Clean Technologies and Environmental Policy

, Volume 20, Issue 7, pp 1697–1719 | Cite as

Optimization of multi-pathway production chains and multi-criteria decision-making through sustainability evaluation: a biojet fuel production case study

  • Eduardo Vyhmeister
  • Gerardo J. Ruiz-Mercado
  • Ana I. Torres
  • John A. Posada
Original Paper


Selection of optimal technologies for novel biobased products and processes is a major challenge in process design, especially when are considered many alternatives available to transform materials into valuable products. Furthermore, such technological alternatives vary in their technical performances and cause different levels of economic and environmental impacts throughout their life cycles. Additionally, selection of optimal production pathways requires a shift from the traditional materials management practices to more sustainable practices. This contribution provides a method for optimizing multi-product network systems from a sustainability perspective by applying the GREENSCOPE framework as a sustainable objective function. A case study is presented in which the four GREENSCOPE target areas (i.e., efficiency, energy, economics, and environment) are evaluated by 21 preselected indicators as part of a multi-objective optimization problem of a biojet fuel production network. The biojet fuel production network evaluated in this study consists of four main elements: (1) feedstocks management, (2) conversion technologies, (3) co-products upgrading, and (4) auxiliary sections for in situ production of raw materials and utilities. For the sustainability objective function, the 21 indicators are analyzed considering multiple perspectives of stakeholders to study their influence on the decision-making process. It is, different sets of weighting factors are assigned to each of the four target areas. Hence, this sustainability evaluation from different stakeholders’ perspectives allows identifying optimal networks, specific target areas with great potential for improvements, and processing steps with great influence in the entire network performance. As a result, diverse optimal network arrangements were obtained according to the multiple stakeholders’ perspectives. This evidences that a win–win situation for all sustainability aspects considered can hardly be reached. Finally, this contribution demonstrated the applicability of the proposed methodology for sustainability evaluation, optimization, and decision-making in the context of a multi-product material facility by developing a multi-objective optimization model.


Biojet fuel biorefinery Multi-criteria decision-making Multi-objective optimization Multi-stakeholder analysis Sustainability assessment Materials management 

List of symbols



Multi-criteria decision-making


Life cycle assessment


Life cycle costing


Greenhouse gas


Global warming potential


Gauging reaction effectiveness for the environmental sustainability of chemistries with a multi-objective process evaluator


GREENSCOPE perspectives (efficiency, energy, environmental, and economics)


Total production cost


EPA’s toxic release inventory



Reaction mass efficiency


Mass intensity


Effective mass yield


Carbon efficiency


Renewability-material Index


Fractional water consumption


Health hazard, irritation factor

HHchronic toxicity

Health hazard, chronic toxicity factor

SHacute tox.

Safety hazard, acute toxicity


Specific toxic release


Global warming potential

WPO2 dem.

Aquatic oxygen demand potential

ms, spec.

Specific solid waste mass

Vl, spec.

Specific liquid waste volume


Specific energy intensity


Renewability energy index


Renewability-exergy index


Discounted payback period


Turnover ratio


Specific raw material cost

CE, spec.

Specific energy cost



Steam methane reforming process


Hydrothermal liquefaction process


Gasification followed by Fischer–Tropsch process



Stakeholder GREENSCOPE perspective weight (i = 1, 2, 3, 4; i.e., efficiency, energy, environmental, and economic)


Relative importance of a j index within the same GREENSCOPE perspective i


ith GREENSCOPE indicator of the efficiency perspective


ith GREENSCOPE indicator of the energy perspective


ith GREENSCOPE indicator of the environmental perspective


ith GREENSCOPE indicator of the economic perspective


ith node


Arc connecting nodes I and j. Starting in i and ending in j

\(N_{{{\text{inlet}} .st - i}}\)

Number of Inlet streams in technology i


Technology matrix


Flowrate at the j position






Total number of technology nodes


Number of units (e.g., nST number of service technology nodes)

Type of nodes


ith service technology


ith reactant


ith intermediate node


ith market service technology


ith feedstock


jth technology of the ith category of grouped technologies


Nonexistence of the jth technology of the ith category of grouped technologies


Decades of unsustainable industrial practices and excessive consumption patterns have resulted in significant environmental pollution and degradation of natural resources. For example, it has been estimated that if everyone on the planet consumed as much as the average U.S. citizen, 3.9 Earths would be needed to sustain the global population.1 Therefore, there is a strong urgency to implement sustainable development goals as worldwide policies to address not only environmental challenges but also social and economic ones.2 For example, in the manufacturing sector, sustainability assessment methods and indicators can play an important role in strategic decision-making since such indicators can provide useful (quantitative or qualitative) information to stakeholders to identify the most sustainable options and to avoid (or improve) unsustainable alternatives. Such sustainability-based decisions can be even more impactful when sustainability criteria, local context, and stakeholders’ perspectives are considered from the early design stages of new projects and high-level alternatives are evaluated (Palmeros Parada et al. 2018). However, selecting sustainability indicators is, by no means, an easy task. While many indicators are applicable (or easily adaptable) to multiple types of systems with different scales and properties, the same indicators may be irrelevant for defining sustainability aspects of stakeholders’ crucial interest. Therefore, selecting indicators and assessment methodologies from the extensive pool of options available in the literature (Palmeros Parada et al. 2017), that are sensitive enough to both relevant process changes and stakeholders’ perspectives, is quite relevant. Such selection is especially sensitive when the assessment method requires the implementation of holistic multi-criteria decision-making (MCDM) principles.

Well-accepted methodologies for assessing environmental and economic impacts of a product or service are the Life Cycle Assessment (LCA) and the Life Cycle Costing (LCC) (Cheali et al. 2016; Gargalo et al. 2016a, b; Luu and Halog 2016). These methods have been used to analyze: (1) a multi-product generation in a lignocellulosic biorefinery by considering economic, environmental, and safety indicators in their objective function (Larragoiti-Kuri et al. 2017); (2) a supply chain profit maximization under sustainability constraints (water requirements and land use) for strategic production and planning (González-Estudillo et al. 2017); (3) a set of possible pathways (multiple products, feedstocks, and technologies) for optimal biorefineries while considering net profit and minimization of global warming potential (GWP) (Murillo-Alvarado et al. 2013); and (4) a supply chain design (Yue et al. 2016) by combining, in a hybrid model, a bottom-up [i.e., process level, see also (Smith et al. 2017)] with a top-down [i.e., input–output method, see also (Cashman et al. 2016)] approach to minimize both total project costs and GWP. However, it is important to mention that these approaches require considerable efforts in data collection and processing. In addition, these approaches mostly focus on optimizing profitability and GWP of the entire life cycle rather than on specific manufacturing stages that is where process designers and stakeholders possess more influence to modify and improve the overall performance of the process (Ruiz-Mercado et al. 2014). In addition, some recent contributions describe the techno-economic evaluation and greenhouse gas (GHG) emissions assessment of biomass to energy supply chains. Mirkouei et al. (2017a) studied the economic and GWP of a mixed supply chain, including uncertainty and sensitivity analysis. Results indicate that mixed supply chains have the potential of cost and GWP reduction against traditional supply chains. Diederichs et al. (2016) performed a techno-economic analysis and comparison of different feedstocks (first and second generation) and pathways for biojet fuel production processes. It was found that feedstock and capital costs are the most influential aspects of the product selling price. Mirkouei et al. (2017b) and Gutiérrez-Antonio et al. (2017) performed some qualitative and quantitative state-of-the-art reviews on biomass to bioenergy supply chains and production processes of biojet fuel. It was concluded that novel pretreatment technologies and multiple processing routes are needed for achieving more profitable supply chains. However, the manufacturing stage is the key part of a system sustainability assessment and optimization methodology in which the designers and decision-makers can implement changes to modify and improve the entire supply chain sustainability. Also, this supply chain stage provides more reliable, practical, relevant, and important data values and models for evaluating sustainability at early development phases.

Unlike traditional process design approaches (i.e., a sequence of inventory of alternatives, evaluation of techno-economic indicators, and multi-criteria comparison of a set of indicators; generally defined by the design team), the design of sustainable processes requires a broader analysis of not only processing alternatives but also the simultaneous consideration of multiple sustainability objectives (e.g, MCDM) and constraints (e.g., environmental, energy, efficiency, and economic indicators). The Gauging Reaction Effectiveness for the ENvironmental Sustainability of Chemistries with a multi-Objective Process Evaluator (GREENSCOPE) methodology (Ruiz-Mercado et al. 2012a, b, 2013) was proposed as an effective decision-making support tool for sustainability evaluation in the development of chemicals and their manufacturing stages. GREENSCOPE evaluates sustainability in four main areas (four E’s): environment, energy, efficiency (material), and economic. A set of almost 140 indicators are proposed for numerical evaluation of process performance and on how those indicators are related to sustainability by using independent defined boundary values for each indicator that refer to a potential best-case and a worst-case scenario. Therefore, this approach employs a dimensionless sustainability scale that provides indicator results as percentage scores (0–100% of sustainability). This standardization allows stakeholders to compare multiple options for manufacturing the same good(s) (i.e., a superstructure-based approach) or when multiple changes or variations are evaluated in a process/technology of interest (i.e., through optimization tools).

Hence, a combination of the sustainability evaluation framework together with a multi-objective optimization tool would allow decision-makers to evaluate multiple alternatives and identify the most sustainable solutions. Furthermore, if sustainability impacts were considered as design objectives rather than just performance constraints, the design outcomes would likely lead to the discovery of novel sustainable alternatives (Cano-Ruiz and McRae 1998). However, in such a multi-variable optimization, a variety of optimal designs can be generated according to (interests of) the stakeholders and the relative weight of their opinions.

This contribution demonstrates and details the successful adaptation and implementation of GREENSCOPE sustainability assessment during the design and optimization of multi-pathway multi-product manufacturing systems as a mechanism to assist stakeholders in the multi-criteria decision-making step for achieving a more sustainable multi-pathway manufacturing route. To do so, a multi-objective optimization model was formulated and applied to optimize the structure of a multi-product materials management facility. A sustainability objective function is proposed by applying the GREENSCOPE framework that is able to consider multiple sustainability criteria for four aspects of sustainability (i.e., environmental, energy, efficiency, and economics) into the model. In addition, this sustainability objective function can identify optimal pathways by considering multiple priorities and requirements of the stakeholders. Finally, the developed optimization and decision-making for sustainability framework is applied to a relevant case study to produce biojet fuel and other second-generation key biofuels. Three different renewable biomass feedstocks (eucalyptus, pine, and macauba) and three conversion technologies (fast pyrolysis or hydrothermal liquefaction followed by a product-upgrade stage, and gasification followed by Fischer–Tropsch process) are considered. Therefore, stakeholders can implement the developed multi-criteria decision-making framework to select an optimally sustainable production chain that considers all processing stages: raw material selection, feedstock conversion into intermediate products (e.g., biocrude), and upgrading to more valuable products (e.g., biojet fuel). Lastly, this contribution aims to demonstrate the practicability of the proposed methodology for sustainability evaluation, optimization, and decision-making in the context of a multi-product chemical facility by developing a multi-objective optimization model which can lead to increased sustainability.

The remaining of the manuscript is divided into three main sections. “Problem formulation and solution approach” section covers the GREENSCOPE approach and explain its use in the setting of multiple perspective decision-making when multiple processing pathways are possible. Immediately after, in the same section, the previous approach is casted into a generic multiple-input, multiple-output, multiple processing technologies network superstructure, and the associated network optimization problem is formulated. For this prototypical network, the objective functions and restrictions involved are presented. “Case study” section starts by describing the case study under geographical, temporal, and technological context. The network structure particular for this case study is presented in “Network formulation of the biomass to biojet fuel processes” section in order to formulate the multiple perspective optimization problem. Once this is covered, the approach to solve the optimization problem is presented (i.e., software algorithms and other important considerations). Finally, “Approach to solve the optimization problem” section focuses on the discussion of the optimization results for this case study and the conclusions of the developed work.

Problem formulation and solution approach

Multiple perspectives of stakeholders

Indicators are useful in comparing alternative process options to better understand the sustainability-related trade-offs among multiple objectives. Dowling et al. (2016) mentioned that dimensionality and ambiguity are two fundamental problems in multi-stakeholder decision-making and optimization. Dimensionality restricts the number of conflicting objectives for multi-objective optimization. Ambiguity affects the finding of true trade-offs between conflicting objectives and stakeholder priorities. Therefore, different techniques to weighting and aggregation of multiple-criteria (i.e., indicators) for sustainability decision-making have been applied and studied (Wang et al. 2009; Buchholz et al. 2009; Rowley et al. 2012). As examples, Dowling et al. (2016) and Zavala (2016) proposed a framework to compute compromise solutions that balance out conflicting priorities of multiple stakeholders on multiple objectives by minimizing average and worst-case stakeholder dissatisfactions. Smith and Ruiz-Mercado (2014) proposed that stakeholders develop relationships between objectives to determine the weighting factors for different situations, allowing for explicit representations of the implied trade-offs to obtain a compromise solution. Other contributions proposed the use of geometric and Euclidean means for normalizing and aggregating multiple indicators (Sikdar 2009; Sikdar et al. 2012).

In this work, the GREENSCOPE framework is used in a MCDM and implemented for designing and optimizing processes with multiple sustainability objectives. The MCDM implementation method is shown in Fig. 1, for supporting stakeholders in the identification and implementation of sustainability requirements and process design specifications (which can also be used for sustainability assessment and optimization).
Fig. 1

A schematic algorithm describing the implementation of MCDM in sustainable design for a multi-input–output processing network

During the first step of MCDM, on the basis of an already available base case design, a processing network is established by including all relevant design variations of the base case. This results in a multi-feedstock multi-technology multi-product system that can be represented by a network model which connects the different feedstock materials, available technologies, possible product/waste sinks, and technology services. These structural design variations provide the constraints of the multi-objective optimization problem as will be described in “Formulation of the optimization problem” section.

As the sustainability goals are established and prioritized by the stakeholders (i.e., indicators identification, and weighting), the sustainability of the processes as a network structure can be directly assessed using the GREENSCOPE indicators. Alternatively, a structure optimization can be performed (instead of evaluating all possibilities) in which the restrictions and sustainability goals are considered (“Formulation of the optimization problem” section) to build feasible structures.

Since GREENSCOPE provides many indicators to describe sustainability performance, the stakeholders can select those indicators describing the key factors which account for the bulk of their sustainability performance. After reviewing the GREENSCOPE indicators (Ruiz-Mercado et al. 2012a) a total of 21 indicators, as described in Table 1, were selected as the most relevant sustainability criteria for the proposed decision-making approach and its corresponding case study application. The selected criteria cover different aspects of the three dimensions of sustainability (economic, environmental, and social). In the particular case of the social dimension, it is addressed by environmental indicators assessing two relevant social issues (i.e., “human health and safety” and “working exposure conditions”) for biojet fuel production (Pashaei-Kamali et al. 2018), which are: Health hazard—Irritation factor; Health hazard—Chronic toxicity factor; and Safety hazard—Acute toxicity (see Table 1). Once the potential design alternatives are evaluated and the corresponding indicator scores are obtained, the GREENSCOPE results are normalized as described in Eq. (1).
Table 1

Selected key indicators for sustainability performance assessment in terms of efficiency, environmental, energy, and economic areas


4 Indicator


5 Symbol

6 Best-case

7 Worst-case


Eff 1

Reaction mass efficiency




Eff 2

Mass intensity




Eff 3

Effective mass yield




Eff 4

Carbon efficiency




Eff 5

Renewability-material index




Eff 6

Fractional water consumption



2.95 m3/kg


Env 1

Health hazard, irritation factor



1.00E + 06 m3/kg

Env 2

Health hazard, chronic toxicity factor

HHchronic toxicity


1.00E + 07 m3/kg

Env 3

Safety hazard, acute toxicity

SHacute tox.


1.00E + 05 m3/kg

Env 4

Specific toxic release



All waste is TRIa

Env 5

Global warming potential



All waste is at least 1 CO2e

Env 6

Aquatic oxygen demand potential

WPO2 dem.


All non-water waste is at least 1 kg O2 demand

Env 7

Specific solid waste mass

m s, spec.


All types of solid waste are released

Env 8

Specific liquid waste volume

V l, spec.


All liquid releases are rated as waste


En 1

Specific energy intensity



1949 MJ/kg

En 2

Renewability energy index




En 3

Renewability energy index





Econ 1

Discounted payback period



11 year

Econ 2

Turnover ratio




Econ 3

Specific raw material cost


0.1b TPCbest-case


Econ 4

Specific energy cost

C E, spec.

Consumed energy from cheapest source (coal) @ $1.72 × 10−6/kJ/TPCbest-case

Consumed energy from expensive source (electricity) @ $1.68 × 10−5/kJ/TPCworst-case

More details regarding indexes and their limits/assumptions (best-case and worst-case) can be found in Ruiz-Mercado et al. (2012a, b). The assumption of DPBP is explained in the “Problem formulation and solution approach” section

aTRI: EPA’s toxic release inventory

bTPC: total production cost

$${\text{Percent Score}} = \frac{{{\text{Actual}} - {\text{Worst}}}}{{{\text{Best}} - {\text{Worst}}}} \times 100$$

In here, it should be highlighted that the limits used as the best and worst-case scenarios can be assigned by the nature of the predominant variables or parameters (e.g., some cannot be lower than 0) represented by each indicator. These default limits are described in Table 1; for each indicator, these boundary values or references can be customized per stakeholder needs and process characteristics (e.g., a project is evaluated for 10 years, therefore the discounted payback period (DPBP) has a limit of 11 years). Such tricks allow the normalization of the target dimensions on the interval [0,1].

For comparing and ranking alternatives and identifying the most favored design variations or arrangements from the group of non-dominated alternatives, a multi-objective function is then created. Decision-making methodologies such as weighting and aggregation are useful in building a single objective function that can be implemented in an optimizer. As GREENSCOPE groups indicators in the four E’s areas, it is natural to aggregate indicators based on these areas and to consider each of the areas as one of the objective functions to be optimized. Hence, the vector-valued objective function (Of) that will be considered in this work is expressed as shown in Eq. (2), where \(a_{1,} a_{2} ,a_{3}\), and \(a_{4}\) are the individual weightings of the sustainable objectives related to the preferences of the stakeholders; Eff, En, Env, and Econ, are the aggregation of the different indicators considered from the GREENSCOPE methodology in the efficiency, energy, environmental, and economic perspectives, respectively.
$${\text{Of}} = - (a_{1} {\text{Eff}} + a_{2} {\text{En}} + a_{3} {\text{Env}} + a_{4} {\text{Econ}})$$
As described by Eqs. (3)–(6), aggregation within the areas is performed by linearly combining the GREENSCOPE indicators included in them. In these equations, the i index is referred to the indicators described in Table 1 and ai,j represents the relative importance of one index between the others in the corresponding criteria (E’s):
$${\text{Eff}} = \mathop \sum \limits_{i = 1}^{6} a_{1,i} {\text{Eff}}_{i}$$
$${\text{En}} = \mathop \sum \limits_{{{\text{i}} = 1}}^{3} a_{2,i} {\text{En}}_{i}$$
$${\text{Env}} = \mathop \sum \limits_{{{\text{i}} = 1}}^{8} a_{3,i} {\text{Env}}_{i}$$
$${\text{Econ}} = \mathop \sum \limits_{i = 1}^{4} a_{4,i} {\text{Econ}}_{i}$$

For superstructure optimization, the final arrangement is obtained (together with the objective function scalar value) by minimizing Eq. (2). By considering different combinations of the individual weights, different problem solutions according to multiple stakeholder opinions are found. If desired, construction of the Pareto frontier can be used to analyze the trade-offs among the different objective functions.

However, even if only a single combination of the weights \(a_{i}\) is used, many alternative solutions may be of interest to the stakeholders. These alternative solutions may come from pathways that result in the same aggregate sustainability score or pathways that have a relatively lower score than the optimal but, it is not significantly distinct. As an example, consider two pathways whose sustainability scores differ by only 1% or less, there is not a significant difference from the sustainability point of view, and yet stakeholders may prefer the non-mathematically optimal pathway. From these alternative solutions, stakeholders can make the final decision and provide the design specifications that improve the initial base case design.

Formulation of the optimization problem

Figure 2 presents a superstructure depicting a generalized production chain. As shown, different feedstock sources (F1 to Fk) can be used and upgraded into different final products through a sequence of processing technologies. Following the work by Kim et al. (2013), alternative technologies are grouped in categories (C1 to Cn) where each category is roughly defined by the equivalence or not of their main outlet streams (e.g., similar generated product). Technologies in downstream categories upgrade the products from main technologies in upstream categories into other (more) valuable products. This upgrading ends up in the production of market products (main and secondary products in Fig. 2) or other intermediate products. In this work, a special category “Service Technologies” (ST1 to STs) is considered to account for the possibility of using a priori residual streams to satisfy utility needs (e.g., steam, energy, hydrogen, etc.). If these service technologies are not used, then the required service stream would need to be purchased from the market, and the residual stream is considered as an emission. In addition, emissions are also generated by any technology in the processing of their corresponding feeds and intermediate streams that cannot be processed by any of the available technologies into valuable products. In other words, residual streams generated by the technologies account as emissions only if they cannot be processed by downstream or service technologies.
Fig. 2

Generalized superstructure for a multiple-input–multiple-output processing network

For the case study here, i.e., biojet fuel production, all technologies have already been designed and simulated offline (i.e., with the aid of a process simulator) for comparable scales and products qualities (Cornelio da Silva 2016; Santos et al. 2017). Hence, each feedstock inlet (or main product outlet mass flowrates) and the activity of each technology block are considered parameters and not variables of the problem. The variables of the problem are binaries that indicate whether technologies and the connections between them participate in the processing network or not.

Under the previous assumptions, the superstructure in Fig. 2 can be cast into a network in which nodes are used to represent: (1) feedstocks, reactants, service streams, and intermediates; (2) different technology options for upgrade of the feedstocks and intermediates; (3) technology options for provision of service streams; (4) sinks for main (target) products, secondary (market) products, and emissions (streams that have to be disposed). Arcs connecting the nodes carry out the information of the connections that are allowed in the superstructure.

Figure 3 shows a prototypical network that can be built from a process superstructure as the one in Fig. 2. This network contains all the elements that are of interest for this study and is used as an example to show how mass balances and node connections can be formulated and, at the same time, to establish the notation that will be used throughout the manuscript. As shown, the network is subdivided into six subsections: Start, Feedstocks, Main Processing Technologies (Category 1 and 2), Market and Service Technologies, and End. Square nodes are used to represent technologies; round nodes are used to represent feedstocks or intermediate compounds; and oval nodes represent feed, emissions, and products. In the Start section, a dummy input node (Y0) establishes a formal inlet for the feedstocks. In the Feedstocks section, two different feedstocks (F1, F2) are considered for conversion. These feedstocks are connected to inlet intermediate nodes (I1 and I2 nodes in the figure), which are nodes that are only connected to one technology. These nodes are not required in the formulation of the optimization problem but help in organizing the information. After the inlet intermediate nodes, the processing technologies nodes are placed. In the Main Processing Technologies (Categories 1 and 2), different technologies for the different categories (C1T1, C1T2, C2T1) can be used to convert the feedstocks into Main Product, Secondary Products, and Emissions.
Fig. 3

The prototypical network used as an example for the problem formulation

Technology C1T1 requires reactants R1 that can be either produced in-house by processing by-product I3 in Service Technology ST1 or directly purchased from the market MT1. Service Technologies are modeled similarly to the Main Processing Technologies; each service technology is assigned one and only one Market counterpart in the Market and Service Technologies subsection. Furthermore, one service technology (or market) is considered per Main Process Technologies (independent if the reactants are the same). If, for example, R1 is purchased from the market, then by-product I3 should be disposed of, and an emission is generated (X15–9). A similar approach may be used to include requirements from service streams. In a similar example, Technology C1T2 produces the Main Product (I4), a Secondary Market Product (I6), a by-product (I5) and emissions (I7). The by-product can be either disposed of, thus generating an emission, or processed by a Technology in Category 2 (C2T1), which, in this example, also results in the production of the main product (I4). To evaluate the emissions once a Main Processing Technology in a secondary (or further) level is not used (i.e., there is no further processing of flows from previous processing technologies, therefore they are considered emissions), a node that represents the inexistence of the considered processing technology, directly connected to the emission intermediate node, is included (NO–C2T1). A node of this type is included per each process that uses previous technologies outputs as input streams.

For computational bookkeeping, nodes are numbered sequentially with the following scheme: Feed nodes, Intermediate nodes, Reactant nodes, Category 1 Technologies, Category 2 Technologies, Service Technologies, Corresponding Market Technologies, Main Product End node, Secondary Products End nodes, and Emissions End node. Each node is assigned a binary variable \(y\left( {\text{node}} \right)\) where \(y\left( {\text{node}} \right) = 1\) if the node is active in the network and \(y\left( {\text{node}} \right) = 0\) if not.

Technology, feeds, ends, and component nodes are linked through arcs which are also assigned a binary \(x\left( {{\text{outlet node}} - {\text{inlet node}}} \right)\), for which \(x\left( {{\text{outlet node}} - {\text{inlet node}}} \right) = 1\) if both outlet and inlet nodes are active, otherwise \(x\left( {{\text{outlet node}} - {\text{inlet node}}} \right) = 0.\)

If the optimization problem will retrieve a solution in which only one feedstock and one technological pathway are chosen (i.e., mixing feedstocks or parallel pathways are not allowed), the Eqs. (7)–(27) can be used to model the connections between arcs and nodes of the whole network (equations for the example depicted in Fig. 3 are given in parentheses).
  • One and only one of the \(F\) feedstocks is allowed (no trivial solution)
    $$\begin{aligned} & y_{0} = \mathop \sum \limits_{f = 1}^{F} y_{f} = 1 \\ & \left( {y_{1} + y_{2} = 1} \right) \\ \end{aligned}$$
  • If \(y_{f}\) is active, then it is connected to the inlet port

    $$\begin{aligned} & x_{0 - f} = y_{f} \forall f\quad (F\,{\text{equations}}) \\ & \left( {x_{0 - 1} = y_{1} \,\&\, x_{0 - 2} = y_{2} } \right) \\ \end{aligned}$$
  • Only one technological pathway is allowed

    This condition is imposed by: (1) restricting the number of outlets from the active feed node to be one (Eq. 9) and (2) by allowing only one inlet intermediate to the technologies (Eq. 10). In these equations, \(nC1\) is the amount of the technologies included in category 1:
    $$\begin{aligned} & \mathop \sum \limits_{{{\text{int}} = 1}}^{nC1} x_{{f - {\text{int}}}} = y_{f} \quad \forall f\quad (F\, {\text{equations}}) \\ & \left( {x_{1 - 3} + x_{1 - 4} = y_{1 } \,\&\, x_{2 - 3} + x_{2 - 4} = y_{2} } \right) \\ \end{aligned}$$
    $$\begin{aligned} & \mathop \sum \limits_{f = 1}^{F} x_{{f - {\text{int}}}} = y_{\text{int}} \forall\, {\text{int }} \in \left[ {1, nC1} \right]\quad (nC1\, {\text{equations}}) \\ & \left( {x_{1 - 3} + x_{2 - 3} = y_{3 } \,\&\, x_{1 - 4} + x_{2 - 4} = y_{4 } } \right) \\ \end{aligned}$$
    By definition, these nodes have only one outlet: which is active if the intermediate node is active
    $$\begin{aligned} & y_{\text{int}} = x_{{{\text{int}} - {\text{tech}}}} \quad \forall\, {\text{tech}} \in \left[ {1, nC1} \right]\quad ( nC1\,{\text{equations}}) \\ & (y_{3} = x_{3 - 11} \,\&\, y_{4} = x_{4 - 12} ) \\ \end{aligned}$$
  • Technology nodes

    A node is active only if the arc connecting the technology with a key intermediate is active (Eq. 12). If a technology node is active, all its additional inlet streams are active (Eq. 13), all its outlet streams are active (Eq. 14), and there can be only one technology between competitive ones for a class after the first, if any of them exist (i.e., a given technology or a nonexistence of a given technology can exist as nodes; Eq. 15). In these equations, \(T\) represents the total number of technology nodes in the network (including services), \(N_{{{\text{inlet}}.{\text{st}} - {\text{tech}}}}\) is the number of the inlet streams required by a technology, and the summation is carried over the inlets (both reactants/services or intermediates).
    $$\begin{aligned} & y_{\text{tech}} = x_{{{\text{key int}} - {\text{tech}}}}\quad \forall {\text{tech}} \in \left[ {1, T} \right]\quad ( T\,{\text{equations}}) \\ & (y_{11} = x_{3 - 11} \,\&\, y_{12} = x_{4 - 12} \,\&\, y_{13} = x_{7 - 13} \,\&\, y_{14} = x_{5 - 14} ) \\ \end{aligned}$$
    $$\begin{aligned} & N_{{{\text{inlet}}.{\text{st}} - {\text{tech}}}} y_{\text{tech}} = \mathop \sum \limits_{\text{inlet}}^{{}} x_{{{\text{inlet}} - {\text{tech}}}} \,\forall {\text{tech}} \in \left[ {1, T} \right](T\,{\text{equation}}) \\ & 2y_{11} = x_{10 - 11} + x_{3 - 11} \,\&\, y_{12} = x_{4 - 12} \,\&\, y_{13} = x_{7 - 13} \,\&\, y_{14} = x_{5 - 14} \\ \end{aligned}$$
    $$\begin{aligned} & N_{{{\text{outlet}}.{\text{st}} - {\text{tech}}}} y_{\text{tech}} = \mathop \sum \limits_{\text{outlet}}^{{}} x_{{{\text{outlet}} - {\text{tech}}}} \forall {\text{tech}} \in \left[ {1, T} \right] (T\,{\text{equations}}) \\ & \left( \begin{aligned} 3y_{11} = x_{11 - 5} + x_{11 - 6} + x_{11 - 7} \,\&\, 4y_{12} = x_{12 - 6} + x_{12 - 7} \,\&\, \hfill \\ \quad +\, x_{12 - 8} + x_{12 - 9} \,\&\, y_{13} = x_{13 - 6} \,\&\, y_{14} = x_{14 - 9} \hfill \\ \end{aligned} \right) \\ \end{aligned}$$
    $$\begin{aligned} & y_{\text{inlet}} = \mathop \sum \limits_{\text{tech}}^{{}} x_{{{\text{outlet}} - {\text{tech}}}} \forall {\text{compet}}\_{\text{tech}} \in \left[ {1, J} \right]({\text{varying}}\,{\text{number}}\,{\text{of}}\,{\text{eqs}} .) \\ & \left( {x_{7 - 13} + x_{7 - 14} = y_{7} } \right) \\ \end{aligned}$$
  • Intermediate nodes

    The node is active if one of the possible inlets is active (Eq. 16). Furthermore, Eq. (17) prevents the node to be active if none of the possible arcs connecting with the intermediate node is active.
    $$y_{\text{int}} \ge x_{\text{tech}}-{\text{int}}\quad \forall {\text{int}} \in \left[ {1, nI} \right] \quad \forall \; {\text{tech}}\;{\text{producing }}\;\text{int} \;\left( {{\text{varying}}\;{\text{ number}}\;{\text{ of }}\;{\text{eqs}} .} \right)$$
    $$y_{\text{int}} \le \mathop \sum \limits_{{{\text{tech s}} . {\text{t tech inlet to y}}_{\text{int}} }} x_{\text{tech}}-{\text{int}} \quad \forall int \in \left[ {3, nI} \right]\;(nI - 2\;{\text{equations}})$$
    $$\begin{aligned} & ({\text{Int}} . {\text{ 3 Node 5}}{:}\,y_{5} \ge x_{11 - 5} \,\&\, y_{5} \le x_{11 - 5} ; \\ & {\text{Int}} . {\text{ 4 Node 6}}{:}\,y_{6} \ge x_{11 - 6} \,\&\, y_{6} \ge x_{12 - 6} \,\&\, y_{6} \ge x_{13 - 6} \,\&\, y_{6} \le x_{11 - 6} + x_{12 - 6} + x_{13 - 6} ; \\ & {\text{Int}} . {\text{ 5 Node 7}}{:}\,y_{7} \ge x_{11 - 7} \,\&\, y_{7} \ge x_{12 - 7} \,\&\, y_{7} \le x_{11 - 7} + x_{12 - 7} ; \\ & {\text{Int}} . {\text{ 6 Node 8:}}\,y_{8} \ge x_{12 - 8} \,\&\, y_{8} \le x_{12 - 8} ; \\ & {\text{Int}} . {\text{ 7 Node 9:}}\,y_{9} \ge x_{12 - 9} \,\&\, y_{9} \ge x_{14 - 9} \,\&\, \\ & \quad y_{9} \ge x_{15 - 9} \,\&\, y_{9} \ge x_{5 - 9} \,\&\, y_{9} \le x_{12 - 9} + x_{14 - 9} + x_{15 - 9} + x_{5 - 9} ) \\ \end{aligned}$$
    An intermediate node terminates in the main product node, one of the secondary market product nodes, or is further upgraded in other technology. If the intermediate node is not further upgraded, then it generates an emission. These cases are considered by activating the corresponding arcs and nodes:
    $$y_{{{\text{int}}\, =\, {\text{product}}}}\, =\, x_{{{\text{int}} = {\text{product}} - {\text{product node}}}} \;\left( {\text{one equation}} \right)$$
    $$y_{{{\text{int}}\, =\, {\text{by}} - {\text{product}}}}\, =\, x_{{{\text{int}} = {\text{byproduct}} - {\text{by }}\;{\text{product }}\;{\text{node}}}} {\times}\;\left( {{\text{varying}}\;{\text{ number }}\;{\text{of}}\;{\text{ eqs}} .} \right)$$
    $$y_{{{\text{int}}\, = \,{\text{all}}\; {\text{others}}}} = x_{\text{int}} - {\text{emission }}\;{\text{node}} + \mathop \sum \limits_{tech} x_{int - tech} \;\left( {{\text{one}}\;{\text{ equation}}} \right)$$
    $$\begin{aligned} & ({\text{Int}} . {\text{ 3 Node 5 is processed in a service technology or not:}}\,y_{5} = x_{5 - 9} + x_{5 - 15} ; \\ & {\text{Int}} . {\text{ 4 Node 6 is the main product:}}\,y_{6} = x_{6 - 17} ; \\ & {\text{Int}} . {\text{ 5 Node 7 is processed:}}\,y_{7} = x_{7 - 13} + x_{7 - 14} ; \\ & {\text{Int}} . {\text{ 6 Node 8 is a secondary product:}}\,y_{8} = x_{8 - 18} ; \\ & {\text{Int}} . {\text{ 7 Node 9 sends all the inputs to the emission node:}}\,y_{9} = x_{9 - 19} ) \\ \end{aligned}$$
  • Reactant/services nodes

    The node is active if the arc connecting it to a technology is active. As in intermediate nodes, an equation is added to prevent its activation if none of the technologies that it requires is active.
    $$y_{R} \ge x_{{R-{\text{tech}}}} \forall R \in \left[ {1, nR} \right]\quad \forall \;{\text{tech}}\;{\text{requiring}}\; R\quad ( nR\;{\text{equations}})$$
    $$\begin{aligned} & y_{R} \le \mathop \sum \limits_{{{\text{tech}}\; {\text{s}} . {\text{t}}\; {\text{tech }}\;{\text{requires }}R }} x_{{R-{\text{tech}}}} \quad \forall R \in \left[ {1, nR} \right]\quad (nR\;{\text{equations}}) \\ & (y_{10} \ge x_{10 - 11} \,\&\, y_{10} \le x_{10 - 11} ) \\ \end{aligned}$$
    A reactant can be provided from the market or a service technology but not both
    $$\begin{aligned} & y_{R} \le x_{STech - R} + x_{MTech - R} \quad \forall R \in \left[ {1, nR} \right]\quad (nR\;{\text{equations}}) \\ & y_{10} = x_{15 - 10} + x_{16 - 10} \\ \end{aligned}$$
    If service is required, all its outputs and inputs related to the active technology are activated
    $$\begin{aligned} & N_{{{\text{outlets}},\,ST}} y_{s} + N_{{{\text{inputs}},\,ST}} y_{s} = \mathop \sum \limits_{\text{outlets}} x_{{ST - {\text{int}}}} \quad \forall ST \in \left[ {1, nST} \right]\quad \left( {nST\;{\text{equations}}} \right) \\ & 3y_{15} = x_{15 - 10} + x_{15 - 9} + x_{5 - 15} \\ \end{aligned}$$
  • Market Node: is active if the arc connecting to the node is active
    $$\begin{aligned} & y_{M} = x_{MT - R} \quad \forall R \in \left[ {1, nR} \right]\quad (nR\;{\text{equations}}) \\ & (y_{16} = x_{16 - 10} ) \\ \end{aligned}$$
As mentioned above, offline process simulations are available for all cases under study. The results of these simulations are cast in the technology matrix \(A\), where each column of \(A\) represents a technology-feed-arc combination (e.g, gasification-Macauba-x11–5) and each row of \(A\) a compound \(e\). The compounds included are those that belong to the feedstocks, products, services, intermediates, reactants/services, and emissions; the elements of \(A\left( {e\left( {Tech - F - X_{i - j} } \right)} \right)\) are the mass flow of each component. The total amount of product, by-products, and emissions is obtained by combining the information in A with the results of solving Eqs. 725 in the corresponding node (e.g, the total emission is the sum of the existing inputs arcs, and the respective elements, in the node 9). The rationale is as follows: if an arc is active, the flow information of a given compound e in that arc (f(e)ij; xij arc) is obtained by multiplying a technology-feedstock-arc combination, the input flow base calculation (B), and the existence of the corresponding arc. The total outputs are estimated in function of the sum of each element in the active arcs over each compound at the corresponding intermediate node (in this case m) (Eq. 26).
$$f\left( {\text{all}} \right)^{\text{out}} = \mathop \sum \limits_{e} \mathop \sum \limits_{i} x_{i - m} \cdot A\left( {e\left( {{\text{Tech}} - F - {\text{arc}}} \right)} \right)_{i - m} \cdot B$$
At this point, it must be highlighted that GREENSCOPE indexes are primarily functions of input/output stream information, therefore arcs that are not connected to an ending intermediate node (such as nodes 6, 8, and 9 in Fig. 3) are not needed for objective function evaluation. Cost indexes are strongly dependent on the node information, i.e., the existence of a given technology. However, environmental indexes strongly depend on the components of a given output stream, therefore information on the individual components of each stream is relevant. In the following equations, \(f\left( e \right)^{\text{in}}\) represents the inlet flowrate of each compound, which is used to compute the input stream as described in Eq. (27); \(f\left( {e,p} \right)^{\text{out}}\) indicates the number of compounds that are part of the product produced by the network; \(f\left( {e,sp} \right)^{\text{out}}\) indicates the number of compounds that are part of other market product(s) produced by the network; and \(f\left( {e,{\text{em}}} \right)^{\text{out}}\) the number of compounds that are part of the emissions produced by the network.
$$f^{\text{in}} = \mathop \sum \limits_{e} \mathop \sum \limits_{F}^{{}} f\left( e \right)^{\text{in}} y\left( F \right)\quad \left( {1{\text{ equation}}} \right)$$
\(f\left( e \right)^{\text{in}} , f\left( {e,p} \right)^{\text{out}} ,f\left( {e,sp} \right)^{\text{out}},\) and \(f\left( {e,em} \right)^{\text{out}}\) are the inputs required by GREENSCOPE for evaluation of the indicators in Table 1. Thus, the optimization problem is formally stated as:
$$\begin{aligned} & \mathop {\hbox{max} }\limits_{x,y,z} a_{1} Eff\left( {f\left( e \right)^{\text{in}} ,f\left( {e,p} \right)^{\text{out}} ,f\left( {e,sp} \right)^{\text{out}} , f\left( {e,em} \right)^{\text{out}} } \right) \\ & \quad \quad + a_{2} En\left( {f\left( e \right)^{\text{in}} ,f\left( {e,p} \right)^{\text{out}} ,f\left( {e,sp} \right)^{\text{out}} , f\left( {e,em} \right)^{\text{out}} } \right) \\ & \quad \quad + a_{3} Env\left( {f\left( e \right)^{\text{in}} ,f\left( {e,p} \right)^{\text{out}} ,f\left( {e,sp} \right)^{\text{out}} , f\left( {e,em} \right)^{\text{out}} } \right) \\ & \quad \quad + a_{4} Econ\left( {f\left( e \right)^{\text{in}} ,f\left( {e,p} \right)^{\text{out}} ,f\left( {e,sp} \right)^{\text{out}} , f\left( {e,em} \right)^{\text{out}} } \right) \\ & {\text{s}}.{\text{t}}\quad {\text{Eqs}}.\, \left( 7 \right) - \left( {27} \right)\;\& \;y_{i} \wedge x_{i - j} \in {\mathbb{Z}}:\text{ }0 \le y_{i} \le 1 \wedge 0 \le x_{i - j} \le 1 \\ \end{aligned}$$
where \({\text{Eff}}\), \({\text{En}}\), \({\text{Env}}\), \({\text{Econ}}\), and the weights \(a\) were defined in “Multiple perspectives of stakeholders” section. The approach used to solve the optimization problem is developed in “Case study” section, after an overview of the case study.

Case study

Geographical and temporal context

The multi-criteria optimization problem formulated in this contribution (see “Formulation of the optimization problem” section), for sustainable biobased production chains, is exemplified with the production chains of biojet fuel synthesis aiming to supply 10% of the 2020 demand of two major Brazilian airports in São Paulo and Rio de Janeiro, i.e., Guarulhos and Galeão. The Brazilian case study is well-established in the bio-economy literature due to its long history with the biofuels sector (i.e., bioethanol) along with both a strong national policy promoting biofuels and renewable energy and a mature sugarcane industry (Herreras Martinez et al. 2013; Alves et al. 2017). In relation to the aviation sector, Brazil has the third largest domestic flights market and it is also one of the fastest growing markets. This set of conditions makes Brazil an attractive country to explore the development of sustainable biojet fuel production chains at commercial scale (Santos et al. 2017).

The projected annual production of biojet fuel to cover 10% of the 2020 demand for these two airports is 210 kton (Jansen 2012; Cortez 2014; Alves et al. 2017; Santos et al. 2017). Furthermore, two regions are identified in the literature as high-potential locations for the biorefining facilities based on biomass (see “Description of processes and technologies for biojet fuel production” section for details on feedstocks) and infrastructure availability, i.e., Minas Gerais state (MG) and Rio Grande do Sul state (RG) (Alves et al. 2017; Cornelio da Silva 2016). The pine-based biorefinery is assumed to be in the Minas Gerais state, while the macauba-based biorefinery is expected to be in the Rio Grande do Sul state; eucalyptus-based biorefinery is the only one suitable for both locations. Finally, all processing facilities are assumed to have an operating time of 330 days per year.

Description of processes and technologies for biojet fuel production

The biojet fuel production supply chains considered in this study consist of four main technological elements: (1) biobased feedstocks, (2) thermochemical conversion of biomass to biojet fuel (and to biobased co-products), (3) co-products upgrading, and (4) auxiliary sections for in situ production raw materials and utilities. The specifics of each technological element are selected based on the literature review and on previous studies for ex-ante techno-economic and environmental assessment of biojet fuel production technologies as described below.

Biomass residues from forestry, wood, and oil crops are consistently reported to lead to lower production costs and GHG emissions (per unit of energy output of biojet fuel), when thermochemical processes are used, as compared to agricultural residues (de Jong et al. 2015; Alves et al. 2017; Tzanetis et al. 2017; Santos et al. 2017). Within the category of forestry residues, pine and eucalyptus have been found to have the best techno-economic and environmental performances, while for the case of oil crops residues, macauba (a palm tree) has been reported as a promising feedstock (Alves et al. 2017; Cornelio da Silva 2016).

Thermochemical conversion of biomass residues for biojet fuel production is possible through three widely discussed technologies, namely: Gasification Fischer–Tropsch, FP, and HTL. GFT and FP are mature technologies and are ASTM approved for 50% blending with fossil jet fuel (Radich 2015) while HTL is a rather novel technology, still under development, but with a large potential given its higher yields and expected modest capital costs (de Jong et al. 2015; Tzanetis et al. 2017).

GFT (see Fig. 4.) is a two-step process where biomass is first gasified at high temperatures in presence of oxygen to produce synthesis gas (syngas, a stream rich on H2 and CO) which subsequently undergoes the Fischer–Tropsch reaction to convert syngas into new hydrocarbons (in general a paraffinic kerosene-like mix) and water. Data for modeling of biomass gasification, i.e., pine, eucalyptus, and macauba (Seo et al. 2015; Nipattummakul et al. 2011), as well as for Fischer–Tropsch synthesis on Co catalyst (de la Ree 2011) are based on data adapted from experimental studies (as shown in Cornelio da Silva 2016; Tzanetis et al. 2017; Santos et al. 2017). Additional technological requirements of the GFT process include biomass preparation (i.e., drying and preheating), syngas cleaning, and refining (i.e., gas, liquid and solids separation: cyclone, tar scrubber, methylethylamine (MEA) extraction, granular activated carbon (GAC), Steam Methane Reforming (SMR), Pressure Swing Adsorption (PSA)), hydrocracking, and hydrocarbons fractionation into light naphtha (C5–C8), biojet fuel (C9–C14), diesel (C15–C18), and wax (C19+) (Tijmensen et al. 2002; Hamelinck and Faaij 2006; Guettel et al. 2008; Bouchy et al. 2009; Lubwama 2009; Swanson et al. 2010; König et al. 2015; Tzanetis et al. 2017; Santos et al. 2017).
Fig. 4

Process flow diagram for the gasification Fischer–Tropsch process

The FP (see Fig. 5) process has three main sections: fast pyrolysis, hydrotreating, and hydrocracking. In the first section, dried biomass is thermochemically decomposed, in the pyrolysis reactor (at high temperatures, for a short time -in the order of seconds- and in absence of air), into non-condensable gases (NCG), bio-oil, and solids (Jones et al. 2013; Tews et al. 2014; Tzanetis et al. 2017). Data on yields and compositions for each phase from eucalyptus, pine, and macauba are adapted from experimental data reported in the literature (Garcia-Perez et al. 2008; Oasmaa et al. 2009; Kim et al. 2010; Abnisa et al. 2011; Jin 2014). Solids (including char, ash, and sand) are used for combustion and heat recovery. The NCG stream is partially recirculated to the reactor and the rest is sent to the H2 plant to produce the hydrogen required in upgrading steps. The pyrolysis bio-oil is submitted to a hydrotreating unit to decrease the O/C ratio (by hydrogen addition). The obtained hydrocarbons are then distilled into 5 cuts: offgas, light naphtha, biojet, diesel, and wax. The latter fraction can be hydrocracked to further increase the overall yield to biojet fuel (Huber et al. 2006), while the offgas is used to produce the hydrogen required in the SMR, water–gas shift reaction, and in the partial oxidation of other light organic molecules.
Fig. 5

Process flow diagram for the fast pyrolysis process

HTL (see Fig. 6) is a more flexible process that is carried out in presence of water, thereby making use of biomass moisture to avoid the high-energy and capital-intensive drying step required in other thermochemical technologies (Tzanetis et al. 2017). The HTL reaction takes place with water at subcritical conditions decomposing the biomass into three phases gas, liquid, and solids, similarly as for FP. Data on yields and compositions are adapted from the literature (Zhu et al. 2014; Singh et al. 2015; Chan et al. 2015; Tzanetis et al. 2017). The gas phase is used for H2 production while the solids (char and ash) are combusted in the co-generation unit for heat and electricity production. The liquid fraction is composed of an organic phase (i.e., bio-oil) and an aqueous phase. After the liquid–liquid separation of these two phases, the bio-oil undergoes a catalytic hydrotreating process (Jones et al. 2014; Tews et al. 2014; Zhu et al. 2014) to lower the O/C ratio and increase the heating value of the HTL-based bio-oil. The hydrotreated bio-oil is first flashed to remove water and light hydrocarbons (C1–C3) and then fractionated into five cuts: offgas, light naphtha, biojet fuel, diesel, and heavy compounds (wax) (DES, DESC 2009). Similarly, as for FP, the heavy compounds fraction can further be cracked using H2 to maximize the biojet yield.
Fig. 6

Process flow diagram for the Hydrothermal Liquefaction process

It is important to note that both biofuels from FP and HTL have cyclic hydrocarbons, both saturated and aromatic, besides paraffinic compounds, which may limit the blending percentage with oil-based jet fuel since the ASTM D1655 standard limits aromatic content to 25% (ASTM D1655 2018). Furthermore, biojet fuel from these two technologies is yet to be approved (Elgowainy et al. 2012).

The three thermochemical processes above described require two major auxiliary processing units, namely: SMR for H2 production, and co-generation for heat and electricity production. The SMR process supplies the hydrogen needs only for the FP and HTL routes via SMR and water–gas shift reactions (Jones et al. 2014; and Elliott et al. 2015), while in the case of the GFT technology, hydrogen requirements are covered by the syngas stream obtained in the biomass gasification. The co-generation section combusts char and offgas with a 20% air excess to supply both steam [high pressure (33 bar), medium pressure (10 bar), and low pressure (2 bar)] and electricity needs (Smith 2005).

Network formulation of the biomass to biojet fuel processes

The processes described in “Description of processes and technologies for biojet fuel production” section can be casted in a network such as the one in Figs. 7 and 8 by first grouping the different unit operations in suitable technology groups and then arrange them into different categories. For doing this, the following assumptions were made:
Fig. 7

FP and HTL processing trains

Fig. 8

GFT processing train

  1. 1.

    For all the technologies under consideration, intermediate products from biomass up to fuels have no commercial value. This implies that all unit operations (reaction, separations) up to the first production of fuels are lumped together and regarded as a technology belonging to Category 1.

  2. 2.

    Operations that are required for cracking long-chain hydrocarbons left over from Category 1 technologies to fuels are lumped into Category 2 technologies.

  3. 3.

    Operations for production of the hydrogen required for conversion of the FP’s and HTL’s oil and cracking are assumed as a service technology in these two (FP and HTL) processing trains. Hence, the option of purchasing it from the market is included. Due to data availability, this is not an option for the GFT processing train, thus H2 is assumed to be always produced “in the house” and the corresponding operations lumped were suitable.

  4. 4.

    Similar considerations are made for the (energy) co-generation operations which are assumed as a service technology in FP and HTL trains.

Figure 9 shows a simplified block diagram containing the division of the processes into different technologies, for the cases under study. Notice that the above divisions are subjective and problem-specific; as an example, another valid network is obtained if the syngas produced in the gasification processing train is considered a marketable product or if several options for its upgrade are included.
Fig. 9

The simplified network model for the processes considered in this study. Main processing technologies 2 and 3 connections have been simplified for a better visual understanding. Similarly, no-cracking processes have also been simplified (NO–C2–T1 and NO–C2–T2)

Based on these diagrams (Figs. 7, 8), the network model in is constructed by assigning a node to each technology and by adding appropriate market nodes for each service technology. The connecting lines are shown in correspond to the streams of compounds as defined in “Formulation of the optimization problem” section; each of these compounds is assigned an intermediate node. Furthermore, instead of directly connecting the technologies in different categories, technologies are connected through these intermediate nodes, and mass balances are performed at these nodes. As three different feedstocks are considered, extra nodes are added for each; the main product is biojet fuel, thus, the node corresponding to this intermediate is linked to the Main Product exit node. The two-secondary products, Diesel and Light Naptha nodes, are linked to the Secondary Product exit nodes. Any other intermediate node is linked to the Emissions exit node.

Formulation of the optimization problem follows the rationale explained in “Formulation of the optimization problem” section. Input–output data for the specific technologies (and some of the relevant information for evaluation of GREENSCOPE indicators) is included in the supplementary material.

Approach to solve the optimization problem

The problem is formulated as a mixed integer nonlinear programming problem (MINLP) given the nonlinearity of the objective function (Eq. 28). Restrictions are linear restrictions involving both continuous and binary variables. In total, the final structure involves the use of 106 binary values (corresponding to arcs and nodes) and 65 linear restrictions. The full combination of the problem leads to a total of 52 feasible configurations. The problem was formulated in Matlab (R2017a) and solved with Tomlab by the branch and bound algorithm minlpbb. The solver was started using the base case (feasible solution) configuration as the initial point, objective function minimum set as − 1 and all tolerances were set to 1E−3. Further information regarding the optimization is given in the supplementary material, General Topics sections. The results reported in the following section correspond to global optimum.

Results and discussion

Sustainability performance for the base case study: FP with eucalyptus feed

As described in “Multiple perspectives of stakeholders” section, a set of 21 GREENSCOPE indicators (listed in Table 1) has been employed for assessing the sustainability performance of potential arrangements of feedstocks, technologies, auxiliary facilities, and products. In addition, these sustainability indicators have provided directions for finding the optimal sustainable structures of such systems under different stakeholder points of view. The indicators calculation requires data entries containing process system data (i.e., equipment cost and utility needs, operating conditions, input/output streams) as well as physicochemical and thermodynamic properties of substances involved in the systematic structure under evaluation. The structured combination under evaluation is delimited for collecting data from all streams entering or leaving the studied subnetwork (see Fig. 9). For comparison of the optimization result, a base case was implemented. In this work, the FP with eucalyptus feed that includes SMR and co-generation of electricity was considered the base case study. This configuration was selected since it was found as the most promising sustainable network from previous studies (da Silva et al. 2016). The input and output stream data, and other stream data such as enthalpy and entropy flow, volumetric flowrate, and stream vapor fraction, are introduced into the GREENSCOPE tool.

Additionally, some compound property data as described in Ruiz-Mercado et al. (2012b) must be entered and the indicator scores can be calculated. Furthermore, stakeholders can identify and select the best-case (100%) and worst-case (0%) scenarios for each indicator to establish the sustainability scale for each indicator as provided in Table 1. Figure 10 represents on a radar diagram the individual scores of all four E’s sustainability indicators in function of the predominant perspective used in the optimization. The center of each radar graph denotes a zero-sustainability value (worst-case scenario) and the external border of the graph signifies a 100% sustainability value (best-case scenario). The continuous green line series represent the individual scores for the above-mentioned base case. The actual values of the different sustainability indicators (in standardized and non-standardized form) are given as supplementary material (Table S5 and Table S6). These were obtained by using an identical calculation base of biomass feed stream (193,056 kg/h) for all case studies. This calculation base was obtained from the base case feed stream data. Table S5 also includes a short description of the non-standardized values reported. A full description of the indicators is in Ruiz-Mercado et al. (2012a, b, 2013).
Fig. 10

GREENSCOPE efficiency (RME, MI, EMY, CE, RIM, FWC), environmental (HHirritation, HHchronic toxicity, SHacute tox., TRs, GWP, WPO2 dem., ms,spec., Vl,spec.), energy (RSEI, RIE, RIEx), and economic (DPBP, TR, CSRM, CE, spec.) indicator results for the solutions that optimize each E area (dashed blue line series). In each diagram, the green line represents the indicator scores for the FP based process with eucalyptus feed as the network base case and as the benchmark. a Environmental, b economic, c energy and d efficiency (material). In addition, e represents a balanced solution (i.e., same relative importance for all four E’s.), and f is an industrial perspective solution (i.e., 50% to economic E, and 25% for both material efficiency and energy E’s, while the environmental E is considered as a constraint to comply with government regulations

Overall, the base case performs well in efficiency (mean average 66%), environmental (average 58%) and energy indicators (average 62%) and poorly in economic ones (27%). A couple of comments are worth mentioning to help to interpret these results. Regarding the economic indicators, we found that Econ 1, the discounted payback period (DPBP), results in 0% score and indicator Econ 2, the turnover ratio (TR), in 8.5%. These figures can be better understood when also considering the criteria selected for best and worst-case scenarios. For example, for DPBP, GREENSCOPE allows 10 years for recovery of the fixed capital investment; then the 0% means that more than 10 years are required. Traditional techno-economical evaluations normally consider a large number of years to evaluate economical feasibilities of large-scale projects. However, promising biobased projects are expected to be economically sustainable in considerable shorter periods, given their high-risk nature, especially at early stages of the project study. As reported in the literature, evaluation limits considerably lower have recently been used (Posada et al. 2016; Santos et al. 2017). Given these considerations, the 10 years DPBP limit was used to identify outperforming processes for projects economically sustainable in the short term.

For the TR indicator, GREENSCOPE considers a value of 4 as the best-case, and 0.2 as the worst-case, however, an absolute value of 0.5, which translates to an 8% score) is considered a benchmark for a chemical industry. Conversely, since biomass is the main feedstock and its relatively low cost, generates a 100% value for the Econ 3, specific raw material cost (CSRM), which is less than the 10% of total production cost (TPC) best-case scenario. Nevertheless, because of the high-energy demand and its corresponding cost, the Econ 4, specific energy cost (CE, spec.), is almost 3.9 times the worst-case scenario, which is the energy cost if all consumed energy is obtained from electricity divided by the TPC worst-case scenario. Regarding the environmental indicators, the values of 0% in Env 7, specific solid waste mass (ms,spec.), and Env 8, specific liquid waste volume (Vl, spec.) have to be interpreted as all generated solid and liquid waste is released to the environment without considering any recovery, recycling, or reuse processing steps. For further work, the absolute values in here can then be used as the worst-case scenario when performing process modifications and technologies and using their initial computed sustainability scores as a basis.

In the next subsection, it will be discussed the sustainability performance for optimized networks under different objective goals from different stakeholder perspectives by assigning different weighting distributions as introduced in “Multiple perspectives of stakeholders” section.

Sustainability performance for optimized networks

The biojet fuel production networks were optimized considering six different priority levels for stakeholders (i.e., Perspective A–F) as described in “Formulation of the optimization problem” section. Results obtained considering different sets of weighting factors for each of the four GREENSCOPE target areas in the six perspectives are discussed below.

Perspective A: maximization of the environmental benefits

Figure 10 shows the individual indicator scores for the solutions that optimize each E (dashed blue line series). The process network that maximizes the environmental benefits, achieving an aggregated score of 72.3% of the best-case scenario, is the one where pine is converted to fuels by GFT reaction with further cracking of long-chain hydrocarbons. Figure 10a shows a comparison with the base case process as described above. The GWP, SH, and ms, spec. indicator scores are improved, while the Env 6, aquatic oxygen demand (WPO2 dem.), presents a slight decline. All the other indicator scores can be considered to remain virtually the same.

All the other processing options considering gasification, regardless of the biomass type of the inclusion (or not) of the cracking unit differ only by 0.05% (aggregated score) from the pine-based option, thus, these can be considered as equally performing processes from the environmental point of view.

In addition, six other options resulted in an aggregated environmental indicator above 70% of the best-case scenario (i.e., within 2% of the optimal solution), thus, these would be regarded as equally good options from the environmental point of view. These options correspond to processes that consider HTL coupled with co-generation and SMR. This can be broadly interpreted as that adding the co-generation and SMR options have lower environmental impacts while adding the cracking option is neutral from the same point of view. The effect of the SMR and co-generation options can be further interpreted as (gaseous) emissions generated by these sub-processes are less pollutant than the intermediate emissions that are used as their feed.

Perspective B: maximization of economic benefits

Figure 10b shows the indicator scores from the process network that maximizes the economic benefits. The top performer with an aggregated score of 29.4% is the option processing pine by FP which does not include the cracking step, co-generation, and SMR units. This result clearly establishes that most of the additional operating units produce a considerable burden in the economic perspective. Nevertheless, the cracking step is the one that produces the lowest effect. The alternative that employs the pine FP followed by the cracking step ranks second with only a 0.18% of difference with respect to the top performer. When compared to the base case scenario indicator scores, the largest improvement is in the Econ 2 (TR, 17.7 vs 8.5%) which indicates the ratio between the total annual revenues and the annualized fixed capital investment (FCI). Although the economic performance of the network is improved, the indicator for the profitability of the system over time (Econ 1) is still zero, signifying that more than 10 years are required to recover the invested capital no matter what potential network arrangement can be achieved.

By evaluating the maximum distance (worst system) with respect to the optimum, the difference in the aggregated score is about 4%, implying that most of the systems score approximately the same for the indexes considered. This trend is highly affected by the fact that all systems score the same for three out of four of the economic indicators (e.g, all systems considered have a DPBP over the limit of 10 years) and therefore this percentage can be assumed biased by this neutral value (i.e., this 4% is relevant independent of the small differences). One important consideration is that FP scored the best results in the economic perspective since the top six system configurations are based on this technology. Additionally, it is observed that options that do not include co-generation and in-house H2 production, rank better than the other ones that include them, suggesting that it is, overall, cheaper to buy these two products from the market rather than their in-house production.

Perspective C: maximization of energy benefits

In terms of optimizing the energy benefits, the best option results from processing eucalyptus by gasification and excluding the extra upgrade. This option results in an aggregated score of 73% of the best-case scenario (100% sustainability). Again, all other gasification options can be considered as equivalent since they differ by less than 1% of the best option. Figure 10c describes the indicator scores from the process network that maximizes the energy benefits and their comparison with the base case network results. It can be noticed that all indicators are improved. However, the renewability energy index indicator (RIE) still exhibits a low score, suggesting an opportunity for improvement in this area (i.e., utilization of more energy from renewable resources).

If all options are ranked it is noticed that processes including FP are always more energetically efficient than the HTL ones. In addition, and as expected, in all cases, the inclusion of co-generation option improves the energy-sustainability indicator scores.

Perspective D: maximization of material efficiency

The process that maximizes the material efficiency is the one that processes eucalyptus by HTL and includes both cracking and in-house production of hydrogen. Such configuration results in an aggregated score of 76%. The pine fed option can be considered as equivalent. As described in Fig. 10d, all indicator scores from the process network that maximizes the material efficiency performance are improved in comparison with the base case network results.

With respect to the general trends, HTL options are always more material-efficient than their FP based equivalents, while gasification options are by far the least efficient, the best among them results in 52% of the best-case scenario).

Perspective E: balanced solution

This case represents the option of giving the same importance to all four E’s. The optimal solution, scoring 58% of the best-case scenario, processes pine by HTL with cracking, co-generation, and SMR. The alternative with no SMR can be considered equivalent since it produces a score difference of only 0.07% with respect to the optimal solution. Interestingly, once all the perspectives are agglomerated, the optimal solution does not correspond to any of the individual evaluations previously analyzed. Furthermore, once it is compared with the base case analysis, the solution shows improvements in the efficiency, energy, and economic areas and a detriment in the environmental area (see Fig. 10e). These previous statements evidence that a win–win situation for all E’s can be rarely achieved.

The in-between solutions show that the HTL-based options perform overall better than the FP based options. However, there is no a clear trend when adding cracking, co-generation or reforming options.

Perspective F: industrial stakeholders perspective

In general, stakeholders do not have an equal preference for the different sustainability areas and their corresponding indicators. In case of industry stakeholders, it is natural to assume that the emphasis is on the economic indicators first, and then on material and energy efficiency. Environmental indicators are usually treated as constraints that need to be satisfied to comply with government regulations. This difference in multiple perspectives of stakeholders can be studied by considering different weights in the objective function. Following a hypothetical scenario, we assumed weights corresponding to 50% contribution to the economic E and 25% contribution for each material efficiency and energy E’s (Labuschagne et al. 2005; Buchholz et al. 2009; and Tahir and Darton 2010). The environmental E was included as a constraint in the problem, this is an average environmental performance greater than 50% of the best-case scenario, was requested for a solution to be optimal. Figure 10f describes the indicator scores from the process network matching the hypothetical industrial perspective scenario and their comparison with the base case network results.

Under these assumptions, the best network solution from the hypothetical industrial perspective is the option of processing pine using fast pyrolysis with no co-generation or steam reforming. In addition, this solution coincides with the second-best option in terms of economic aspects and ranks high in the efficiency solutions. Once again, all options that include gasification can be considered as equivalent. Given the stakeholder consideration, any process ranking high in economic terms (such as in this case) and well positioned in the efficiency and energy terms, would have a high possibility of being selected as a suitable alternative. In fact, the alternative ranking second only has a difference of 0.2% with respect to the optimal configuration (48.3%) and scored in second place with respect to the efficiency perspective.

Regarding the environmental performance constraint, the process identified as the best solution from this industrial perspective results in an average of ~ 56% of the best-case scenario. In here, it is worth mentioning that all the possible processes (i.e., combinations of operations as schematized in Fig. 9) satisfy the greater than 50% constraint. However, if individual indicators instead of averages are considered, none of the processes would satisfy the constraint. In all cases, the most critical indicator is the Aquatic Oxygen Demand, whose largest value among the process combinations is ~ 15%, a value achieved for the combinations: pine-HTL with cracking and SMR options and no-co-generation (second-best industrial option) or pine FP with cracking and SMR options and no-co-generation. Therefore, given the better environmental performance and the small overall difference with optimal solution, it might be argued that the second-best option could in fact be a better processing option.


This contribution provides a method for optimizing a multi-product network system concerning economic, environmental, efficiency, and energy areas (4 E’s) for sustainable processes. This approach gives a broad overview of measuring sustainability and process performance indicators and links every indicator with an element of the multi-product process network. The elements of the facility network are based on a multi-objective optimization model that has been adapted to include the four sustainability objective costs. New objective functions were formalized and integrated into the optimization model. Additionally, the new optimization model for the evaluation and optimization of sustainability in multi-product networks was tested in a biorefinery case study for the management of biomass feedstocks.

The results obtained for this simulated multi-feedstock-technology-product network were illustrated on radar diagrams for each of the four E’s. This clear sustainability dimension scale allows stakeholders to identify optimal trends and network areas with great potential for sustainability improvements. In addition, a systematic sustainability evaluation provides results to aid in understanding the most relevant network aspects, which are significant in influencing the sustainability performance.

The case study results here presented show that using pine biomass to produce aviation biofuels by different technologies allows maximizing environmental and economic benefits. However, for energy and material efficiency, a maximum score was achieved with eucalyptus by GFT and HTL, respectively. By considering an even importance of the E’s, the methodology that processes pine by HTL with secondary processing units and service technologies (i.e., cracking, co-generation, SMR) showed to excel over the different options. Overall, there is a clear tradeoff between technologies at the moment of choosing the one that fits all the E’s requirements. This evidences that a win–win situation for all E’s can be hardly ever reached. In general, adding the co-generation and SMR options has a positive effect on the environmental performance. Moreover, options without co-generation and H2 production show better economic scores than the ones including those processes, suggesting that, overall, it is cheaper to buy these utilities from the market rather than in-house production. In addition, processes including FP are always more energetically efficient when HTL is used. The inclusion of the co-generation process also improves the energy-sustainability indicators.

Meanwhile, many potential combinations fall within 5% difference with respect to the optimal solution and thus, these options can be considered as virtually equivalent and worth for further analysis and improvements through, for example, process optimization. In addition, it should be recognized that the current GFT processes are not modeled at the same level of detail and complexity than the FP and HTL technologies. These differences might be an important aspect to be considered by decision-makers since many assumptions, simplifications, and approximations, at early process design stages, might cause an underestimation of potentially negative economic, environmental, efficiency, and energy impacts.

Finally, this contribution demonstrated the practicability of the proposed methodology for sustainability evaluation, optimization, and decision-making in the context of a multi-product chemical facility by developing a multi-objective optimization model to increase its sustainability.


The views expressed in this contribution are those of the authors and do not represent the views or policies of the U.S. EPA. Mention of trade names, products, or services does not convey, and should not be interpreted as conveying official U.S. EPA approval, endorsement, or recommendation.


Supplementary material

10098_2018_1576_MOESM1_ESM.docx (48 kb)
Supplementary material 1 (DOCX 48 kb)


  1. Abnisa F, Wan Daud WMA, Husin WNW, Sahu JN (2011) Utilization possibilities of palm shell as a source of biomass energy in Malaysia by producing bio-oil in pyrolysis process. Biomass Bioenerg 35:1863–1872CrossRefGoogle Scholar
  2. Alves CM, de Jong S, Bonomi A, van der Wielen LAM, Mussatto SI (2017) Techno-economic assessment of biorefining technologies for aviation biofuels supply chains in Brazil. Biofuels Bioprod Biorefin 11:67–91CrossRefGoogle Scholar
  3. ASTM D1655 (2018) Standard specification for Aviation Turbine Fuels. ASTM International, West Conshohocken. CrossRefGoogle Scholar
  4. Bouchy C, Hastoy G, Guillon E, Martens JA (2009) Fischer–Tropsch waxes upgrading via hydrocracking and selective hydroisomerization. Oil Gas Sci Technol Rev IFP 64:91–112CrossRefGoogle Scholar
  5. Buchholz T, Luzadis VA, Volk TA (2009) Sustainability criteria for bioenergy systems: results from an expert survey. J Clean Prod 17(Supplement 10):S86–S98CrossRefGoogle Scholar
  6. Cano-Ruiz JA, McRae GJ (1998) Environmentally conscious chemical process design. Annu Rev Energy Environ 23:499–536CrossRefGoogle Scholar
  7. Cashman SA, Meyer DE, Edelen AN, Ingwersen WW, Abraham JP, Barrett WM, Gonzalez MA, Randall PM, Ruiz-Mercado G, Smith RL (2016) Mining available data from the united states environmental protection agency to support rapid life cycle inventory modeling of chemical manufacturing. Environ Sci Technol 50:9013–9025CrossRefGoogle Scholar
  8. Chan YH, Yusup S, Quitain AT, Tan RR, Sasaki M, Lam HL, Uemura Y (2015) Effect of process parameters on hydrothermal liquefaction of oil palm biomass for bio-oil production and its life cycle assessment. Energy Conversat Manag 104:180–188CrossRefGoogle Scholar
  9. Cheali P, Posada JA, Gernaey KV, Sin G (2016) Economic risk analysis and critical comparison of optimal biorefinery concepts. Biofuels Bioprod Biorefin 10:435–445CrossRefGoogle Scholar
  10. Cornelio da Silva C (2016) Lignin valorisation for integrated biorefineries in Brazil: explorative analysis for biojet fuel and chemicals production. Design report: Professional Doctorate in Engineering—Bioprocess Engineering Design. Delft University of TechnologyGoogle Scholar
  11. Cortez LAB (2014) Roadmap for sustainable aviation Biofuels for Brazil: a Flightpath to aviation biofuels in Brazil. Editorial Edgard Blücher, Sao PauloGoogle Scholar
  12. da Silva C, Mussatto MI, Osseweijer P, van der Wielen L, Posada JA (2016) Techno-economic and GHG emissions analysis of oil crops and forestry residues as feedstocks for integrated biojet fuel biorefineries in Brazil. In: Bioresource technology conference, 23–26 October, Sitges, Barcelona, SpainGoogle Scholar
  13. de Jong S, Hoefnagels R, Faaij A, Slade R, Mawhood R, Junginger M (2015) The feasibility of short-term production strategies for renewable jet fuels—a comprehensive techno-economic comparison. Biofuels Bioprod Biorefin 9:778–800CrossRefGoogle Scholar
  14. de la Ree AB (2011) Fischer–Tropsch catalyst for aviation fuel production. In: IECEC-9, San Diego, CAGoogle Scholar
  15. DESC (Dynamotive Energy Systems Corporation) (2009) Dynamotive upgrades BioOil from multiple biomass sources to a uniform hydrocarbon product. Accessed 16 March 2016
  16. Diederichs GW, Mandegari MA, Farzad S, Görgens JF (2016) Techno-economic comparison of biojet fuel production from lignocellulose, vegetable oil and sugar cane juice. Biores Technol 216:331–339CrossRefGoogle Scholar
  17. Dowling A, Ruiz-Mercado G, Zavala VM (2016) A framework for multi-stakeholder decision-making and conflict resolution. Comput Chem Eng 90:136–150CrossRefGoogle Scholar
  18. Elgowainy AJH, Wang M, Carter N, Stratton R, Hileman J, Malwitz A, Balasubramanian S (2012) Life cycle analysis of alternative aviation fuels in GREET, June 30Google Scholar
  19. Elliott D, Hallen R, Schmidt A (2015). Hydrothermal processing of biomass. In: DOE bioenergy technologies office (BETO)—2015 project peer review project peer review. Pacific Northwest National Laboratory, March 26Google Scholar
  20. Garcia-Perez M, Wang XS, Shen J, Rhodes MJ, Tian F, Lee WJ, Wu H, Li CZ (2008) Fast pyrolysis of oil mallee woody biomass: effect of temperature on the yield and quality of pyrolysis products. Ind Eng Chem Res 47:1846–1854CrossRefGoogle Scholar
  21. Gargalo C, Cheali P, Posada JA, Carvalho A, Gernaey KV, Sin G (2016a) Assessing the environmental sustainability of early stage design for bioprocesses under uncertainties: an analysis of glycerol bioconversion. J Clean Prod 139:1245–1260CrossRefGoogle Scholar
  22. Gargalo C, Cheali P, Posada JA, Gernaey KV, Sin G (2016b) Economic risk assessment of early-stage designs for glycerol valorization in biorefinery concepts. Ind Eng Chem Res 55:6801–6814CrossRefGoogle Scholar
  23. González-Estudillo JC, González-Campos J, Nápoles-Rivera F, Ponce-Ortega JM, El-Halwagi MM (2017) Optimal planning for sustainable production of Avocado in Mexico. Process Integr Optim Sustain 1:109–120CrossRefGoogle Scholar
  24. Guettel R, Kunz U, Turek T (2008) Reactors for Fischer–Tropsch synthesis. Chem Eng Technol 31:746–754CrossRefGoogle Scholar
  25. Gutiérrez-Antonio C, Gómez-Castro FI, Lira-Flores JA, Hernández S (2017) A review on the production processes of renewable jet fuel. Renew Sustain Energy Rev 79:709–729CrossRefGoogle Scholar
  26. Hamelinck CN, Faaij AP (2006) Outlook for advanced biofuels. Energy Policy 34:3268–3283CrossRefGoogle Scholar
  27. Herreras Martinez S, van Eijck J, da Cunha MP, Guilhoto JJM, Walter A, Faaij A (2013) Analysis of socio-economic impacts of sustainable sugarcane-ethanol production by means of inter-regional input–output analysis: demonstrated for Northeast Brazil. Renew Sustain Energy Rev 28:290–316CrossRefGoogle Scholar
  28. Huber GW, Iborra S, Corma A (2006) Synthesis of transportation fuels from biomass: chemistry, catalysts, and engineering. Chem Rev 106:4044–4098CrossRefGoogle Scholar
  29. Jansen RA (2012) Second generation biofuels and biomass: essential guide for investors, scientists and decision makers. Wiley, New YorkCrossRefGoogle Scholar
  30. Jin F (2014) Application of hydrothermal reactions to biomass conversion. Springer, BerlinCrossRefGoogle Scholar
  31. Jones S, Meyer P, Snowden-Swan L, Padmaperuma A, Tan E, Dutta A, Jacobson J, Cafferty K (2013) Process design and economics for the conversion of lignocellulosic biomass to hydrocarbon fuels: fast pyrolysis and hydrotreating bio-oil pathway. National Renewable Energy Laboratory (NREL), GoldenGoogle Scholar
  32. Jones S, Zhu Y, Anderson D, Hallen R, Elliott D, Schmidt A, Albrecht K, Hart T, Butcher M, Drennan C, Snowden-Swan L, Davis R, Kinchin C (2014) Process design and economics for the conversion of algal biomass to hydrocarbons: whole algae hydrothermal liquefaction and upgrading. Pacific Northwest National Laboratory, RichlandCrossRefGoogle Scholar
  33. Kim SJ, Jung SH, Kim JS (2010) Fast pyrolysis of palm kernel shells: influence of operation parameters on the bio-oil yield and the yield of phenol and phenolic compounds. Biores Technol 101:9294–9300CrossRefGoogle Scholar
  34. Kim J, Sen SM, Maravelias CT (2013) An optimization-based assessment framework for biomass-to-fuel conversion strategies. Energy Environ Sci 6(4):1093–1104CrossRefGoogle Scholar
  35. König DH, Baucks N, Dietrich RU, Worner A (2015) Simulation and evaluation of a process concept for the generation of synthetic fuel from CO2 and H2. Energy 91:833–841CrossRefGoogle Scholar
  36. Labuschagne C, Brent AC, van Erck RPG (2005) Assessing the sustainability performances of industries. J Clean Prod 13(4):373–385CrossRefGoogle Scholar
  37. Larragoiti-Kuri J, Rivera-Toledo M, Cocho-Roldán J, Maldonado-Ruiz EK, Le Borgne S, Pedraza-Segura L (2017) Convenient product distribution for a lignocellulosic biorefinery: optimization through sustainable indexes. Ind Eng Chem Res 56:11388–11397CrossRefGoogle Scholar
  38. Lubwama M (2009) Technical assessment of the functional and operational performance of a fixed bed biomass gasifier using agricultural residues. MSc thesis in Energy Technology, KTH School of Industrial Engineering and ManagementGoogle Scholar
  39. Luu LQ, Halog A (2016) Chapter fourteen—life cycle sustainability assessment: a holistic evaluation of social, economic, and environmental impacts. In: Ruiz-Mercado Gerardo, Cabezas Heriberto (eds) Sustainability in the design, synthesis and analysis of chemical engineering processes. Butterworth-Heinemann, Oxford, pp 327–352CrossRefGoogle Scholar
  40. Mirkouei A, Haapala KR, Sessions J, Murthy GS (2017a) A mixed biomass-based energy supply chain for enhancing economic and environmental sustainability benefits: a multi-criteria decision making framework. Appl Energy 206:1088–1101CrossRefGoogle Scholar
  41. Mirkouei A, Haapala KR, Sessions J, Murthy GS (2017b) A review and future directions in techno-economic modeling and optimization of upstream forest biomass to bio-oil supply chains. Renew Sustain Energy Rev 67:15–35CrossRefGoogle Scholar
  42. Murillo-Alvarado PE, Ponce-Ortega JM, Serna-González M, Castro-Montoya AJ, El-Halwagi MM (2013) Optimization of pathways for biorefineries involving the selection of feedstocks, products, and processing steps. Ind Eng Chem Res 14:5177–5190CrossRefGoogle Scholar
  43. Nipattummakul N, Ahmed II, Gupta AK, Kerdsuwan S (2011) Hydrogen and syngas yield from residual branches of oil palm tree using steam gasification. Int J Hydrog Energy 36:3835–3843CrossRefGoogle Scholar
  44. Oasmaa A, Solantausta Y, Arpiainen V, Kuoppala E, Sipila K (2009) Fast pyrolysis bio-oils from wood and agricultural residues. Energy Fuels 24:1380–1388CrossRefGoogle Scholar
  45. Palmeros Parada M, Osseweijer P, Posada JA (2017) Sustainable biorefineries, an analysis of practices for incorporating sustainability in biorefinery design. Ind Crops Prod 106:105–123CrossRefGoogle Scholar
  46. Palmeros Parada M, Asveld L, Osseweijer P, Posada JA (2018) Setting the design space of biorefineries through sustainability values, a practical approach. Biofuels Bioprod Biorefin 12:29–44CrossRefGoogle Scholar
  47. Pashaei-Kamali F, Borges JA, Osseweijer P, Posada JA (2018) Towards social sustainability: screening potential social and governance issues for biojet fuel supply chains in Brazil. Renew Sustain Energy Rev 92:50–61CrossRefGoogle Scholar
  48. Posada JA, Brentner LB, Ramirez A, Patel MK (2016) Conceptual design of sustainable integrated microalgae biorefineries: parametric analysis of energy use, greenhouse gas emissions and techno-economics. Algal Res 17:113–131CrossRefGoogle Scholar
  49. Radich T (2015) The flight paths for biojet fuel. U.S. Energy Information Administration, WashingtonGoogle Scholar
  50. Rowley HV, Peters GM, Lundie S, Moore SJ (2012) Aggregating sustainability indicators: beyond the weighted sum. J Environ Manag 111:24–33CrossRefGoogle Scholar
  51. Ruiz-Mercado GJ, Smith RL, Gonzalez MA (2012a) Sustainability indicators for chemical processes: I. Taxonomy. Ind Eng Chem Res 51:2309–2328CrossRefGoogle Scholar
  52. Ruiz-Mercado GJ, Smith RL, Gonzalez MA (2012b) Sustainability indicators for chemical processes: II. Data needs. Ind Eng Chem Res 51:2329–2353CrossRefGoogle Scholar
  53. Ruiz-Mercado GJ, Gonzalez MA, Smith RL (2013) Sustainability indicators for chemical processes: III. Biodiesel case study. Ind Eng Chem Res 52:6747–6760CrossRefGoogle Scholar
  54. Ruiz-Mercado GJ, Gonzalez MA, Smith RL (2014) Expanding GREENSCOPE beyond the gate: a green chemistry and life cycle perspective. Clean Technol Environ Policy 16:703–717CrossRefGoogle Scholar
  55. Santos CI, Silva CC, Mussatto SI, Osseweijer P, van der Wielen LAM, Posada JA (2017) Integrated 1st and 2nd generation sugarcane bio-refinery for jet fuel production in Brazil: techno-economic and greenhouse gas emissions assessment. Renew Energy. CrossRefGoogle Scholar
  56. Seo Y, Ly HV, Kim J, Kim SS, Cho W, Baek Y (2015) Gasification characteristics of Pinus rigida (Pitch Pine) and Quercus variabilis (Oriental Oak) with Dolomite Catalyst in a Fluidized Bed Reactor. Energy Sources Part A Recovery Util Environ Eff 37:1637–1646CrossRefGoogle Scholar
  57. Sikdar SK (2009) On aggregating multiple indicators into a single metric for sustainability. Clean Technol Environ Policy 11:157–161CrossRefGoogle Scholar
  58. Sikdar SK, Sengupta D, Harten P (2012) More on aggregating multiple indicators into a single index for sustainability analyses. Clean Technol Environ Policy 14:765–773CrossRefGoogle Scholar
  59. Singh R, Prakash A, Balagurumurthy B, Singh R, Saran S, Bhaskar T (2015) Hydrothermal liquefaction of agricultural and forest biomass residue: comparative study. J Mater Cycles Waste Manag 17:445–452CrossRefGoogle Scholar
  60. Smith R (2005) Chemical process design. Wiley, New YorkGoogle Scholar
  61. Smith RL, Ruiz-Mercado GJ (2014) A method for decision making using sustainability indicators. Clean Technol Environ Policy 16:749–755CrossRefGoogle Scholar
  62. Smith RL, Ruiz-Mercado G, Meyer DE, Gonzalez MA, Abraham JP, Barrett WM, Randall PM (2017) Coupling computer-aided process simulation and estimations of emissions and land use for rapid life cycle inventory modeling. ACS Sustain Chem Eng 5:3786–3794CrossRefGoogle Scholar
  63. Swanson RM, Satrio JA, Brown RC, Platon A, Hsu DD (2010) Techno-economic analysis of biofuels production based on gasification. National Renewable Energy Laboratory (NREL), Golden, ColoradoCrossRefGoogle Scholar
  64. Tahir AC, Darton RC (2010) The process analysis method of selecting indicators to quantify the sustainability performance of a business operation. J Clean Prod 18(16):1598–1607CrossRefGoogle Scholar
  65. Tews I, Zhu Y, Drennan CV, Elliott DC, Snowden-Swan LJ, Onarheim K, Solantausta Y, Beckman D (2014) Biomass direct liquefaction options: TechnoEconomic and life cycle assessment. Pacific Northwest National Laboratory, RichlandCrossRefGoogle Scholar
  66. Tijmensen MJ, Faaij APC, Hamelinck CN, van Hardeveld MRM (2002) Exploration of the possibilities for production of Fischer Tropsch liquids and power via biomass gasification. Biomass Bioenergy 23:129–152CrossRefGoogle Scholar
  67. Tzanetis KF, Posada JA, Ramirez A (2017) Analysis of biomass hydrothermal liquefaction and biocrude-oil upgrading for renewable jet fuel production: the impacts of reaction conditions on production costs and GHG emissions performance. Renew Energy 113:1388–1398CrossRefGoogle Scholar
  68. Wang JJ, Jing YY, Zhang CF, Zhao JH (2009) Review on multicriteria decision analysis aid in sustainable energy decision making. Renew Sustain Energy Rev 13:2263–2278CrossRefGoogle Scholar
  69. Yue D, Pandya S, You F (2016) Integrating hybrid life cycle assessment with multiobjective optimization: a modeling framework. Environ Sci Technol 50(3):1501–1509CrossRefGoogle Scholar
  70. Zavala VM (2016) Chapter seven—managing conflicts among decision makers in multiobjective design and operations. In: Ruiz-Mercado Gerardo, Cabezas Heriberto (eds) Sustainability in the design, synthesis and analysis of chemical engineering processes. Butterworth-Heinemann, Oxford, pp 169–180CrossRefGoogle Scholar
  71. Zhu Y, Biddy MJ, Jones SB, Elliott DC, Schmidt AJ (2014) Techno-economic analysis of liquid fuel production from woody biomass via hydrothermal liquefaction (HTL) and upgrading. Appl Energy 129:384–394CrossRefGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Universidad Central de ChileSantiagoChile
  2. 2.National Risk Management Research LaboratoryU.S. Environmental Protection AgencyCincinnatiUSA
  3. 3.Instituto de Ingeniería Química, Facultad de IngenieríaUniversidad de la RepúblicaMontevideoUruguay
  4. 4.Department of BiotechnologyDelft University of TechnologyDelftThe Netherlands

Personalised recommendations