Energy Fuels 2009, 23, 5134–5143 Published on Web 08/27/2009
: DOI:10.1021/ef9004779
Spatially Explicit Static Model for the Strategic Design of Future Bioethanol Production Systems. 2. Multi-Objective Environmental Optimization Andrea Zamboni,† Fabrizio Bezzo,† and Nilay Shah*,‡ †
Dipartimento di Principi e Impianti di Ingegneria Chimica (DIPIC), Universit a di Padova, via Marzolo 9, I-35131, Padova, Italy, and ‡Centre for Process Systems Engineering (CPSE), Imperial College London, South Kensington Campus, SW7 2AZ London, United Kingdom Received May 18, 2009. Revised Manuscript Received August 4, 2009
In developing an optimization framework to assist in the design process of biofuel systems, the economic effectiveness of the supply network should not be adopted as the sole criterion to focus on. In fact, there has recently been growing attention in including environmental concerns at the strategic level of supply chain management. In this part 2, the spatially explicit multi-echelon mixed integer linear program (MILP) modeling framework described in part 11 of this work has been extended by including environmental issues along with the traditional economic ones within a more comprehensive multi-objective optimization tool. The economics have been assessed by means of supply chain analysis techniques, focusing on biomass cultivation site locations, ethanol production capacity assignment and facilities location, as well as transport system optimization. The environmental performance of the system has been evaluated in terms of greenhouse gas (GHG) emissions, by adopting a well-to-tank (WTT) approach to consider the supply network operating impact on global warming over the entire life cycle. The strategic design tool as developed has been applied and solved in assessing the emerging corn-based Italian ethanol system. The resulting outcomes demonstrate the valuable support that the model may provide in formulating a welladvised strategic policy to promote the market penetration as well as to reduce the social and environmental impacts of biomass-based fuels.
performance (e.g., sales may be positively affected by a better public perception of the product, or operating cost reduction may be achieved by improving the process efficiency2). In view of above, there has been a paradigm shift to understand the design approach as well as the scope of the analysis of process systems. In fact, as pointed out by Cano-Ruiz and McRae3 in reviewing the state of the art of environmentally conscious design of chemical plants, a new approach that considers the environmental performance as a design objective rather than a design restriction may lead to the discovery of unexplored solutions that not only minimize ecological damage but also lead to overall economic profits. However, limiting the scope of the analysis to a company-centric view of the production system may result in misleading solutions, which may occur when a decrease of the local impact determines an increase in the overall ecological damage. These drawbacks have been overcome by including environmental responsibility principles within a more comprehensive approach analyzing the performance of a production system across the entire supply chain to generate a new branch of SCM, namely, the green supply chain management (GrSCM). In an extensive review4 of over 200 scientific contributions encompassing various research areas, Srivastava remarks the importance of a more extensive use of mathematical programming (MP) to contribute to a major advance in an environmentally conscious SCM.
Introduction 1
Part 1 of this paper addressed the development of a spatially explicit mixed integer linear program (MILP) for the strategic design of biofuel production systems. The integrated management of the whole supply chain (SC) including agricultural practice, biomass supply, fuel production, and logistics of transport was taken into account to develop a specific tool providing strategic decision support for supply chain management (SCM). Cost minimization was adopted as the optimization criterion to configure the system according to the traditional approach assuming the economic benefits to be the main drivers that motivate and promote the development of decision-making modeling tools for SCM. However, the economics of a system should not be the only issue to focus on in pursuing a detailed assessment of a multifaceted problem where conflicting aspects may concur in determining the optimal configuration of the network. In the past decade, ever growing attention has been directed toward the inclusion of pollution mitigation as part of the optimization criteria driving the design of process systems. This was a direct consequence of the policies that the process industry had to implement for complying with the ever more restricting environmental regulation imposed by governments. In addition, this trend has been recently accelerated by the growing awareness that improving the environmental sustainability of a production process may also improve its economic
(2) Azapagic, A.; Perdan, S. Indicators of sustainable development for industry: A general framework. Trans. IChemE 2000, 78, 243–261. (3) Cano-Ruiz, J.; McRae, G. Environmentally conscious chemical process design. Annu. Rev. Energy Environ. 1998, 23, 499–536. (4) Srivastava, S. K. Green supply-chain management: A state-of-theart literature review. Int. J. Manage. Rev. 2007, 9, 53–80.
*To whom correspondence should be addressed. Fax: þ44-(0)207594-6606. E-mail:
[email protected]. (1) Zamboni, A.; Shah, N.; Bezzo, F. Spatially explicit static model for the strategic design of future bioethanol production systems. 1. Cost minimization. Energy Fuels 2009, DOI: 10.1021/ef900456w. r 2009 American Chemical Society
5134
pubs.acs.org/EF
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
The main driver in using MP lies in the possibility of performing a simultaneous optimization of different issues,5 thus enabling the exploration of a balanced trade-off between conflicting objectives. This requires the incorporation of multiple criteria decision-making techniques within the modeling framework, namely, multi-objective mathematical programming (MoMP). According to Mavrotas et al.,6 the methods for solving MoMP problems can be classified into three categories according to the phase in which the decision maker is involved in the decision process: the “a priori”, the interactive, and the “a posteriori” methods. In both the “a priori” and the interactive methods, the objectives are weighted and grouped together in a single optimization criterion, thus imposing to the decision makers to express their preference before the solution procedure. This represents a limitation in the MoMP capabilities because the whole set of different solutions may not be analyzed and assessed. This drawback is overcome adopting the “a posteriori” method, which provides a full set of feasible (non-inferior or Pareto optimal) solutions, so that the decision makers can evaluate and decide the most viable alternative according to their needs. The application of MoMP within the specific field of GrSCM is further motivated by the approach that it adopts: the evaluation of the SC performance in terms of ecological damage covers all stages of the life cycle of the product, thus fitting with the needs of life cycle assessment (LCA) techniques, broadly acknowledged as the best methodology to rigorously quantify the environmental burdens and their potential impact of a process, product, or activity.7 However, including LCA techniques within a MoMP framework poses the problem to find the most appropriate approach to evaluate the ecological damage of the system. This issue asks for the adoption of appropriate indicators capable of measuring the environmental performance. To date, much effort has been directed toward the analysis of the problem2,7,8 to define a broadly accepted standard. However, especially in applying the LCA theory to the SC optimization, the best viable option is determined by analyzing the scope of the problem of interest as well as by a trade-off between a detailed and thorough comparison among alternative products (as required by LCA) and the minimization of the calculation effort (needed to carry out the SC optimization). One of the most controversial issues in alternative energy systems is indeed the assessment of the effective benefits coming from the use of biomass-based forms of energy and, particularly, fuels for automotive purposes. The use of biofuels, such as bioethanol and biodiesel, is broadly acknowledged as one of the most readily available options answering the global energy supply question:9 converting biomass into liquid fuels ensures some undeniable ecological benefits (i.e., reduction of CO and VOC emissions) in addition to other
strategic ones, such as enhancement of the internal supply security, support to the agricultural sector, and stimulus to the development of new technologies. However, concerns related to land competition10 with food production (and other biomass-based energy sectors, too) together with the modest capabilities in mitigating the global warming effect in the transport sector11 pose the issue of a more comprehensive approach to design, assess, and optimize the production SC. This is even more topical within the European Union (EU) scenario. In fact, according to the latest EU standards laid out during the European Council held in Brussels in December 2008, biofuels are required to have a minimum of 35% (percentage that should increase up to 50% in 2017) of greenhouse gas (GHG) emissions saving, at least with respect to the renewable 10% energetic quota defined as the 2020 target. MP-based design turns out as a necessary tool to optimize the use of constrained resources, such as land use, as well as the deployment of economic resources to support the novel biofuels sector to explore the best alternatives complying with the European standards. The objective of part 2 of this work is to deliver an environmentally conscious decision making tool for the strategic design level of biofuel systems. Specifically, it includes an environmental objective function within the spatially explicit modeling framework described in part 11. The problem is formulated as a MILP, which accounts for the minimization of the overall SC operating costs as well as for the minimization of the global environmental impact in terms of GHG emissions. The identification of the set of non-inferior or Pareto optimal solutions of the resulting multi-objective MILP is achieved through the implementation of a multiparametric “a posteriori” method12 solved via a dedicated algorithm.13 The emerging Italian corn-based ethanol production system is chosen as a case study to demonstrate the model outcomes. In particular, northern Italy is considered as a geographical benchmark, and the dry grind process was assumed as the standard technology for ethanol production. The suitability of the modeling framework for supporting strategic decisions is demonstrated through a real-world case study regarding the bioethanol supply chain. As in part 11, the economics are assessed by means of supply chain analysis (SCA) techniques, focusing on biomass cultivation site locations, ethanol production capacity assignment and facilities location, as well as transport system optimization. The environmental performance of the system is evaluated in terms of GHG emissions, by adopting a well-to-tank (WTT)14 approach to consider the supply network operating impact on global warming over the entire life cycle. This paper is organized as follows. A general description of the LCA application to biofuel systems is first presented. ~ez, E. E.; (10) Escobar, J. C.; Lora, E. S.; Venturini, O. J.; Yan Castillo, E. F.; Almazan, O. Biofuels: Environment, technology and food security. Renewable Sustainable Energy Rev. 2009, 13, 1275– 1287. (11) Dale, B. Biofuels: Thinking clearly about the issues. J. Agric. Food Chem. 2008, 56, 3885–3891. (12) Papalexandri, K.; Dimkou, T. A parametric mixed-integer optimization algorithm for multi-objective engineering problems involving discrete decisions. Ind. Eng. Chem. Res. 1998, 37, 1866–1882. (13) Dua, V; Pistikopoulos, E. N. An algorithm for the solution of multi-parametric mixed integer linear programming problems. Ann. Oper. Res. 2000, 99, 123–139. (14) CONCAWE, EUCAR, and JRC. Well-to-Wheels Analysis of Future Automotive Fuels and Powertrains in the European Context: WELL-TO-TANK Report. Version 2c. European Commission: Brussels, Belgium, 2007.
(5) Guillen-Gos albez, G.; Grossmann, I. E. Optimal design and planning of sustainable chemical supply chains under uncertainty. AIChE J. 2009, 55, 99–121. (6) Mavrotas, G.; Diakoulaki, D.; Florios, K.; Georgiou, P. A mathematical programming framework for energy planning in services’ sector buildings under uncertainty in load demand: The case of a hospital in Athens. Energy Policy 2008, 36, 2415–2429. (7) Azapagic, A. Life cycle assessment and its application to process selection, design and optimisation. Chem. Eng. J. 1999, 73, 1–21. (8) Pre’ Consultants. The Eco-indicator 99, a Damage Oriented Method for Life Cycle Impact Assessment: Methodology Report and Manual for Designers, 2nd ed.; Pre' Consultants: Amersfoort, The Netherlands, 2000. (9) Solomon, B. D.; Johnson, N. H. Valuing climate protection through willingness to pay for biomass ethanol. Ecol. Econ. 2009, 63, 2137–2144.
5135
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
Figure 1. Life cycle stage of a biomass-based fuel SC: Well-to-tank approach.
tion plants needs considering, too. Hence, the set of life cycle stages s considered in evaluating the environmental performance of the bioethanol production system are s ∈ S fbc, bds, bt, fp, ftg
Next, the mathematical formulation associated with environmental modeling is explained in detail. In the successive section, the environmental aspects of the case study are presented in terms of parameter definitions and modeling assumptions relating to each life cycle stage. The multi-objective optimization is then carried out considering the 2010 demand scenario. Different suitable biomass supply options are compared relating to a short-term demand scenario. In addition, the use of valuable subproducts (specifically the DDGS coming from the corn conversion into ethanol) is investigated as a suitable alternative to improve the environmental performance of the system. Some final remarks on the model capabilities and shortcomings conclude the paper.
where bc, bdc, bt, fp, and ft represent biomass cultivation, biomass drying and storage, biomass transport, fuel production, and fuel distribution, respectively. Issues such as potential differences in vehicle conversion efficiency (fuel energy to mechanical energy) as well as in vehicle technology related to the substitution of gasoline, the so-called tank-to-wheel (TTW) contribution to the overall impact, were not dealt with in accordance to two assumptions: (i) in a novel biofuel system (such as the bioethanol industry in Italy over the time horizon considered), the new fuel must be used in blends that do not need specific engines or equipment, and (ii) carbon dioxide emissions resulting from the combustion of the biofuel are assumed to be comprised in the carbon dioxide captured during crop growth and are therefore not included in the total count. The last step in outlining the general features of the LCA analysis for inclusion within the optimization framework is the emissions inventory. Here, we list the set of environmental burdens to be counted in evaluating the total ecological damage associated with the SC operation within the boundaries previously defined. In particular, the GHG contribution on global warming was captured by inventorying the following set of burdens: b ∈ B fCO2 , CH4 , N2 Og
Problem Description The objective is the development of a general modeling framework to design strategic SCs for biofuel systems. In particular, the design process is conceived as an optimization problem in which the production system is required to comply with both costs and GHG emissions minimization criteria. In incorporating environmental consciousness as part of the design process, especially when considered through the extended view of SCM, it is commonly advised to adopt a global approach capable of assessing the impact of the system under investigation over the entire life cycle. The methodology adopted in this work refers to the classical LCA techniques, whose principles and standards are laid out by the International Standards Organization.15 As mentioned in these guidelines, the goal and scope definition is a preliminary step of utmost importance in which decisions are taken about the precision and the representative value of the assessment that is going to be approached. A LCA analysis within the scope of mathematical programming has to be carried out with a double purpose. The first one is to perform the environmental optimization: LCA is needed as an endogenous tool to compare alternative topologies of the same production network. The second goal is to use the results obtained through optimization to compare the environmental performance of the obtained production system with that of an exogenous supply network (e.g., the conventional process that is aimed to replace). The performance measure that the overall environmental assessment must refer to, i.e., the functional unit of the system, needs to capture the nature of the service provided by products. Accordingly, when referring to alternative fuels, this should be defined as an absolute benchmark, such as kilometers driven or gigajoules of energy, provided using those fuels in a combustion engine. On the other hand, to obtain a satisfactory estimation of the emissions, special attention has to be given to the choice of the life cycle stages to be included, i.e., the system boundaries definition. According to the WTT approach adopted, the biomass-based fuel production system can be generally described as illustrated in Figure 1. With respect to part 11 of this paper, note that the drying and storage process of biomass before delivery to the produc-
that were grouped together in a single indicator in terms of carbon dioxide equivalent emissions (CO2 equiv). The derivation of carbon dioxide equivalent emissions is based on the concept of 100 year global warming potentials (GWPs) as specified by the International Panel on Climate Change.16 On the basis of the assumptions adopted in implementing the environmental assessment of the biofuel production system, the design problem is reformulated as follows. Given the following inputs: (i) geographical distribution of demand centers, (ii) fuel demand over a fixed time horizon, (iii) biofuel market characteristics, (iv) biomass geographical availability, (v) biomass production costs, (vi) biofuel production facilities capital and operating costs, (vii) transport logistics (modes, capacities, distances, availability, and costs), (viii) environmental burden of biomass production, (ix) environmental burden of biofuel production, and (x) transport means emissions, the global objective is now to determine the set of optimal system configurations resulting from the tradeoff between operating costs and GHG emission for the entire SC. Therefore, the key variables to be optimized are (i) geographical location of biomass production sites, (ii) biomass production for each site, (iii) supply strategy for biomass to be delivered to production facilities, (iv) biofuel production facilities location and scale, (v) distribution processes for biofuel (16) Intergovernmental Panel on Climate Change (IPCC). Climate Change 2001: The Scientific Basis; Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X., Maskell, K., Johnson, C. A., Eds.; Cambridge University Press: Cambridge, U.K., 2001.
(15) International Organization for Standardization (ISO). ISO 14040: Environmental Management;Life Cycle Assessment;Principles and Framework; ISO: Geneva, Switzerland, 1997.
5136
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al. 18
to be sent to blending terminals, (vi) supply chain management costs, and (vii) supply chain impact on global warming.
biomass crop has been established. In particular, the actual environmental performance is affected by fertilizer and pesticides usage, irrigation techniques, and soil characteristics. The factor may differ strongly from one production region to another. Accordingly, the form of eq 3 for the biomass production stage is defined as follows: X Ibc ¼ fbc, g Fbc, g ð4Þ
Mathematical Formulation The general modeling framework was formulated as a MILP problem according to the mathematical features outlined in part 11. The environmental frame as well as the multi-objective MILP solution algorithm are based on the approach proposed by Hugo and Pistikopoulos17 following the works by Papalexandri and Dimkou12 and Dua and Pistikopoulos.13 Objective Function. The mathematical formulation of the MoMP problem being investigated commences with the definition of the environmental criterion to be coupled with the economic one. The objective is the minimization of the total daily impact (TDI) (kg of CO2 equiv/day) resulting from the operation of the biofuel SC. Thus, the definition of TDI needs to consider each life cycle stage contribution, as expressed by the following equation: X TDI ¼ Is ð1Þ
g
where fbc,g is the carbon dioxide emissions equivalent per unit of biomass produced in element g (kg of CO2 equiv/ton) and Fbc,g is the daily biomass production in element g, i.e., PTbiomass,g (tons/ day) [see part 11]. Biomass Drying and Storage. Unlike part 11, where biomass drying and storage costs were not considered because they are already included in the biomass production costs, here, the impact deriving from the drying and storage of biomass cannot be neglected and has to be analyzed separately from the production stage. The environmental performance of this stage has no relation to the geographical location of the dedicated facilities but rather depends upon the technology adopted to process the biomass. This last issue was simplified by considering an average emission factor, fbds (kg of CO2 equiv/ton], estimated with reference to the performance of the most common practices adopted. Therefore, the total emission of the drying and storage stage is only influenced by the amount of biomass processed X Ibds ¼ fbds Fbds, g ð5Þ
s
The environmental impact Is (kg of CO2 equiv/day) resulting from the operation of the single stage s is calculated as follows: X db cb, s Fs " s ∈ S ð2Þ Is ¼ b
where the reference flow Fs (units/day), specific for each life cycle stage s, is multiplied by the emission coefficient cb,s, representing the quantity of substance b emitted at stage s per unit of reference flow, and by the damage factor db, characterizing the contribution of each burden b to the global warming in terms of carbon dioxide emissions equivalent per unit of burden emitted, i.e., the GWPs. This formulation, although broadly recognized as a rigorous and comprehensive practice, may nonetheless turn out to be too onerous in terms of both calculation effort and data collection. For this reason, the mathematical formulation was simplified by grouping the emission coefficient, cb,s, together with the damage factor, db, thus devising a global emission factor fs, which represents the carbon dioxide emissions equivalent at stage s per unit of reference flow. Accordingly, eq 2 takes the form Is ¼ f s F s " s ∈ S ð3Þ
g
Transport System. The global warming impact related to both biomass supply and product distribution is due to the use of different transport means fueled with fossil energy, typically either conventional oil-based fuels or electricity. The resulting GHG emissions of each transport option depend upon both the distance run by the specific means and the freight load delivered. As a consequence, the emission factor fs,t represents the total carbon dioxide emissions equivalent released by transport unit t per kilometer driven and ton carried. Thus, Is is evaluated as follows: X Is ¼ fs, t Fs, t " s ∈ fbt, ftg ð6Þ t
with the reference flow Fs,t now representative of the delivery distance (ADDt,g,g0 ) and the load of goods transported (Qi,g,t,g0 ), as defined by the equation XX Qi, g, t, g0 ADDt, g, g0 " s ∈ fbt, ftg ð7Þ Fs, t ¼
As will be further detailed in the following sections, both fs and Fs might be either grid- or transport-dependent according to the specific life cycle stage s that they refer to. As a consequence, eq 3 can be expressed as either X Is ¼ fs, g Fs, g " s ∈ fbc, bds, fpg ð3aÞ
g
g
or Is ¼
X
fs, t Fs, t
" s ∈ fbt, ftg
g0
Fuel Production. The environmental impact of the biofuel production stage is related to raw materials (other than biomass) and utilities required in operating the conversion facilities. Accordingly, the GHG emissions resulting from this life cycle stage were assumed to be proportional to the total daily amount of biofuel produced, PTfuel,g (tons/day) (taken as reference flow Ffp,g) and independent of location, as shown in the following expression: X Ifp ¼ ffp Ffp, g ð8Þ
ð3bÞ
t
Life Cycle Stages Impact. The stage-related environmental impacts as represented in eq 3 are generally defined for the entire set of life cycle stages. However, the reference flows as well as the impact factors may depend upon either the specific location (grid element g) or the transport mode t. Thus, it is necessary to uniquely define the reference flows for each individual life cycle stage and express them explicitly as a function of the design variable controlling the optimization problem. Biomass Production. GHG emissions resulting from the production of biomass notoriously depend upon the cultivation practice adopted as well as the geographical region in which the
g
Emission Credits. The effect of byproducts, some of which are valuable products in other markets, is essential to allocate the total impact associated with a particular production chain. Currently, there is no accepted best method to cope with this issue. In this work, allocation by substitution was chosen (18) Romero Hernandez, O.; Salas-Porras, E. D.; Rode, M. I. Panorama of the Social, Environmental, and Economic Conditions of Corn Ethanol in Mexico; Instituto Tecnologico Autonomo de Mexico (ITAM): Mexico City, Mexico, 2008.
(17) Hugo, A.; Pistikopoulos, E. N. Environmentally conscious longrange planning and design of supply chain networks. J. Cleaner Prod. 2005, 13, 1471–1491.
5137
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
19
imposed by current Italian regulations for 2010 was assumed as the only demand scenario to design a corn-based ethanol SC considering both operating costs and GHG emission minimization criteria. As a first instance, we considered that the DDGS would be used as animal feed with the corresponding allocation of SC operating costs and GHG emissions. Results were compared to a second instance, where DDGS is assumed to fuel a CHP station providing the utility requirements of the conversion facilities. To analyze the system, the environmental frame previously described was tailored to the case study being investigated to define the emission factors. This is a crucial point to apply the modeling framework in a sensible and reliable way. In fact, given the simplifications in the final formulation, it is only through a rigorous set of parameters that the needs of accuracy and thoroughness required by the exogenous comparison can be met. To comply with these requirements, we used an interactive spreadsheet-based tool specifically developed to investigate the GHG emission related to wheat to ethanol production in the U.K.22 This tool uses default values based on a typical production chain. The default input values were modified and adapted to calculate the specific emission factors for corn-based ethanol production in Italy. In the following sections, assumptions and approaches used in deriving the model parameter set will be described with reference to each life cycle step. Biomass Production. Actual data regarding the Italian corn cultivation practice and, in particular, crop yields, mineral (N, K, and P) and organic (cattle manure) fertilizer requirements, seeds and pesticides usage, and diesel fuel for irrigation were retrieved from both literature and governmental institution databases.23-29 Because important differences exist between wheat and corn cultivation practices, the actual values of the global emission factors fbc,g were calculated adopting the equations recommended by the IPCC guidelines.30 Global emission factors are calculated with reference to 1 ha of cultivated land. To match the needs of the mathematical formulation adopted, the set of fbc,g is assumed to be grid-specific (and to refer to the units of biomass produced). This is not a trivial unit conversion exercise, because just a
following the recommendations of Rickeard et al.: this method assigns the primary product the total GHG emissions minus the credits derived by the emissions avoided because of displacements of alternative goods by the byproduct. In first-generation bioethanol systems based on grains, the main byproduct is a high-protein meal coming from the solid fraction of the postprocess residues (DDGS). This is a valuable substitute for cattle feed and may also be used as a fuel for CHP generation.20 The modeling framework was developed to take into account these two alternative options to calculate credits for emissions avoided through displacement of equivalent amounts of cattle feed production (sm) or electricity and heat generation (en). Following Delucchi,21 no credits were assigned for land usage. In fact, the conversion from crop for food to crop for fuel generates a gap in the market that has to be filled by either importing corn (resulting in a higher impact) or cultivating other lands (resulting in a lower impact in the case of setaside land). Therefore, it is advised to consider an average situation in which no credits arise from changes in the land usage. The byproduct credits allocation was included in the mathematical formulation by considering the emission credits as a negative contribution to the life cycle stage impact calculation. This means that the sum on the right side of eq 1 needs to comprise one more competitive contribution that can be alternatively X Ism ¼ -fsm Fsm, g ð9Þ g
where fsm is the carbon dioxide emissions equivalent credit assigned to cattle feed displacement per unit of fuel produced and Fsm,g is the daily fuel production in element g, PTfuel,g (tons/ day), or X Ien ¼ -fen Fen, g ð10Þ g
with fen representing the CO2 equivalent emission assigned to energy production displacement per unit of fuel produced and Fen,g still indicating the daily fuel production in element g. Note that, according to the formulation adopted in this work, the two alternatives are assessed independently as a pseudo-life cycle stage set C {sm, en}. Accordingly, eq 3 is reformulated as follows: Is ¼ fs Fs " s ∈ S ∪ C ð11Þ
Case Study (22) Brown, G.; Wood, J.; Estrin, A. Bioethanol Greenhouse Gas Calculator: Users’ Guide; Home-Grown Cereals Authority (HGCA): London, U.K., 2005. (23) Istituto Nazionale di Statistica (ISTAT) database, www.istat.it. (24) Grignani, C.; Zavattaro, L. A survey on actual agricultural practices and their effects on the mineral nitrogen concentration of the soil solution. Eur. J. Agron. 2000, 12, 251–268. (25) Grignani, C.; Zavattaro, L.; Sacco, D.; Monaco, S. Production, nitrogen and carbon balance of maize-based forage systems. Eur. J. Agron. 2007, 26, 442–453. (26) Guerini, G.; Maffeis, P.; Allievi, L.; Gigliotti, C. Integrated waste management in a zone of northern Italy: Compost production and use, and analytical control of compost, soil, and crop. J. Environ. Sci. Health, Part B 2006, 1, 1203–1219. (27) Locatelli, L. Analisi LCA (life cycle assessment) della produzione di bioetanolo. M.S. Thesis, University of Padova, Padova, Italy, 2007. (28) Marchetti, R.; Ponzoni, G.; Spallacci, P. Simulating nitrogen dynamics in agricultural soils fertilized with pig slurry and urea. J. Environ. Qual. 2004, 33, 1217–1229. (29) Sacco, D.; Bassanino, M.; Grignani, C. Developing a regional agronomic information system for estimating nutrient balances at a larger scale. Eur. J. Agron. 2003, 20, 199–210. (30) Intergovernmental Panel on Climate Change (IPCC). 2006 IPCC Guidelines for National Greenhouse Gas Inventories; Prepared by the National Greenhouse Gas Inventories Programme; Eggleston, S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K., Eds.; Institute for Global Environmental Strategies (IGES): Hayama, Japan, 2006.
The emerging corn-based ethanol production system in northern Italy is chosen as an illustrative example. In part 11, it was shown that importing corn allowed for a more economical design of the overall SC. However, importing corn from eastern European countries to meet biomass demand poses the question on the effective environmental performance of the system as conceived. In fact, one may ask how the best design in terms of cost reduction performs from an environmental standpoint. Thus, the case study allowing corn importation described in part 11 was taken as a reference to formulate a new case study for the multi-objective modeling framework addressed here. The ethanol market penetration (19) Rickeard, D. J.; Punter, G.; Larive, J. F.; Edwards, R.; Mortimer, N. D.; Horne, R.; Bauen, A.; Woods, J. WTW Evaluation for Production of Ethanol from Wheat; Low Carbon Vehicle Partnership (LCVP): London, U.K., 2004. (20) Morey, R. V.; Tiffany, D. G.; Hatfield, D. L. Biomass for electricity and process heat at ethanol plants. Appl. Eng. Agric. 2006, 22, 723–728. (21) Delucchi, M. A. Lifecycle Analyses of Biofuel; Draft report; University of California: Davis, CA, 2006.
5138
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al. Table 1. Global Emission Factors fs,t for Different Transport Modes mean t
fs,t (kg of CO2 equiv ton-1 km-1)
small truck truck rail barge ship trans-ship
0.591 0.123 0.021 0.009 0.007 0.006
Fuel Production. Given the existing similarity between the wheat- and corn-based technology to convert starch into ethanol, the global emission factors were calculated from the spreadsheet. Default values related to raw materials and utilities for wheat-based production were replaced by cornspecific inputs obtained from a detailed process model.33 Accordingly, the global value of fft resulting from the spreadsheet calculation is equal to 1052.23 kg of CO2 equiv/ton. Emission Credits. Two different instances were considered as viable alternatives to exploit DDGS side production. In particular, DDGS was considered as either a suitable soymeal substitute in the animal feed market or a fuel to feed a CHP station providing steam and power to the ethanol plant as well as an electricity surplus to be sold to the national grid. In Italy, soy meal for animal feed is usually imported from Brazil. Unfortunately, very scarce data are available on the life cycle emissions from production and importation of Brazilian soy meal to Italy. Therefore, data regarding imported soy meal from the U.S. were used:14 each kilogram of DDGS is supposed to replace 0.78 kg of soy meal (in terms of protein content). Moreover, production in the U.S. and transport to EU of a kilogram of meal result in emissions of 0.46 kg of CO2 equiv. Consequently, assuming a fixed yield33 of 0.954 kg of DDGS/kg of ethanol produced, it is straightforward to obtain the global emission credit fsm value of 342.22 kg of CO2 equiv/ton. If DDGS is used for CHP generation, credits were calculated under the assumption20 that burning the whole production of DDGS allows us to satisfy the production process utilities needs and to deliver a surplus of electricity to the grid. This corresponds to a displacement of 12.3 GJ of natural gas and 5.2 GJ of electricity for each ton of ethanol produced. In this second instance, the global emission credits fen related to energy substitution were equal to 1427.38 kg of CO2 equiv/ton. It is also noteworthy to assess the effects on the economics when either one of these two options is adopted. As already stated in part 11, selling DDGS as an animal feed substitute entails a cost allocation of 20% to discount from the entire SC overheads. The situation changes if it is used for CHP production because the cost allocation decreases to about 15%;33 however, also the unit production costs UPCfp reduce substantially thanks to the energy savings. This required the redefinition of the UPCfp for each plant capacity range. Table 2 summarizes the model parameters related to the biofuel production in the second instance.
Figure 2. Global emission factors for corn cultivation (fbc,g): the emission factor depends upon the actual crop yield.
subset of the input parameters depend upon the corn yield. As a consequence, the conversion was based on the following assumptions: (1) mineral fertilizers usage was described as linearly dependent upon the corn yield (this is not generally true, but it is a reasonable compromise in the range of corn yields considered in this study); the larger the local yield, the larger the amount of mineral fertilizers per unit of land (and the emissions too because of fertilizer production); (2) organic fertilizer usage per unit of cultivated land was set constant; the larger the corn yield, the lower the soil emissions because of manure usage; and (3) diesel usage for irrigation per unit of cultivated land was set constant; the larger the corn yield, the lower the emissions because of diesel usage. As a result, a grid-dependent set of parameters was generated, as illustrated in Figure 2, where every diamond represents the global emission factors fbc,g as a function of the crop yield. In this way, it was possible to represent the real situation adequately; the optimal impact per unit of biomass produced comes from a trade off between usage of resources and effective corn yield. The global emission factor identifying the foreign supplier (g = 60) was calculated by assuming a hypothetical corn yield that was set equal to the average CYg value weighted on the Italian data. Under this assumption fbc,60 turns out to be equal to 359.9 kg of CO2 equiv/ton. This approximation is needed to overcome the lack of actual data for the supplier countries considered in the SC analysis. However, our analysis indicates that this assumption does not affect the quality of the results in a significant way. Biomass Drying and Storage. With respect to the GHG emissions related to biomass drying and storage, the set of parameters was derived from the spreadsheet results by introducing country-specific data. In particular, the emission related to diesel and electricity usage in Italy were taken from DEFRA31 and EME,32 respectively. Accordingly, the value of fbds was set equal to 63.34 kg of CO2 equiv/ton. Transport System. Global emission factors specific to each transport option were taken from DEFRA.31 Table 1 reports the resulting transport-related emission factors, fbt,t and fft,t.
Results and Discussion The 2010 scenario described in part 1 (10.1021/ef900456w) and defined according to the market penetration set by biofuel regulation was taken as the bioethanol demand benchmark. Initially, the instance according to which DDGS is sold as in
(31) Department for Environment, Food and Rural Affairs (DEFRA). Guidelines to Defra’s GHG Conversion Factors: Methodology Paper for Transport Emission Factors; DEFRA: London, U.K., 2008. (32) URS Corporation. Greenhouse Gas Emission Factor Review: Final Technical Memorandum; Edison Mission Energy: Irvine, CA, 2003.
(33) Franceschin, G.; Zamboni, A.; Bezzo, F.; Bertucco, A. Ethanol from corn: A technical and economical assessment based on different scenarios. Chem. Eng. Res. Des. 2008, 86, 488–498.
5139
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
Table 2. Ethanol Production Costs when DDGS Is Used To Provide the Process Energy Needs plant size p 1 2 3 4
UPCfp (euros/kg) 0.060 0.049 0.041 0.029
the animal feed market is considered. The two objective functions problem is solved through the CPLEX solver in the GAMS modeling tool.34 Figure 3 shows the resulting trade-off set of non-inferior solutions. The shape of the curve reveals the expected conflict existing between environmental and economic performance. The optimum in terms of economic performance (case A as reported in Figure 3) involves a marginal operating cost value of 23.03 euros/GJethanol against an overall environmental impact of 79.15 kg of CO2 equiv/GJethanol corresponding to a GHG emissions reduction of about 8% compared to gasoline (the GHG emissions factor for gasoline was assumed equal to 85.8 kg of CO2 equiv/GJ22). Table 3 resumes the details of the optimization outcomes of case A, while Figure 4 shows the corresponding network configuration. The graphical representation reveals the SC general structure. Biomass demand is met by importing corn from eastern European countries. Corn is directly shipped to the two production plants of the maximum capacity (about 250 kilotons/year) and located within the industrial areas close to the main ports of Venice (g = 32) and Genoa (g = 46). This configuration allows for the best economic performance in terms of both biomass supply costs, because of the lower price of the imported corn and ethanol production costs, positively affected by the scale factor. In terms of environmental performance, 8% of GHG reduction in the emissions is not enough to meet the latest EU standards, which require biofuels to have a minimum of 35% of GHG emissions savings. However, even by minimizing the impact (point B of Figure 3), the resulting GHG emissions are still too high to meet the minimum requirements, albeit the substantial economic effort required to reach the target; reducing the marginal impact down to 74.88 kg of CO2 equiv/ GJethanol (equal to 13% of GHG reduction) results in an increase of the overall operating costs up to 25.80 euros/ GJethanol. In fact, when moving along the Pareto curve (Figure 3) from point A to point B, we see a gradual transition toward a network configuration (illustrated in Figure 5), proposing a more decentralized fuel production system that requires the establishment of four conversion plants: one of large capacity (p = 4) located in the neighborhood of Venice (g = 32) and three other plants of smaller size (p = 1) sited in the most convenient area in relation to the domestic biomass production (g = 26, 43, and 52). In comparison of the related costs details reported in Table 4 to those of case A, it is evident that the supply solution outlined in B would ensure a better environmental performance in terms of biomass distribution and corn production impact but also a clear deterioration of the system economics because of the negative scale factor on ethanol production costs as well as the unprofitable biomass supply conditions. Furthermore, even achieving a more sustainable supply system does not
Figure 3. Pareto curve: simultaneous optimization under operating costs and GHG emissions minimization criteria.
satisfy the EU standard requirements in terms of GHG emissions saving. The second instance considers DDGS as a fuel for CHP stations. This alternative use of DDGS would entail a production costs reduction, because of substantial savings on utilities supply costs, but also a considerable capital investment for the power station installation. The surplus of electricity production (globally amounting to 3.2 MJ for each ton of ethanol produced) can be sold to the national grid to gain some emission credits assigned for electricity displacement. The Pareto curve is similar to the one in Figure 3 and is not reported here. In fact, the optimal network configuration defined by the costs optimization is identical to the one illustrated in Figure 4; this is quite expected, because cost minimization would still favor low cost corn (imported from abroad) and large plants. However, the marginal operating costs value is now lower and equal to 22.40 euros/GJethanol against an overall environmental impact of 38.85 kg of CO2 equiv/GJethanol. As reported in Table 5, this reduction on the entire SC overheads is due to the decrease in ethanol production costs, notwithstanding a slight reduction in the allocation credits (now accounting for the 15% of the overall SC operating costs). On the other hand, the substantially improved environmental performance occurring with this system configuration is attributable to the larger emission credits coming from the alternative use of byproducts. This solution, indeed, would allow for a GHG reduction of about 55% compared to gasoline. If the optimization is forced toward the minimization of the environmental impact, the SC performance in terms of GHG emissions is even more promising. As reported in Table 6, the optimization results in an estimated environmental burden reduction of about 60% (corresponding to a marginal value of 34.58 kg of CO2 equiv/GJethanol). However, in this situation, the marginal operating costs increase up to 26.81 euros/ GJethanol, thus exceeding the value calculated in the first instance. This depends upon the lower costs allocation that is not balanced by the utilities supply costs reduction. Additional remarks are drawn by analyzing the social costs of bioethanol production. As mentioned in part 11 of this paper, Italian regulation provides for a taxation discount on inland duties amounting to 4.2 euros/GJ with respect to other conventional automotive fuels. This involves a reduction of the breakeven point with gasoline down to 74 $/bbl. However, if the objective is to promote maximum GHG
(34) Rosenthal, R. E. GAMS;A Users’ Guide (Version 22.5); GAMS Development Corporation: Washington, D.C., 2006.
5140
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
Table 3. Costs-Optimal Solution: Results of the SC Optimization (Case A: Cost Minimization) biomass production biomass drying and storage transport system fuel production allocation credits total
operating costs (euros/GJ)
GHG emissions (kg of CO2 equiv/GJ)
13.12
41.21 7.25 4.34 39.04 12.70 79.15
3.16 12.51 5.76 23.03
mitigation, then either additional subsidies would be needed or a higher breakeven point is to be expected. The cost to bridge this gap amounts to about 3 euros/GJethanol (0.10 euros/Lethanol) when DDGS is used as an animal-feed substitute. If DDGS is used for CHP fuel, the difference is about 4 euros/GJethanol (0.12 euros/Lethanol). However, note
the second instance allows for a breakeven point of about 72 $/bbl when optimized under costs minimization. This last situation would allow for better use of financial resources; the system might be supported with the same amount of subsidies to ease the market penetration, and still it would be possible to match the EU regulation in terms of
Figure 4. Costs-optimal solution: supply network configuration. Case A: cost minimization.
Figure 5. Emissions-optimal solution: supply network configuration. Case B: environmental optimization.
5141
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
animal fodder or CHP fuel has been discussed (and optimized) without assessing the potential variations in terms of the production profitability [e.g., in terms of net present value (NPV)]. However, the great uncertainty on the DDGS selling price as well as on the subsidies (green credits), deriving from selling “renewable” electricity to the national grid, persuaded us that a first evaluation tool should be based on costs. Future work will be dedicated to a more comprehensive financial analysis. Finally, a deeper assessment may be required to evaluate the effects of the uncertainty related to parameter definitions. Although our assumptions of the GHG emissions concerning corn in eastern European countries appears to be sensible enough, it is important to note that importing corn from countries characterized by uncertain environmental standards in cropping practice may significantly affect the environmental results and put at risk the achievement of the EU goals (albeit through an improvement of the overall economics).
Table 4. Emissions-Optimal Solution: Optimization Results (Case B: Environmental Optimization)
biomass production biomass drying and storage transport system fuel production allocation credits total
operating costs (euros/GJ)
GHG emissions (kg of CO2 equiv/GJ)
15.21
39.64 7.25 1.64 39.04 12.70 74.88
3.84 13.20 6.45 25.80
Table 5. Costs-Optimal Solution: Results of the SC Optimization Considering DDGS as CHP Fuel
biomass production biomass drying and storage transport system fuel production allocation credits total
operating costs (euros/GJ)
GHG emissions (kg of CO2 equiv/GJ)
13.12
41.21 7.25 4.30 39.04 52.96 38.85
3.18 10.05 3.95 22.40
Acknowledgment. A.Z. gratefully acknowledges the financial support of the University of Padova under Progetto di Ateneo 2007 (cod. CPDA071843): “Bioethanol from lignocellulosic biomass: Process and equipment development”.
Table 6. Emissions-Optimal Solution: Results of the SC Optimization Considering DDGS as CHP Fuel
biomass production biomass drying and storage transport system fuel production allocation credits total
operating costs (euros/GJ)
GHG emissions (kg of CO2 equiv/GJ)
15.17
39.63 7.25 1.62 39.04 52.96 34.58
3.44 12.93 4.73 26.81
Nomenclature Sets s ∈ S ∪ C = set of life cycle stages (S) and pseudolife cycle stages accounting for emissions credits solutions (C) b ∈ B = set of burdens (CO2, CH4, or N2O) g ∈ G = set of square regions g0 ∈ G = set of square regions different from g i ∈ I = set of products (biomass or biofuel) t ∈ T = set of transport modes (truck, rail, barge, ship, or trans-ship)
GHG reduction for biofuel production processes. Therefore, in the particular case of the Italian corn-based ethanol production, a well-advised strategy would address the design process under economic criteria, especially adopting a system configuration in which byproducts are used to provide the energy needs of the production facilities.
Parameters
Conclusions
db = damage factor for each burden b (kg of CO2 equiv/ unit of b) cb,s = emission coefficient of burden b at stage s (unit of b/ unit of reference flow) fs = global emission factor for stage s (kg of CO2 equiv/ unit of reference flow) UPCfp = unit production costs for biofuel through plants of size p (euros/ton)
The spatially explicit modeling tool for the strategic design of biofuel supply networks addressed in part 11 of this work has been improved here by adding an environmental analysis within the optimization framework. The aim of the study has been to incorporate a GHG emissions reduction objective within the costs minimization criteria to create a multiobjective optimization tool that may be helpful in driving more conscious policies to support the biofuel sector. The bioethanol production system of northern Italy outlined in part 11 has been taken as a reference to define the case study used to illustrate the model application and suitability for strategic design. Both the options to sell DDGS as an animalfeed substitute and to exploit DDGS to fuel CHP stations have been analyzed and discussed. The non-inferior set of viable solutions indicates that the most interesting alternative proposes: (i) the design of the bioethanol supply system under costs minimization and (ii) the usage of DDGS as fuel to produce the heat and power required by the production facilities. The optimization outcomes demonstrate the effectiveness of the system in reaching the GHG mitigation (by 55%) necessary to meet the EU standards. Note that, in this study, the economic performance of the system has been defined in terms of the ethanol production costs, only. As a consequence, the effect of using DDGS as
Continuous Variables TDI = total daily impact (kg of CO2 equiv/day) Is = daily impact for stage s (kg of CO2 equiv/day) Fs = reference flow for stage s (units/day) PTi,g = production rate of product i in region g (tons/ day) Qi,g,t,g0 = flow rate of product i via mode t between g and g0 (tons/day) ADDt,g,g0 = actual delivery distance between grids g and g0 via mode t (km) Acronyms bc = biomass cultivation bds = biomass drying and storage bt = biomass transport CHP = combined heat and power 5142
Energy Fuels 2009, 23, 5134–5143
: DOI:10.1021/ef9004779
Zamboni et al.
DDGS = distiller’s dried grains with soluble en = energy fp = fuel production ft = fuel transport GHG = greenhouse gas GrSCM = green supply chain management GWP = global warming potential LCA = life cycle analysis MP = mathematical programming
MILP = mixed integer linear programming MoMP = multi-objective mathematical programming PSE = process systems engineering SC = supply chain SCA = supply chain analysis SCM = supply chain management sm = soy meal WTT = well to tank TTW = tank to wheel
5143
