Strategic Optimisation of Biomass

8/18/2019 Strategic Optimisation of Biomass

1/14

Computersand Chemical Engineering 87 (2016) 68–81

Contents lists available at ScienceDirect

Computers and Chemical Engineering

journal homepage: www.elsevier .com/ locate /compchemeng

Strategic optimisation of biomass-based energy supply chains forsustainable mobility

Federico d’Amore, Fabrizio Bezzo∗

CAPE-Lab – Computer-Aided Process EngineeringLaboratory, Department of Industrial Engineering, Universityof Padova, viaMarzolo 9, I-35131 Padova,

Italy

a r t i c l e i n f o

Article history:

Received 9 November 2015Received in revised form

17December 2015

Accepted 5 January 2016

Available online 13 January 2016

Keywords:

Alternative fuel vehicle

Bioethanol and bioelectricity supply chain

First and second generation

Indirect land use change

Multi-objective optimisation

a b s t r a c t

The identification of alternative and sustainable energy sources hasbeenoneof the fundamental research

goals of the last two decades, and the transport sector plays a key role in this challenge. Electric cars and

biofuel fed vehiclesmay contribute to tackle this formidable issue. According to this perspective, a multi-

echelon supply chain is here investigated considering biomass cultivation, transport, conversion into

bioethanol or bioelectricity, distribution, and final usage in alternative bifuel (ethanol and petrol) and

electric vehicles. Multiperiod and spatially explicit features are introduced in a Mixed Integer Linear

Programming (MILP) modelling framework where economic (in terms of Net Present Value) and envi-

ronmental (in terms of Greenhouse Gases emissions) objectives are simultaneously taken into account.

The first and second generation bioethanol production supply chain is matchedwith a biopower produc-

tion supply chain assessing multiple technologies. Both corn grain and stover are considered as biomass

sources. In the environmental analysis, the impact on emissions caused by indirect Land Use Change

(iLUC) effects is also assessed. Results will show the efficacy of the methodology at providing stake-

holders with a quantitative tool to optimise the economic and environmental performance of different

supply chain configurations.

© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The global energy consumption by transport has grown by 2%

per year since 2000 and accounted for 28% of the overall energy

consumption in 2012 (IEA, 2015). Considering that the road trans-

port almost totally relies on petroleum derived fuels, diminishing

the mobility dependency on fossil fuels may represent not only a

strategic decision, but also an environmental necessity. One pos-

sibility to reach that goal is the establishment of the production

of biofuels and bioelectricity for alternative fuel vehicles (AFVs).

On the one hand, biofuels have played a highly significant role in

the search for alternatives as they have seemed tomany the only

feasible approach to replace petroleum-based traditional fuels in

the transport sector. On the other hand, the recent introduction of

the electric vehicles (EVs) in the private fleet market offers a new

possibility to reduce the petroleumdependency.

No parts of this paper may be reproduced or elsewhere used without theprior

written permission of the authors.∗ Corresponding author. Tel.: +390498275468; fax: +39049 8275461.

E-mail address: [email protected] (F. Bezzo).

In regards to both biofuel and bioenergy, many Process Sys-

tems Engineering (PSE) approaches focusing on the Supply Chain

(SC) design and optimisation through mathematical program-

ming (typically Mixed Integer Linear Programming – MILP) have

been recently proposed. With concern to biofuels (in fact, mainly

bioethanol), contributionshavedealteitherwiththemaximisation

of the economic performance (e.g. Dunnett et al., 2008; Zamboni

et al., 2009a), also by considering uncertainty effects (e.g. Dal-

Mas et al., 2011; Kim et al., 2011), or the interaction between

different players (Bai et al., 2012; Yue and You, 2014a,b) or with

the minimisation of the environmental impact (Garcia and You,

2015), typically through a multi-objective optimisation approach

(e.g. Zamboni et al., 2009b; You andWang, 2011). For a more com-

prehensive review, see also Yue et al. (2014c). The design of SCs

for bioenergyproduction has also been optimised in a similar way.

Many mathematical models for biomass production centres and

conversion facilities location have been carried out (e.g. Fiorese

et al., 2005; Freppaz et al., 2004; Voivontas et al., 2001), also com-

bining a detailedenergy conversion optimisationwithenergy/heat

transportationcosts(SödermanandPettersson,2006). For instance,

BruglieriandLiberti (2008)proposedamathematical programming

approach for planning and running an energy production system

process based on burning different biomasses. Contributions have

http://dx.doi.org/10.1016/j.compchemeng.2016.01.003

0098-1354/© 2016 Elsevier Ltd. All rightsreserved.

http://localhost/var/www/apps/conversion/tmp/scratch_5/dx.doi.org/10.1016/j.compchemeng.2016.01.003http://www.sciencedirect.com/science/journal/00981354http://www.elsevier.com/locate/compchemengmailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_5/dx.doi.org/10.1016/j.compchemeng.2016.01.003http://localhost/var/www/apps/conversion/tmp/scratch_5/dx.doi.org/10.1016/j.compchemeng.2016.01.003mailto:[email protected]://crossmark.crossref.org/dialog/?doi=10.1016/j.compchemeng.2016.01.003&domain=pdfhttp://www.elsevier.com/locate/compchemenghttp://www.sciencedirect.com/science/journal/00981354http://localhost/var/www/apps/conversion/tmp/scratch_5/dx.doi.org/10.1016/j.compchemeng.2016.01.003


2/14

F. d’Amore, F. Bezzo / Computers andChemical Engineering 87 (2016) 68–81 69

List of symbols

Sets

g ∈G grid squares, G={1,. . .,60} g’ ∈G set of square regions different than g

i∈ I biomass types, I ={corn, stover }

j∈ J product types, J = {ethanol, DDGS, power }

k∈K production technologies,K = {1,2,3,4,11,22,33}

l∈ L transportmeans, L= {truck, rail, barge, ship, tship} p∈P discretisation intervals for plant size linearisation,

P ={1,. . .,6}

s∈ S lifecyclestages,S = {bg ,bpt ,bt , fp, epow, fd, fdist , ebat ,

ebifuel, ec }

t ∈ T timeperiods, T ={1,. . .,5}

Subsets

elec (k) ⊂ K subset of pure power production technologies,

fratio (k) = {11, 22, 33}

Scalars

ı conversion factor specific for DDGS, 0.954tonDDGS/tonEtOH

fixed costs % over incomes, 0.15 emission inbattery production,3046.924kg ofCO2-

eq/EV

emission in bifuel car driving, 0.005515kg of CO2-eq/kmbifuel

LHV e ethanol lower heating value, 26.952 GJ/tonEtOH ethanol density, 0.7891 tonne/l MWh to tonne ethanol conversion, 0.133570792

tonethanol/MWh

MWh/year to number of EVs conversion,

1.896918157 MWh/EV/year

charg domestic electric charger 1.4kW cost, 59.055

D /newEV

inc differential EVs purchasing cost, 5000D /newEV

KMcost differential EVs driving cost, 0.03 D /kmEVkmCAR average daily trip in Italy, 45km/day

2 conversion of tee to km driven,64825.42357km/tonne

Parameters

̊g averageethanol-petrol distribution diameter, km

AD g arable land density (km2arable/km2grid surface)

BCDmax g maximum cultivation density in region g ,

km2cultivation/km2arable land

dfTCI t discount factor for investments at time t

dfCF t discount factor for cash flow at time t

CFdfCARt discount factor for cash flows at time t for EVs

etperc t ethanol blendingpercentage at time t

t differential EVs purchasing cost reduction at time t

gasolTOT t total number of traditional petrol fleet at time t renewCAR1t relative number of old EVs to be substituted

with new ones at t =4

renewCAR2t relative number of old EVs to be substituted

with new ones at t =5

ωk exceeding electricity production specific for eachconversion technology k, kWhel/lEtOH

ER p ethanol production rate for each plant size p,

tonEtOH/month

PR p electricity production rate for each plant size p,

MWh/month

i,k conversion factor specific for each biomass type i,

tonEtOH/tonbiomass

GS g grid surface, km2

MP j market price for product j

BA g,i biomass i availability for ethanol production in

region g , tonne/time period

ˇi,k fractionof ethanol rate from biomass type i foreachtechnology k

burni,k fractionof biomass i fed to CHP for each technology

k

BY i biomass yield of product i in region g ,tonbiomass/timeperiod/km

2

CI p,k capital investment at each linearisation interval p

and for technology k, MD

c k,cc coefficients for the linear regression of production

costs for each technology k, slope [D /tonEtOH] and

intercept [D ]

fbg i,g emission factor for biomass i growth in grid g and

biomass, kg CO2-eq/tonbiomass fbpt i emissionfactor forbiomass ipre-treatment,kg CO2-

eq/tonbiomass fbt l emission factor for biomass supply via mode l, kg

CO2-eq/tonbiomass km2

ffpi emission factor for ethanol production from

biomass i, kg CO2-eq/tonEtOH fppi,k emission factor forpower production from biomass

i, kg CO2-eq/MWh

ffdl emission factor for ethanol distribution viamode l,

kgCO2-eq/tonEtOH km2

fec k emission credits for each technology k, kg CO2-

eq/tonEtOHLD g,g local delivery distance between grids g and g ’ ,km

PC p,t production costs linearisedfor size pandconversion

technology k, D /timeperiod

PCap p plant capacity of size p used for cost linearisation,

tonne/time period

r k power factor for capital cost estimation for conver-

sion technology k

g,l,g tortuosity factor of transportmode l between g and g ’

UPC i,g unit production costs for biomass type i in grid g ,

D /tonbiomass z i,k biomassconversion intoelectricity,MWh/tonbiomass

Continuous variables

bifuelCARS t number of bifuel vehicles at time t

bifuelKM t total distance travelledbybifuelvehicles at time t ,

km/month

CapElec i,k,g,t supply of biomass i to plant of technology k in

region g at time t , tonne/month

BPC t biomass production constant time t , D /time period

CCF discounted Cumulative Cash Flow,D

CF t Cash Flowat time t , D /time period

Dt Depreciation at time t , D /timeperiodEPC t ethanol production cost at time t , D /timeperiod

ELtot k,g,t energy produced at time t by plant k in region g ,

MWh/month

Etot g,t ethanol produced at time t , tonne/month

EVmt EVs market share at time t

exCOt extra costs for EVs fleet, D /time period

FCC discounted facilities capital costs, D

FCC t facilities capital costs at time t , D /timeperiod

FixC t fixed cost at time t , D /time period

Impact s,t impact forlifecycle stage s at timet ,kgCO2-eq/time

period

Inc t gross earnings at time t , D /time period


3/14

70 F. d’Amore, F. Bezzo / Computers andChemical Engineering 87 (2016) 68–81

p,k, g ,t linearisation variables for TCI at interval p and fortechnology k, in region g at time t

plan p,k,g,t

linearisation variables for TCI at interval p and for

technology k, in region g at time t

NPV net present value,D

NPV chain net present value for SC profit, D

NPV car net present value for EVs fleet, D

nCARS t number of EVs at time t

newCARS t number of new EVs at time t objo objectivefunctionexpressedas thenegative of NPV ,

D , or as overall impact, tonne CO2-eq

PBT t profit before taxes for production technology k at

time t , D /timeperiod

Pbi,g,t production rate of biomass i in region g at time t ,

tonne/time period

Ppi,k,g,t powerproduction ratefrombiomass i throughtech-

nology k in region g at time t , tee/month

PPC t power production cost at time t , D /timeperiod

P TOT j,k,g,t total production rate forproduct j through technol-

ogy k in region g at time t , tonne/time period or

MWh/timeperiod

Potot g,t energyproduced in region g at time t , tonne/month

powerKM t total distance travelled by EVs at time t ,km/month

RISP t money saved driving EVs rather than petrol ones,

D /time period

TAX t tax amount at time t , D /timeperiod

TCbt biomass transport cost at time t , D /timeperiodTCpt products transport cost at time t , D /timeperiod

TCf ,t ethanol transport cost at time t , D /timeperiod

TCI t total capital investment at time t , D

TDt total ethanol and power demand at time t ,

tonne/month

TDetht total ethanol demandat time t , tonne/month

TDpowt total power demandat time t , tee/month

TGHG total GHG impact, kg of CO2-eq

TI t total impact at time t , kg CO2-eq/time periodTP t total ethanol and power production at time t ,

tonne/month

TPetht total ethanol production at time t , tonne/month

TPpowt total power production at time t , tee/month

VarC t variable costs at time t , D /time period

objective objective selection variable

Binary variables

y p,k,g,t 1ifaproductionfacilitykofsize p is tobeestablished

in region g at time t , 0 otherwiseY k,g,t 1 if a production facility k is already established in

region g at time t , 0 otherwise

Y plank,g,t

1 if theestablishment of a newconversion facility k

is to be planned in region g during time period t , 0otherwise

Y start k,g

1 if the establishment of a new conversion facility

k is to be planned in region g at the beginning, 0

otherwise

Acronyms

AFV Alternative Fuel Vehicle

CHP Combined Heat andPower

C +R Combustion andRankine Cycle

DDGS Distiller’s DriedGrainswith SolublesDGP Dry-Grind Process

EU EuropeanUnion

EV Electric Vehicle

GHG GreenhouseGas

G+MCI Gasification and Internal Combustion Engine

G+ TG Gasification andTurboGas cycleIGSP Integrated Grain-Stover Process

LCA Life Cycle Analysis

LCEP Ligno-Cellulosic Ethanol Process

MILP Mixed Integer Linear Programming

NPV Net Present ValueSC Supply Chain

dealt either with the maximisation of the economic performance

(e.g. Patel et al., 2011; Moonet al., 2011; Kumar et al., 2010; Wang

et al., 2009) orwith theminimisationof the environmental impact

(Muench, 2014; Rickeard et al., 2004). ˇ Cuˇ cek et al. (2010)proposed

an MILP model for the economic optimisation of a bioenergy net-

work in a regional context; the generation of different types of

energyproducts, heat, electricity, biofuels and food was taken into

account. Later on, ˇ Cuˇ cek et al. (2012) also proposed a framework

where the design of regional biomass SCs is obtained under the

simultaneousmaximisation of the economic performance and theminimisation of the environmental and social footprints.

Other studies havemainly focused on a well towheel life cycle

analysis (LCA) of alternative vehicles. Most contributions have

focused on EVs, usually demonstrating the potential benefits in

terms of GHGemissionsreduction (e.g. Sandy, 2012), although the

effective impact is shown to be highly dependent on the technol-

ogyused forelectricityproduction (Onatet al., 2015; Tessumet al.,

2014). A study by Campbell et al. (2009) suggested that bioelec-

tricity is more energetically and environmentally sustainable than

ethanol, especially if first generation technologies are taken into

account.

In general, as discussed above, available studies focused on the

optimisation of upstream SCs for ethanol or electricity production

or analysed the environmental effects of the technologies. How-ever, to our knowledge no study have been presented where both

theproduction SC andvehicle utilisation aresimultaneously taken

into account and optimised for a strategic assessment of biomass

exploitation. This work aims at bridging this gap by introduc-

ing a modelling framework where the whole production SC for

ethanol and/or electricity and the final customer needs are opti-

mised according to both economic and environmental objectives.

Bothcorngrainand stoverwillbe consideredasbiomass choices

and several technological options will be taken into account to

produce either ethanol and/or electricity. Site location and scale,

logistic infrastructure definition (biomass or bioethanol transport)

and end user demand evolution for AFVs (bifuel or EVs) will

be simultaneously incorporated within the optimisation model

according tothespatiallyexplicitNorthernItalyframeworkalreadypresentedby Zamboniet al. (2009a,2009b)andGiarolaetal.(2011).

The economic performance of the entire network will be assessed

in terms of the SC Net Present Value (NPV) and of the end user

potential savings inpurchasinganddriving anAFVinstead of a tra-

ditional one. Theenvironmental performanceof thesystemwill be

evaluated in terms of GHG emissions, by considering the impact

of each single life cycle stage and also incorporating the potential

consequences related to indirect Land UseChange (iLUC).

This study is organised as follows. First, the general modelling

frameworkof theSC ispresented; thesubsequentsectiondescribes

themathematical formulation of the model. Themain case studies

are then introduced and, eventually, the results of the bi-objective

optimisation arepresentedanddiscussed.Somefinalremarks con-

clude thework.


4/14


2. Assumptions andproblem statement

This paper deals with the strategic design and planning of an

industrial SC for the production of bioethanol and bioelectricity in

North Italy over a 15-years’ time horizon. The design process is

conceived as a multi-objective optimisationproblem aiming at: (i)

the maximisation of the financial performance of the business (in

terms of global NPV), and (ii) the minimisation of the impact on

global warming (in terms of overall GHGemissions). The problem

is formulatedas a spatiallyexplicitmulti-periodandmulti-echelon

modelling framework devised for the strategic design and invest-

ment planningof biofuels and biopower supply networks.

Theentirenetworkcanbe divided into twomain substructures:

(i) the upstream network, dealing with biomass growth, biomass

pre-treatment and transport to the conversion facilities, and (ii)

the downstream network, dealing with products production, dis-

tribution and final usage by end user. This study integrates the

multi-objectiveMILPmodellingframeworkproposedbyGiarola et

al. (2011), representing the dynamic evolution of a bioethanol SC

localised in North Italy, with biopower production and the imple-

mentationof enduser-related stages (in terms of AFVs fleet).

As depicted in Fig. 1, the set of LCA stages s considered in

the evaluation are given by biomass growth (bg ), biomass pre-

treatment (bpt ), biomass transport (bt ), bioethanol production ( fp),biopower production (epow), bioethanol transport ( fd), fuel dis-

tribution ( fdist ), bifuel vehicles usage (ebifuel), EVs usage (ecars),

batteries production (ebat ) and, finally, emission credits (ec ) in

terms of GHG saving (as a result of goods or energy displacement

by process by-products end-use).

The environmental performance of the system, including issues

such as potential differences in vehicle conversion efficiency, as

well as vehicle technology for petrol (gasoline) progressive sub-

stitution, is based on two main assumptions: (i) carbon dioxide

emissionsresultingfrom thecombustionof biofuels (inbifuel vehi-

cles)areassumedtooffsetthe carbondioxidecapturedduring crop

growth; and (ii) carbondioxide emissions resulting from the com-

bustionof biomassorsyngas(inpowerplants)are assumedtooffset

thecarbondioxidecapturedduringcropgrowth.However,CH4 andN2O emissions are taken into account.

The GHG overall impact is evaluated in terms of carbon diox-

ide equivalent emission (CO2-eq), using the same assumptions as

Giarola et al. (2011) and considering the contribution of the same

gases (CO2, CH4, N2O). In regards toAFVs, their environmental per-

formance is calculated with respect to traditional petrol vehicles.

On the one hand, by assuming that bifuel vehicles have the same

engineandidentical energy efficiency as traditionalpetrol cars, the

emissionsof bifuelvehiclesonly dependon thebiofuel quota com-

bustion while driving (ebifuel in Fig. 1). On the other hand, EVs

(ecars) manufacturing and usage is assumed not to produce any

extraCO2-eqemissions,apart fromthoserelatedtobatteryproduc-

tion(ebat ),whichshould beadjustedbyremovingthecontribution

related to theconstruction of the combustion engine (which is not

installed in EVs). However, several studies in the literature (Faria

et al., 2013) show that: (i)the battery contribution in termsofGHG

emission is almost comparable with the impact deriving by the

wholevehicle production; and (ii) there is no significant difference

in the environmental impact between EV and traditional vehicle

manufactures. Thus, the impact of the combustion engineproduc-

tion can bereasonably neglected.With concernto the finalproduct

distribution, it is assumed that the already existing infrastructure

for both electricity and liquid fuels may be exploited; accordingly

no environmental impact is assigned to the infrastructure distri-

bution stages. However, we incorporated in the environmental

balance thebioethanol CO2-eqemissionquota ( fdist ) related to the

fuel transportationby truck (fromblendingterminals toendusers).

Conversely, the costs related to the bioethanol distribution do not

need to be accounted for; this depends on the way the overall NPV

is represented. As will be detailed in Section 3, the NPV related to

theproduction SC (fromcorn cultivationdownto bioethanol distri-

butionto thedepots) ishandled separately fromtheNPV concernedwith theAFVs.In particular, thelatter is formulatedasa differential

cost between an EV anda bifuel vehicle. Since ethanol distribution

costs are comprised in the fuel liquid price (as much as electricity

distribution costs are within the electricity market price) and the

distribution configuration from depots to final gas station is not

included in the optimisation problem, there is no need to have an

explicit representation.

On the basis of the above mentioned assumptions, the design

problem canbe formulated as follow. Given the following inputs:

• geographical distribution of ethanol demand centres;• bioethanol and bioelectricity demand over the entire time hori-

zon;• biomass geographical availability;• biomass production costs as function of geographical region;• technical and economic parameters as function of biomass type,

conversion technology and plant scale;• environmental burdens of biomass production as a function of

biomass type and geographical region;• environmental burdens of bioethanol and bioelectricity produc-

tion as a function of biomass type andconversion technology;

Fig. 1. Global SC network.


5/14


• transport logistics (e.g. typology, costs, emissions) and allowed

links;• fuel distribution from terminal to enduser;• electricity distribution network efficiency;• ethanolmarket price;• electricitymarket price;• AFVs features (e.g. efficiency, costs, consumptions, average dis-

tances, emissions);

the objective is to study, both economically (in terms of global

NPV maximisation) andenvironmentally (in terms of carbon foot-

printminimisation)the entire SC of biorefineriesandpower plants.

Therefore, the keyvariables tobeoptimisedover theplanningtime

horizon are:

• geographical location of biomass production sites;• biomass production rate and feedstock mix to the plant;• bioethanol facilities technology selection, location and scale;• biopower facilities technology selection, location and scale;• characterisation of transport logistics;• financial performance of the industrial SC over the time horizon;• financial performance of end user economy over the time hori-

zon;• demands quota evolutionover the time horizon (Instance B);• system impact on globalwarming.

The overall time horizon has been divided into five time inter-

vals (each three-years long), starting from 2015.

3. Mathematical formulation

The general modelling framework was formulated as a MILP

problemaccording tothemathematical featuresoutlinedin Giarola

et al. (2011). In particular, we retained themathematical formula-

tion for:

• objective functions general definition;• bioethanol SC economics;• cost linearisation;• logical constraints andmass balances;• environmental issues related to bioethanol SC.

On the other hand, the following new features were imple-

mented:

• biopower SC economics;• end user AFVs-related economics;• environmental issues related to biopower SC;• environmental issues related to AFVsutilisation.

New formulations and adjustments are discussed in the fol-

lowing. Further details can be found in Section 4 and in the

Supplementarydata.

3.1. Objective functions

The first objective function is the maximisation of the NPV [D ]

of theentirebusiness,whichis here expressedas theminimisation

of its opposite value:

Objeco = −NPV (1)

One important difference with respect to Giarola et al. (2011)

is that here the NPV is calculated by summing the industrial SC

profit (NPV chain [D ]) and the end user savings or costs (NPV car [D ])

in driving EVs instead of bifuel vehicles1:

NPV = NPV chain +NPV car (2)

In other words, on the one side we separate the economics of

theproductionSC from the economicinterest of thefinalcustomer.

On the other side, the impact on final customer is represented as

a cost difference with respect to a bifuel car. Note that the over-

all optimisation results in terms of theSC configuration would notchange if NPV car were described as the actual cost of buying and

using bifuel or EVs. Also note thatNPV car is not an actual and tan-

gible profit; it represents an economic advantage or burdenwhen

moving from traditional cars to EVs and can be interpreted as a

market assessingmetric to forecast AFVs penetration. TheNPV chainis calculated by summing up thediscounted cumulative cash flows

(CCF [D ]) minus the capital investment required to establish both

biofuels andbiopower production facilities (FCC [D ]). Accordingly:

NPV chain = CCF − FCC (3)

while NPV car is calculated by summing up the potential saving

indriving EVsinsteadof bifuelones(RISP [D ])minus essentialextra

costs occurred to buy EVs (exCO [D ]):

NPV car = RISP − exCO (4)

The second objective function Objenv aims at minimising the

total GHG impact (TGHG [kg of CO2-eq]) which results from the

operationof thebioethanol and the biopower SC over the 15-years

time horizon.Accordingly:

Objenv = TGHG (5)

Similarly to what was proposed by Zamboni et al. (2009b), the

value of TGHG is estimatedby summing up the total impacts TI t [kg

ofCO2-eq/timeperiod],whichin thiscaseresultfromtheoperation

of theproductionchainand theAFVsutilisation foreachtimeperiod

t :

TGHG =

t

TI t (6)

3.2. Economics

Thefollowingsub-sectionswillmainlydiscussthedifferences in

themathematical formulation withrespect to Giarola et al. (2011),

concerning thecharacterisationof thebiopowerSC andof theAFVs

fleet.

3.2.1. Modelling the biopower SC

The terms CCF and FCC for the NPV chain calculationof Eq. (3) are

evaluated as in Giarola et al. (2011):

CCF =

t

CF t · dfCF t (7)

FCC =

t

TCI t · dfTCI t (8)

where CF t [D ] is the cashflow for eachtime period t , TCI t [D ] isthe

total capital investmentanddfCF t anddfTCI t arethetimedependent

discount factors. Both CF t and TCI t are discounted through the fac-

tors whichare collectedin thetwo differentarrays dfCF t and dfTCI t .

The futureinterest rate has been assumed tobe constantandequal

1 Since in this study we are comparing AFVs end user-related economics with

respect to bifuel vehicles in terms of costs per km driven (through the differential

variable NPV car ) there is no need to describe bifuel vehicles economics through a

specific formulation.


6/14


to 10% as resulting from the application of the CAPM (Capital Asset

PricingModel) rule.

The value of TCI t inEq. (8) is calculated as inGiarola et al. (2011)

by summing up the expenditures needed to establish the produc-

tionfacilitiesplannedat eachtimeperiod t according totheir capital

investment CI p,k (see Section 4):

TCI t =

p,k,g

plan p,k,g,t

· CI p,k (9)

where plan p,k,g,t is a continuousplanningvariablewhich isassigned

a non-zerovalue only forthe time period t inwhich the investment

decision occurs.

The value of CF t in Eq. (7) is given by the following relation:

CF t = PBT t + Dt − TAX t (10)

where Dt [D /time period] and TAX t [D /time period], which are

respectively the depreciation charge and the tax amount for each

time period t , are unchanged with respect to Giarola et al. (2011);PBT t [D /timeperiod] represents the profitbeforetaxesand iscalcu-

lated by summing up the business incomes (Inc t [D /time period])

minus the overall operatingcosts, both fixed (FixC t [D /timeperiod],

evaluated as in Giarola et al., 2011) and variable (VarC t [D /time

period]), andminus thedepreciationcharge foreach time period t :

PBT t = Inc t − VarC t − FixC t − Dt (11)

The business incomes for each time period t (Inc t ) come from

the sumof the total annual revenues earnedthroughthesale of the

product j (i.e. ethanol, electricityor DDGS)obtainedfroma conver-

sion facility of the technology k at the time period t . Accordingly:

Inc t =

j,k,g

P TOT j,k,g,t

· MP j (12)

where P TOT j,k,g,t [tonne/timeperiodorMWh/timeperiod] is thepro-

duction rate of theproduct j obtained from a conversion facility of

the technology k in the region g at the timeperiod t , andMP j is the

marketpriceof theproduct j [D /tonneorD /MWh].Theproduct j setincludesethanol,powerandDDGSwhoseamountsandproportions

dependonprocessing technologyk. Thedetailsontechnologiesand

their formal representation for the economicanalysiswill be given

in Section 4. Conversely, the modelling approach to describe the

electricity production is discussed in Section 3.2.3.

The term VarC t o f Eq. (11) is calculated by summing up the

main costs involved in the operation of the SC and, differently

fromGiarola et al. (2011), alsoaccounts forpower generation costs

through the variable PPC t [D /tee]. Accordingly:

VarC t = EPC t + BPC t + PPC t + TCbt + TCf t (13)

where EPC t [D /tonne] represents ethanol production costs, BPC t [D /tonne] represents biomass production costs, TCbt [D /tonne]

represents biomass transport costs and TCf t represents ethanoltransportcost [D /tonne].Powerproduction costs (PPC t ) aredefined

as a linear function of the total production rate of electricity from

powerplants(ELtot k,g,t [MWh]) anda fixed quotadependingon the

production technology k adopted:

PPC t =

k,g

coef k,"slope" · ELtot k,g,t + coef k,"intercept " · Y k,g,t

∀k ∈ elec k (14)

Further details about electricity production ELtot k,g,t can be

found in Section 3.2.3. In Eq. (14), coef k,“slope [D /MWh] and

coef k,“intercept [D /time period] are the arrays of linear coefficients

specific for each technology k, and Y k,g,t is the binary variable

accounting for whether a facility is operatingwith the conversion

technology k in the region g at the time period t .

3.2.2. Modelling AFVs economics

AsregardsNPV car inEq. (4), thetermRISP isevaluatedas follows:

RISP =

t

RISP t · CFdfCARt (15)

where RISP t [D /time period] represents the potential savings byend users in driving EVs instead of bifuel cars for each time period

t , which is discounted through the CFdfCARt factors for each time

period t . RISP t is evaluated by multiplying the global average

distance covered by EVs ( powerKM t [km/time period]) for the dif-

ferential travelling cost with respect to a bifuel vehicle (KMcost

[D /km]):

RISP t = powerKM t · KMcost (16)

Thevalue of KMcost is assumed tobe0.03D fortravelling1 km

(Renault, 2015; Van Vliet et al., 2010; Kay et al., 2013), consider-

ing an average consumers electricity market price of 170 D /MWh

(Autorità per l’energia elettrica e il gas, 2015). On the other hand,

thediscount factorCFdfCARt is calculated as follows:

CFdfCARt =1

(1+ i)t (17)

where i represents the timeperiod interest rate (for 3 years) and is

evaluated from the yearly interest rate i0, which is set equal to 5%

as results from the application of the CAPM rule:

i = (1+ i0)3− 1 (18)

With concern to theNPV car calculation of Eq. (4), the term exCO

is calculated by summing up the time dependent variable exCOt [D /timeperiod] foreach timeperiodt , discountedthroughthesame

CFdfCARt factors utilised for Eq. (15):

exCO =

t

exCOt · CFdfCARt (19)

where exCOt represents the additional investment for end user to

buy an EVwith respect to a bifuel one:

exCOt = newCARSt · (charg + inc ·t ) (20)

In Eq. (20), newCARS t [new EVs/time period] represents the

globalamountofnewEVspurchasedforeachtimeperiod t , depend-

ingon the electricitymarket demandevolution. Theconstantcharg

(set equal to 59 D /new EV according to Peterson and Michalek,

2013) represents the average cost of a domestic electric charger,

while the constant inc [D /new EV] evaluates the differential pur-

chasing cost of an EVwith respect to a bifuel one. This parameter

mostlydependsonhigh production costsof batteries forelectricity

storage, set nowadays just below 1000 D /kWh by Van Vliet et al.

(2010, 2011). This result is quite far from the U.S. Advanced Bat-teryConsortium longterm commercialisationcost, set equalto 150

$/kWh (USABC, 2007). By taking into account also the Renault EVs

catalogue (2015) and Perujo and Ciuffo (2010), constant inc is set

equal to5000D pernewEVpurchased bythe enduser.The conver-

sion of kWh of energy into number of EVs is operated through the

constant , which will be described in the following sub-sections.Theconstantinc isdecreasedforeach timeperiodt , accordingto the

valueof thet parameter. Heinet al. (2012) suggesta 74%decrease

forbatteriesproductioncostby 2030(from12,000D /kWhat t =1to

3000 D /kWh at t =5). An evenmore optimistic forecast is given by

Weiss et al. (2012), where an identical purchasing cost is foreseen

for EVs and bifuel cars in the Europeanmarket in 2030. Averaging

out those results, the t value was set equal to 0.125, in order to

express a lineardecreaseof EVsdifferential purchasingcost at2030


7/14


until 12.5%of thecurrent value. Thenumber of newEVspurchased

(newCARS t ) for each time period t , is calculated as follow:

newCARSt = (nCARSt − nCARSt −1)+ renewCARSt =4 + renewCARSt =5

(21)

where renewCARS t represents the substitution of obsolete EVs for

new ones. Battery lifetime is supposed to be 10 years (Hein et al.,

2012).Onthe otherhand,thevariablenCARS t representsthe cumu-lative amount of EVs for each time period t , depending on market

dynamics.

3.2.3. Modelling the economics of electricity generation

Details about biomass cultivation (bg ), biomass pre-treatment

(bpt ), biomass transport (bt ) and ethanol production ( fp) can be

found in Giarola et al. (2011). In this work, however, biomass

can be directly converted into electricity rather than ethanol. The

subset elec k of pure power generation is described inside the facil-

ity typology set k. Similarly to what was proposed for ethanol in

Giarola et al. (2011), power production ELtot k,g,t [MWh/month] by

facility k in region g at time t is set by means of the following

inequality:

ELtot k,g,t ≤

p

PR p · p,k,g,t ∀k ∈ elec k (22)

where PR p [MWh/month] represents output rates according to the

plant scale p and p,k,g,t is a binary variable representing the exist-

ence of the facility k of the scale p in the region g at the time t . Eq.

(22) definedas an inequality allowsfor partial loadworkingplants.

Facilities output variability from t to t +1 is limited by:

ELtot k,g,t ≥ 0.8 ·

p

PR p · p,k,g,t ∀k ∈ elec k (23)

In other words, a facilities output maximum variability from

t to t +1 is set equal to 20% (for power generation), according tothe coefficient 0.8 imposed in Eq. (23). Biomass feedstock2 is eval-

uated through the variable CapElec i,k,g,t [tonbiomass/month], which

represents the biomass i input to facility k in region g at time t .

Accordingly, the power generation Ppi,k,g,t [MWh/month] through

biomass i by plant k in region g at time t is set as follows:

Ppi,k,g,t = z i,k · CapElec i,k,g,t (24)

where z i,k [MWh/tonbiomass] evaluatesthe biomass i conversioninto

electricity by the power plants elec k. By using Ppi,k,g,t , it is possible

to evaluate variable ELtot k,g,t used in Eqs. (22) and (23) through

parameterbetaei,k, describing thebiomass i fractioninput to facilityk:

Ppi,k,g,t = ELtot k,g,t · betaei,k (25)

Power production technologies k, assumptions and related

parameters can be found in Section 4 and in the Supplementary

data.

It is important to note that some biorefinery technologies for

ethanol production also result in electricity generation through

the exploitationof DDGS for CHPproduction (see Section 4). Thus,

the overall power generation Ptot power k,g,t from a generic plant k in

2 Inthe presentstudy, thestoveris theonly suitable feedstockforpurepowergen-

eration technologies. Nevertheless, a general feedstock-dependent formulation is

herepresented, in order to permit the future implementationof additionalbiomass

typologies.

region g at time t is evaluated by summing up the contribution of

the biorefineries and of the power plants. Accordingly:

Ptot power k,g,t

= ωk ·Ptot ethanol

k,g,t

+ ELtot k,g,t (26)

where ELtot k,g,t represents, as described in Eq. (22), power plants

electricity output, while the biorefineries contribution to the

power generation is taken into account through the parameter

ωk [kWh/lethanol], representing the electric energy in excess withrespect to ethanol production, according to conversion technology

k. Constant [tonne/l] stands for the ethanol density.

3.2.4. Demand evolution

In this study, market mathematical formulation is based on

two main assumptions: (i) the total transport energy demand is

assumed to be constant during the period of interest; and (ii) the

renewabletransport energy isassumedtogrowaswillbe described

in Section 4. Assuming to not store production, it is imposed that

the global ethanol (TPetht [tonethanol/month]) and power (TPpowt [teepower/month]) production rates match exactly the ethanol

(TDetht [tonethanol/month]) and power (TDpowt [teepower/month])

demands:

TPetht = TDetht (27)

TPpowt = TDpowt (28)

Productions TPetht and TPpowt are given by:

TPetht =

k,g

Ptot ethanolk,g,t (29)

TPpowt =

k,g

Ptot power k,g,t

· (30)

where constant in Eq. (30) allows the conversion of electricity

production Ptot power k,g,t from MWh to ethanol equivalent tonnes

(tee)so astoallowsummingupthecontributionsofdifferentgoods.Withconcernstoelectricity,an immediate andregion-independent

distribution is assumed. Therefore, there is no need for describing

the regional biopower demand, according to its spatially implicit

usability.

AFVs penetration in traditional carfleet depends on bioethanol

and biopower production evolutions. Bifuel vehicles bifuelCARS t (i.e. fuelled by an ethanol and petrol blending) market penetration

is calculated by summing up traditional vehicles fleet ( gasolTOT t [traditional vehicles/time period]) minus EVs (nCARS t [EVs/time

period]) for each time period t :

bifuelCARSt = gasolTOT t − nCARSt (31)

Assuming an average trip distance of 45km per day per vehicle

(ISFORT,2015; JRC,2015), it ispossibleto evaluatethe totaldistancecovered by bifuel vehicles (bifuelKM t [km/time period]) for each

time period t :

bifuelKM t = bifuelCARSt · 45 (32)

On the other hand, the number of EVs (nCARS t ) for each time

period t is calculated as follow:

nCARSt =TDpowt

(33)

where parameter =1.897 MWh/EV/year represents the elec-tric energy required to fuel an EV for 1 year. Its value was derived

fromPerujo andCiuffo (2010), describingtheEVspenetration inthe

province of Milan (Italy). Assuming again an average trip distance


8/14


of 45km per day per vehicle, it is possible to evaluate the total dis-

tance covered by EVs ( powerKM t [km/time period]) for each time

period t . Accordingly:

powerKM t = nCARSt · 45 (34)

3.3. Environmental impact

Thedefinition of TI t (Eq. (6)) is as follows:

TI t =

s

Impact s,t (35)

where Impact s,t [kg of CO2-eq/time period] is the GHG emission

rate resulting from the operation of each single stage s at time t .

TheGHGemission rate is generally definedas follows:

Impact s,t = f s · F s,t (36)

where the reference flow F s,t [units/time period], specific for each

lifecyclestage sandtime t ismultipliedbyaglobalemissionfactor f s[kg ofCO2-eq/unit],which representsthe carbondioxideemissions

equivalent at the stage s per unit of reference flow.

Detailsabout emissionsrelated tobiomassgrowth(bg ), biomass

pre-treatment (bpt ) andbioethanol production ( fp) can befoundinGiarola et al. (2011).

3.3.1. Transport system

The resulting GHG emission of each transport option depends

onboth the distance run by the specific means and the freight load

delivered. As a consequence, the emission factor represents the

total carbondioxide emission equivalent released by thetransport

unit l perkmdrivenandtonne carried. Theglobalwarming impact

related to both the biomass supply (bt ) and the ethanol distribu-

tion to the blending terminals ( fd) is mathematically defined as in

Giarola et al. (2011).

On the other hand,emissionrelated to the final biofuel distribu-

tion to the end users ( fdist ) is a new feature of this study.The liquid

fuel (i.e. ethanol plus petrol) is assumed to be delivered by truckto petrol stations. Therefore, the total amount of ethanol Etot g,t [tonne/time period] blended in terminal g at the time t is deliv-

eredto thefinal usersaccordingtoanaveragedistributiondiameter

̊g [km]. The resulting GHG emission is obtained by multiplying

the transportation emission factor for road transport ( ffdtruck [kg of

CO2-eq/time period]) and the ethanol distribution flow occurring

from each terminal g . Accordingly:

Impact fdist ,t = ffdtruck ·

g

̊g · Etot g,t (37)

Distribution diameters ̊g are evaluated for each blending ter-

minal g according to the following equation:

̊g =

g (LD g,g · Dterm g )

2 · N (38)

where LD g,g [km]represents thedistancebetweena terminal g and

agenericregion g ’,whileDterm g isabinaryvariablewhichidentifies

thepresenceofaterminalintheregion g considered.By multiplying

up LD g,g andDterm g , it ispossibletodefinethedistance betweenan

established terminal g and a generic demand region g ’ . In Eq. (38),

N represents thenumber of distribution regions g ’. By summing up

thedistances for eachdistribution region g’ anddividingby 2N , it is

possible to calculate the average distribution diameter ̊g for each

terminal g . By adding up the circles of diameter ̊g , the resulting

area is equal to 1.3 times the distribution surface (which turns up

to be identical to the average value of the regional tortuosity g,l,g

used in Zamboni et al., 2009a).

3.3.2. Impact in electricity generation

TheGHGemissionsresulting from the bioelectricity generation

(epow) are estimated according to the methodology proposed by

IPCC (2006, 2013), whose results were compared also with those

obtained by Corti (2004), Carpentieri et al. (2005) and Mann et al.

(1996). Considering an average conversion efficiency for each pro-

duction technology k, the emission factors for power production

fppi,k [kg of CO2-eq/MWh] were implemented in the model for-

mulation (the emission factors canbe found in theSupplementary

data). Those factors are multiplied by electricity outputs for the

calculationof this stage impact for each time period t :

Impact epow,t =

i,g,k

fppi,k · Ppi,k,g,t (39)

where Ppi,k,g,t is the power rate from the biomass i at the facility k

in the region g at the time t .

3.3.3. Impact related to EV battery production

Each new EV purchased by the end users (newCARS t [new

EVs/time period]) was selected as the functional unit for this stage

emission evaluation. Accordingly:

Impact ebat ,t = · newCARSt (40)

The final emission factor value [kg of CO2-eq/new EV] was

fixed equal to 3046.9kg of CO2-eq/new EV, by averaging out some

studies in the literature (Zackrisson et al., 2010; Faria et al., 2013;

Li et al., 2014; Majeau-Bettez et al., 2011; Notter et al., 2010). In

this study, is thedifferential emission factor between traditional

petrol andEVsmanufacturing,thusboththebatteryproductionand

the lack of the internal combustion engine are taken into account

through this parameter.

3.3.4. Impact on AFVs usage

With concern to EVs, it is assumed that there are no emissions

(apart frombatteriesproduction stage ebat ). Therefore,EVsutilisa-

tion impact (ecars) is null for each timeperiod t :

Impact ecars,t = 0 (41)

On the other hand, referring to bifuel vehicles-related emis-

sions (ebifuel), a proper study is needed. Just like power plants

emission stage formulation (epow), bifuel vehicles emission factor

has been estimated following the methodology proposed by IPCC

(2006, 2013), under the following assumptions:

• vehicle running with hot engine;• exclusion of “low emission vehicles” (IPCC, 2006);• average vehicle in the USA car fleet.

The resulting emission factor [kg of CO2-eq/kmbifuel car] wasthereforeset equalto 0.005515andthe functionalunitfor theemis-

sion stage calculationwas set equal to 1km (distance travelled by

bifuel vehicles). The GHG emission is obtained bymultiplying the

bifuel vehicle emission factor with the total distance travelledbifuelKM t for each time period t . Accordingly:

Impact ebifuel,t = ς · bifuelKM t (42)

3.3.5. iLUC impact

iLUC occurs when pressure on agriculture due to the displace-

ment of previous activity or use of the biomass induces land-use

changes on other lands, in order to maintain the previous level

of production. Therefore, it has consequences in the GHG balance

of the proposed production systems. According to the European

Commission (2009, 2012), all the facilities (i.e. biofuels plants) that

use land will get an iLUC factor if [kg of CO2-eq/GJethanol]; how-

ever, feedstock that does not require any land for its production


9/14


(i.e. waste, residues, algae) is exempt from iLUC effects. Thus, in

this studybioelectricity production does not produce iLUC-related

emissions, because the conversion facilities input is considered a

residual feedstock (corn stover). Theestimated iLUC emissionsare

introduced as follows into the modelling framework:

Impact iLUC ,t =

k,g

if · Pf k,g,t (43)

where the iLUC factor for the feedstock group (if ), in the case of this work the corn utilised as a biomass to produce ethanol, corre-

sponds to a value of 12kg of CO2-eq/GJethanol produced (European

Commission, 2009, 2012).

3.3.6. Emission credits

Following Zamboni et al. (2009b), the substitution procedure

has beenchosenin order todealwith theeffect ofbyproductalloca-

tionon emissiondiscount.Accordingto thisapproach, theemission

credits earnedby thedisplacementof alternativegoods alongwith

byproducts are subtracted from the primary product total GHG

emissions.

4. Case study

Northern Italy is chosen as a demonstrative case study. The

bioenergy production system is optimised over a 15-year horizon

by setting the transport energy demand so as to find the best mix

of power and ethanol (and accordingly of bifuel and EVs) meet-

ing the economic and environmental objectives. Market prices for

ethanol, DDGSandpowerwere fixed equal to710D /tonethanol, 300

D /tonDDGS and90D /MWh(greencreditswerenot considered).Fur-

ther details can be found in the following subsections and in the

Supplementarydata.

4.1. Spatially explicit features

Northern Italy was discretised according to the grid approachdescribedbyZamboni et al.(2009a), consisting of 59homogeneous

squares of equal size (50kmof length). One additional cell ( g =60)

isused to allow for importingbiomass from foreign suppliers. As in

Zamboni et al. (2009a), bioethanol is sent to the blendingterminals

existing at given locations.

Two main instances were formulated: in Instance A both

bioethanol and bioelectricity demand variations are set, whereas

in Instance B only the global demand is set, allowing the solver to

reach the optimum quota agreement.

4.1.1. Instance A

In this design configuration, both bioethanol and bioelectricity

demands are assumed to be fixed a priori so that their respective

productions,whichare increasing along the time horizon, arepre-set for each time period t .

On the one hand, dealing with EVs penetration, Shepherd et al.

(2012) suggest a market share of 3% by t =5 (2030), which is sim-

ilar to that indicated in a forecast by the Boston Consulting Group

(2009) (3.015%) and ina study bytheBERR(2008) (3.5%).Consider-

ing anaverage value of those results, itwas hereassumed amarket

share for EVs of 3.26% for the entire car fleet by2030. Thus, a linear

growth-rate (starting from zero) was implemented in the study to

describe the bioelectricitymarket evolution. The actual number of

EVswasconvertedinto electricity demandaccording to theparam-

eter [MWh/EV/year] described in the previous Section 3. Thecirculating traditional car fleet in the Northern Italy was assumed

to be constant along the time horizon, consistently to the statistics

of the last years (Automobile Club d’Italia, 2015).

Table 1

Ethanol (TDetht ) and power (TDpowt ) demands, ethanol blending (etperc t ) and EVs

market share (EVmt ) for each time period t (Instance A) and global demand TDt (InstanceB).

t Instance A Instance B

TDetht TDpowt etperc t EVmt TDt [kton/year] [ktee/year] [%vol] [%] [kton/year]

1 557 36 10.20 0.65 593

2 659 73 12.10 1.30 732

3 761 109 14.00 1.95 870

4 857 146 15.80 2.61 1003

5 953 182 17.60 3.26 1135

On the other hand, dealing with the bifuel vehicles penetration

(i.e. petrolplus bioethanol), theirmarketsharedecreasesalong the

15-years time horizon (because of the progressive substitution of

the traditional vehicles with EVs). However, the ethanol blending

(represented by the etperc t parameter) grows during the 5 time

periods, assuming as mandatory the future EU targets on biofuels

already described in Giarola et al. (2011).

The consequence of this design configuration is a forced drop

of petrol utilisation, which is due to: (i) the progressive substitu-

tion of the traditional vehicles with the EVs, and (ii) the increasing

ethanol blending in the bifuelvehicles. Theglobal ethanol demand(TDetht [tonne/month]) and the global electricitydemand (TDpowt [tee/month]),as well as the ethanolblending(etperc t [%

vol]) and the

EVs market share (EVmt [%]) evolutions, are reported in Table 1 f or

each time period t .

4.1.2. Instance B

In this design configuration, the global demand (TDt [tonne/month]) is set, allowing the solver to find the optimal

ethanol/electricity quotas. The global demand TDt is evaluated

by summing up the ethanol demand TDetht and the electricity

demand TDpowt as they resulted in InstanceA for each timeperiodt . The global demand TDt is reported in the last columnof Table 1.

Clearly in this case the AFVs market evolution is a result of the

optimisation.

4.2. Biomass growth, biomass pre-treatment and transport

systems

Withconcerns tocorncultivation, thespatiallyspecificdatasets

(i.e. BCDmax g , AD g , BY i,g and UPC i,g ) were taken fromZamboni et al.

(2009a),whilethe stoveryields (BY “stover , g ) and costs (UPC “stover ,g ),

aswell as thedatatoestimate the impacton globalwarming( fbg i,g )

of biomass production, were derived from Giarola et al. (2011).

Biomass pre-treatment dealswith thedryingand storage oper-

ations after the biomass harvesting and collection. Costs are not

considered because already included in the biomass production

costs, whereas the environmental impact deriving this stage is

taken from Zamboni et al. (2009a).The distribution infrastructure includes trucks, rail, barges and

ships as possible deliverymeans.Trans-shippingwas included as a

viable transport option for biomass importation.All the transport-

relatedparametershavebeenbasedon actualgeographicdistances

between regions g and g’ according to the procedure described in

Zamboni et al. (2009a).

4.3. Bioethanol production

Following the results of Giarola et al. (2011), three main

processing technologies were identified: (i) theDry Grind Process

(DGP), where corn is converted into ethanol through a biologi-

cal process; (ii) the Ligno-Cellulosic Ethanol Process (LCEP), where

stover only is converted into ethanol; and (iii) the Integrated


10/14


Table 2

Technological option for ethanol production, identification and products

description.

k Process Input Output

DGP LCEP IGSP Grain Stover Ethanol CHP DDGS

1 X X X X

2 X X X X

3 X X X X X

4 X X X X X

Grain-Stover Process (IGSP), where both corn grain and stover are

processed to obtain ethanol. Table 2 summarises the main fea-

tures of the four technological options, in particular the feasible

input/output combinations are pointed out. Among IGSP possibili-

ties,twodifferenttechnologies (k=3,4)areimplementedaccording

to different biomasses mixtures in terms of corn grain and stover

as described in Giarola et al. (2011).

4.4. Bioelectricity generation

In order to analyse the potential technological options for

the conversion of biomass into electricity, it was decided to

exclude those technologies that currently do no exhibit eithera proven maturity or a sound economic viability (e.g. Stirling

engines, Organic Rankine Cycles, Fuel cells) (Lora and Andrade,

2009). Accordingly, three technological options were considered

in this study: (i) biomass direct combustion for Rankine steam

cycle (C+R, k=11); (ii) biomass gasification for Turbo Gas cycle

(G+TG, k=22); and (iii) biomass gasification for Internal Combus-

tion Engine (G+MCI, k=33). It was assumed to consider only pure

power generation, excluding from the calculations the combined

heat and power potential production. Thus, the electricity genera-

tion (PR) was assumed to be proportional to the biomass (i.e. corn

stover only) input (Capelec ), according to the following equation:

Capelec =PR · f c

k · LHV fuel(44)

where f c represents the load factor (assumed to be of 8000h/year

independently from the technology k), while k is the conversionefficiency foreachtechnology k and LHV fuel is a constant represent-

ing the lower heating value of the feedstock introduced (which for

corn stoveris 15.9MJ/kg, accordingto the Jenkinscatalogue,1998).

Electricity generationratesPR p are discretised according to the set

p (Table 3). The capital investment CI and the production costs

PC of the technologies elec k are evaluated through the following

equations:

CI = a · PR−b (45)

PC = c · PR−d (46)

where thecoefficients a, b, c and d (Table 4) dependon theconver-

sion technology k.With concern to thebiomass direct combustion within a Rank-

inecycle, their capital investment and theproductioncosts, aswell

as the conversion efficiency, were evaluated from the literature

Table 3

Production capacity, nominal values foreach plant size p, ER p and PR p.

p ER p PR p[kton/year] [MW]

1 96 1

2 110 5

3 150 10

4 200 15

5 250 30

6 276 60

Table 4

Facilities coefficients a, b, c and d for thecalculation of costs,CI and PC.

k a b c d

11 137,720 0.407 0.263 0.280

22 76,007 0.433 0.202 0.240

33 26,414 0.377 4.768 0.730

(Caputo et al., 2005; Patel et al., 2011; Weiss et al., 2012; Fulmer,1991). According to theliterature, theplant scale for theconversion

technologywas limited between 3MWand 60MW.

For the gasificationplants, both the technical and the economic

parameters were also derived from the literature (Arena et al.,

2010; Kinoshita et al., 1997; Craig and Mann, 1996; Wu et al.,

2002, 2008). The results were also compared with the Biomass

CHP catalogue (2007). According to the literature, the plant scale

for the Turbo Gas cycles was limited between 100kWand 60MW,

while for the Internal Combustion Engines the possible range was

set between 100kW and 10MW (the latter is the largest existing

size for stationary generation according to General Electric, 2015).

Similarly to the mathematical formulation adopted for biore-

fineries, linear equations were obtained by the regression of the

capital investment CI p,k (Table 5) and the production costs PR p,k(Table 6) values, related to several capacity intervals p per each

technology k.

The biomass conversion factors (i.e. corn stover converted

into electricity which is then distributed to end user), z i,k[MWhdist/tonstover], are set equal to 1.0890 for k=11, 1.6326 fork=22 and 1.5306 for k=33. According to statistics made by Terna

(2014), the average efficiency of the Italian National grid is equal

to 0.935 (defined as the electricity distributed to end users with

respect to the electricity produced at the facility). The parameter z i,k wasevaluatedbymultiplyingthatvalue (0.935)andtheaverage

conversion efficiencyk for each technology k.With regard to the environmental aspects, the GHG emission

from the power production stage was assumed to be proportional

to the total annual amount of biomass i processed by the facility k,accordingto themethodologyproposed by IPCC (2006). The global

emission factors assigned to elec k stover-basedprocesses (i.e. fppi,k[kg ofCO2-eq/MWh]), setequal to1.97 fork=11, 1.47for k=22 and

1.53 for k=33, were calculated through a spreadsheet tool (IPCC,

2013) accounting only for CH4 and N2O contributions.

Table 5

Capital Investment, valuesof the linearisation parametersCI p,k [MD ].

p k 1 2 3 4 11 22 33

1 62 396 187 81 – 4 2

2 70 434 204 90 22 10 5

3 91 535 252 117 32 14 8

4 115 648 305 149 49 21 –5 139 753 354 179 62 26 –

6 151 804 379 195 94 39 –

Table 6

ProductionCosts, valuesof the linearisation parameters c k,c .

k slope intercept

[D /tonne or D /MWh] [D /month]

1 140.83 169,906

2 202.88 891,755

3 143.36 507,404

4 17.746 388,525

11 10.71 64,377

22 13.71 45,894

33 2.91 19,805


11/14


B2

B3

B1

iLUC

iLUC

A2

A3

A1

iLUC

iLUC

-8

-6

-4

-2

0

2

4

6

8

0 10 20 30 40 50 60 70 80 90 100 110 120

N P V - N e t P r e s e n t

V a l u e [ € / G J

o u t p u t ]

TGHG -

Total GHG Emission [kg of CO2-eq/GJoutput]

Instance B Instance A

Fig. 2. Paretocurves under bi-objectiveoptimisation forInstanceA andInstanceB.

5. Results and discussion

Thedesign variables (about 300k continuous variables and 15kdiscrete ones per iteration) were optimised in about 60hours by

using the CPLEX solver in the GAMS® modelling tool on a 3.40GHz

processor.

As expected, the Pareto set of sub-optimal solutions (see Fig. 2)

resulting from the bi-objective problem solution reveals the con-

flict between environmental and economic performance. First of

all, theresultswill bediscussedwithoutconsidering theiLUCeffect

on environmental performance. In Instance A, the optimal config-

uration in terms of economic performance (case A1 as reported in

Fig. 2) entails a marginal NPV of 2.08 D /GJoutput against a global

environmental impactof 94.8kg ofCO2-eq/GJoutput, which infactis

equivalent to a GHG increase of about 10% compared topetrol (the

GHGemissionfactorforpetrolwasassumedequalto85.8kgofCO2-

eq/GJ, according to HGCA, 2005). In such a configuration theusageof biomass (at least, the biomass considered in this study) does not

lead to any environmental advantage. The environmental impact

is almost identical in Instance B, where the optimal configuration

in terms of economic performance (case B1 as reported in Fig. 2)

entails a marginalNPV of 5.55D /GJoutput (significantly higher than

in InstanceA)anda globalenvironmental impact of 95.2kgof CO2-

eq/GJoutput, equivalent toa GHG increaseof about 11% compared to

petrol. In both Instance A and InstanceB, the systemdesignwould

involve the establishment of standardDGPbiorefineries (k=1) and

a significant corn importation fromabroad.With regardto theelec-

tricity production, G+TG (k=22) and G+MCI (k=33) facilities are

establishedandthefeedstockfor theelectricity conversion (stover)

is completely produced in Northern Italy.

Fig. 3a illustrates the optimal economic configuration in thecase of Instance B. Fig. 4a shows the contributions of NPV chain and

NPV car to the global NPV under the economic optimisation. With

regards to NPV chain, both instances perform similarly (440MD vs.

407).The good SC economicperformance ismainlyrelated to tech-

nology choicesandbiomasssupply costs.On theotherhand,NPV car is quite different in the two cases. In InstanceA it isworth 198MD ,

while it reaches 1294MD inInstance B. This is due tothe larger EV

penetration in Instance B (about 12% by t =5). In fact, that design

configuration can produce a NPV car of about 1.3GD in 15 years,

against of about 200MD in Instance A. Thus, it would seem that

EVs would acquire a higher than expected market share in an

unconstrained condition. However, Fig. 5 suggests a more com-

plexsituation. AlthoughthefinalNPV is largely positive, InstanceB

requires that the final users accept a period of negative economic

performance,which isneeded topaybackthe investment forestab-

lishing the production facilities. In other words, due to the initial

high EV cost, initially EV buyers would not get any return from

their investment (note from Fig. 5 that the initial payback time is

about 10years,whichis also the average battery lifetime)and sim-

plywould be instrumental to set inmotion the production SC. This

is unlikely to occur unless some incentives are promoted through

dedicatedpolicies. Inview of the above, Instance A, despite a lower

value of NPV car , exhibits a payback time of 7 years and probablyrepresents a more sensible designoption.

The best configurations in terms of global warming mitigation

potential are achieved through the establishment of different con-

version technologies for ethanol production. In Instance A, the

optimalconfigurationin termsof environmentalperformance(case

A2 as reported in Fig. 2) entails a global environmental impact

of 34.6kg of CO2-eq/GJoutput corresponding to a GHG decrease

of about 60% compared to petrol (compliant with long terms EU

environmental targets). The environmental optimum (case B2 as

reported in Fig. 2) is even better in Instance B, in which the global

environmental impact is of 16.0kg of CO2-eq/GJoutput, correspond-

ing to a GHG decrease of about 81% compared to petrol. However,

theoperationof suchasystemwouldbefeasibleonlyundera strong

support policy (governmental subsidy should account for about

3.85 D /GJoutput in Instance A, corresponding to about 1.2GD par-

celled out over the 15 years horizon, and for about 5.92 D /GJoutputin Instance B, corresponding to about 1.8GD over 15 years).

In Instance A, the strategic design involves the establishment

of the expensive LCEP technology for the ethanol production

(k=2) and, similarly to the economic optimum, of gasification

technologies (k=22, 33) for electricity generation. On the other

hand, in Instance B the SC configuration results do not suggest

the establishment of pure power generation facilities to reach

the environmental goals (see Fig. 3b). It appears that adopting

LCEPtechnology toproducebothethanolandelectricity represents

the best environmental option. Fig. 4b shows the contributions

of NPV chain and NPV car to the global NPV under the environ-

mental optimisation. In Instance A NPV chain is worth −1377MD

against −1799MD in Instance B, whereas in Instance A the NPV car is worth 198MD against −15MD in Instance B. The Instance

B negative value of NPV car is related to the fact that second

generation ethanol technologies are preferred and the payback

time is not reached within the time horizon considered in this

study.

An intermediate situation is representedby thecasesA3and B3

reportedinFig. 2,whichrepresentthe thresholdbetweenprofitable

andunprofitableproduction SC (NPV chain≈0).RegardingInstanceA,

theoptimisationentailsamarginalNPV of0.50D /GJoutput (thanksto

the contribution of the AFVs penetration in the market) against a

global environmental impact of 40.6kg of CO2-eq/GJoutput equiv-

alent to a GHG decrease of about 53% compared to petrol. The

economic result is better in Instance B,with amarginalNPV of 2.93

D /GJoutput (again given only by NPV car ) against an environmentalperformance of 41.9kg of CO2-eq/GJoutput corresponding to a GHG

decrease of about 51% compared to petrol. In both Instance A and

Instance B, those results are achieved through the installation of

IGSP biorefineries (k=3, 4), which could represent an interesting

halfway technology between the high economic performance of

DGP process and the low environmental impact of LCEP process.

Again, the power generation depends on gasification technologies

(k=22, 33) in both Instance A and Instance B.

Theeffectsof iLUC aremerely environmental,as showninFig.2.

As a consequence, in cases A1 and B1 the environmental results

are worsened significantly after the incorporation of iLUC effects:

theoverall GHGemissions increase in Instance A to about 106.8kg

of CO2-eq/GJoutput (+25% with respect to petrol) and in Instance B

to about 103.5kg of CO2-eq/GJoutput (+21% with respect to petrol).


12/14


Fig. 3. SC configuration at the end of timehorizon (t =5) in Instance B under (a)economic and (b) environmental optimisation.

There isno iLUCeffect in cases A2and B2 since in bothcases stover

is theonly feedstock.

5.1. Taxation effects

In our analysis, NPV car was calculated considering the actual

pricea consumerwouldpayforfuelandelectricityin Italy. Toassess

the effect of taxation on fuel and electricity, duties were excluded

in the calculation of the parameter KMcost . In other words, the

market prices are now assumed to be equal to the selling prices

at the facility plus the distributioncosts. Theprice for electricity is

set equal to 110 D /MWh and the price for the bioethanol is 0.709

D /l. The latter was calculated by summing up an average market

price for the ethanol (0.56 D /l according to Giarola et al., 2011)

and by assuming an average distribution cost, evaluated as for the

petrol in 0.15D

/l (Unione Petrolifera, 2015). As a consequence,


13/14


-2000

-1500

-1000

-500

0

500

1000

1500

2000

NPVchain NPVcar NPV

N e

t P r e s e n t V a l u e [ M € ]

(a) Instance A Instance B

-2000

-1500

-1000

-500

0

500

1000

1500

2000

NPVchain NPVcar NPV

N e t P r e s e n t V a l u e [ M € ]

(b) Instance A Instance B

Fig. 4. NPV chain, NPV car and NPV [MD ] under: (a) economic or (b) environmental

optimisation.

parameter KMcost is now worth 0.05 D /km and the effect isto make EVs more advantageous with respect to bifuel ones. In

Instance A the NPV increases up to 6.00 D /GJoutput (+191%) while

in Instance B it becomes 35.35 D /GJoutput (+543%). Clearly the

elimination of taxation would have positive consequences on the

economics of EVs and it shows how current Italian taxation donot

favour EVspenetration in themarket.

5.2. A sensitivity analysis on price decrease in EVs

Thepotential savings fora final customermainly dependon the

dynamicsconcerning thedifferential purchasing costbetween EVs

and bifuel cars over the timehorizon. Thus, for InstanceB weveri-

fiedtheeffect ontheeconomicoptimumby consideringdifferential

pricevariationdynamics,assuming thefinal differentialpurchasing

costby2030 froma null toa maximumvalue of2500D /newEV.On

the one hand, the results show that the differential cost of EVs has

hardly any consequence in the NPV chain (i.e. theSC configuration is

notaffectedby thisparameter).Ontheother hand,notsurprisingly,

the consequences onNPV car are remarkable (see Fig. 6).

-1500

-1000

-500

0

500

1000

1500

2000

2018

2021

2024

2027

2030

N e t P r e s e n t V a l u e f o r E V s - N P V c a r

[ M € ]

Years

Instance A_NPVcar Instance B_NPVcar

Fig. 5. Actualisation of NPV car through the years [MD ] for Instance A and Instance

B.

-1000

0

1000

2000

3000

0 500 1000 1500 2000 2500 N e t P r e s e n t

V a l u e - N P V [ M € ]

EVs differential cost by 2030

Instance A_NPVchain Instance A_NPVcar

Instance B_NPVchain Instance B_NPVcar

Fig. 6. EVs differential purchasing cost by 2030: economic consequences.

In Instance A, there is no convenience for the end users in pur-

chasing an EV if the differential purchasing cost is still of 1200

D /new EV by t =5 (2030). Nevertheless, this value seems quite

pessimistic with respect to the one that was implemented in the

formulation according to the literature (625 D /EV by t =5). In

Instance B, the initially not affordable economic situation for end

users, due to a high differential purchasing cost, generates a drop

in the EVs penetration during the first years.

6. Final remarks

A multi-objective MILP modelling framework for the strate-

gic optimisation ofmulti-echelonbioethanol andbiopower supply

chains, intended to support the alternative fuel vehicles penetra-

tionin theNorthern Italymarket, hasbeenpresentedanddiscussed.

All simulation studies show that both bifuel and electric vehicles

are needed for complying with themarket demand.

Froman economicstandpointof thesupply chain infrastructure,

the results show that a correct mix of first generation biorefineries

for ethanol production and of gasification facilities for electricity

generation represents a viable industrial optionandpermits a fea-

sible penetration of the alternative fuel vehicles in the traditional

market. Nevertheless, this network configuration does not repre-

sent a suitable answer tomatch the EU targets on global warming

mitigation, due to the high environmental impact resulting from

the handling of the first generation technologies, especially after

incorporating iLUC effects. On the other hand, second generation

biorefineries are still rather expensive and requiring some sort of

incentives. However, coupled with same gasification technologies

for power generation, they represent the best option for a signifi-

cant reductionof global GHGemissions.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,

in theonline version,athttp://dx.doi.org/10.1016/j.compchemeng.

2016.01.003.

References

ACI. Statistiche annuali delle immatricolazioni [Internet]; 2015, Available from:http://www.aci.it/.

ArenaU,DiGregorioF, SantonastasiM. A techno-economiccomparisonbetweentwodesign configurations for a small scale, biomass-to-energy gasification basedsystem. Chem Eng J 2010;162:580–90.

Autorità per l’energia elettrica e il gas. Condizioni economiche di fornitura per unafamiglia con 3kW di potenza impiegata e 2700 kWh di consumo annuo; 2015.

Bai Y,OuyangY, PangJS.Biofuel supplychain design undercompetitive agriculturalland use and feedstockmarket equilibrium. Energy Econ2012;34:1623–33.

Biomass CHP catalog USEP. Biomass combined heat andpower catalog of technolo-

gies;2007.

http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://www.aci.it/http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0025http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0020http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0015http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0010http://www.aci.it/http://www.aci.it/http://www.aci.it/http://www.aci.it/http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003http://dx.doi.org/10.1016/j.compchemeng.2016.01.003


14/14
http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0360http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbref0355http://refhub.elsevier.com/S0098-1354(16)00006-5/sbr

Documents

Strategic Optimisation of Biomass