optimization of pre-treatment selection for the useof

ARTICLE IN PRESSCHERD-1463; No. of Pages 24

Oo

MLa

Bb

1

Tshapcdlwctn

A

0h

chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxx

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

j ourna l h omepage: www.elsev ier .com/ locate /cherd

ptimization of pre-treatment selection for the usef woody waste in co-combustion plants

ar Prez-Fortesa,, Jos M. Lanez-Aguirreb, Aarn David Bojarskia,uis Puigjanera,

Department of Chemical Engineering, Universitat Politcnica de Catalunya, Av. Diagonal 647, PG-2, 08028arcelona, SpainSchool of Chemical Engineering, Purdue University, 480 Stadium Mall Dr., West Lafayette, IN 47907, USA

a b s t r a c t

This work is focused on the use of biomass waste to feed already existing coal combustion plants as a part of paving

the way toward the reduction of the environmental impact. The biomass waste supply chain optimization is critical

to conceive long-term viable projects and deal with the biomass heterogeneous nature and drawbacks to be used

with coal, i.e. principally high moisture content and low bulk density. This paper studies biomass transportation,

storage and change of properties (moisture content and hence dry matter, energy density and bulk density) through

the use of different pre-treatments: (i) torrefaction, (ii) torrefaction combined with pelletization, (iii) pelletization,

(iv) fast pyrolysis and (v) fast pyrolysis combined with char grinding, which produce a range of very different pre-

treated biomass. The optimization problem is formulated as a mixed integer linear program (MILP) that evaluates

the net present value and the environmental impact through a life cycle assessment (LCA). The results propose

locationallocation decision together with the selection/capacity of pre-treatment technologies for each scenario.

The scenarios contemplate different biomass characteristics, availability and distribution for a supply chain case

study located in Spain: forest and agricultural woody residues used to replace at least 10% of the total thermal inlet

power provided by coal in the existing network of thermal plants. 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Biomass supply chain; Biomass waste; Biomass seasonality; Multi-objective MILP; Co-combustion; Super-

structure of pre-treatments

of biomass properties along it, to suitably evaluate the biomass. Introduction

he energy sector aims at increasing the use of renewableources and of efficient processes. Countries all over the worldave become conscious of the biomass potential as a renew-ble source to contribute to the satisfaction of energy needs,romote development in rural areas and alleviate climatehange. Biomass can be of different origins and, among theifferent types, organic wastes are of special concern to avoidand use controversy and to propose a safe final disposalhile generating electricity, fuels and/or heat/cool. Biomassan contribute in centralized or large scale, as well as in dis-ributed or small scale projects, by satisfying different userPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

eeds. Biomass use in centralized energy systems (CES) needs

Corresponding authors at: Department of Chemical Engineering v. Diagonal 647, PG-2, 08028 Barcelona, Spain. Tel.: +34 93 245 13 94.E-mail addresses: [email protected] (M. Prez-Fortes), luis.Received 2 August 2013; Received in revised form 30 December 2013; A

263-8762/$ see front matter 2014 The Institution of Chemical Engittp://dx.doi.org/10.1016/j.cherd.2014.01.004from well-established channels to receive raw material ontime at the required characteristics, and to distribute the finalproduct. This final output can be any bio-product: chemi-cal compounds, electricity, heat, hydrogen, synthetic naturalgas and/or liquid fuels (Higman and van der Burgt, 2003).As an emerging sector, there is a need to define its basis tosecure that the biomass processing project as a whole is eco-nomically, environmentally and socially sustainable with thecurrent and future tendencies and policies, from the renew-able source to the grid (Cherubini and Stromman, 2011; Goldand Seuring, 2011).

The biomass SC configuration has to consider the changesre-treatment selection for the use of woody waste in co-combustion01.004

CEPIMA, Universitat Politcnica de Catalunya (UPC-ETSEIB),

[email protected] (L. Puigjaner).ccepted 2 January 2014

amounts to be transported and used to produce the final

neers. Published by Elsevier B.V. All rights reserved.

dx.doi.org/10.1016/j.cherd.2014.01.004http://www.sciencedirect.com/science/journal/02638762www.elsevier.com/locate/cherdmailto:[email protected]:[email protected]/10.1016/j.cherd.2014.01.004

Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pre-treatment selection for the use of woody waste in co-combustionplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.01.004

ARTICLE IN PRESSCHERD-1463; No. of Pages 242 chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxx

Superscripts and subindices

ar as received basisdaf dry and ash free basise electricth thermal

AcronymsAWR agricultural woody residuesBD bulk densityBM biomassCCS carbon capture and storageCES centralized energy systemsd dayDM dry matterEcO economic optimizationEnvO environmental optimizationEff efficiencyFU functional unitFWR forest wood residuesGHG greenhouse gasesGIS geographic information systemIGCC integrated gasification combined cycleLCA life cycle assessmentLCI life cycle inventoryLCIA life cycle impact assessmentLHV lower heating valueMC moisture contentMILP mixed integer linear programMINLP mixed integer non-linear programmoMILP multi-ojective mixed integer linear program-

mingNPV net present valueOIL bio-oilO&M operation and maintenancePC pulverized coalPEL pelletpts points (from LCA methodology, Impact 2002+)SC supply chainSCN scenarioSLU bio-oilSOTA state-of-the-artSTN state task networkTOP pellet of torrefied biomassTOR torrefied biomassUTM Universal Transverse Mercatoryr year

Mathematical formulation: Indicesa mid point environmental impact categoriese suppliersf, f facility locationsg end point environmental impact categoriesi tasksj equipment technologyk intervals for piecewise linear approximation of

economies of scales materials (states)t, t planning periods

SetsAg set of midpoint environmental interventions

that are combined into endpoint damage fac-tors g

Erm set of suppliers e that provide raw materialsEprod set of suppliers e that provide production ser-

vicesEtr set of suppliers e that provide transportation

servicesFP set of materials s that are final productsI set of tasks i with variable inputIj set of tasks i that can be performed in technol-

ogy jJe technology j that is available at supplier eJf technology j that can be installed at location fJi technologies that can perform task iJStor technologies to perform storage activitiesMkt set of market locationsNTr set of production, or non-transport, tasksRM set of materials s that are raw materialsSstor set of materials/biomass that if stored change

their propertiesSup set of supplier locationsTs set of tasks producing material sTs set of tasks consuming material sTr set of distribution tasks

ParametersAsft (kg) maximum availability of raw material s in

period t in location fAdaptLf (kg) lower bound for the capacity of co-

combustion in plant fAdaptUf (kg) upper bound for the capacity of co-

combustion in plant fDemsft (MJ) demand of product s at market f in period tCostCoalf (D ) cost of coal supplied to facility fdistanceff (km) distance from location f to location f

FCFJjft (D /h) fixed cost per unit of technology j capacityat location f in period t

FElimitjfk (h) increment of capacity equal to the upper limitof interval k for technology j in facility f

ir (adim.) discount rateHVCoalf (MJ/kg) heating value of coal supplied to facility fM (adim.) a big numberNormFg (adim.) normalizing factor of damage category

gPricesft (D /MJ) price of product s at market f in period tPricelimitjfk (D /MJ) investment required for an increment of

capacity equal to the upper limit of interval kfor technology j in facility f

Tortuosity (adim.) tortuosity factorWaters (adim.) moisture for material sWatermaxij (adim.) maximum moisture for task i per-

formed in equipment j

Greek symbolssij (adim.) mass fraction of task i for production of

material s in equipment jsij (adim.) mass fraction of task i for consumption of

material s in equipment j

dx.doi.org/10.1016/j.cherd.2014.01.004

Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

ARTICLE IN PRESSCHERD-1463; No. of Pages 24chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxx 3

jf (adim.) minimum utilization rate of technology jcapacity that is allowed at location f

ag(adim.) g end-point damage characterization factorfor environmental intervention a

ij (adim.) efficiency of task i performed in equipment j

plantf

(adim.) efficiency of combustion in plant fijff (h/kg) capacity utilization rate of technology j by

task i whose origin is location f and destinationlocation f

treff t (D /kg) unitary transportation costs from location f

to location f during period tut1ijfet

(D /kg) unitary cost associated with task i performedin equipment j from location f and payable toexternal supplier e during period t

ut2sfet

(D /kg) unitary cost associated with handling theinventory of material s in location f and payableto external supplier e during period t

est (D /kg) unitary cost of raw material s offered by exter-nal supplier e in period t

ijff a (points/kg) a environmental category impact CFfor task i performed using technology j receiv-ing materials from node f and delivering it atnode f

Coalaf

(points/kg) a environmental category impact CF forcoal supplied to facility f

Tija

(points/(kg km)) a environmental category impactCF for the transportation of a mass unit ofmaterial over a length unit

Binary variablesVjft 1 if technology j is installed at location f in

period t, 0 otherwise

SOS2 variablesjfkt variable to model the economies of scale for

technology j in facility f at period t as a piece-wise linear function

Continuous variablesAdaptf installed capacity for co-combustion in plant fDamCgft (point) normalized endpoint damage g for loca-

tion f in period tDamCSCg (point) normalized endpoint damage g along

the whole SCEPurchet (D ) economic value of purchases executed in

period t to supplier eESalest (D ) economic value of sales executed in period tFAssett (D ) investment on fixed assets in period tFCostt (D ) fixed cost in period tFjft (h) total capacity of technology j during period t at

location fFEjft (h) capacity increment of technology j at location f

during period tHVs (MJ/kg) lower heating value for material sICaft (point) midpoint a environmental impact associ-

ated to site f which rises from activities inperiod t

Impact2002f (point) total environmental impact for site f

Impact2002overall (point) total environmental impact for thewhole SC

NPV (D ) economic metric, net present value

Pijff t (kg) specific activity of task i, by using technologyj during period t, whose origin is location f anddestination is location f

Profitt (D ) profit achieved in period tPvsijft (kg) input/output material of material s for activity

of task i with variable input/output, by usingtechnology j during period tin location f (thismust be a production activity)

Profitt (D ) profit achieved in period tPurch

pret (D ) amount of money payable to supplier e in

period t associated with production activitiesPurchrmet (D ) amount of money payable to supplier e in

period t associated with consumption of rawmaterials

Purchtret (D ) amount of money payable to supplier e inperiod t associated with consumption of trans-port services

Salessff t (D ) amount of product s sold from location f inmarket f in period t

Ssft (kg) amount of stock of material s at location f inperiod t

TCostCoalt (D ) total cost of replaced coal during period tbioproduct required by the user (Gold and Seuring, 2011;Rentizelas et al., 2009b). Biomass as a commodity shouldmeet the adequate quality parameters and standards, whilepromoting the development of the most biomass productiveareas. Legislation for this purpose is underway and needsdecision-making tools to support the correct policies andappropriate subsidies (e.g. grants, tax incentives, bonuses,feed-in tariffs, emissions trading) to establish new markets(Dasappa, 2011; Lamers et al., 2012).

Co-firing biomass with fossil fuels is regarded as an alter-native of transition toward a carbon free society (Berndes et al.,2010; Faaij, 2006; Gmez et al., 2010b). Co-combustion or co-firing can be defined as the simultaneous combustion of twoor more fuels in the same combustion plant (Berndes et al.,2010), using at least one type of fossil fuel (Sami et al., 2001). Itis focused on existing installations, originally performed touse 100% fossil fuels to generate from tens to hundreds ofMW of power (Rodrigues et al., 2007). Reduction of CO2 (takinginto account a life cycle assessment LCA), sulfur and nitro-gen emissions (regarding biomass composition) are directlyderived from the coal fraction substitution with biomass. Co-combustion is the most direct application among biomass andrenewable sources use (Faaij, 2006). It may cost 25 times lessthan any other biomass process to produce electricity (Berndeset al., 2010). Coal is a primary resource that is predicted tolast as a major fuel for power generation for decades, there-fore, efforts are directed toward increasing the plant efficiencywhile decreasing its environmental impact (Giddey et al., 2012;Royo et al., 2012).

The biomass properties and quantity should mimic andmatch as much as possible the replaced fossil fuel. There-fore, biomass should prevent high moisture content (MC) or,low dry matter (DM), low grindability, low bulk density (BD)and low lower heating value (LHV). Biomass heterogeneity andinherent characteristics, as well as a seasonal production, maybe overcome using biomass densification, stabilization and anre-treatment selection for the use of woody waste in co-combustion01.004

appropriate storage planning (Puigjaner, 2011, Chapter 2). Forthose reasons pre-treated or enhanced biomass is crucial for

dx.doi.org/10.1016/j.cherd.2014.01.004

ARTICLE IN PRESSCHERD-1463; No. of Pages 244 chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxxthe development of the bioenergy sector. Enhanced biomass,as known nowadays, ranges from solids like pellets, torrefiedbiomass or torrefied pellets to liquids like pyrolysis oil or bio-oil (Uslu et al., 2008).

Several studies evaluate the effect of the coal replacementfraction on power plant efficiency, showing that there is arange for coal substitution without either efficiency penaltynor technical problems in the plant. The work by Dai et al.(2008) exhaustively revise the issues that may arise frombiomass use with coal, ranging from slagging and fouling, tocatalyst poisoning during flue gas cleaning. The work by Liuand Balasubramanian (2014) propose a pre-treatment step toavoid ashes issues. The paper by Gmez et al. (2010b) pro-poses coal substitution between 5% and 10% on energy basis,being the 5% the minimum fraction that may produce a valu-able change in plant performance. The affordable shares onan energy basis according to Berndes et al. (2010) are 15%for fluidized bed boilers and 10% for pulverized coal (PC) andgrate-fired boilers.

The energy efficiency of a coal fired plant oscillates between30% and 45% on a LHV basis, depending on the boiler technol-ogy, subcritical or supercritical water, and the antiquity of theplant (Berndes et al., 2010), being 36% a common average value.In Van Den Broek et al. (1996), the efficiency ranges from 39%to 44% for co-firing in PC power plants, up from 100 MW. Thework by Gmez et al. (2010a) uses a value of 38% as co-firingefficiency. Some other works, such as Sami et al. (2001), showsthe synergetic effect between coal and biomass. It means that,not only the efficiency of the coal combustion plants can bemaintained, but it may increase when using biomass.

This work deals with the strategic and tactical levels ofthe SC problem at the design step. As in a classical SCproblem, processing sites, distributors, transportation, ware-houses, raw material suppliers and customer markets areidentified. Specifically important for bioenergy problems isthat raw materials can be highly distributed, with the excep-tion of energy crops, so logistic issues have to be properlyconsidered (Caputo et al., 2005); biomass production, stor-age and transportation may represent a significant fractionof the whole SC cost (Panichelli and Gnansounou, 2008) andof the environmental impact. The approach used here for-mulates a large-scale mixed integer linear program (MILP)that captures the relevant information for each facility andflow (Graves and Tomlin, 2003). The bio-based SC problem for-mulation and resolution face new challenges to capture (i)biomass properties change, (ii) different types of biomass com-bination, (iii) biomass seasonality, (iv) demand seasonality, (v)multi-site production of biomass and the subsequent, and (vi)size of the problem. This leads to a highly computationallydemanding model. The starting point is the mathematicalformulation developed in Puigjaner (2011, Chapter 3), andPrez-Fortes et al. (2012): it has been adapted and extendedto consider a new real complex scenario, a whole country,contemplating a substantial number of sites and variables.A superstructure of pre-treatments and monthly periodsare considered. As a result, the model allow to (i) locateand identify the most appropriate pre-treatment technolo-gies and feedstock suppliers to replace a certain amount ofcoal in the already existing Spanish park of thermal powerplants, (ii) to evaluate matter flows between nodes and ech-elons and (iii) to give a preliminary approach to the monthlysequence, i.e. inputs and outputs by month, production ratesPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

and inventory, for all the units and biomass productionsites.2. Literature review and the SC

The biomass SC problem may be addressed using a wide rangeof decision-maker outlooks. As example, Caputo et al. (2005)evaluate the net present value (NPV) of 100% biomass projects,focusing on transportation. Bowling and Dale (2011) look foran optimal SC for a biorefinery, considering overall sales andcosts optimization to discern between a distributed or central-ized structure with special attention on transportation costs.Ayoub et al. (2009) focus on costs and environmental impactthrough emissions to air, water pollutants and solid wastes.Damen and Faaij (2006) perform a life cycle inventory to com-pare co-combustion and combustion of only coal and Perryand Rosillo-Calle (2008) focus on CO2 emissions along thewhole SC. A more recent work from Mele et al. (2011) combinesthe use of mathematical programming with LCA, to performa multi-objective optimization based on the NPV and the LCA,to produce bioethanol from sugar cane in Argentina. Environ-mental evaluations often take into account a LCA (Cherubiniand Stromman, 2011). Other attempts have been recently doneto add the social criterion to the economic and environmentalpoints of view, as the creation of places of job (Prez-Forteset al., 2012; You et al., 2012).

The most state-of-the-art works combine multi-objectiveoptimization and mathematical programming (MILP, mixedinteger non-linear program, MINLP, with and without uncer-tainty and risk consideration) or scenario-based optimizationwith geographic information systems (GIS) for spatial dataanalysis. The literature review from An et al. (2011a) exposesthat bioenergy is approaching to an important grow and needsto integrate strategic, tactical and operational decisions (i.e.the operations research point of view) to enhance and securetheir viability, even if planning models have not been fullyrequired (and therefore, developed) yet. The review by Scottet al. (2012) points out that the most used method for multi-criteria decision-making in bioenergy systems is optimizationusing few alternatives (i.e. a limited number of scenarios),followed by mathematical optimization (i.e. coverage of allpossible options). Usually, the performed case studies considerone single biomass type. An et al. (2011b) demonstrate theeconomic viability of a biofuel SC (lignocellulosic biomass tobioethanol) in Texas through an strategic-tactical mathemati-cal model. The works by Zamboni et al. (2009, 2009) formulate aMILP model to minimize operating costs and GHG emissions ofa biofuel SC. The second work performs a multi-objective opti-mization by taking into account the whole SC, from biomasscultivation to fuel distribution. The region under study is dis-cretized into a grid of square regions to map the networkand to estimate the biomass cultivation potential for eachregion. Using as starting point the formulation developed inZamboni et al. (2009), Akgul et al. (2011) also minimize costs,while decreasing calculation time including spatial restric-tions through a neighborhood flow limitation. The startingpoint in Giarola et al. (2012) is Zamboni et al. (2009). Theapproach here is extended to consider more than one type ofraw biomass to produce bioethanol, and deals with biomassavailability limitation mainly due to food and land contro-versies. Kin et al. (2011) develop a MILP model to decide thelocation and capacity of fast pyrolysis and FischerTropschplants to produce biofuel from distributed biomass sources. AMILP determines optimal sizes and locations of biofuel plantsthat use an energy crop as raw material for already existingre-treatment selection for the use of woody waste in co-combustion01.004

gas stations in Austria, in Leduc et al. (2008). The optimiza-tion deals with investment and operating costs. The paper by

dx.doi.org/10.1016/j.cherd.2014.01.004


Fig. 1 A generic bio-based SC that produces and distributes electricity. Transportation and storage may occur whenn

EptuableaoaTrmotstGtlc

Sam(b(SactiamIioosddaabt

o

eeded.

lia et al. (2011) uses a MILP multi-objective optimization toropose a sustainable SC for a novel hybrid concept for a planthat produces liquid fuels, i.e. gasoline, diesel and kerosene,sing coal, biomass or natural gas as main feedstocks. Theuthors conclude that the hybrid configuration (fossil fuels

biomass) allows for competitive biofuels prices. The worky Tavares et al. (2011) use GIS to identify the most suitableocations for municipal solid waste plants, according to socio-conomic, technical and environmental criteria. Panichellind Gnansounou (2008) solve the locationallocation problemf two torrefaction plants and a gasification unit, by means of

costs criterion, for a specific region in the north of Spain.he exact spatial distribution of biomass waste resourcesoads map are used. Zhang et al. (2011) adopt a GIS-basedethod to find the best location for a biofuel plant, basedn the distributed nature of the woody biomass sources andhe associated transportation costs. The methodology con-ists of two steps: location of site candidates and selection ofhe cost-optimal area. The paper by Chiueh et al. (2012) usesIS to identify the biomass waste resources and the specificransportation routes for a co-combustion retrofitting prob-em, evaluating different levels of pre-treatment (torrefaction)entralization.Different works can be found that go in depth into a specific

C echelon. Fig. 1 represents a generic bio-based SC to producend distribute electricity, with multiple biomass providers andultiple biomass processing plants: (i) feedstock production

growing, harvesting and collection), or waste generation, (ii)iomass pre-treatment, (iii) storage, (iv) biomass treatment,v) electricity distribution and (vi) electricity consumption.torage and transportation can be needed in any momentlong the SC, depending on the specific distances to beovered, the supply and the demand characteristics. Never-heless, it is more convenient to store and transport biomassn an upgraded state, i.e. after a pre-treatment, to avoid costsnd non-desired effects. Transportation and storage are theost critical echelons for a bio-based SC (Caputo et al., 2005;

akovou et al., 2010; Rentizelas et al., 2009b). The electric-ty distribution level may include the desired level of detailf a traditional grid, a smart grid or a microgrid, dependingn each case study. Storage is imperative when consideringeasonal biomass supplies and electricity demand. Biomassifferent origins, high MC, low BD, low LHV, and biomassegradation during storage, lead to a needed improvementnd homogenizing step to optimize transportation, handlingnd processing, to obtain a standard product. Therefore, aio-based SC may consider changes on biomass characteris-ics into each SC echelon.Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

Covering all biomass states and pre-treatments, the meansf transport used can be truck, trailer, train, ship or pipeline.The study by Cherubini and Stromman (2011) reviews theworks performed in the field of LCA for bio-based systems,showing a general reduction of GHG emissions and of fos-sil fuel consumption when using biomass instead of fossilfuels. Even if the LCA follows a well-established methodology,the selection of the functional unit (FU) changes according toauthors criteria (for instance, power produced or energy/massflowrates of biomass introduced into the system) and it makesdifficult the comparison among different works. Furthermoremany LCAs have been conceived under different assumptions,hence their comparison is also difficult.

According to the reviewed works, no other work has takeninto account a superstructure of pre-treatments solved by aMILP formulation that considers simultaneously the changeof biomass properties along the different echelons, to evalu-ate large scale retrofitting co-combustion plants projects. Thegaps to cover to successfully implement bionenergy projectsare: (i) integration of decision levels, (ii) multi-objective opti-mization, (iii) biomass properties (mainly MC and BD) changealong the SC, (iv) pre-treatments evaluation, (v) seasonalraw matter and demand, and (vii) limitation of calculationtime. According to that, the approach of this work pro-poses a sustainable SC by supporting the decision-makingtask regarding production/distribution and storage network,selection of most appropriate technologies, units sizing, allo-cation of matter flows, allocation of products and monthlyplanning of activities, while providing costs and environ-mental impact shares. The co-combustion echelon in themathematical formulation is modeled as a transformationcoefficient to estimate electricity production, i.e. the pro-cess efficiency on LHVar basis, 36% as a conservative value(Berndes et al., 2010). The problem assumes a maximumcoal substitution of 15% in thermal basis and by powerplant. It deals with monthly periods for a given amountof identical years. It considers solid and liquid biomasstransportation. The developed tool is used for a specificcase study, but can be adapted to handle any other typeof case, with wider (international) or narrower (regional)frontiers.

The article is organized as follows. The superstructure ofbiomass pre-treatment technologies is characterized, togetherwith biomass storage, as echelons that change biomass prop-erties. After this section, the major features of the approachare described. Then, the problem statement sums up the maininput and output data. The mathematical model part depictsthe current model from the previous formulation developed(Prez-Fortes et al., 2012). Following that, the paper focuseson the specific case study, and finally on the results and dis-re-treatment selection for the use of woody waste in co-combustion01.004

cussion of the economic and environmental results of thescenarios performed.

dx.doi.org/10.1016/j.cherd.2014.01.004

ARTICLE IN PRESSCHERD-1463; No. of Pages 246 chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxx3. Biomass storage and superstructure ofbiomass pre-treatments

The presence of pre-treatment technologies is needed toenhance biomass properties to convert biomass into a com-modity in energy projects. Thus, pre-treatment requirementis one of the drawbacks of using biomass. The work by Balsand Dale (2012) remarks that as the pre-treatment operationsincrease considerably the cost of the final biomass product,it is crucial to determine if the tasks of homogenization anddensification are needed to have the desired results. As alter-native, it can be determined if the project can be locallyperformed, that is the promotion of distributed projects ratherthan CES. Nowadays, biorefinery installations have not beenwidely established since there is a lack of sustainable SCfor lignocellulosic feedstock. Biorefineries (centralized) needdistributed pre-treatment installations. However, current pre-treatment technologies are expensive and inefficient, reasonswhy this field needs more research and development (Kurianet al., 2013). Wu et al. (2010) put into manifest that a bioslurry-based SC would be competitive if more than 1500 dry tons perday are being transported. On the other side, for less than 500dry tons processed per day, a conventional SC is favored.

Fig. 1 depicts five main steps, that modify biomasscharacteristics: biomass collection, pre-treatment, storage,transportation and treatment. Pre-treatment and storage ech-elons may change biomass characteristics, as described in thefollowing lines. Torrefaction and fast pyrolysis are the state-of-the-art technologies considered, concretely fast pyrolysisin an early stage of development, while the well-known pel-letization is also included by itself, or in combination withtorrefaction, as pre-treatment options.

Storage. This echelon may change MC, DM and LHV dueto microbiological degradation. It is mainly affected by thetemperature of the stored biomass, biomass properties andstate (solid or liquid). The type of storage depends on theclass of biomass to be stored, i.e. (i) open-air storage in trop-ical summer or during hot periods for raw solid biomass,which even contributes to biomass drying, (ii) silos andbunkers (closed storage) for pre-treated biomass. Fermenta-tion can develop hot spots, which can set on fire the biomasspiles, so storage facilities must be carefully designed. Liq-uid biomass, like bio-oil, rarely changes its properties whilestored if convenient conditions are set. To alleviate degra-dation, the stored biomass should be as homogeneous aspossible and have a MC usually under 20% on a mass basis(Rentizelas et al., 2009a,b).

Torrefaction. It is a thermal process that takes place atrelatively low temperatures, 225300 C, and uses a lowheating rate 50 C/min, performed at atmospheric pres-sure in an inert atmosphere (Prins et al., 2006a). The solidproduct is a uniform solid with lower MC and higher LHVthan the raw material, as well as highly hydrophobic Usluet al. (2008). The by-products obtained are a condensableliquid and a non-condensable gas. It is a relatively newtechnique applied to biomass, but the benefits have beenalready perceived via higher BD values and better grind-ability (Prins et al., 2006b; Uslu et al., 2008) that enhancebiomass gasification efficiency (Couhert et al., 2009; Prinset al., 2006a), produce a less reactive char with steam aftergasification (Couhert et al., 2009) and limit biological degra-Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

dation (Bergman et al., 2005). The paper by Deng et al. (2009)evaluates the performance of torrefied agricultural residuesin co-gasification with coal: torrefied biomass is more simi-lar to coal than raw biomass, as well as more homogeneous.

Pelletization. It is a physical process that dries, mills, densi-fies and homogenizes biomass, frequently using only ligninas binding agent. The BD is higher, biological degrada-tion is restricted, and the matter is easier to handle thanraw biomass, with the subsequent benefits in transporta-tion and storage. Nevertheless, pellets, which are cylindersof 68 mm diameter, may acquire moisture, have a lowermechanical resistance toward crushing and may form dust.The market share of pellets is growing and this is the mostcommon form of biomass commercialized. Wood is the pri-mary raw material used (Sultana et al., 2010; Maciejewskaet al., 2006).

Pelletization of torrefied biomass (TOP pellets). The combi-nation of both techniques can overcome the drawbacks oftorrefied biomass and pellets separately, with subsequentadvantages in transportation and storage. TOP pellets havelow MC with limited predisposition to uptake humidity,biological degradation is almost completely inhibited, den-sity and calorific value are higher. Moreover, TOP pelletsallow a wider range of raw biomass types and yield simi-lar physical properties to the final product. Pelletization oftorrefied biomass is less energy consuming than pelletiza-tion of untreated biomass (Uslu et al., 2008; Maciejewskaet al., 2006). According to Uslu et al. (2008), TOP pellets arethe most suitable option for international trade.

Fast pyrolysis. Pyrolysis is a thermal process that decom-poses biomass at temperatures in the range of 400800 C inan inert atmosphere. Slow and fast pyrolysis differs in theheating rate, and therefore, in the proportion of char, gasand liquid products obtained. Products range also dependson the biomass type and reaction parameters (Uslu et al.,2008). Fast pyrolysis takes place at 450550 C and countswith heating rates as high as 60,000 C/min. Its main prod-uct is in liquid phase, the so-called bio-oil or pyrolysis oil.It is a mixture of 70% oxygenated organics and 30% water,on a mass basis (Maciejewska et al., 2006). Bioslurry is themixture of bio-oil and biochar, which is the solid productfrom the pyrolysis process, milled into fine particles. Hence,the efficiency of the process is higher. As a liquid fuel, trans-portation and handling can be easier and cheaper (Abdullahet al., 2010; Wu et al., 2010).

The different biomass properties transformationcoefficients for each technology and echelon are sum-marized in Section 7, in accordance with linear models forMC, DM, BD and/or LHV modification. Table 5, in AppendixA, compiles the biomass properties that change into each SCechelon considered.

4. Problem statement

The bio-based SC configuration and operation should be tailor-made. The approach described in this work can be adapted toany other decision-maker needs, for example, the fraction ofcoal to be replaced, different technologies share and biomasstypes, with the subsequent differences on seasonality andintrinsic properties.

The bio-based SC assumes fixed located sites for biomasscollection, as well as with established points for demand, elec-tricity generation plants that request a specific amount of inletre-treatment selection for the use of woody waste in co-combustion01.004

energy supplied by biomass waste. A list of potential locationsfor pre-treatment sites and distribution centers or storage is

dx.doi.org/10.1016/j.cherd.2014.01.004


pdmoeabcoaSt

1

2

3

1

2

roposed taking into account the highest biomass waste pro-uction sites. Processing sites are limited by minimum andaximum capacities. They can be supplied by more thanne raw matter producer. Investment costs take into accountconomies of scale. Distribution centers and transportationctivities are modeled by considering upper and lower boundsased on their biomass handling capacity. Distribution centersan be supplied from more than one biomass producing siter pre-treatment plant. A material flow between facilities mayppear if selecting such flow improves the performance of theC. The market demand for energy can be satisfied by morehan one site.

Inputs:

. Process data Amount of raw biomass waste that is available. Matter

characteristics and seasonality. SC echelons and inlet biomass properties required, i.e.

conditions of humidity, shape and/or energy content, forthe specified activities. A set of pre-treatment technolo-gies, types of storage and means of transportation.

Activities (pre-treatment, storage, transportation andtreatment) efficiencies, to linearly model each processand determine the outlet matter characteristics in termsof MC, DM, BD and LHV. Utilities consumption and life-time for each one of the activities.

A set of matter states that quantifies MC, DM, BD andLHV for each flow of mass between activities.

A set of demands, i.e. amounts of seasonal energyrequired by the co-combustion plants.

A set of providers, intermediates and consumers loca-tions.

Time period, planning horizon, project lifetime and aver-age working hours for a specific interval of time.

. Economic data Investment, fixed and variable costs associated to all the

technological options involved: pre-treatments, storageand co-combustion plants adaptation for biomass firing.Unit transportation costs per km and volume of biomassto move.

Base escale and the associated economies of scale fortechnologies capacity.

Price of raw biomass, coal, utilities, consumables andelectricity.

Interest rate.. Environmental data Environmental interventions of the biomass waste and

coal production. Environmental interventions of the electricity produc-

tion from coal. Activities environmental interventions. Utilities environmental interventions.

Outputs:

. The network structure that shows the optimal level ofcentralization/de-centralization: selected pre-processingunits with their capacities and locations; types of trans-portation, storage sites and dimensions, percentage ofsubstitution of coal for each power plant. Connectionsamong sites.Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

. Feedstock utilization and schedule, i.e. suppliers operationper month.3. Flows of biomass among sites at each time period.4. Inventory levels at each time period.5. Breakdown of costs, revenues, cost saving and investment,

and environmental impacts per echelon or equipment.Impact categories share per each one of the options con-sidered, biomass usage vs. coal replaced. Economic andenvironmental criteria values for the whole time horizon.

The multi-objective optimization integrates economic andenvironmental criteria, taking into account the NPV, in D ,and a LCA in Impact 2002+points (pts). The assumed objec-tive functions quantify the difference between the only-coalwith the coal-biomass SC.

Economic optimization. The NPV evaluates the investmentand the operational costs belonging to the biomass SC andthe feedstock savings related to the SC with only coal (dif-ference between the cost of the coal replaced and the wasteused), at the same level of incomes.

Environmental optimization. The LCA expresses the envi-ronmental impact difference between 100% coal powerplants and plants that replace a fraction of feedstock bybiomass.

5. Mathematical model

The mathematical model of the biomass SC considers dif-ferent decisions among them: (i) activation of SC nodes andlinks among them (a node can be represented by processingor distribution activities), (ii) the capacity to carry out eachperiod, (iii) their capacity utilization levels, (iv) selection of themost appropriate pre-treatment technologies, (v) transporta-tion links, (vi) fraction of coal replaced by biomass, and (vii)the amount of matter moved along the different SC nodes. Asperformance indicators, the overall NPV as well as the envi-ronmental impact savings associated with the whole SC areutilized.

The resulting model is solved using a multi-objective MILPapproach, which allows to assess the trade-off between theenvironmental damage categories and the economic indica-tor. The variables and constraints of the model are classifiedinto three groups: (i) process operations constraints given bythe SC topology (the so-called design-planning model), (ii) theeconomic metric and (iii) the environmental metric formula-tion. They are described in the following paragraphs.

Fig. 1 is a general outline of the different components ofthe bio-based SC model: sourcing, pre-treatment, final prod-uct processing and distribution. The SC is defined as a numberof potential locations where processing sites or distributioncenters, or both of them can be installed. Suppliers are atfixed locations where biomass is available. The product can beprocessed at several sites. The properties of raw biomass maybe modified/improved by means of the pre-treatment units soas to allow intermediates to meet the characteristics requiredby subsequent steps in the SC. Pre-treatments may be alsoconvenient to induce savings in transportation costs.

5.1. Design-planning model

The design-planning model selected to deal with the biomassSC network is adapted from the works of Lanez-Aguirre et al.(2009), Prez-Fortes et al. (2012). This model translates there-treatment selection for the use of woody waste in co-combustion01.004

State-Task-Network (STN) formulation (Kondili et al., 1993),which is a widely known approach for scheduling, to the

dx.doi.org/10.1016/j.cherd.2014.01.004

ARTICLE IN PRESSCHERD-1463; No. of Pages 248 chemical engineering research and design x x x ( 2 0 1 4 ) xxxxxxSC context. The main advantage of this formulation is thatit collects all the SC nodes activities information through asingle variable set. This feature eases the economic and envi-ronmental metrics formulation as well as it facilitates theconsideration of the different pre-treatment and their possiblecombination to obtain the required outputs. The SC materialbalances can be modeled by means of a single equations setfor all materials and facilities. This is possible since processingnodes and distribution centers, as well as final products, rawmaterial and intermediates are handled indistinctively. Themost relevant variable of the model is Pijff t [kg], which rep-resents the magnitude of particular task i, performed usingtechnology j during period t, whose origin is location f anddestination is location f. In the case of production activities,they must receive and deliver material within the same site(Pijfft). In contrast, in a distribution activity, facilities f and f

must be different. This mathematical formulation assumesthat an activity consumes and produces certain materialswith determined properties and can be performed in differ-ent equipments. By using the activities as the core of theformulation rather than products-materials, it results in aflexible formulation that can easily incorporate new technolo-gies/processes. The equations comprising this formulation aredescribed in the following paragraphs.

The mass balance must be satisfied at each node of thenetwork. The expression for the mass balance for each type ofmaterial s (i.e. raw material, pre-treated biomass, or final prod-uct) processed at each potential site f in every time period t ispresented in Eq. (1). Parameter sij [dimensionless] is definedas the mass fraction of material s that is produced by task iusing technology j. Ts is a set that refers to tasks that produces, while sij and Ts are associated with tasks that consume s.

Ssft Ssft1 =f

iTs

j(JiJf )

sijPijf ft

f

iTs

j(JiJf )

sijPijff ts, f, t (1)

The change in inventory of material s for consecutive plan-ning periods and for each location is given by the differencebetween the amount of material produced by those tasksbelonging to set Ts and the amount consumed by those tasksincluded in set Ts. The model assumes that process parame-ters such as conversions, separation factors or temperatures,are fixed for each activity in order to enforce the linearity ofthe problem. In this sense, parameters sij [dimensionless] andsij [dimensionless] provide the recipe for a specific activity.Nevertheless, there are activities (I) for which it is desirable tolet the model specify the mixture of inputs in order to achievea given value of a specific biomass property, for instance, aspecific moisture content or calorific value. For such activities,the combination of feedstocks and, therefore, the proportionof each feedstock is variable. In order to take into account suchactivities, the mass balance is modified as shown in Eq. (2).Note that Eq. (1) is a particular case of Eq. (2).

Ssft Ssft1 =f

iTs

j(JiJf )

sijPijf ft f

iTs

j(JiJf )

sijPijff t

+

Pvsijft

Pvsijft s, f, tPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

i(TsI)j(JiJf ) i(TsI)j(JiJf )(2)With regard to the variables Pijff t [kg] and Pvsijft [kg], notethat the former is used in the mass balance (Eq. (2)) coupledwith the parameter sij or sij which specify a fixed proportionof material produced or consumed for a task i. On the otherhand, Pvsijft is modeling flexible tasks which allow the propor-tion of the material produced or consumed to vary so as toprovide alternatives for biomass mixing to the model. For thisreason, Pvsijft is not multiplied by such parameters in Eq. (2).

For the activities for which biomass properties are fixed,the energy balance is satisfied directly by the definition of thestreams. However, it is necessary to verify that the energybalance is satisfied for the flexible activities (I). The energybalance for such activities is represented by Eq. (3). Here, HVs[MJ/kg] is the heating value of material s and ij is the efficiencyof process i performed in equipment j. Each different type ofbiomass has a different heating value. Therefore, a specificactivity changes the heating value of the output stream if (i)it is a pre-treatment task that modifies explicitly the calorificvalue of the biomass, or (ii) it is a task whose main objective isthe change of shape, but it is receiving a mixture of biomassesas input.

ij

sTs

HVsPvsijft =sTs

HVsPvsijft i I, j, f, t (3)

Another important extension of this model is the consid-eration that storage is capable of changing biomass properties(e.g. MC, DM, LHV). In order to do so, storage must be consid-ered as an actual activity. Subsets Jstor and Sstor will representthe storage equipment and those materials that when keptin storage change their properties, respectively. Notice that themass balance has been decomposed in this case so as to dealwith the one period delay necessary for the properties changeto occur. As expressed by Eqs. (4) and (5), a storage activityplaces inventory in the current period t and takes inventoryfrom the previous period t 1.

Ssft =f

iTs

j(JiJf Jstor )

sijPijf ft s Sstor, f, t (4)

Ssft1 =f

iTs

j(JiJf Jstor )

sijPijff t s Sstor, f, t (5)

In order to keep the model linear Eqs. (2) and (3) arelimited to combine different materials in proportions whichare defined by the optimization just for the activities rightbefore the energy generation. The states feeding any otheractivity and their corresponding proportions and properties(e.g. MC, DM, BD, LHV) must be defined a priori.

Eqs. (6) and (7) represent the temporal change in the equip-ment technology installed in a potential facility location. Wewill consider economies of scale by using a piecewise linearapproximation in K different intervals and a so-called SOS2variable type (jfk). Such variables are positive and at mosttwo consecutive variables are non-zero. FElimitjfk [h] is the limitof capacity for interval k. Vjft is a binary variable indicatingwhether or not capacity of technology j is expanded at site fin period t. This formulation will be recalled in the economicperformance metric section for computing the investmentsassociated with capacity expansions. Eq. (8) is used for thetotal capacity Fjft [h] bookkeeping taking into account there-treatment selection for the use of woody waste in co-combustion01.004

capacity expansion during the planning period t (FEjft [h]) forequipment technology j in facility f. This equation considers

dx.doi.org/10.1016/j.cherd.2014.01.004


tr

F

ttautrijf

drdt

cbr

cemEtb

itboc

A

S

trhe case of the initial design of a SC (FEjf0 = 0) as well as a SCetrofit scenario (FEjf0 /= 0).

k

jfktFElimitjfk = FEjft f, j Jf , t (6)

k

jfkt = Vjft f, j Jf , t (7)

jft = Fjft1 + FEjft f, j Jf , t (8)

Eq. (9) is used to ensure an utilization greater than or equalo a minimum value established by the decision-maker andhat the utilized capacity is lower than or equal to the avail-ble one. Parameter jf [dimensionless] defines a minimumtilization rate of technology j in site f as a proportion of theotal available capacity. Parameter ijff [h/kg] represents theesource utilization factor. This is the capacity utilization rate,n terms of capacity units (e.g. machine-hours), of technology

by task i whose origin is location f and destination location.

jf Fjft1 f

iIj

ijff Pijff t Fjft1 f, j Jf , t (9)

The capacity is expressed as equipment j available timeuring one planning period, then ijff represents the timeequired to perform task i in equipment j per unit of pro-uced material. Thus, once operation times are determined,his parameter can be readily approximated.

Eq. (10) guarantees that the amount of raw biomass s pur-hased from site f at each time period t is lower than an upperound given by suppliers availability Asft [kg] (e.g. seasonality,esidues availability in a specific region).

f

iTs

jJi

Pijff t Asft s RM, f Sup, t (10)

Adaptf is a variable which represents the capacity of a co-

ombustion plant to process biomass fuels, while plantf

is thefficiency of plant f. Eq. (11) expresses the feasible limits for theodification of a combustion plant to process biomass fuels.q. (13) expresses that inlet energy flows must be lower thanhe available capacity at the co-combustion plant f to processiomass fuels.By Eq. (12) sales of final product s FP carried out from facil-

ty location f to market f < Mkt are estimated. Eq. (14) is usedo express that the demand can be partially satisfied due toiomass production or supplier capacity limitations: the salesf product s carried out in market f during the time period tould be less than or equal to the demand.

daptLf Adaptf AdaptUf f Mkt (11)

alessf ft = plantf HVs

Pijf ft

Please cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

i(TsTr)j(JiJf )

s FP, f Mkt, f / Mkt, t (12)sFP

f /Mkt

Salessf ft Adaptf f Mkt, t (13)

f /Mkt

Salessf ft Demsft s FP, f Mkt, t (14)

It is important to emphasize that for this sort of networksthe final product is the energy delivered to the differentregions (Mkt). Thus, for those tasks that carry out the energygeneration their corresponding parameter sij is a function ofthe heating value of the input materials and the efficiency ofthe equipment.

5.2. Economic formulation

The expressions required to compute the operating revenue,the operation costs, the total capital investment, and NPV arenext described.

The operating revenue is expressed in Eq. (15) as the prod-uct sales during period t.

ESalest =sFP

f Mkt

f /(MktSup)

Salessf ftPricesft t (15)

The operating costs include fixed and variable costs: Eq.(16) describes the total fixed costs of operating the SC network.FCFJjft [D /h] is the fixed unit capacity cost of using technologyj at site f.

FCostt =

f /(MktSup)

jJf

FCFJjftFjft t (16)

In turn, as variable costs, the cost of purchases from sup-plier e, includes raw material procurement, transportation andproduction resources, as shown in Eq. (17). The purchases ofraw materials (Purchrmet [D ]) made to supplier e are evaluatedin Eq. (18). We will assume a different supplier for each com-ponent of the variable costs. This assumption can be easilyrelaxed to account for the specific characteristics of the prob-lem being dealt with. The variable est [D /kg] represents thecost associated with raw material s purchased to supplier e.Transportation and production variable costs are determinedby Eqs. (19) and (20), respectively. tr

eff t [D /kg] denotes the eprovider unit transportation cost associated with material dis-tribution from location f to location f during period t. ut1

ijfet

[D /kg] represents the unit production cost associated withperforming task i using technology j, whereas ut2

sfet[D /kg] rep-

resents the unit inventory costs of material s storage at sitef. The parameter ut1

ijfetentails similar assumptions to the ones

considered for sij and sij, namely, the amount of utilities andlabor required by an activity are proportional to the amountof material processed.

EPurchet = Purchrmet + Purchtret + Purchprodet e, t (17)

Purchrmet =sRM

f Fe

iTs

jJi

Pijfftest e Erm, t (18)

re-treatment selection for the use of woody waste in co-combustion01.004

Purchet =iTr jJiJe f f

Pijff ttreff t e Etr, t (19)

dx.doi.org/10.1016/j.cherd.2014.01.004


subject toPurchprodet =

f

i/Tr

j(JiJf )

Pijfftut1ijfet

+s

f /(SupMkt)

Ssftut2sfet e Eprod, t (20)

The total capital investment on fixed assets is calculatedby means of Eq. (21). These equations include the investmentmade to expand the technologys capacity j in facility site f inperiod t. Recall that economies of scale for technologies capac-ity is considered using a piecewise linear approximation in Kintervals. Here, Pricelimitjfk [D ] is the investment for a capacityexpansion equal to the limit of interval k (FElimitjfk [h]).

FAssett =f

j

k

Pricelimitjfk jfkt t (21)

Eq. (22) calculates the profit in period t, as operating rev-enues minus fixed and variable operating costs. We take intoaccount the cost related to the coal that is replaced bybiomass which is expressed in Eq. (23). The NPV is calculatedas in Eq. (24). Note that the NPV only for the biomass-based-energy generation and considers the opportunity costsassociated with the fraction of replaced coal.

Profitt = ESalest (FCostt +

e

EPurchet

)t (22)

TCostCoalt =f Mkt

f /Mkt

s

CostCoalf Salessf ft

f HVCoalf

(23)

NPV =t

(Profitt FAssett + TCostCoalt

(1 + ir)t)

(24)

5.3. Environmental formulation

The application of the LCA methodology to the SC modelrequires four steps according to ISO documents (ISO14040):goal definition and scope, life cycle inventory (LCI), life cycleimpact assessment (LCIA) and results interpretation. LCI val-ues are retrieved from Ecoinvent (2006) database using PRConsultants (2004).

Environmental interventions are translated into metricsrelated to environmental impact as end-points or mid-pointsmetrics using characterization factors. Eq. (25) calculates ICaft[pts] which represents the mid-point environmental impact aassociated with site f, as a consequence of carrying out activi-ties in period t. In turn, ijff a [pts/kg] is the a characterizationfactor of the environmental category impact for task i per-formed using technology j, receiving materials from node fand delivering them to node f.

ICaft =jJf

iIj

f ijff aPijff t a, f, t (25)

Likewise to sij and sij, the value of ijff a is fixed and con-stant, provided that all environmental impacts are consideredlinearly proportional to the activity performed in the node(Pijfft [kg]) (Heijungs and Suh, 2002). Environmental impactsPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

associated with transportation, have as FU the amount ofkg of material transported over a given distance (kg km).Consequently, the value of the mid-point environmentalimpact ijff a associated with transport, is calculated as inEq. (26), where T

ijff a [pts/(kg km)] represents the a charac-

terization factor of the environmental category impact forthe transportation of a mass unit of material over a unit oflength. Note that the impact is assigned to the origin node.The environmental impacts associated with production (Eq.(25)) or transportation (Eq. (26)), can be calculated by settingthe indices summation over the corresponding tasks (i Tr ori NTr). A tortuosity factor (Tortuosity [dimensionless]) may beemployed to correct the estimated (linear) distance betweennodes.

ijff a = Tijadistanceff Tortuosity i Tr, j Ji, a, f, f (26)

Eq. (27) introduces DamCgft [pts] which is a weighted sumof all mid-point environmental interventions. They are com-bined using g end-point damage factors ag [dimensionless],normalized with NormFg factors. Moreover, Eq. (28) computesthe g normalized end-point damage along the SC (DamCSCg[pts]).

DamCgft =aAg

NormFgagICaft g, f, t (27)

DamCSCg =f

t

DamCgft g (28)

Similarly to the economic formulation, we consider theimpact that is avoided by replacing the coal by biomass inthe combustion plants. Eq. (29) expresses the impact of thereplaced coal.

ImpactCoal =g

aAg

f Mkt

f /Mkt

t

Coalaf

NormFgagSalessf ft

f LHVCoalf

(29)

The impact of coal considers (i) coal supply, whereimported coal, moreover, considers the impact of the coverageof longer distances, and (ii) electricity generation as a whole,including (iii) transportation impacts. Eq. (30) collects the end-point environmental damages for the whole SC consideringthe savings in terms of impact due to the replacement of coal.Biomass SC impact distinguishes between solid and liquidtransportation. The impact associated with each processingunit is proportional to the utilities consumption, i.e. tons ofdiesel consumed or GJ of electricity used.

Impact2002overall = ImpactCoal f

g

t

DamCgft (30)

For further details regarding the operational and environ-mental formulation, and the specific points distributed byactivities, the interested reader is referred to Bojarski et al.(2009) and Prez-Fortes (2011).

The overall optimization problem can be posed mathemat-ically as follows:

MaxX,Y

{NPV, Impact2002overall

}

re-treatment selection for the use of woody waste in co-combustion01.004

Eqs.(1)(30);

X {0, 1}; Y R+

dx.doi.org/10.1016/j.cherd.2014.01.004


Fig. 2 Specific mathematical formulation nomenclature. The scheme points out how to deal with biomass mixtures andp tates

wtrb

6a

Tfinpcaam

aeTs(t(speciLtit

ad

roperties changes in the linear program: combinations of s

here, X denotes the binary variables set, while Y correspondso the continuous variable set. Note that NPV and Impact2002overallepresent the differences between the SC with coal and theiomass SC, being negative and positive, respectively.

. Major features in the bio-based SCpproach

he model intends to elucidate strategic decisions like thenal distribution network and selection of pre-treatment tech-ologies, as well as tactical determinations like the planning ofrocesses, flows between nodes and allocation of products. Itontributes by integrating a superstructure of pre-treatmentsnd considering biomass properties MC, DM, LHV and BDlong the SC. The model is flexible to incorporate/update infor-ation or change the case of study.The linearity of the mathematical formulation does not

llow the multiplication of variables, like the flow of differ-nt inlet biomass that are combined conveniently to create anew biomass with different properties MC, DM, LHV and BD.his happens in the SC proposed when considering (i) biomasstorage, (ii) biomass combinations before pre-treatment, andiii) biomass combinations before treatment. To deal with that,he current model proposes the definition of specific statess) and tasks (i) that assumes and defines a priori the pos-ible results of the storage activity or of each one of there-treatments. See in Fig. 2 an outline of the different ech-lons and the related mathematical nomenclature. Since theo-combustion plants needs a specific amount of energy (flownlet condition), the only important characteristic here is theHV multiplied by the mass flowrate. Moreover, the output ofhe co-combustion plants is not a biomassic state defined byts properties. For that reason, this echelon is modeled usinghe so-called flexible units, explained in Section 5.

This problem assumed a maximum of two months at openPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

ir storage for raw material, and its subsequent biomass fromifferent periods mixtures. No combination between forest and tasks.

wood residues (FWR) and agricultural wood residues (AWR)is allowed. It creates a total of 55 states along the whole SC.Nevertheless, after preliminary model runs and due to the spe-cific conditions of the problem, the model decides to store onlypre-treated matter, since its storage does not change biomassproperties. So, for the reported case studies results, we sim-plified the number of states to 15.

The model is solved along a period of 10 years, assumingyearly strategical decisions and monthly planning solutions.The MILP is implemented in GAMS, and counts with 59,057equations, 5,925,346 continuous variables and 589 discretevariables. It takes a maximum of 13,971 CPU seconds to reacha solution with a integrality gap of 2.5% on a desktop computerdual four core Intel Core i7 920, with 2.66 GHz CPU and 24 GBRAM with 8 processors, using the MIP solver CPLEX 9.0.

7. Case study: a BSC located in Spain usingco-combustion

The case study is a retrofitting proposal for coal combustionpower plants in Spain that contemplates the use of biomass toreplace a fraction of coal. Given a set of biomass collection sitesand the list of power plants in Spain, the SC model will helpon the decision-making task by providing solutions for thelocationallocation problem and the flows of matter betweensites, while quantifying them in monetary and environmen-tal terms. Universal Transverse Mercator (UTM) geographiccoordinate system is used; linear distances between sites arecalculated and corrected by a tortuosity factor. The bound-aries of the problem are from cradle-to-gate: the final echelonis the co-combustion plant. The distribution and use of elec-tricity will be the same irrespective on their source, thusselecting a boundary limit at the co-combustion plant seemsreasonable. On the other hand the environmental interven-tions associated with the co-firing are not distinguished andre-treatment selection for the use of woody waste in co-combustion01.004

the environmental impact of co-firing is mainly associatedwith the use of coal. Several issues might arise when co-firing,

dx.doi.org/10.1016/j.cherd.2014.01.004


Table 1 Feedstock characteristics (Gmez et al., 2010a).

Biomasswaste

MC (wt%) LHVar(MJ/kg)

BD (kg/m3) Yearlyavailable (kton)

Adjustedavailability (kton)

Seasonality Cost (D /ton)

FWR 30 12.5 140 7748 1162 None 56AWR 40 10.8 100 3883 2718 Summer and winter 52

raw material. MC decreases due to natural drying and DM is

0 200 400 600 800 1000

4000

4200

4400

4600

4800

UTM coordinate x (zone 30 no rth) (km)

UT

M c

oord

inat

e y

(zon

e 30

nor

th)

(km

)

FWRAWRCoComb. plantsPretreat., storage

Fig. 3 Feedstock sites and the power plants location map.They are represented according to their relative GJ/yrproduction for sources, and according to their relative GJ/yrthermal input demand for power plants. Potential sites forpre-treatment and storage selection are selected amongfeedstock sites that produce higher amounts of biomass.

4

3

3.5

4

2.5

3

onth FWR

1.5

2

PJ/m

o

AWR

Co-combuson plants

0.5

1o co bus o p a s

(10% inlet ener gy)

0

see for example (Dai et al., 2008), one of them is ashes forma-tion, which is increased by biomass use. In this approach weassume the use of biomass while co-firing does not changesdramatically the environmental interventions of coal powerproduction. In this sense these interventions are not signifi-cant when plant operates at low co-firing ratios, and/or whenhigh-quality biomass is used. The currency used is D 2010.1

7.1. Raw materials and coal combustion plants

The types of biomass waste used in this case study are FWRand AWR from Gmez et al. (2010a,b). The amount of biomassavailable is estimated by a hierarchy of potential approachthat integrates physical, geographical and technical limita-tions, providing an upper bound for the potential estimation.Gmez et al. (2010a) use geo-referenced information to esti-mate the biomass production. The starting point of the presentwork are the UTM coordinates of selected collection areas andtheir biomass production, determined in Gmez et al. (2010a).The size of the areas are decided to provide enough biomass toallow a minimum installed capacity of hypothetical combus-tion or gasification plants at small scale. The reference areasare of 60 km 60 km for AWR and 80 km 80 km for FWR,regarding their territorial distribution and LHV.

We assume that no transportation cost is charged forbiomass assembly inside the collection areas. Only those areasthat produce more than 50 kt compiled are considered for theco-combustion network. The amount of available biomass isreduced to take into account a certain competition with othermatter applications. Fig. 3 depicts the location and the rela-tive size, in energy terms, of the selected biomass providers.Table 1 sums up the main characteristics of FWR and AWRused for modeling purposes.

The selected technique that allows biomass usage in a com-bustion plant is co-firing. According to Basu et al. (2011), theinvestment is 192 D /kWth. Operation and maintenance (O&M)costs are 4% the investment (Gmez et al., 2010b). The existinggroup of coal combustion power plants in Spain is consideredby its used capacity during the year 2010, with an averageworking hours of 2800 h/yr (Red Elctrica de Espana, 2010).See in Fig. 3 the power plants represented by their maximumpercentage of inlet thermal power that can be replaced bybiomass, i.e. 15%, without any efficiency penalty. The type ofcoal used into each plant is specified in Lpez-Vilarino et al.(2003). See in Appendix A Tables 7 and 6 for further detail aboutthe power plants, their coal origin and power produced.

7.2. Seasonality and storage

The total amount of energy contained in the biomass wastethat can be used in the co-combustion power plants is set inPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

44 PJ/yr; 15 PJ/yr coming from FWR and 29 PJ/yr from AWR. It is

1 Chemical Engineering Plant Cost Index (CEPCI) from ChemicalEngineering.assumed that the amount of coal that has to be replaced is a10% of the total inlet thermal power, 22 PJ/yr. The yearly energyprofile is represented in Fig. 4: the period starts in March inorder to allow biomass storage for the upcoming months. It isimportant to remark that between the biomass potential andthe thermal demand, the SC efficiency, i.e. matter and energylosses in storage and pre-treatment echelons, decreases theinitial available energy that have to match with the powerplant requirement (see Fig. 5).

Biomass storage is allowed after harvesting/collection andafter pre-treatment sites, according to the general formula-tion. However, as pointed out in the previous section, thiscase study only considers storage or pre-treated matter. Thepresent case study considers open air covered storage forre-treatment selection for the use of woody waste in co-combustion01.004

3 4 5 6 7 8 9 10 11 12 1 2Mont h

Fig. 4 Biomass seasonal availability and power plantsdemand, according to a 10% of inlet thermal energyreplacement.

dx.doi.org/10.1016/j.cherd.2014.01.004


Chipp er

Chipp er

Chipper

Resi due piles

Resi due pile s

Resi due pile s

BM

BM

BM

Covered

storage and

drying

Dryer

Dryer

Torre fac tion + pelletizati on

Torrefaction

Sil ostorage

TO POW ER PLANT

Silo

stora ge

Silo

storage

TOR

PEL

TOP

Tank

stora ge

OIL

Tank

storag e

SLU

Pelletization

Fast pyrolysis

Fast pyrolysis + char gri ndi ng

Dryer

Resi due pil esBM

MC (%w t) = 30LHV (MJ/ kg) = 12.5DM (%wt ) = 70BD (kg/ m3) = 140

MC (%wt) = 30LHV (MJ /kg) = 12 .25DM (%wt) = 70

BD (kg/m3) = 240

MC (%wt) = 15

LHV (MJ/ kg) = 14.5DM (%wt ) = 85BD (kg/ m3) = 198

MC (%wt) = 10

LHV (MJ/kg) = 15.5

DM (%wt) = 90BD (kg/m

3) = 187

MC (%wt) = 5

LHV (MJ/kg) = 16.5

DM (%wt ) = 95BD (kg/ m3) = 177

MC (%wt) = 3LHV (MJ/kg) = 21.75

DM (% wt) = 97BD (kg/m

3) = 230

MC (%wt) = 1.5

LHV (MJ/kg) = 22.51

DM (%wt) = 98.5

BD (k g/m3) = 800

MC (%wt) = 9

LHV (MJ/kg) = 12.33

DM (%wt) = 91

BD (kg/m3) = 575

MC (%wt) = 20

LHV (MJ/kg) = 12.17

DM (%wt) = 80

BD (k g/m3) = 1200

MC (%wt) = 20LHV (MJ/kg) = 17.0DM (%wt) = 80

BD (kg/m3) = 1200

Eff = 96.04%

Eff = 97.54%

Eff = 98.4 5%

Eff = 99.2 7%

Eff = 91.46%

Eff = 91.71%

Eff = 89.85%

Eff = 61 .08%

Eff = 86 .08%

on a

rc2msdItea

7

FlscctbicT

Fig. 5 Properties for a FWR stream along the SC. LHV is

educed due to degradation: losses of 2% MC and 0.25% DMan be accounted (Rentizelas et al., 2009a; Maciejewska et al.,006). Raw biomass waste can be stored a maximum of twoonths. Storage of raw biomass allows for mixtures of the

ame type of biomass, gathering it from diverse sites and atifferent months. Mixtures of FWR and AWR are not allowed.t is assumed that storage of pre-treated biomass, in silos oranks, does not change biomass properties. Therefore, thischelon, that preceeds power plants, is modeled as a flexiblectivity.

.3. Pre-treatment units and transportation

ig. 5 depicts the general network considered. After being col-ected, the biomass waste may be transported to differentites to be stored. Following, there are two mandatory pro-esses before biomass pre-treatment to obtain the mandatoryonditions of MC and shape: chipping and drying. Thereafter,orrefied biomass (TOR), torrefied pellets (TOP), pellets (PEL),io-oil (OIL) or bioslurry (SLU) are produced before being storedf needed, and processed in the power plant. Trucks, adapted toarry solids or liquids are in charge of biomass transportation.he average working hours are 2400 h/yr.

Chipping. Based on a roll crusher, this unit increases the BDof raw biomass through conversion into chips to 240 kg/m3.A loss of 2% of raw matter is homogeneously distributedamong MC and DM (Hamelinck et al., 2003). The capacityranges 10 and 80 ton/h.

Drying. A rotatory drum dryer is used to decrease the MC ofthe inlet mixture to 15%, 10% or 5%, depending on the pre-treatment unit requirement. Chipped biomass (from theprevious echelon) changes its LHV and DM content accord-ingly (Hamelinck et al., 2005). The capacity ranges 40 and100 ton/h.

Torrefaction. Biomass with 15% MC is feeded to this unit.Torrefied biomass represents 70% of its initial weight andPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

90% of its early energy content on a daf (dry and ash free)basis. The remaining humidity is 6% the inlet amount. Ther basis. The efficiency (Eff) is defined in terms of LHVar.

final BD is 230 kg/m3 (Uslu et al., 2008). The capacity ranges10 and 160 MW of the inlet energy content.

Pelletization. This activity processes biomass with 10% MC.The final bulk density is 575 kg/m3. The remaining humid-ity is a 17% of the inlet value. The energy loss is assumedto be 94% (Uslu et al., 2008). The capacity to be installed isbetween 10 and 160 MW of the inlet energy content.

Pelletization of torrefied biomass. TOP process increases theBD to 800 kg/m3. MC decreases to 5% of the inlet biomasshumidity (Uslu et al., 2008). The capacity is within 10 and160 MW of inlet energy content.

Fast pyrolysis. The unit processes biomass with a maximum5% MC. Bio-oil represents 73% of the initial weight and 66%of the inlet energy content on a daf basis. The remaininghumidity is 64% the inlet value; the yield of water producedis between 5% and 15% on a dry basis. The BD results in1200 kg/m3 (Magalhes et al., 2009; Uslu et al., 2008). Thecapacity is within 20 and 160 MW of inlet energy content.

Fast pyrolysis and char grinding. The bioslurry is 74% of theinitial weight and 92% of the inlet energy content on a dafbasis. The final MC is 53% of the inlet humidity. The BD is1200 kg/m3 (Magalhes et al., 2009; Uslu et al., 2008). Thisunit capacity ranges between 1 and 20 ton/h.

Transportation. Raw and pre-treated biomass is moved bytrucks. Trucks for solid biomass can handle a maximum of130 m3, while trucks for liquid biomass, 33 m3. Moreover,trucks that move pyrolysis oil are more expensive than nor-mal carbon steel tanks (Hamelinck et al., 2003). Distances(in km) are assumed lineal from site to site with a tortuosityfactor of 1.4 (Gmez et al., 2010b; Yu et al., 2009).

See in Fig. 5, as example, the changes in properties of astream of FWR. The network with the highest efficiency corre-sponds to pelletization of torrefied biomass, while the lowestvalue is for fast pyrolysis.

The total capital requirement is spent at the beginningof the project. The scale factor is 0.7 (Uslu et al., 2008;re-treatment selection for the use of woody waste in co-combustion01.004

Yu et al., 2009). Table 8 (see Appendix A) compiles invest-ment, O&M costs and utilities consumption for pre-treatment

dx.doi.org/10.1016/j.cherd.2014.01.004


200

250

300

350

400

450

1.0

1.2

1.4

1.6

1.8

2.0

ings

(Tho

usan

dsof

pts/yr

)

ses(Billion

s)

NPV losses Impact ann ual savings

0

50

100

150

200

0.0

0.2

0.4

0.6

0.8

EcO-SCN1 EnvO -SCN1 EcO-SC N2 EnvO-SCN2 EcO-SCN3 EnvO -SCN3

Impa

ct20

02+savi

NPV

loss

Fig. 6 Objective functions results for each SCN and

extreme of the PF.

units, storage and transportation echelons. Utilities priceare 1393 D /t (Ministerio de Industria, 2010) for diesel and0.04463 D /kWh (Comisin Nacional de Energa, 2010) for elec-tricity. The electricity sold is bought at 0.03701 D /kWh (OMEL,2011). The depreciation period is 10 years: on the one hand,the combustion power plants have been constructed before2008, and on the other, the minimum average life for a pre-treatment unit is 10 years (Red Elctrica de Espana, 2010). Thecandidate sites to allocate pre-treatment units and storage areselected among the locations with higher amounts of wasteproduced, above 95 kt/yr, to avoid transportation costs (seeFig. 3).

8. Results

The following results show the capability of the developedmodel to propose optimal bio-based SC according to theeconomic and environmental criteria defined in Section 5. Eco-nomic optimization (EcO) and environmental optimization(EnvO) provide accordingly the results that correspond to thetwo ends of the Pareto frontier. The results cover: (i) objec-tive function values, (ii) breakdown of monetary spends, (iii)breakdown of environmental impact, (iv) technology installed,dimensions and flowrates among SC sites and echelonsby month, and (v) elucidation about motivating policies oractions. Three scenarios are proposed to evaluate the trade-offs among the state-of-the-art (SOTA) pre-treatments, i.e.torrefaction, pelletization, pelletization of torrefied biomass,fast pyrolysis and fast pyrolysis combined with char grinding.The scenarios have been proposed to discern the best pre-treatment option, the best option for the case study (includingthe no use of any pre-treatment) and the best liquid-producerpre-treatment:

Scenario 1 (SCN1). This option analyzes all the SOTApre-treatments proposed as option before biomass co-combustion, looking for the optimal technology selection.

Scenario 2 (SCN2). This scenario considers the optimizationof the whole superstructure, including the no use of SOTApre-treatments (however, chipping and drying are manda-tory).

Scenario 3 (SCN3). This alternative optimizes the SC consid-ering as SOTA pre-treatments only fast pyrolysis and fastpyrolysis with char grinding.

8.1. EcO and EnvOPlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

Fig. 6 depicts the values of the two objective functions byscenario. The selected pre-treatments are, (i) torrefaction andpelletization of torrefied biomass for the economic criterion,only torrefaction for the environmental one, in SCN1, (ii) theno use of any SOTA pre-treatment for SCN2, and (iii) the use offast pyrolysis combined with char grinding in SCN3 for bothcriteria. The NPV losses ranges between 0.7 and 2 D billions,while the environmental impact saving oscillates between270 and 425 thousands of pts/yr. At a first sight, SCN2 isthe scenario that provides less NPV losses and higher envi-ronmental impact savings, followed by SCN1 and SCN3. It isremarkable that all the solutions derive into an environmen-tal impact reduction if compared to the use of only coal, evenwith the production of liquid biomass, which uses small trucksfor product transportation, hence with higher environmentalimpact than the trucks used for solid transportation (see asreference Table 8 in Appendix A). As the biomass providersare highly dispersed along the territory in this case study, thetrade-off between the investment in a pre-treatment technol-ogy vs. savings in transportation (due to the higher BD of thepre-treated biomass) favors the simplest choice. To verify anddiscuss this point, we analyze Figs. 7 and 8.

In Fig. 7 the breakdown of initial investments and annualoperation costs are depicted. From Fig. 7(a) it is recognized thatfor SCN1 and SCN3, almost half of the investment comes fromthe adaptation of combustion plants to fire biomass, whilein SCN2 this is the main item. For SCN1 and SCN3 chippingand drying contribute to the investment less than the half ofthe costs of SOTA pre-treatments. Note that in environmentaloptimum solutions the investment in storage is higher thanin economic ones. Fig. 7(b) shows that the difference betweenscenarios due to operational costs is less than the differenceperceived for the investment. The major contribution comesfrom the acquisition of raw material. However, this item isslightly smaller for SNC2; the difference comes from the effi-ciency of the chains (see Table 9 in Appendix B), that takesinto account the loss of matter and energy along the differ-ence processes from the collection step up to the entranceto the co-combustion plant. Differences among scenariosare coming from the installed technologies: a summation ofutilities, O&M and transport. Concerning transportation differ-ences and storage investment differences observed previouslybetween economic and environmental optimum configura-tions, it is deduced that the environmental optimum reduceimpact by reducing transportation, and therefore, increasingstorage.

Fig. 8 depicts the environmental impacts resulting from thedifferent bio-based SC activities. Fig. 8(a) reveals that the elec-tricity generation activity by biomass combustion is the itemthat contributes the most to the overall impact. This is fol-lowed by SOTA pre-treatment options. The lowest contributioncomes from chipping, followed by transportation and biomasscollection. According to Fig. 8(b), the main changes betweeneconomic and environmental optimum configurations are intransportation impact, followed by minimum differences inpre-treatment for SCN1, in transportation and drying for SCN2and just in transportation for SCN3. In SCN1 the differencesin pre-treatment impact are due to the use of torrefaction andTOP pellets, or just torrefaction. In SCN2, the decrease in trans-portation impact is directly related to an increase of dryersinstallation. The contribution of drying and pre-treatmentsto the environmental impact in SCN3 (use of fast pyrolysiscombined with char grinding) is larger than SCN1.

Overall, SCN2 is more advantageous in investment andre-treatment selection for the use of woody waste in co-combustion01.004

environmental impact, while the difference among scenariosis less important just regarding O&M costs. As suspected at

dx.doi.org/10.1016/j.cherd.2014.01.004


0.6

0.8

1.0

1.2nt

(Billion

s)

Chipper Dryer Pre-treatment Storage Power plants

0.0

0.2

0.4

EcO-SCN1 EnvO-SCN 1 EcO -SC N2 EnvO-SCN2 EcO -SC N3 EnvO-SCN3

Investme

(a) Investment needed at the beginning of the project.

150

200

250

300

costs(Millions)

Transport Raw material Ulies O&M

0

50

100

EcO-SCN1 EnvO-SCN1 EcO-SC N2 EnvO-SCN2 EcO-SC N3 EnvO-SCN3

Annu

alc

(b) Annual costs share.

Fig. 7 Breakdown of costs for economic and environmental optimal networks for each SCN.

300

400

500

600

(Tho

usan

dsof

pts/yr)

Transporta on Chi ppe r DryerPre-treatment Biomass Electricity generaon

0

100

200

EcO-SC N1 EnvO-SCN1 EcO-SC N2 EnvO-SC N2 EcO-SCN3 EnvO-SCN3

Impa

ct20

02+(

(a) Environmenta l impact distribution among al l th e SC activities.

60

80

100

120

Thou

sand

sofp

ts/yr)

Transportaon Chip per Drye rPre-treat ment Biomass

0

20

40

EcO-SC N1 EnvO-SC N1 EcO-SCN2 EnvO-SCN2 EcO-SC N3 EnvO-SCN3

Impa

ct20

02+(T

(b) Share of th e environmenta l impact distribution without considering electricity generation.

Fig. 8 Distribution of the environmental impact by echelon of the bio-based SC for each SCN and economic ande

tet

ttnefitats

nvironmental optimal networks.

he beginning of the section, for this specific case study, thextra-investment in SOTA pre-treatments is not justified byhe environmental impact of transportation.

See in Table 2 the comparison, in terms of environmen-al impact, between the contribution of the biomass SC andhe coal replaced SC for each optimum configuration and sce-ario. It is seen that the coal impact is different even if inletnergy replaced is the same in all the cases. This is due to theact that the inlet coal replaced per power plant is differentnto each configuration (see next subsection). The most con-ributing impact categories are climate change for biomass,nd human health followed by climate change for coal. NotePlease cite this article in press as: Prez-Fortes, M., et al., Optimization of pplants. Chem. Eng. Res. Des. (2014), http://dx.doi.org/10.1016/j.cherd.2014.

hat the replaced coal SCs that have higher impacts corre-pond to environmental optimizations. Accordingly, biomass

Table 2 End-point impact categories for each optimum configuSC corresponding to the replaced coal.

End-point impact category EcO-SCN1 EnvO-SCN1

BiomassHuman health 56,202.98 52,397.64 Ecosystem quality 9027.57 8294.78 Climate change 421,785.87 419,046.10 Resources 32,233.37 28,636.16 Impact 2002+/yr 519,249.79 508,374.69

Coal replacedHuman health 379,305.62 385,769.57 Ecosystem quality 20,438.46 22,539.02 Climate change 232,909.63 240,757.59 Resources 176,593.30 185,578.79 Impact 2002+/yr 809,247.01 834,644.97 SCs that optimize the economic criteria have higher impactthan environmental optimum SCs.

8.2. Optimum networks description

Figs. 911 illustrate the optimum networks obtained for eachcriterion and scenario. Biomass providers are characterizedby their relative GJ/yr used from each production site to sup-ply the demand. Co-combustion plants are represented bytheir replaced thermal input in GJ/yr. The lines reproducetransportation of biomass between sites and echelons, implic-itly, in terms of ktons/month. It distinguishes transportationre-treatment selection for the use of woody waste in co-combustion01.004

of raw matter from suppliers toward intermediate (pre-treatment) sites, transportation of pre-treated matter between

ration, taking into account the only biomass SC and the

EcO-SCN2 EnvO-SCN2 EcO-SCN3 EnvO-SCN3

27,350.03 27,095.71 58,077.93 56,683.346815.16 6591.74 9927.48 9613.97365,294.84 357,574.40 442,714.77 441,636.2317,254.91 19,029.81 37,123.09 35,931.63416,714.94 410,291.66 547,843.27 543,865.17

383,845.49 385,774.88 383,567.11 385,771.9121,914.52 22,541.43 21,824.01 22,540.08238,417.47 240,760.41 238,079.73 240,758.83182,928.67 185,607.99 182,540.19 185,591.66827,106.16 834,684.71 826,011.04 834,662.48

dx.doi.org/10.1016/j.cherd.2014.01.004


Table 3 Total capacity installed in the optimum configurations.

EcO

optimization of pre-treatment selection for the useof

Documents

biomass use

use of biomass waste

biomass potential

different pretreated

biomass heterogeneous

energy density

different userplease

different origins