w ¼

X

CostUnitTransi;l;w $DFLeavBagCentral i;l;t;w ;

t2T

c i2I; l2L; w2W

(56)

(67)

Then, at most one transportation option w for each segment can be selected as follows:

yTDIS i;j;w

 1;

c i2I; j2J

(57)

Central DFLeavBagCentral i;l;t;w  FLeavBag i;l;t;w

The continuous variables involved in the disjunction are disaggregated as follows:

X

FLeavBagDistributed ¼ i;j;t

DFLeavBagDistributed ; c i2I; j2J; t2T i;j;t;w

w2W

(58) X

DCostTransi;j;w ;

c i2I; j2J

(59)

DCostTransi;j;w ¼

X

$yTCEN i;l;w ;

TCEN DCostTransi;l;w  CostTransMAX i;l;w $yi;l;w ;

(68)

c i2I; l2L; w2W (69)

Transportation cost for biomass from tequila industry k to distributed plants j. In the modeling of the transportation costs for the biomass from the tequila industry k to distributed biorefineries TDIS and disjunction are used: j, the following Boolean variable Yk;j;w

2

w2W

And the relationships are stated in terms of the disaggregated variables:

MAX

c i2I; l2L; t2T; w2W

w2W

CostTransi;j ¼

(66)

w2W

t2T

X

(65)

w2W

2

c i2I; j2J

DFLeavBagCentral i;l;t;w ; ci2I; l2L; t2T

6 ∨ 4

w2W

CostTransk;j ¼

X

3

TDIS Yk;j;w

CostUnitTransk;j;w $FBagTequilaDistributed k;j;t

t2T

c k2K; j2J

Distributed CostUnitTransi;j;w ,DFLeavBagi;j;t;w ;

t2T

(70)

ci2I; j2J; w2W (60) Upper bounds are defined for the disaggregated variables:

DFLeavBagDistributed i;j;t;w

7 5



MAX FLeavesBagDsitributed $yTDIS i;j;t;w i;j;w ;

TDIS DCostTransi;j;w  CostTransMAX i;j;w $yi;j;w ;

X (61)

c i2I; j2J; t2T; w2W

This disjunction is also reformulated in the same way using the binary variables yTDIS as follows: k;j;w

yTDIS k;j;w  1;

c k2K; j2J

(71)

w2W

¼ FBagTequilaDistributed k;j;t

X

DFBagTequilaDistributed ; k;j;t;w

w2W

c i2I; j2J; w2W

(72)

c k2K; j2J; t2T (62)

Transportation cost from cultivation areas i to central biorefineries TCEN is used in the following disjunction for l. The Boolean variable Yi;l;w modeling the transportation costs for the bagasse from the cultivation fields i to the centralized biorefineries:

CostTransk;j ¼

X

DCostTransk;j;w ;

c k2K; j2J

(73)

w2W

DCostTransk;j;w ¼

X

CostUnitTransk;j;w $DFBagTequilaDistributed ; k;j;t;w

t2T

2

TCEN Yi;l;w

c k2K; j2J; w2W

3

(74)

6 7 X ∨ 4 5 CostTransi;l ¼ CostUnitTransi;l;w $FLeavBag Central i;l;t

w2W

MAX

 FBagTequilaDistributed DFBagTequilaDistributed k;j;t;w k;j;t;w

t2T

(75)

c i2I; l2L (63) Following a similar procedure as before, the disjunction is reformulated using the binary variables yTCEN into a set of algebraic i;l;w equations as follows:

X w2W

yTCEN i;l;w  1;

c i2I; l2L

$yTDIS k;j;w ;

(64)

c k2K; j2J; t2T; w2W TDIS DCostTransk;j;w  CostTransMAX k;j;w $yk;j;w ;

c k2K; j2J; w2W (76)

Transportation cost for bagasse from tequila industry k to central biorefineries l. For modeling the transportation costs for bagasse from tequila industry k to central biorefineries l, the Boolean variTCEN and the following disjunction are used: able Yk;l;w

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

2 6 ∨ 4

w2W

X

CostTransk;l ¼

3

TCEN Yk;l;w

7 5 CostUnitTransk;l;w $FBagTequilaCentral k;l;t

DCostTransl;m;w ¼

X

9

CostUnitTransl;m;w $DgprodBioet Central l;m;t;w ;

t2T

c l2L; m2M; w2W

t2T

c k2K; l2L

(88) (77)

which is reformulated using the binary variables yTCEN in the same k;l;w

Central gprodBioet Central l;m;t;w  gprodBioet l;m;t;w

yTCEN k;l;w  1;

c k2K; l2L

(78)

TCENMIX ; DCostTransl;m;w  CostTransMAX l;m;w $yl;m;w

FBagTequilaCentral ¼ k;l;t

DFBagTequilaCentral k;l;t;w ; c k2K; l2L; t2T

w2W

(79) CostTransk;l ¼

X

DCostTransk;l;w ;

c k2K; l2L

(80)

DCostTransk;l;w ¼

Transportation cost for bioethanol from distributed biorefinery j to mixing center m. For modeling the transportation costs for bioethanol from distributed biorefinery j to market m, the Boolean TDISMIX and the following disjunction are used: variable Yj;m;w

2 6 ∨ 4

w2W

w2W

CostTransj;m ¼

X

3

TDISMIX Yj;m;w

CostUnitTransj;m;w $gprodBioet Distributed j;m;t

c j2J; m2M

CostUnitTransk;l;w $DFBagTequilaCentral k;l;t;w ;

(91)

c k2K; l2L; w2W (81)

The disjunction is reformulated using the binary variables yTDISMIX as follows: j;m;w

X DDFBagTequilaCentral k;l;t;w



MAX FBagTequilaCentral $yTCEN k;l;t;w k;l;w ;

(82)

c k2K; l2L; t2T; w2W T DCostTransk;l;w  CostTransMAX k;l;w $yk;l;w ;

yTDISMIX  1; j;m;w

gprodBioet Distributed ¼ j;m;t

the following disjunction are used:

CostTransl;m ¼

X

DgprodBioet Distributed ; j;m;t;w

(93)

c j2J; m2M; t2T CostTransj;m ¼

X

DCostTransj;m;w ;

c j2J; m2M

(94)

w2W

DCostTransj;m;w ¼

X

CostUnitTransj;m;w $DgprodBioet Distributed ; j;m;t;w

c j2J; m2M; w2W

7 Central 5

(95)

CostUnitTransl;m;w $gprodBioet l;m;t

t2T

MAX

 gprodBioet Distributed gprodBioet Distributed j;m;t;w j;m;t;w (84)

This disjunction is also reformulated using the binary variables yTCENMIX into algebraic equations as follows: l;m;w

yTCENMIX  1; l;m;w

X

t2T

3

TCENMIX Yl;m;w

c l2L; m2M

X

(92)

w2W

c k2K; l2L; w2W

Transportation costs for products from central biorefineries l to markets m. To calculate the transportation costs of bioethanol from TCENMIX and central biorefinery l to market m, the Boolean variable Yl;m;w

2

c j2J; m2M

w2W

(83)

w2W

7 5

t2T

X t2T

6 ∨ 4

(89)

(90)

c l2L; m2M; w2W

w2W

X

$yTCENMIX ; l;m;w

c l2L; m2M; t2T; w2W

way as before as follows:

X

MAX

c l2L; m2M

$yTj;m;w ;

c j2J; m2M; t2T; w2W TDISMIX DCostTransj;m;w  CostTransMAX j;m;w $yj;m;w ;

c j2J; m2M; w2W

(96)

(97)

(85)

w2W

3.2. Economic objective function

¼ gprodBioethanolCentral l;m;t

X

DgprodBioet Central l;m;t;w ;

w2W

(86)

c l2L; m2M; t2T

CostTransl;m ¼

X w2W

DCostTransl;m;w ;

c l2L; m2M

(87)

The economic objective function corresponds to the maximization of the overall profit, which accounts for the sales minus the costs. The total cost takes into account all the costs involved in the proposed model, including the operating cost, the capital cost and the transportation cost for distributed and central biorefineries (considering the best transportation way for the raw materials and products obtained). The following equation is used to calculate the total cost:

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

10

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

0 TotCost ¼ @CostDist þ

OP

XX

þ CostDist

CAP

CostTransi;j þ

i2I j2J

þ

XX

XX

CostTransi;l þ

XX

XX

þ

CostTransk;j þ

XX

CostTransk;l þ

XX

CostTransl;m þ

X X

þ

ImpacTranspBagasse ¼

CostFixTransk;l

X X

CostTransj;m þ

X X

þ

CostFixTransl;m

þ

X XX

dTransport FBagTequilaCentral k;l;t

CostFixTransm;n1 A

In this equation, ImpacTranspBagasse represents the environmental impact associated with the transportation of the raw materials, while dTransport is a parameter that represents the damage caused per unit of mass transported between the source and the origin. This parameter is retrieved from the eco-invent database. Note that the information taken from the eco-invent database is used to assess the impact of the main activities in the supply chain. In transportation, it is assumed that the environmental impact is a function of the amount of material transported and the fuel used by the transportation mode. The calculation of the impact of ethanol production considers the environmental impacts associated to the processing steps, which include milling and distillation. Finally, for the gasoline, both the production and combustion stages are considered. The Eco-invent database was used to determine the emissions from the activities involved in the supply chain operation. When possible, data from processes in Mexico were considered (i.e., the emissions of the Mexican electricity mix). On the contrary, for other general technologies for which specific information is missing in the Eco-invent database, standard values available in the database as proxy of the impact of the technologies implemented in Mexico were used. This is a standard approach widely followed in the literature. The environmental impact associated with the transportation of products from distributed and central biorefineries to markets is calculated as follows:

GBioethanolm;t $bBioethanol m

X X

GSolidFueln2;t $bSolidFuel n2

n22N2 t2T

X X

dTransport FBagTequilaDistributed k;j;t

(101)

m2M t2T

þ

X XX

1

On the other hand, the total sales include the sales for bioethanol, solid fuel and biogasoline. In this case, the amount sold in each market over each time period is multiplied by the factor bm, which represents the unit price for each product (bioethanol, solid fuel and gasoline):

þ

dTransport FLeavBagCentral i;l;t

k2K l2L t2T

(98)

X X

XXX

k2K j2J t2T

CostFixTransj;m

m2M n12N1

TotSales ¼

þ

CostTransm;n1

X

dTransport FLeavBagDistributed i;j;t

i2I l2L t2T

j2J m2M

X

XXX i2I j2J t2T

l2L m2M

m2M n12N1

X

calculations are equivalent for every damage category). The overall subsystem has been divided into: transportation tasks, production of biofuel and production of gasoline. Damage associated with transportation. The environmental impact associated with the transportation of raw materials to distributed and central biorefineries is determined from the corresponding material flows and unitary damage parameters as follows:

k2K l2L

j2J m2M

þ

CostFixTransk;j

k2K j2J

k2K l2L X X

X

CostFixTransi;l

XX

l2L m2M

þ

CostFixTransi;j

i2I l2L

k2K j2J

þ

þ CostCent

CAP

i2I j2J

i2I l2L

þ

þ CostCent

OP

(99) !

TotalGasolinem;t $bGasoline m

m2M t2T

Finally, the economic objective function is to maximize the total annual profit, which is determined by the difference between total sales minus total costs:

PROFIT ¼ TotSales  TotCost

(100)

3.3. Environmental objective function

ImpacTranspProducts ¼

X X X

dTransport gprodBioet Distributed j;m;t

j2J m2M t2T

The environmental objective function corresponds to the minimization of the entire environmental impact measured through the Eco-indicator 99 method. The Eco-indicator 99 is a standard method for evaluating the global environmental impact of a process, product and/or activity (see Geodkoop and Spriensma, 2001). This method can be applied either as a standalone tool or combined with an optimization model. The proposed model integrates the Eco-indicator 99, whose calculation has been carried out considering the specific activities taking place in the operation of the considered biofuel supply chain. Particularly, this paper considers the damages to the ecosystem quality, human health and resources. The main sources of impact are the transportation tasks (for different segments and considering the best type of transport), and the production of gasoline and bioethanol (including all the related upstream processes, like those associated with the raw materials extraction and utilities generation). The calculation of the environmental impact in a generic category is described next (the

þ

X X X

dTransport gprodBioet Central l;m;t

l2L m2M t2T

þ

X X

dTransport TotalGasolinem;t

m2M t2T

(102) where ImpacTranspProducts represents the environmental impact associated with the transportation of the final products, that is, the transport of bioethanol obtained in the central and distributed plants, and the gasoline used to satisfy the demand in the markets. The environmental impact of bioethanol production is calculated as follows:

ImpactPlantPlant ¼

X X

dPlant FStockPlant p;t

(103)

p2P t2T

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

where ImpactPlantPlant represents the environmental impact of bioethanol production, which considers the impact for all the stages involved in the life cycle of the process (milling, hydrolysis reaction, fermentation, distillation). To obtain this information, the energy required by any processing stage is determined and then the corresponding environmental impact is evaluated using data retrieved from the eco-invent database. Note that the environmental impact associated with the biomass production has been neglected, since biomass is regarded as a byproduct of the tequila industry that is typically disposed if no further processing is carried out. The environmental impact of the gasoline accounts for the damage associated with its production and transportation as follows:

ImpactGasoline ¼

X X

dGasoline TotalGasolinem;t

m2M t2T

þ

X X

dTransport TotalGasolinem;t

(104)

m2M t2T

The total environmental impact takes into account the main sources of impact described above:

EnvImpact ¼ ImpacTranspBagasse þ ImpacTranspProducts þ ImpactPlantPlant þ ImpactGasoline

(105)

To calculate the environmental impact in each damage category (i.e., human health, ecosystem quality and resources), the same equations described above are used (Eqs. (101)e(105)), choosing the appropriate value of d according to the damage category being assessed in each case. Presently, there are several LCA methods for quantifying the environmental impact for any process or activity, including the EcoIndictor 99, CML-IA as well as ReCiPe 2008 methods. In a recent publication, it was shown that many environmental impacts tend to be highly correlated, as the same or similar substances ultimately
lez et al., 2015). Hence, the outcome of cause them (Pascual-Gonza the optimization may change depending on the particular impact being optimized, but no significant differences might be observed if the same impact is quantified following one LCA metric or the other. It should be noted that the proposed optimization formulation is general enough to incorporate any LCA method for determining the environmental impact. It is therefore the choice of the decision-maker to select a specific impact assessment method depending on the main environmental issues of concern. The optimization approach includes all the relationships previously described (Eqs. (1)e(105)), which are optimized simultaneously by the algorithm. This model can be solved using any multi-objective optimization algorithm available in the literature. Without loss of generality, this paper uses the epsilon-constraint method to generate the Pareto solutions of the problem. 4. Case study For applying the proposed model, first the places where there are available residues from the tequila industries are identified (tequila factories and harvesting places). For these places, the

11

amount of biomass residues in the different seasons of the year is first calculated. Then, the potential places for the location of the central and distributed bioethanol processing facilities are defined along with the potential transportation routes. Then, all the technical information for the bioethanol processing is obtained from the pilot plant (including unit conversion factors, needed external resources and unit costs). Furthermore, market information on bioethanol demand is obtained from the literature (demands and prices for the bioethanol in the different markets). With this information at hand, an LCA is conducted using the eco-invent database and the Eco-indicator 99 method. The proposed optimization formulation is coded in the software GAMS, which calculates the Pareto solutions through the epsilon constrain approach. Each point of the Pareto curve corresponds to a mixed-integer linear programming model that is solved with the CPLEX solver (Brooke et al., 2014). The case study used to illustrate the capabilities of the proposed approach is shown in Fig. 1. A generic supply chain is considered that accounts for different central and distributed biorefineries as well as cultivation areas and existing tequila factories. All the information on agave cultivation was obtained from the Department of Agriculture of Mexico (SAGARPA-SIAP, 2013). Several transportation modes were considered, namely, free road, toll road and train. The environmental impact was quantified following the Ecoindicator 99 method, and using information retrieved from the ecoinvent database. Table 1 summarizes the results of the LCA analysis. Several options for transporting raw materials and products were considered, including using toll and free highways, existing and new pipes as wells as existing and new roads. The transportation unit costs were obtained from information available. Gasoline demands in the different markets were obtained from SENER (2013). The minimum and maximum demands are 1.325% MAX for GSolidFuelMIN n2;t and 2.65% forGSolidFueln2;t of the total demand of the market. The fuel demand can be satisfied by using either pure gasoline or a mixture of 80% gasoline 20% ethanol. The MILP model contains 36,410 continuous variables, 1,224 binary variables and 35,414 constraints. Different scenarios were considered to illustrate the capabilities of the model. Particularly, to facilitate the analysis of the results, a set of bi-criteria problems in which the economic performance is traded-off against each single impact category separately are solved. Each bi-criteria model was calculated via the epsilon-constraint method by keeping the economic indicator in the objective function and transferring the environmental criterion to an auxiliary constraint. In addition, the calculations were repeated for several cases reflecting different realistic scenarios. Scenario 1. Supply chain considering the economic aspect and minimizing the environmental impact in ecosystem quality. In this scenario, two objectives are prioritized; the economic objective function and the environmental objective function associated to the ecosystem quality. Two options were considered: (i) satisfying the entire demand of fuels, and (ii) leaving free the demand satisfaction level. Scenario 1a (total demand of fuels satisfied). This option considers that the total demand is fully satisfied. Fig. 4 shows the Pareto points obtained via the epsilon constraint method. Three solutions

Table 1 Eco Invent data for the case study considering the Eco-indicator 99 method. Activity (unit)

Ecosystem quality (PDF m2 year)

Human health (DALYs)

Resources (MJ)

Transport (TKM) Bioethanol production (kg bioethanol) Gasoline (kg gasoline)

0.001598 0.0005665 0.001685

0.0050235 0.007843 0.021228

0.0082522 0.031235 0.18986

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

12

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

Fig. 4. Pareto curve for the minimization of the damage to ecosystem quality considering variable and fixed demand in markets. In the points A, B and C is satisfied the total demand of biofuel. The demand satisfaction level attained by points A0 , B0 and C0 are depicted in the figure.

that might be particularly appealing for decision-makers are discussed in further detail next. First, point A represents the solution with lowest damage to ecosystem quality but also with the lowest overall profit. Fig. 5 shows the supply chain configuration for point A, while Table 2 shows the associated materials flows. In this solution, central and distributed biorefineries produce 1.35
108 kg/ year of bioethanol. Pareto point B represents an intermediate solution. In this point, the bioethanol production decreases while the amount of gasoline produced increases. Fig. 6 shows the supply chain associated to this point, while Table 3 shows the material flows. The total profit in this solution is 7.5573
108 USD/year, while the total amount of bioethanol produced is 4.84
107 kg/year. Finally, solution in point C yields the maximum profit but also the highest damage to ecosystem quality. In this solution, the fuel demand is fully covered using pure gasoline, which leads to a total profit of 7.9346
108 USD/year. Scenario 1b (variable demand). In this section, the model was solved considering upper and lower limits on the demand. Fig. 4 shows the Pareto curve considering the variable demand. In this case, in the point A0 of the Pareto curve the demand was satisfied

mainly with gasoline and a small amount of bioethanol. In the final section (at the left hand side), the Pareto curve shows an elliptic behavior, which indicates the installation of central and distributed biorefineries; for this reason, the profit decreases at the same time that the environmental impact associated to the damage to the ecosystem quality decreases due to the use of bioethanol. Three solutions in this Pareto curve of Fig. 4 are further analyzed. Point A0 shows the lowest profit (3.43
108 USD/year), but also the lowest environmental impact. In this point, the demand satisfaction is 50%, with a total bioethanol production of 6.54
107 kg/year. In contrast, solution C0 shows the best economic performance. This point covers the demand using only gasoline, which leads to the highest profit (7.89
108 USD/year) with a demand satisfaction of 2.3% of the total demand of fuel in the markets. Fig. 7 represents the supply chain associated to solution of point B0 , which presents a profit of 3.901
108 USD/year, the total bioethanol produced is 4.81
106 kg/year, Table 4 shows the flows associated to the supply chain determined. Scenario 2. Supply chain considering the economic aspect and the environmental impact associated to the damage to the human health. This scenario considers the maximization of the profit and the minimization of the environmental impact associated to the damage to human health, leaving aside the damages to the ecosystem quality and resources. Two options are considered, the first one considering a fixed satisfied demand and the second one leaving free the demand satisfaction level. Scenario 2a: Total demand of biofuels satisfied. In this option, the multi-objective optimization model is solved considering a fixed demand in all the markets. Fig. 8 shows the Pareto curve for this option. Three solutions are identified in the Pareto curve of Fig. 8. Point A represents the solution with the lowest environmental impact associated in human health but at the same time with the lowest economic benefit. Fig. 5 shows the supply chain associated to Point A, which has a profit of 6.6775
108 USD/year and a total bioethanol production of 1.68
108 kg/year. Notice that this is the same configuration than the one obtained when minimizing the ecosystem quality damage. This supply chain installs the maximum number of central and distributed plants.

Fig. 5. Supply chain for the solution of point A of scenario 1.

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

13

Table 2 Flows of the supply chain for solution of the point A in scenario 1. Central plant

Bagasse from cultivation areas (kg/year)

Bagasse from tequila industry (kg/year)

Bioethanol obtained (kg/year)

Markets considered

1

3. 6.77
107

e

2.13
107

2

7. 5.97
107

e

1.88
107

3

1. 3.54
107, 3. 3.25
107

e

2.13
107

1. Aguascalientes, 6. Colima, 10. Durango, 14. Jalisco, 18. Nayarit, 24. Sinaloa. 7. Chiapas, 20. Oaxaca, 26. Tabasco, 28. Veracruz.
taro, 23. San Luis Potosí, 11. Guanajuato, 22. Quere 27. Tamaulipas.

Distributed plant 1 2 3

2. 2.57
106 4. 2.69
107 5. 6.72
106

e e 5. 1.41
108

8.09
105 8.45
106 4.64
107

4 5 6

6. 1.46
107 8. 1.01
107 9. 1.18
107

e e 6. 1.98
107

4.59
106 3.20
106 1.0
107

12. Guerrero.
xico, 16. Michoaca
n. 15. Me 9. Distrito Federal, 12. Guerrero, 13. Hidalgo,
xico, 17. Morelos, 21. Puebla. 15. Me 24. Sinaloa. 21. Puebla, 28. Veracruz. 19. Nuevo Leon, 30. Zacatecas

Fig. 6. Supply chain for solution of point B of scenario 1.

Table 3 Flows of the supply chain for solution of the point B in scenario 1. Central Plant

Bagasse from cultivation areas (kg/year)

Bagasse from tequila industry (kg/year)

Bioethanol obtained (kg/year)

Markets considered

1

3. 6.87
105 6. 2.06
106 1. 3.54
107

1. 1.67
107 2. 1.23
107 e

9.99
106

6. Colima, 14. Jalisco, 18. Nayarit.

1.11
10

107 106 106 107 107

e 5. 2.87
107

8.45
106 1.51
107


xico, 16. Michoac

taro. 15. Me an, 22. Quere
xico, 17. Morelos. 9. Distrito Federal, 15. Me

e

3.72
106

1. Aguascalientes, 30. Zacatecas

3 Distributed plant 2 3

6

4. 2. 5. 8. 9.

2.69 2.57 6.72 1.01 1.18








7


taro. 1. Aguascalientes, 11. Guanajuato, 22. Quere

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

14

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

Fig. 7. Supply chain for solution of the point B0 of scenario 1.

Table 4 Flows of the supply chain for solution of point B0 for ecosystem damage. central plant

Bagasse from cultivation areas (kg/year)

Bagasse from tequila industry (kg/year)

Bioethanol obtained (kg/year)

Markets considered

2 Distributed plant 1 4 6

e

4. 4.75
106

1.49
106

20. Oaxaca.

2. 2.57
106 6. 1.55
106 9. 5.44
106

e 6. 9.86
105

8.09
105 4.86
105 2.02
106

12. Guerrero. 18. Nayarit. 1. Aguascalientes, 30. Zacatecas

Point B represents an intermediate solution, whose configuration is presented in Fig. 9. This configuration installs only two central biorefineries. The total profit of this configuration is 7.29
108 USD/year and the amount of produced bioethanol is 9.46
107 kg/year. Table 5 shows the flows associated with the supply chain corresponding to point B0 .

Fig. 8. Pareto curve for scenario 2.

Finally, solution C attains the maximum profit using only pure gasoline, which leads to the highest environmental impact. Scenario 2b: Variable satisfied demand (50e10% of the demand of fuel). Fig. 8 shows the Pareto curve (line of red points, in the web version, representing the variable demand case) for the case in which maximum and minimum limits are imposed on the satisfied demand. The Pareto curve shows a similar behavior than the one of scenario 1. This is because the environmental objective function considers the same sources of environmental impact (transport, bioethanol production and gasoline use), which are quantified using different damage factors. Scenario 3. Supply chain considering the economic aspect and the environmental impact associated to the damage to resources. This Scenario 3 presents the solution considering the maximization for the profit and the minimization of the environmental impact associated to the damage to the resources, and Fig. 10 shows the Pareto curve associated to this scenario with a fixed (line of blue points, in the web version) and variable (line of red points, in the web version) demand in markets. Scenario 3a: The Pareto curve of Fig. 10 (fixed demand) shows a pseudo linear behavior due to the environmental impact considering the damage to the resources for the used gasoline and the bioethanol production are similar. This way, the demand of each

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

15

Fig. 9. Supply chain for solution of the point B0 of scenario 2.

Table 5 Flows of the supply chain for solution B0 of scenario 2. Central plant

Bagasse from cultivation areas (kg/year)

Bagasse from tequila industry (kg/year)

Bioethanol obtained (kg/year)

Markets considered

1

e

1. 2.18
107 2. 4.84
107 3. 4.25
107

2.21
107

1. Aguascalientes, 6. Colima, 10. Durango, 14. Jalisco, 18. Nayarit, 30. Zacatecas. 9. Distrito Federal, 11. Guanajuato, 12. Guerrero, 13. Hidalgo,
xico, 16. Michoaca
n, 17. Morelos, 21. Puebla, 15. Me
taro, 23. San Luis Potosi, 28. Veracruz. 22. Quere

3

7

1. 3.54
10 3. 1.53
108

market is satisfied with gasoline, but it is possible to slightly reduce the environmental impact when the bioethanol is produced in central or distributed biorefineries. In most of the Pareto curve, the solutions state the installation of one central biorefinery, but only changing the amount of produced bioethanol for reducing the environmental impact. Fig. 11 shows the supply chain for solution B of the Pareto curve of Fig. 10, with a profit of 7.2829
108 USD/year.

7.25
10

7

The central plant of the SC produces 9.68
107 kg of bioethanol per year. Table 6 shows the material flows for the supply chain. Scenario 3b: Finally, the model was solved considering a variable satisfied demand. Fig. 10 shows the Pareto curve (line of red points, in the web version), which behaves almost linearly. When the profit is maximized, the demand is satisfied only with gasoline, while the solution with minimum damage to resources satisfies the demand with bioethanol only. Table 7 summarizes the results in all the scenarios. In the variable demand case, the maximum percentage of reduction is higher because the supply chain does not fully satisfy the demand in the markets. For this reason, the use of gasoline and bioethanol decreases, and so does the environmental impact. Finally, Fig. 12 shows three scatter plots that depict the Pareto points in the space of every two impact categories. As observed, the minimization of each impact results in the minimization of the others, since all the impact categories are highly correlated. 5. Conclusions

Fig. 10. Pareto curve for scenario 3 considering a fixed demand satisfied.

This paper has proposed a novel multi-objective optimization model for designing a biorefinery system in Mexico based on residues from the tequila industry. The development of such optimization model represents a scientific challenge because all the involved tasks in the entire supply chain have to be considered to

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

16

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

Fig. 11. Supply chain for solution A of scenario 3 considering a fixed satisfied demand.

Table 6 Flows of the supply chain for solution A in scenario 3. Central plant

Bagasse from cultivation areas (kg/year)

Bagasse from tequila industry (kg/year)

Bioethanol obtained (kg/year)

Markets considered

3

1. 3.54
107 3. 1.57
108

3. 1.15
108

9.68
108

All markets

determine simultaneously the configuration and operation of the supply chain. Furthermore, the optimization formulation accounts for the simultaneous maximization of the profit and minimization of the environmental impact in human health, ecosystem quality and resources. The environmental impact was quantified through the Eco-indicator 99 method. The proposed approach identifies the optimal biorefinery location, and distribution of residues and products, including the selection of the best transportation options. In addition, this paper investigated the trade-offs between contradicting objectives in the design of a biofuel supply chain, thereby providing valuable insight for decision makers. These contributions represent a clear step forward in the literature. The optimization formulation was applied to design the future biofuels supply chain from tequila residues in Mexico, where several options were analyzed. The results (shown through Pareto

curves) indicate that the installation of a biorefinery system in Mexico based on the use of residues from the tequila industry can provide appealing solutions in terms of economic and environmental performance. The use of residues as feedstock leads to significant environmental benefits in comparison with the traditional use of fossil fuels. The results show that the solutions with the lowest environmental impact require the installation of several processing plants that use the available feedstock. However, the best economic solution nowadays is to use gasoline, at the expense of showing worst environmental damage. Furthermore, intermediate solutions involve the installation of several central and distributed bioethanol processing facilities using residues from the tequila industry. These intermediate solutions represent the optimal tradeoff between economic and environmental aspects. Finally, the results from this research can be useful for the decision makers in the tequila industry in Mexico as well as for the Mexican government to choose as an option the production of biofuels from the residues of the tequila industry, because this represents attractive economic and environmental solutions. Furthermore, this research can be as a basis for different regions around the world to consider the available biomass to establish a biorefinery system, because this way there can be obtained economic and environmental benefits.

Table 7 Summary of solutions discussed in the case study. Fixed demand in markets (total demand satisfied) Ecosystem quality point A B C Human health point A B C Resources point B

Environmental impact (PDF m2 year) 1.4672
106 1.5076
106 1.548
106 Environmental impact (DALYs) 1.6407
107 1.7154
107 1.8234
107 Environmental impact (MJ) 1.45
108

Profit (USD/year) 6.8689
108 7.5573
108 7.9346
108 Profit (USD/year) 6.6775
108 7.2919
108 7.9267
108 Profit (USD/year) 7.2829
108

Percent of impact reduced from point AeC (%) 5.2196

Environmental impact (PDF m2 year) 7.295
105 7.7
105 1.54
106

Profit (USD/year) 3.4344
108 3.9011
108 7.8937
108

Percent of impact reduced from point AeC (%) 52.629

Percent of impact reduced from point AeC (%) 10.0197

Percent of impact reduced from point AeC (%) 15.625

Variable demand in markets Ecosystem quality point A0 B0 C0

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

Sets I J K L M N2 P W T

17

set for agave cultivation areas i set for distributed biorefineries j set for tequila industries k set for central biorefineries l set for markets m for bioethanol set for markets n2 for solid fuel set for central and distributed biorefineries alternative w of transportation set for time periods t

Parameters aDehydratedPlant efficiency for dehydration p efficiency for mill process in plant p aMillPlant p aJuicePlant efficiency for mill process in plant p p aDistillPlant efficiency for distillation p aFermentationPlant efficiency for fermentation p efficiency for filtration at first stage aFiltred1Plant p efficiency for second section of filtration at first stage aFiltred2Plant p efficiency for second filtration process at aFilteredsecond1Plant p second stage to obtain the fermentable material aFilteredsecond2Plant efficiency for second filtration to obtain solid p fuel efficiency for reaction at first stage of hydrolysis aReactorPlant p efficiency for reaction at second stage aReactor2Plant p bioethanol price in market m bBioethanol m solid fuel price in market n2 bSolidFuel n2 bGasoline m

gasoline price in market m

dEQ Transport Eco-Indicator 99 for the transport considering ecosystem quality dEQ Plant

Eco-Indicator 99 for the bioethanol production considering ecosystem quality

dEQ Gasoline

Eco-Indicator 99 for the use of gasoline considering ecosystem quality

CFixPlant Column fixed cost for distillation column p Fig. 12. Correlations between impact categories.

CFixPlant Column2 fixed cost for dehydrated column p fixed cost for milling CFixPlant Mill p fixed cost for reactor CFixPlant Reactor p

Acknowledgment Authors acknowledge the financial support obtained from SAGARPA-CONACYT (Grant number 174560).

fixed cost for reactor of the second stage CFixPlant Reactor2 p CFixPlant Sieve fixed cost for sieve p CFixPlant Sieve2 fixed cost for sieve of the second stage p fixed cost for tanks CFixPlant Tank p

Nomenclature

fixed cost for tanks of the second stage CFixPlant Tank2 p fixed cost for stock tank CFixPlant TankA p

Indexes i j k l m n2 t p

unit variable cost for distillation column CVarPlant Column p index for agave cultivation areas index for distributed biorefineries index for tequila industries index for central biorefineries index for markets that demand fuels index for markets that demand solid fuel index for time periods distributed and central biorefineries

unit variable cost for dehydrated column CVarPlant Column2 p unit variable cost for milling CVarPlant Mill p unit variable cost for reactor CVarPlant Reactor p CVarPlant Reactor2 unit variable cost for reactors of the second stage p CVarPlant Sieve unit variable cost for sieve p CVarPlant Sieve2 unit variable cost for sieve of the second stage p

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

18

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

CVarPlant Tank unit variable cost for tanks p unit variable cost for stock tank CVarPlant TankA p Tank2 unit variable cost for tanks of the second stage CVarPlant p CostTransMAX maximum cost of transportation of agave bagasse i;l;w from cultivation areas to central plants CostTransMAX i;j;w maximum cost of transportation of agave bagasse from cultivation areas to distributed plants CostTransMAX k;j;w maximum cost of transportation of bagasse from tequila industry to distributed plants

CostUnitTranspk;l;w unit transportation cost from tequila industry k to central plants l CostUnitTranspl;m;w unit transportation cost from central plants l to markets m CostUnitTranspj;m;w unit transportation cost from distributed plants j to markets m efficiency for agave plant heads in cultivation area i zAHi zALi efficiency for agave bagasse in cultivation area i zCookedk efficiency for agave bagasse in tequila industry k

CostTransMAX k;l;w maximum cost of transportation of bagasse from tequila industry to central plants

Variables

CostTransMAX l;m;w maximum cost of transportation of bagasse from

capital cost for the second columns CPlant Column2 p

CostTransMAX j;m;w

CPlant Column Capital cost for columns p

central plants to markets

CPlant Mill p capital cost for mills

maximum cost of transportation of bagasse from

capital cost for reactors CPlant Reactor p

distributed plants to markets maximum agave available in cultivation area i FAgaveMAX i;t FTotBagPlant p

MIN

minimum flowrate that is processed in plants

FTotBagPlant p

MAX

maximum flowrate that is processed in plants MAX

FLeavBag Central i;l;t;w

maximum flow of bagasse from cultivation area i to central plants l MAX

FLeavesBagDistributed i;j;t;w

MAX FBagTequilaDistributed k;j;t;w

MAX FBagTequilaCentral k;l;t;w

gprodBioet Central l;m;t;w

MAX

maximum flow of bagasse from cultivation

CPlant Sieve capital cost for the sieves p CPlant Sieve2 capital cost for the second stage of sieves p capital cost for tanks CPlant Tank p capital cost of the stock tanks CPlant TankA p operating cost for columns CoperPlant Column p

area i to distributed plants j

operating cost for the second stage of columns CoperPlant Column2 p

maximum flow of bagasse from tequila

operating cost for mills CoperPlant Mill p

industry k to distributed plants j

operating cost for reactors CoperPlant Reactor p

maximum flow of bagasse from tequila

CoperPlant Reactor2 operating cost for the second stage of reactors p

industry k to central plants l

CoperPlant Sieve operating cost for sieves p

maximum flow of bagasse from central plants l to markets m MAX

gprodBioet Distributed j;m;t;w

capital cost for the second stage of reactors CPlant Reactor2 p

maximum flow of bagasse from distributed plants j to markets m

CoperPlant Sieve2 operating cost for the second stage of sieves p operating cost for tanks CoperPlant Tank p operating cost for the stock tanks CoperPlant TankA p

TotalFuelMIN m;t minimum demand of gasoline in market m

operating cost for the second stage of tanks CoperPlant Tank2 p

GSolidFuelMAX n2;t maximum demand of solid fuel in market n2

CostPlant CAP total capital cost for processing plants CostPlant OP total operational cost for processing plants CostTransi;j sost for transportation of agave bagasse from cultivation areas to distributed plants DCostTransi;j;w desegregate variable for cost of transportation of agave bagasse from cultivation areas to distributed plants CostTransi;l cost for transportation of agave bagasse from cultivation areas to central plants DCostTransi;l;w desegregate variable for cost of transportation of agave bagasse from cultivation areas to central plants CostTransk;j cost for transportation of agave bagasse from tequila industry to distributed plants DCostTransk;j;w desegregate variable for cost of transportation of agave bagasse from tequila industry to distributed plants CostTransk;l cost for transportation of agave bagasse from tequila industry to central plants DCostTransk;l;w desegregate variable for cost of transportation of agave bagasse from tequila industry to central plants

GSolidFuelMIN n2;t minimum demand of solid fuel in market n2 KF factor used to annualize the capital costs UCostPlant OpColumn unit operating cost for distillation column p unit operating cost for dehydrated column UCostPlant OpColumn2 p unit operating cost for mills UCostPlant OpMill p unit operating cost for reactors UCostPlant OpReactor p UCostPlant OpReactor2 unit operating cost for reactors of second stage p UCostPlant OpSieve unit operating cost for sieves p UCostPlant OpSieve2 unit operating cost for sieve of second stage of p plants unit operating cost for tanks UCostPlant OpTank p unit operating cost for tanks of second stage UCostPlant OpTank2 p unit operating cost for stock tanks UCostPlant OpTankA p CostUnitTranspi;j;w unit transportation cost from cultivation area i to distributed plants j CostUnitTranspi;l;w unit transportation cost from cultivation area i to central plants l CostUnitTranspk;j;w unit transportation cost from tequila industry k to distributed plants j

DgprodBioet Central l;m;t;w desegregate variable for flow from central plants to markets

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

DgprodBioet Distributed desegregate variable for flow from distributed j;m;t

FJuicePlant p;t flowrate of juice fermentable in mill process

plants to markets CostTransl;m cost for transportation of bioethanol from central plants to markets DCostTransl;m;w desegregate variable for cost of transportation of bioethanol from central plants to markets CostTransj;m cost for transportation of bioethanol from distributed plants to markets DCostTransj;m;w desegregate variable for cost of transportation of bioethanol from distributed plants to markets

flowrate in reaction process in plants FReact Plant p;t

desegregate variable of flow of bagasse from DFLeavBagDistributed i;j;t;w cultivation area i to distributed plants j DFLeavBagCentral i;l;t;w

desegregate variable of flow of bagasse from cultivation area i to central plants l

DFBagTequiladistributed desegregate variable of flow of bagasse from k;j;t;w tequila industry k to distributed plants j DFBagTequilaCentral k;l;t;w desegregate variable of flow of bagasse from tequila industry k to central plants l EQuality

EcoTranspBagasse environmental impact for the transport of bagasse EQuality EcoTranspProducts

environmental impact for the transport of the bioethanol obtained

capacity flowrate in the reactor of first stage FReactCAP Plant p;t flowrate in reaction of the second stage in plants FReact2Plant p;t capacity flowrate in reactor of the second stage FReactCAP2Plant p;t amount of bioethanol obtained in plants FStockPlant p;t capacity of flowrate of bioethanol obtained in plants FStockCAP Plant p;t FTequilaBagk;t total flowrate of bagasse in tequila industry k total flow of bagasse from tequila industry k to FBagTequiladistributed k;j;t distributed plant j total flow of bagasse from tequila industry k to FBagTequilaCentral k;l;t central plant l FBagTequilaPlant k;p;t total flow of bagasse from tequila industry k to plant p total flowrate of bagasse in plants FTotalBagPlant p;t FTotalBagCAP Plant capacity total flowrate of bagasse in plants p;t TequilaIndustry

FTotalHeadsk;t

total flowrate of plant heads in tequila

industry k FTotLeavBagi;t flowrate of agave bagasse in cultivation area i

EcoPlant EQuality environmental impact for the bioethanol Plant production in plants

FLeavBagDistributed flow of bagasse from cultivation area i to i;j;t

EcoGasolineEQuality environmental impact for the use and transport of gasoline

FLeavBagCentral flow of bagasse from cultivation area i to central i;l;t flow of bagasse from cultivation area i to plant p FLeavBagPlant i;p;t

Eco99EQuality total environmental impact considering ecosystem quality GBioethanolm;t total bioethanol flowrate sent to the market m GSolidFueln2;t total solid fuel flowrate sent to the market n2 gprodBioet Plant p;m;t bioethanol flowrate sent from plants to markets bioethanol flowrate sent from central plants to gprodBioet Central l;m;t markets gprodBioet Distributed bioethanol flowrate sent from distributed j;m;t plants to markets gSolidFuelPlant p;n2;t solid fuel flowrate sent from plants to markets FAgavei;t agave flow from cultivation areas i over time t FCol1Plant p;t flowrate inlet to the distillation column capacity flowrate in the distillation column FColCAP1Plant p;t FCol2Plant p;t flowrate inlet to the dehydrated column capacity flowrate in the dehydrated column FColCAP2Plant p;t flowrate inlet to the fermentation process FFerment Plant p;t FFermentCAP Plant p;t

capacity flowrate inlet in the fermentation equipment

flowrate inlet to the filter of the second stage FFIlteredsecondPlant p;t flowrate outlet to the first filter Ffilt1Plant p;t Plant Ffilt2p;t flowrate outlet to the first filter sent to second stage of hydrolysis Plant capacity flowrate in filter of the second stage Ffilt2CAPp;t

flowrate of solid fuel obtained FFuelSolidPlant p;t FPlantHeadsi;t flowrate of agave plant heads in cultivation area i TequilaIndustry

FPlantHeadsi;k;t

flow of plant heads from cultivation area i to tequila industry k

FMillPlant p;t

flowrate in mill process

capacity flowrate in mill equipment FMillCAP Plant p;t

19

distributed j l

FuelMixm;t flowrate of mixed fuel (bioethanol and gasoline) in markets m GasoPurem;t flowrate of pure gasoline in markets m GasoMixm;t flowrate of gasoline in the mix bioethanolegasoline PROFIT total annual profit TotalFuelm;t total flowrate of fuel in market m TotalGasolinem;t total flowrate of gasoline sold in markets m TotCost total annual cost TotSales total annual sales YpPlant

Boolean variable for the existence of plants

yPlant p

binary variable for the existence of plants

TDIS Yi;j;w

Boolean variable for transport selection from cultivation

yTDIS i;j;w

areas i to distributed plants j binary variable for transport selection from cultivation

TCEN Yi;l;w

areas i to distributed plants j Boolean variable for transport selection from cultivation

yTCEN i;l;w

areas i to central plants l binary variable for transport selection from cultivation

TDIS Yk;j;w

areas i to central plants l Boolean variable for transport selection from tequila

yTDIS k;j;w

industry k to distributed plants j binary variable for transport selection from tequila

TCEN Yk;l;w

industry k to distributed plants j Boolean variable for transport selection from tequila

yTCEN k;l;w

industry k to central plants l binary variable for transport selection from tequila

industry k to central plants l TCENMIX Boolean variable for transport selection from central Yl;m;w plants l to markets m yTCENMIX binary variable for transport selection from central plants l;m;w l to markets m

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052

20

P.E. Murillo-Alvarado et al. / Journal of Cleaner Production xxx (2015) 1e20

TDISMIX Boolean variable for transport selection from distributed Yj;m;w

yTDISMIX j;m;w

plants j to markets m binary variable for transport selection from distributed plants j to markets m

References Agostinho, F., Ortega, E., 2013. Energetic-environmental assessment of a scenario for Brazilian cellulosic ethanol. J. Clean. Prod. 47, 474e489. Akgul, O., Zamboni, A., Bezzo, F., Shah, N., Papageorgiou, L.G., 2011. Optimization based approaches for bioethanol supply chains. Ind. Eng. Chem. Res. 50 (9), 4927e4938. Albernas-Carvajal, Y., Corsano, G., Kafarov, V.V., Gonzales-Cortes, M., 2014. Optimal design of pre-fermentation and fermentation stages applying nonlinear programming. Energy Convers. Manag. 87, 1195e1201. Alex-Marvin, W., Schmidt, L.D., Benjaafar, S., Tiffany, D.G., Daoutidis, P., 2012. Economic optimization of a lignocellulosic biomass-to-ethanol supply chain. Chem. Eng. Sci. 67 (1), 68e79. Brooke, A., Kendrick, D., Meeraus, A., Raman, R., 2014. GAMS, a User's Guide. GAMS Development Corporation, Washington, D.C., USA. www.gams.com. Cardona, C.A., Quintero, J.A., Paz, I.C., 2010. Production of bioethanol from sugarcane bagasse: status and perspectives. Bioresour. Technol. 101 (13), 4754e4766. Chouinard-Dussault, P., Bradt, L., Ponce-Ortega, J.M., El-Halwagi, M.M., 2011. Incorporation of process integration into life cycle analysis for the production of biofuels. Clean Technol. Environ. Policy 13 (5), 673e685.
nez-Gutie
rrez, A., Mannan, M.S., ElEl-Halwagi, A., Rosas, C., Ponce-Ortega, J.M., Jime Halwagi, M.M., 2013. Multi-objective optimization of biorefineries with economic and safety objectives. AIChE J. 59 (7), 2427e2434. Geodkoop, M., Spriensma, R., 2001. The Eco-Indicator 99, a Damage Oriented for Life Cycle Impact Assessment: Methodology Report and Manual for Designers. Technical report. PRe Consultants, Amersfoort, The Netherlands. Gong, J., You, F., 2014. Global optimization for sustainable design and synthesis of algae processing network for CO2 mitigation and biofuel production using life cycle optimization. AIChE J. 60 (9), 3195e3210.
n-Gos
Guille albez, G., Grossmann, I.E., 2010. A global optimization strategy for the environmentally conscious design of chemical supply chains under uncertainty in the damage assessment model. Comput. Chem. Eng. 34 (1), 42e58. Kajaste, R., 2014. Chemicals from biomass e managing greenhouse gas emissions in biorefinery production chains e a review. J. Clean. Prod. 75, 1e10. Karlsson, H., Borjesson, P., Hansson, P.A., Ahlgren, S., 2014. Ethanol production in biorefineries using lignocellulosic feedstock e GHG performance, energy balance and implications of life cycle calculation methodology. J. Clean. Prod. http://dx.doi.org/10.1016/j.jclepro.2014.07.029. Kazi, F.K., Fortman, J.A., Anex, R.P., Hsu, D.D., Aden, A., Dutta, A., Kothandaraman, G., 2010. Techno-economic comparison of process technologies for biochemical ethanol production from corn stover. Fuel 89, S20eS28. Kelloway, A., Daoutidis, P., 2014. Process synthesis of biorefineries: optimization of biomass conversion to fuels and chemicals. Ind. Eng. Chem. Res. 53 (13), 5261e5273.
n-Gosa
lbez, G., Mele, F.D., Bagajewicz, M.J., Jime
nez, L., 2012. Kostin, A.M., Guille Design and planning of infrastructures for bioethanol and sugar production under demand uncertainty. Chem. Eng. Res. Des. 90 (3), 359e376. Liew, W.H., Hassim, M.H., Ng, D.K.S., 2014. Review of evolution, technology and sustainability assessments of biofuel production. J. Clean. Prod. 71, 11e29. Lim, J.S., Manan, Z.A., Hishim, H., Alwi, S.R.W., 2014. Synthesis of a sustainable integrated rice mill complex. J. Clean. Prod. 71, 118e127. Maes, D., Dael, M.V., Vanheusden, B., Goovaerts, L., Reumerman, P., Luzardo, M.N., Pasel, S.V., 2015. Assessment of the sustainability guidelines

of EU renewable energy directive: the case of biorefineries. J. Clean. Prod. 88, 61e70. Martinez-Hernandez, E., Campbell, G.M., Sadhukham, J., 2014. Economic and environmental impact marginal analysis of biorefinery products for policy targets. J. Clean. Prod. 74, 74e85.
n-Gos

nez, L., 2009. Optimal planning of supply Mele, F.D., Guille albez, G., Jime chains for bioethanol and sugar production with economic and environmental concerns. Comput. Aided Chem. Eng. 26, 997e1002.
lbez, G., Jimenez, L., 2011. Multiobjective model Mele, F.D., Kostin, A.M., Guillen-Gosa for more sustainable fuel supply chains. A case study of the sugar cane industry in Argentina. Ind. Eng. Chem. Res. 50 (9), 4939e4958. Moncada, J., Tamayo, J., Cardona, C.A., 2014. Evolution from biofuels to integrated biorefineries: techno-economic and environmental assessment of oil palm in Colombia. J. Clean. Prod. 81, 51e59. Mota, B., Gomes, M.I., Carvalho, A., Barvbosa-Povoa, A.P., 2014. Towards supply chain sustainability: economic, environmental and social design and planning. J. Clean. Prod. http://dx.doi.org/10.1016/j.jclepro.2014.07.052. ~ ez-Aguilar, J.E., Ponce-Ortega, J.M., CastroMurillo-Alvarado, P.E., Santiban Montoya, A.J., Serna-Gonz
alez, M., El-Halwagi, M.M., 2014. Optimization of the supply chain associated to the production of bioethanol from residues of agave from the tequila process in Mexico. Ind. Eng. Chem. Res. 53 (13), 5524e5538. Ng, R.T.L., Ng, D.K.S., Ta, R.R., 2015. Optimal planning, design and synthesis of symbiotic bioenergy parks. J. Clean. Prod. 87, 291e302. ~ ez, H.M., Rodríguez, L.F., Khanna, M., 2011. Agave for tequila and biofuels: an Nún economic assessment and potential opportunities. GCB Bioenergy. 3 (1), 43e57.
n-Gos

nez-Esteller, L., Pascual-Gonz
alez, J., Guille albez, G., Mateo-Sanz, J.M., Jime 2015. Statistical analysis of global environmental impact patterns sing a world multi-regional inputeoutput database. J. Clean. Prod. 90, 360e369. Patrizi, N., Caro, D., Pulselli, F.M., Bjerre, A.B., Bastianoni, S., 2013. Environmental feasibility of partial substitution of gasoline with ethanol in the province of Siena (Italy). J. Clean. Prod. 47, 388e395. Patrizi, N., Pulselli, F.M., Morandi, F., Bastianoni, S., 2015. Evaluation of the emergy investment needed for bioethanol production in a biorefinery using residua resources and energy. J. Clean. Prod. 96, 549e556. SAGARPA-SIAP, 2013. Mexican System of Information about Agriculture and Fishing. Advance of Planting and Harvesting for Mexico. Mexico City, Mexico. http:// www.siap.gob.mx/index.php?option¼com_ wrapper&view¼wrapper&Itemid¼350 (accessed 01.04.14.). ~ ez-Aguilar, J.E., Gonza
lez-Campos, J.B., Ponce-Ortega, J.M., SernaSantiban Gonz
alez, M., El-Halwagi, M.M., 2011. Optimal planning of a biomass conversion system considering economic and environmental aspects. Ind. Eng. Chem. Res. 50 (14), 8558e8570. ~ ez-Aguilar, J.E., Gonza
lez-Campos, J.B., Ponce-Ortega, J.M., SernaSantiban Gonz
alez, M., El-Halwagi, M.M., 2014. Optimal planning and site selection for distributed multiproduct biorefineries involving economic, environmental and social objectives. J. Clean. Prod. 65, 270e294. Saucedo-Luna, J., Castro-Montoya, A.J., Martinez-Pacheco, M.M., Sosa-Aguirre, C.R., Campos-Garcia, J., 2011. Efficient chemical and enzymatic saccharification of the lignocellulosic residue from Agave tequilana bagasse to produce ethanol by Phia caribbica. J. Ind. Microbiol. Biotechnol. 38, 725e732. SENER, Mexican Secretary of Energy. Potential and viability for the use of bioethanol and biodiesel for transportation in Mexico. http://sie.energia.gob.mx/ bdiController.do?action¼temas (accessed 01.04.14.). Tay, D.H.S., Ng, D.K.S., 2012. Multiple-cascade automated targeting for synthesis of a gasification-based integrated biorefinery. J. Clean. Prod. 34, 138e148. Vujanovic, A., Cucek, L., Pintaric, Z.N., Pahor, B., Kravnja, Z., 2015. Synthesis of environmentally-benign energy self-sufficient processes under uncertainty. J. Clean. Prod. 88, 90e104. Yue, D., You, F., 2014. Fair profit allocation in supply chain optimization with transfer price and revenue sharing: MNLP model and algorithm for cellulosic biofuel supply chains. AIChE J. 60 (9), 3211e3229.

Please cite this article in press as: Murillo-Alvarado, P.E., et al., Multi-objective optimization of the supply chain of biofuels from residues of the tequila industry in Mexico, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.08.052