You are on page 1of 16

pubs.acs.

org/IECR Article

Integrated Methodology for Optimal Synthesis of Lignocellulosic


Biomass-to-Liquid Fuel Production Processes: 2. Superstructure
MINLP Modeling and Evaluation for Optimal Biofuel Process
Synthesis and Integration
Paola Ibarra-Gonzalez and Ben-Guang Rong*
Cite This: https://dx.doi.org/10.1021/acs.iecr.0c02903 Read Online
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ACCESS Metrics & More Article Recommendations *


sı Supporting Information
Downloaded via UNIV OF WOLLONGONG on August 11, 2020 at 18:02:54 (UTC).

ABSTRACT: For thermochemical conversion routes to BtL fuel production, in part 1, we have developed a framework for the
synthesis of a BtL superstructure that was formulated by the interconnection of processing blocks consisting of unit operations
belonging to five base case process routes along with the data and information prepared for the embedded processing blocks. In part
2, a mixed integer nonlinear programming (MINLP) model is implemented for the superstructure optimization. The superstructure
is defined as an MINLP problem coded in GAMS 24.5.6, which sets the objective to minimize the total cost of manufacturing
(TCOM) of BtL fuels under different cases and integration scenarios. Three cases are demonstrated for three different product
profiles, and three network flowsheets are obtained as optimal technology routes. The three optimal network flowsheets are then
rigorously simulated and compared with the base case process flowsheets. The results demonstrated that this methodology can
generate optimal BtL biofuel production processes under different product scenarios.

1. INTRODUCTION costs. As in process synthesis, a consensus is that, if the optimal


Currently, various process routes to produce liquid and flowsheet is not predefined, it will not be found. Thus, the BtL
gaseous fuels from lignocellulosic biomass have been explored, process configurations cannot be limited to a certain number.
and it has been proven that the properties of these biomass-to- Each biofuel technological route can be considered as an
liquid (BtL) fuels can comply with the regulated fuel quality individual concept, and customized designs can be explored.
standards.1−3 BtL fuel conversion routes and plant config- From process point of view, other alternative to improve the
urations can vary quite extensively and can be adapted processes is the coprocessing combining two or more
according to product specifications. For instance, production feedstocks or technologies. In previous studies, the application
of advanced biofuels via gasification followed by Fischer− of coprocessing has been proposed. For instance, Baliban et
Tropsch (FT) synthesis is receiving increased attention due to al.5 proposed the thermochemical conversion of biomass and
the high-quality fuels produced. This process route produces natural gas to liquid fuels, such as, gasoline, diesel, and
fuel components compatible with conventional fossil fuels.
However, the high investment associated with its commercial Received: June 10, 2020
application makes this option economically unfeasible.4 For Revised: July 22, 2020
this reason, other process routes such as pyrolysis followed by Accepted: July 23, 2020
hydroprocessing and/or catalytic cracking, as well as different Published: July 23, 2020
configurations for the gasification-FT route, should be
explored, primarily aiming to lower investment and production

© XXXX American Chemical Society https://dx.doi.org/10.1021/acs.iecr.0c02903


A Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 1. Systematic procedure for superstructure MINLP modeling and evaluation for optimal biofuel process synthesis.

kerosene with simultaneous heat, power, and water integration. of using lower-cost fossil-based feedstocks, it may not always
The results suggested that these systems can be economically be practical to utilize multiple feedstocks within a single
competitive with petroleum-based processes. Even though refinery. The reason is that the distributed network of biomass
hybrid-feedstock refineries can take advantage of the benefits feedstock locations will contain several points where the
B https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

infrastructure for coal or natural gas delivery is minimal or investment and operation cost of different distillation column
nonexistent.6 Moreover, coprocessing of biomass and natural designs. To our knowledge, this work is the first to consider
gas achieved only 50% emissions reduction because a major simultaneously different thermochemical-based process tech-
component of the life-cycle emissions is attributed to nologies with consistent rigorous simulations as performance
combustion of the liquid fuels.5 and cost sources for the superstructure formulation as well as
The selection of the most promising BtL fuel process for the posterior MINLP optimization to evaluate different
configurations depends on the raw material availability, process designs under different product profile scenarios. The
location, plant capacity, conversion steps for total production simultaneous optimization strategy mathematically guarantees
processes, desired products and relevant by-products, as well as that the algorithm will evaluate all the process alternatives on
the state of technical development, and overall production the same basis for both base cases and new process routes and
costs. Moreover, the interaction of all these variables plays an will generate the process configuration that presents the
important role in why and how certain biofuel production minimal cost of the BtL production and meets product
processes are more successful than others. Santibañez-Aguilar requirements.
et al.7 proposed a mathematical programming model for the This study is to develop a superstructure-based optimization
optimal design of a supply chain of biorefineries to produce algorithm for synthesis of softwood BtL processes using
ethanol and biodiesel, which considers the uncertainty in the rigorous simulation results as performance and cost sources. In
prices of raw materials. In another work by Santibañez-Aguilar part 1, the step-by-step formulation of a superstructure by the
et al.,8 the focus was on a MINLP optimization problem to combination and interconnection of unit operations was
propose the optimal supply chain design considering the presented. The process simulator Aspen Plus was used to
harvesting sites, availability of the biomass sources, processing generate the consistent and rigorous data of the processing
plants geographical location, demand of products in each blocks conforming the BtL process alternatives in the
market, location of storage facilities, and transportation
superstructure. In this part 2, an optimization framework
between these components.
with MINLP modelling of the superstructure is presented,
From the different issues that need to be considered to select
which considers different thermochemical-based processes,
the most promising BtL biorefinery design, in this work, the
feedstock scenarios, product portfolios, and integration
focus is on the technical development and production costs.
This work, instead of focusing on the evaluation of one process scenarios. The MINLP solution is to generate optimal
route at once, proposes the development of a superstructure alternatives that minimize the total overall process cost and
for producing liquid hydrocarbon fuels from lignocellulosic maximize the biofuel yield.
biomass to generate new process systems, which may increase
the feasibility of these technologies by reducing the high- 2. METHODOLOGY FOR SUPERSTRUCTURE MINLP
investment cost associated with its application. MODELING AND EVALUATION FOR OPTIMAL
Process synthesis strategies capable of generating and BIOFUEL PROCESS SYNTHESIS
evaluating promising process alternatives under different In part 1, a softwood BtL superstructure (Figure S1) was
constraints and specifications have been recently developed formulated based on the information collected from the
in a fully automated and computationally efficient way. For rigorous simulation of different technological sections
instance, Martiń and Grossman9 proposed a superstructure for (thermochemical conversion, upgrading section, and separa-
the production of diesel from switchgrass via gasification
tion section) conforming five base case processes. In this part,
followed by FT. The optimization of the system was
the softwood BtL superstructure is formulated as a MINLP
formulated as a mixed-integer nonlinear programming
problem coded in GAMS 24.5.6. Figure 1 presents a systematic
(MINLP) problem for the optimal production of FT-diesel
procedure for the mathematical modeling of the super-
while minimizing the energy and hydrogen consumption.
structure. For the optimization algorithm setup, first, input
Baliban et al.6,10 proposed an optimization-based process
synthesis capable of analyzing distinct process designs of a specifications were given in the form of matrices including
hardwood BtL refinery considering the gasification of biomass processing block numbers and their corresponding feedstocks,
and FT hydrocarbons synthesis. Torres-Ortega et al.11,12 intermediate components, and fluidizing agents flow rates.
defined a synthesis network superstructure based on rigorous Likewise, catalyst and auxiliary reagent ratios and raw material
simulations and a MINLP using simplified model to optimize costs were defined. The input specifications were collected
the integrated production of biodiesel and hydrotreated from literature and the rigorous simulations. Then, processing
vegetable oil diesel. It was observed that the previous works block information (number of unit operations, capital costs,
use estimates from several literature sources with simplified and energy costs) is given in the form of matrices. Mass
models for the unit operation costs and energy cost evaluation. balance equations based on the results from the rigorous
For such process synthesis and evaluation, it is important to simulations in Aspen Plus were specified in the algorithm.
use rigorous and consistent data and information including Binary and fractional variables were used to define the
technical and economic data for both operation units and existence or absence of unit operations and to specify the
process flowsheets. Moreover, integrated rigorous simulation mass and energy integration possibilities within the processing
and mathematical techniques to optimize and synthesize new blocks. The processing block mass balances together with the
process routes have shown to simultaneously determine specification of binary and fractional variables allow the
different design aspects that increase the feasibility of definition of the possible interconnections between processing
processes. For instance, Caballero et al.13 presented a blocks. Moreover, to specify the possible integration scenarios,
superstructure-based optimization that combines integrated the mass balances and capital and energy costs of the
mathematical programming techniques and rigorous simu- wastewater treatment plant and cogeneration plant were
lations to select the design parameters that minimize the total specified.
C https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Concerning the economic calculations, equations for the utilizing part of the wastewater process streams to generate and
TCOM evaluation included capital costs, operating labor costs, provide energy for the system. Thus, in the cogeneration plant,
energy costs, cost of waste disposal, and raw material costs. pure water coming from the processes can be sent to the steam
For implementation of the algorithm toward specified generation section where the water can be directed to a
product cases, product profile constraints defining the pressure boiler to produce steam. Part of the steam can be used
maximum and/or minimum total desired fuel flow rates were for heating purposes, and the other part can be sent to the
given. The initial values for product profile constraints were electricity generation section (described in section 3.4.).
based on the actual gasoline and diesel flow rates obtained Simultaneously, the fuel gases (methane, ethane, and
from the rigorous simulation of the base case processes. Then, propylene) can be compressed and burned to produce the
the objective function is defined, which is to minimize the heat required to increase the water temperature and produce
TCOM of BtL fuels and maximize the productivity under steam.
different constraints and integration scenarios. The boundaries 3.4. Electricity Integration. In the cogeneration plant,
and initial values for the MINLP solution are given for the steam and gas turbines can be employed to produce the
specified case study. electricity required by the processing blocks. For instance, a
Finally, the optimal MINLP solutions were rigorously fraction of the steam produced in the pressure boiler, described
simulated for evaluation and comparison. The comparisons in section 3.3, can be sent to a steam turbine to produce
include product profiles, TCOM with and without integra- electricity. Moreover, the fuel gases used to increase the water
tions, productivity, TCOM per gasoline gallon equivalent, and temperature in section 3.3 can be further sent to a gas turbine
product properties. to generate additional electricity. The total electricity demand
The advantage of the methodology for simultaneous of each of the gasification-based and pyrolysis-based process
optimization is for searching the structure space to obtain routes can be supplied with the electricity generated in the
the optimal process with given constraints and specifications. steam and gas turbines.
This cannot be done with the conventional enumerating
approach to design the individual feasible process flowsheet; it 4. SUPERSTRUCTURE MATHEMATICAL MODELING
is also different from the numerical optimization for a single- From the superstructure and integration possibilities, the
prescribed process optimization. synthesis and evaluation of technological routes and
integration scenarios are determined with algorithms that are
3. SUPERSTRUCTURE MASS AND ENERGY able to handle complicated nonlinear and continuous problems
INTEGRATION with given constraints. The optimization of mass and energy
In part 1, processing blocks were defined and interconnected transfer between processing blocks is addressed by MINLP in
into a superstructure. Besides considering block combination, which the TCOM of the BtL fuel production processes is
the mass and energy integration between the unit operations minimized and the biofuel production is maximized.
conforming the processing blocks can be explored. In a The MINLP model rigorously describes the input−output
previous study by Ibarra-Gonzalez and Rong,14 the liquid, gas, relationships of each processing block within the super-
and solid emissions of the five base case process routes were structure. The MINLP modeling consists of specifying the
reported and it was found that important amounts of gases and input (feedstocks, model compounds, auxiliary streams, flow
water are produced or wasted through the processes. rates, conversion factors, recovery factors, etc.) and output
Therefore, process integration techniques can be implemented (desired products, minimum outlet flow rates, etc.) stream data
to reduce the liquid emissions and overall energy consumption collected from the rigorous simulations and the equations that
of the process routes. Moreover, cogeneration through define the interconnection between the thermochemical
combined heat and power is another integration option. The conversion, upgrading, and separation processing blocks that
integration possibilities are described in the following sections. convert the inputs into the desired outputs, while meeting the
3.1. Process Water Integration. BtL processes require objective function.
high amounts of water. For instance, during bio-oil and syngas Accordingly, the synthesis process can be decomposed into
production, an excessive amount of water is required for two levels. On the first level, the optimization problem selects
product condensation and cleaning. Therefore, in the the technological routes, which implies discrete decisions. On
pyrolysis-based routes, water coming from the quench columns the second level, the decision on how these technologies
employed in the cooling and recovery of the pyrolysis oil could should be interconnected is performed in a continuous space.
be processed and recycled back to the quench columns. The equations used to represent the BtL processing super-
Additionally, in the gasification-based routes, the wastewater structure are mass balances, economics, capacity adjustments,
stream coming from the scrubber can be purified (contains etc., which are sets of linear and nonlinear equations. For
ammonia and CO2) and sent back to the scrubber. instance, some of the nonlinearities are due to the definition of
3.2. Gases Integration. During the pyrolysis reaction, the separation blocks where stream splits are included as well
NCGs are usually employed as fluidizing agents to increase the as in the economic blocks. Considering this, it is clear that the
formation rates of the pyrolysis products. These gases (mainly optimal process route synthesis involves a discrete-continuous
CO, CO2, CH4, H2, C2H4, and C2H6) are also produced nonlinear problem.
through the process and thus can be potentially recovered and 4.1. Superstructure MINLP Algorithm Setup. As
recycled to the pyrolysis reactor. On the other hand, fuel gases described before, the BtL processing superstructure algorithm
such as methane, ethane, and propylene produced in both the consists of the specification of input and output streams and
pyrolysis-based and gasification-based process routes can be processing block parameters, which are defined, based on data
sent to a cogeneration plant for power generation. collected from the simulations (mass and energy balances,
3.3. Heat Integration. The general objective is to reduce emissions, processing blocks capital and energy costs,
heat requirements in the process, which can be achieved by economic constraints, and product profile constraints). After
D https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Table 1. Conversion Factor Convf(i,j,k) Matrix Table

the definition of the overall process inputs and outputs, the minimize TCOM GGE
equations to describe the superstructure are specified in the TCOM
GAMS algorithm. These equations are the objective function, =
productivity
mass component balances, integrations (gases, water and
cogeneration), waste streams, and economic calculations 0.280CCtot + 2.73COL + 1.23(ECtot + C WT + C RM)
=
(capital cost, energy cost, waste disposal cost, operating labor productivity
cost, and raw material costs), which are given as follows. (1)
4.1.1. Parameter Specification. For the algorithm setup, all where TCOMGGE is the TCOM per gasoline gallon equivalent
the input information was specified as a set of matrix tables. (GGE), CCtot is the total capital cost of the plant, COL is the
First, feedstock flow rates and market prices for raw materials, cost of operating labor, ECtot is the total cost of utilities of
utilities, and intermediate chemicals were defined, as presented energy cost, CWT is the cost of waste treatment and disposal,
in Table S1. Then, processing block performance parameters and CRM is the raw material cost. The equations for the
such as conversion and separation factors were defined. In economic calculations will be introduced in section 4.2. The
term productivity stands for the GGE produced per year, as
Table 1, an example of how the matrix tables were given in the
depicted in eq 2. By introducing the term productivity to the
algorithm is presented. The term Convf(i,j,k) refers to the equation, it is guaranteed that the total product flow rate is
conversion factors where, as introduced in part 1, the index i being maximized, and the manufacturing costs are minimized.
refers to the number of processing block (1, 2, 3, ...n), j refers
to the technology option (1, refers to feedstock selection; 2, for productivity = ∑ [fuel product flow rate·GGE]·OY (2)
a reaction block; or 3, for a separation block), and k refers to
the component being consumed or produced. In total 66 The fuel product flow rate values calculated in kilograms per
hour are then converted to gallons. Then, the results are
components were considered among all the process streams. In
multiplied by the gasoline and diesel GGE values and an
Table 1, the term x represents the conversion factor value, operational year (OY) consisting of 8000 h is considered.
which takes negative values (−x) if the component is being 4.1.3. Mass Balances. The mass balances show the key
consumed or positive values (x) if is being produced. For inputs and outputs of the superstructure blocks. In part 1 of
instance, the performance parameters for the gasification-low this study, the processing blocks that were interconnected into
temperature FT (LTFT)-upgrading processing blocks, calcu- the superstructure were identified with specific tags, as
lated in Part 1 of this work, are presented in Tables S2 and S3. presented in Figure S1. For instance, the reaction blocks
Besides these parameters, in another matrix table, auxiliary were tagged as RXN(i,j) and the separation blocks as SEP(i,j).
In these tags, the index i refers to the processing block number
reagent ratios (fluidizing agents, catalysts, solvents, etc.) were
and j to the technology option (1, refers to feedstock selection;
given. For the economic calculations, in other four matrix 2, for a reaction block; or 3, for a separation block). However,
tables, the numbers of unit operations, capital costs, electricity, to describe the performance of each processing block, a new
and thermal costs of each reaction, and separation block were index k is introduced when the mass balances are concerned.
specified. Likewise, the number of unit operations of the The index k represents the components (e.g., k = 1 is the
cogeneration and wastewater treatment plant and their softwood, k = 2 is CO, k = 3 is CO2, etc.). A total of 66
respective capital and energy costs were included. components including feedstock, solvents, fluidizing gas, and
4.1.2. Objective Function. The objective is to determine the so on were considered. The values of the index k are defined in
the matrix tables, and with them, the mass flow rate per hour of
optimal processing route within the superstructure under
each component can be calculated. Moreover, in Figure S1, a
specified constraints and product portfolios, which minimizes subscript LB was given to the lumped-process blocks (set of
the TCOM of BtL fuel systems and maximizes the individual reaction and separation process blocks grouped into
productivity. The objective function was defined as a relation one major block as a functional process section) to
of these two variables in the form of eq 1 differentiate them from the individual process blocks.
E https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 2. Fractional variables (Yn) to define interconnections between the thermochemical conversion blocks and upgrading blocks.

However, in the algorithm setup, there is no need to define the plus the inlet flow rate of the limiting reagent ‘k’ times the
blocks with subscripts. conversion factor, as presented in eq 7.
Furthermore, as can be observed in Figure S1, each lumped- min(RXN(i , j , k)) = moutPB + mnew(RXN(i , j , k)) (6)
process block consists of a reaction block and separation block,
and therefore, separate mass balances for each lumped-reaction
block and lumped-separation block are carried out. The mass mout(RXN(i , j , k)) = min(RXN(i , j , k)) + min(RXN(i , j ,‘k ’))
balance equations for the reaction blocks in the conversion ·Convf (RXN(i , j ,‘k ’))
section (PYR and GAS) are in the form of eqs 3 and 4. (7)

min(RXN(i , j , k)) = m feedstock(RXN(i , j , k)) + mnew(RXN(i , j , k)) Concerning the separation block mass balances, as depicted
(3)
in eqs 8 and 9, the block inlet flow rate min(SEP(i, j, k)) is equal to
the previous block flow rate moutPB plus the auxiliary stream
mout(RXN(i , j , k)) = min(RXN(i , j , k)) + min(RXN(i , j ,‘k ’)) flow rate mnew(SEP(i, j, k)). The term mnew(SEP(i, j, k)) is only defined
when an extra reagent stream is required (e.g., water needed
·Convf (RXN(i , j ,‘k ’)) (4) for scrubbing). On the other hand, the separation block outlet
flow rate mout(SEP(i, j, k)) is equal to the sum of the separation
where min(RXN(i, j, k)) is the reaction block total inlet flow rate in block outlet streams moutn (n = 1,2,3, ...)
kilograms per hour, mfeedstock(RXN(i, j, k)) is the mass flow rate of
spruce and/or pine residues, mnew(RXN(i, j, k)) is the auxiliary min(SEP(i , j , k)) = moutPB + mnew(SEP(i , j , k)) (8)
stream flow rate, such as, catalysts, hydrogen, steam, sand,
fluidizing gas, and so on. The auxiliary stream flow rate mout(SEP(i , j , k)) = mout1(SEP(i , j , k)) + mout 2(SEP(i , j , k)) +
mnew(RXN(i, j, k)) is calculated with eq 5, and it is equal to the
feedstock flow rate times the new stream component ratio ... + moutn(SEP(i , j , k)) (9)
Newf(RXN(i, j, k)). As depicted in eq 4, the term mout(RXN(i, j, k)) In addition, from the separation unit rigorous simulations,
stands for the total outlet flow rate, which is equal to the inlet separation or recovery factors Sepf(SEP(i, j, k)) were calculated.
flow rate plus the limiting reagent ‘k’ inlet flow rate times the These factors allow to calculate the component mass flow rate
conversion factor Convf(RXN(i, j, ‘k’)). This factor represents the of each separation block outlet stream. These factors are
conversion of reagents to products implicit in the separation block outlet stream term
mnew(RXN(i , j , k)) = Newf(RXN(i , j , k))·mfeedstock (5) moutn(SEP(i, j, k)), as presented in eq 10
moutn(SEP(i , j , k)) = min(SEP(i , j , k))·Sepf(SEP(i , j , k)) (10)
For the subsequent individual reaction blocks and lumped-
reaction blocks belonging to an upgrading section, the mass 4.1.4. Thermochemical Conversion and Upgrading Block
balances are in the form of eqs 6 and 7 where the total inlet Interconnection. To define possible interconnections, frac-
flow rate of the reaction block min(RXN(i, j, k)) is equal to the tional variables Yn are introduced. The term Yn defines the
outlet flow rate moutPB coming from a previous block (reaction interconnections between the processing blocks in the
(mout(RXN(i, j, k))) or separation (moutn(SEP(i, j, k)))) plus the conversion section and following upgrading blocks. This
auxiliary stream mnew(RXN(i, j, k)), which refers to catalysts and term indicates the product stream fraction from the
other auxiliary reagents. On the other hand, the outlet flow rate thermochemical conversion section that is to be sent to the
of a reaction block mout(RXN(i, j, k)) is equal to its inlet flow rate upgrading blocks. For example, as shown in Figure 2, Y1
F https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

indicates the fraction of pyrolysis oil that should be redirected values of 1 if it is considered as an emission or 0 if it is being
to the hydroprocessing blocks and Y2 indicates the fraction of recycled, similar to what was previously explained in section
pyrolysis oil going to the catalytic cracking blocks. For the 4.1.5
gasification routes, the terms Y3 and Y4 define the fraction of
syngas being sent to LTFT and/or HTFT, respectively. These min(WWD(i , j , k)) = ∑ mout (SEP(i ,j ,k))·Xn
n (14)
variables can take values from 0 to 1.
4.1.5. Integrations. A term Xn is introduced to define the 4.2. Economic Calculations. The cost associated to the
possible mass and/or energy integrations. The different levels operation of a BtL plant must be estimated to analyze the
of integrations considered for the mathematical modeling are economic feasibility of the proposed process routes. In this
NCG integration, water integration, and cogeneration. The section, the important cost factors related to the calculation of
general equation to represent the integrations is in the form of the TCOM are described.
a mass balance where the extra term Xn considering the 4.2.1. Capital Cost. The total capital cost of the plant was
integration is introduced, as depicted in eq 11 for a reaction calculated from the results of the processing block rigorous
block. For instance, as defined in eq 11, a stream coming from simulations (given as a matrix table in the algorithm) and
a separation block mrec(SEP(i, j, k)) is recycled to a reaction block considering the Lang factor technique. The total cost is
determined by multiplying the total purchased cost of all the
min(RXN(i , j , k)) = moutPB + mnew(RXN(i , j , k)) + mrec(SEP(i , j , k))·X n processing blocks (group of unit operations) involved in the
(11) process by a constant, called the Lang factor. For solid−fluid
The term mrec(SEP(i, j, k)) refers to the outlet stream coming processing plants, the Lang factor is equal to 3.63.15 To
from a separation block, which might be recycled, and the term calculate the total cost of the plant first, the individual
Xn decides whether the stream is being recycled or not. An processing block capital cost should be calculated.
equation of this form can be also defined for a separation block The individual processing block capital cost CCbase(i, j),
if the outlet stream from a separation block mrec(SEP(i, j, k)) is which is given in the algorithm, was calculated in an Aspen
being recycled also to a separation block (e.g., water recovered process economic analyzer for a base-flow rate mbase, and thus,
from a distillation column and sent to a scrubber). an equation to adjust the capital cost CCadj(i, j) depending on
The term Xn was defined as an integer variable because more the real mass inlet flow rate min(i, j, k) is introduced for the cases
than the total flow rate of the recycle stream is needed to fulfill where a different plant capacity is desired. The capital cost
adjustment is depicted in eq 1515
ij min(i , j , k) yz
the actual process requirements. Xn takes a value of 1 if the

= CC base(i , j)jjj zz
j mbase zz
stream is being recycled or 0 if the stream is considered as 0.6

k {
emission. For example, for SEP(1,3)LB in Figure 3, terms like CCadj(i , j)
(15)
Finally, the total capital cost CCtot of a process route is equal
to the sum of the individual processing block capital cost
divided by the plant lifetime times the Lang factor, as

ij CCadj(i , j) yz
presented in eq 16. As plant lifetime, 20 years were considered.

CCtot = 3.63 ∑ jjj z


j plant lifetime zzz
k { (16)
Figure 3. Integer variables (Xn) to define possible integration of water 4.2.2. Energy Cost. The energy cost EC(i, j) described by eq
stream in SEP(1,3)LB.
17 is equal to the utility consumption (amounts of electricity,
steam, and thermal fluids) of each processing block times the
X1 and X2 are considered for the water outlet stream in which utility cost and plant operating hours. The electricity, heating,
the term X1 refers to a possible integration (water recycling to and cooling requirements of each processing block were
quenching) and X2 refers to an emission. Since X1 considers calculated with the Aspen process economic analyzer V8.8 in
the water integration, this term should be specified in the form the form of eq 17 and specified in the algorithm as a matrix
of the integration in eq 11, as shown for this example in eq 12. table. The utility consumption and costs were reported in
On the other hand, the term X2 should be defined in the form kilograms or kilowatts per hour and dollars per kilogram or
of the waste stream disposal as in eq 14. kilowatt, respectively. For the plant operation, 8000 h was
considered per year (OY).
min(SEP(1,3, k)) = mout(RXN(1,2, k)) + mnew(SEP(1,3, k))
EC(i , j) = utility consumption· utility cost ·OY (17)
+ mrec(SEP(1,3, k))·X1 (12)
The total energy cost ECtot of a process route is equal to the
Moreover, an equation to relate the terms X1 and X2 is sum of the individual energy cost of the processing blocks, as
introduced, as depicted in eq 13. This means that, if X1 is equal shown in eq 18
to 1, then the total stream flow rate is being recycled and by
default, there is no emission and X2 is equal to 0 ECtot = ∑ EC(i ,j) (18)
X1 + X 2 = 1 (13)
4.2.3. Operating Labor Cost. The operating labor costs COL
4.1.6. Waste Stream Disposal. The separation block outlet calculation considers the necessary operating labor OL and
streams that are not sent or being recycled to other blocks are their corresponding salary per year, as depicted in eq 19. It was
considered as emissions. The emissions or waste streams can assumed that each operator is paid 25.33 dollars per hour and
be quantified with eq 14. In this equation, the term Xn can take works in total 2000 h per year16
G https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 4. MINLP algorithm implementation and solution.

COL = OL· annual salary (19) Likewise, the major decisions taken by the algorithm according
to the constraints are presented. Then, to prove the flexibility
The operating labor OL is equal to the number of operators of the optimization algorithm, three different case studies were
needed per shift NOL times the number of shifts per day. In eq evaluated. The case studies consider different values for the
20, the NOL calculation is presented.15 In this study, four shifts feedstock and product flow rates.
of 6 h were assumed 4.3.1. Constraints for Optimization Solution. Based on the
NOL = (6.29 + 31.7P 2 + 0.23Nnp(i , j))0.5 input information, the optimization algorithm evaluates the
(20) superstructure and selects the optimal technology within a set
where P is the number of processing steps involving the of technological selection constraints, mass balance constraints,
handling of solids, such as grinding and solid removal, and and techno-economic evaluation constraints. For the MINLP
Nnp(i, j) is the number of unit operations included in each algorithm implementation, boundaries and initial values for the
processing block, such as, compressors, separation units, following variables should be given as follows:
reactors, and heat exchangers. For the operating labor • Feedstock mass flow rate and composition.
calculation, the definition of a matrix table including the • Gasoline minimum or maximum desired flow rate
number of unit operations of each reaction and separation (based on the rigorous simulation results).
block was required. • Diesel minimum or maximum desired flow rate (based
4.2.4. Waste Disposal Costs. The waste disposal costs are on the rigorous simulation results).
calculated based on the emissions stream flow rate • Product component maximum flow rate (based on the
min(WWD(i, j, k)), which are not recycled to the BtL plant. The rigorous simulation results).
total waste disposal cost is the multiplication of the wastes flow • Integer variable values: existence (1) or absence (0) of
rate (kilograms per hour) times the disposal costs DC(dollars integration possibilities.
per kilogram) and considering that an OY consists of 8000 h, • Fractional variable values: fraction of products being
as presented in eq 21 sent to upgrading units.
C WT = min(WWD(i , j , k))·DC ·OY (21) After the definition of the initial values and the
implementation of the algorithm, the major decisions taken
4.2.5. Raw Material Costs. The total raw material cost CRM by the algorithm are as follows:
is equal to the sum of the feedstock flow rate mfeedstock(RXN(i, j, k)) • Feedstock mass flow rate and composition (in case
and the new stream flow rate mnew(RXN(i, j, k)) times their different sources are considered).
corresponding costs (Cfeed(RXN(i, j, k)) and Caux(RXN(i, j, k)), • Processing block interconnection.
respectively) in dollars per kilogram, as presented in eq 22. • NCG integration.
The raw materials, intermediate chemicals, and catalyst costs • Water integration.
are presented in Table S1. For the catalyst costs, the catalyst • Cogeneration.
loading and lifetime were considered in the calculations. In • Fraction of gases and/or water sent to cogeneration for
Table S4, the liquid hourly space velocity (LHSV), the weight steam and/or electricity production.
hourly space velocity (WHSV), and the gas hourly space • Fraction of syngas or pyrolysis oil being sent to the
velocity (GHSV) values as well as catalyst lifetimes considered upgrading alternatives.
for the catalyst cost calculations are presented • Final product flow rates.
(
C RM = mfeedstock(RXN(i , j , k))·Cfeed(RXN(i , j , k)) Some of these decision variables, but not all, are restricted to
be integer. For instance, the integration decision variables are
+ ∑ mnew(RXN(i ,j ,k))·Caux(RXN(i ,j ,k)))·OY (22) restricted to be integer. For this, a branch-and-bound approach
was selected since it is recommended to solve problems in
4.3. MINLP Algorithm Implementation. The proposed which some of the variables are constrained to be integer.
MINLP model has the advantage to find the optimal network Branch-and-bound strategy considers the division of the
flowsheet that fits a given set of constraints. Therefore, if the feasible region into more manageable subdivisions. The
set of constraints is changed, the problem solution might subdivisions are generated solely by the integer variables.
change also. In this section, first, the optimization constraints The general procedure is to subdivide based upon the variable
belonging to a feasible region, which need to be set to define with the highest objective contribution.17 In this study, it was
the limits of performance for the system, are described. subdivided in terms of the possible mass and energy
H https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

integrations. To achieve this, the model uses as a solver, integration possibilities. Concerning the biofuel-desired flow
branch-and-reduce optimization navigator (BARON), which is rates, gasoline and diesel flow rates were set to be greater than
recommended for the global solution of nonlinear program- or equal to 260 and 50 kg/h, respectively. These values were
ming (NLP) and MINLP. Moreover, BARON implements given according to the minimum gasoline and diesel flow rates
deterministic global optimization algorithms of the branch- obtained from the simulations and are given in section 4.4.
and-bound type that are guaranteed to provide the global For this case, the optimization solution proposed the
optima under fairly general assumptions.18 processing blocks depicted in Table 2. The results proposed
On the other hand, the selection of processing blocks
conforming the optimal technological route was performed by Table 2. MINLP Solution: Processing Blocks for Case Study
not specifying the block interconnections as integer variables 1
but as fractional variables (i.e., stream splitting and mass
processing
distribution), and therefore, the model can consider all the block description products
process routes simultaneously.
RXN(6,2) thermochemical conversion: syngas
4.4. MINLP Algorithm Solution. To find the optimal Gasification
route for each case study, the algorithm considers a set of RXN(2,2) combustion CO2
initial values for the feedstock flow rate, feedstock composition, RXN(7,2) LTFT distillate and
types of integrations, fraction of thermochemical products sent wax range
to the different upgrading alternatives, and the maximum and/ RXN(9,2) HTFT naphtha range
or minimum biofuel-desired flow rates, as presented in Figure and gases
4. Boundaries concerning gasoline and diesel production were RXN(10,2) fractional upgrading: cat. reforming and gasoline and
alkylation gases
given based on the actual component flow rates taken from the RXN(8,2) distillate hydrogenation diesel
rigorous simulations. For instance, the minimum diesel flow RXN(13,2)/ wax hydrocracking gasoline and
rate obtained from the simulation of the base case processes (14,2) diesel
was around 50 kg/h and the maximum flow rate was 343 kg/h. SEP(6,3) separation of gases and gasoline
On the other hand, for the gasoline production, the minimum SEP(7,3) separation of syngas and impurities
flow rate was 260 kg/h and the maximum was 830 kg/h. Based SEP(8,3) separation of FT syncrude fractions
on the data collected, initial values were given, which are the SEP(11,3) fractionation column for final fuels
starting point that allows the algorithm to search the optimal recovery
solution within the limits and product profile specifications.
The MINLP problem was solved on an HP EliteBook laptop a combined flowsheet of gasification followed by simultaneous
with Intel(R) Core(TM) i5-6300U, 240 GHz CPU, 8 GB high- and low-temperature FT (HTFT and LTFT, respec-
RAM, and Windows 10 64-bit. The final solution was tively) reactions and FT fractional upgrading blocks (distillate
determined by analyzing which reaction and separation blocks hydrotreating, wax hydrocracking, naphtha hydrotreating and
present inlet flow rates (min(RXN(i, j, k)) and min(SEP(i, j, k)), reforming, and tail gas alkylation) as the global optimal
respectively) and by confirming if there is consistency between technological route. This route was designated as combined
the components produced from the reaction blocks and alternative one (CA1-GLTHT), and it is presented in Figure 5.
reported final fuel product mass component flow rate. As can be observed from Table 2, the fuel components
Moreover, to determine the possible integrations, the Xn produced from each reaction block are consistent with the
resulting values were analyzed and collected. Likewise, to desired fuel targets. Likewise, from Figure 5, it can be observed
quantify the fraction of syngas or pyrolysis oil being sent to the that the waste gases from the lumped-block RXN(10,2)-
different upgrading alternatives, the Ynvalues were collected. SEP(8,3) that could produce more gasoline products were not
To demonstrate the methodology for the optimal synthesis upgraded, confirming that the gasoline production is not the
of network process flowsheets for different scenarios of product only objective for this case study. Concerning the other
and process specifications, three different case studies decision variables, a feedstock composition of 50 wt % spruce
considering different fuel product profiles and constraints and 50 wt % pine to be sent to the gasification section was
were considered, as will be described in the next section. proposed, which is understandable since spruce and pine
present similar cost and composition. Moreover, by analyzing
5. CASE STUDIES FOR SUPERSTRUCTURE MINLP the optimization solution, values for both fractional variables
ALGORITHM IMPLEMENTATION AND SOLUTION Y3 and Y4 were of 0.5, which means that the LTFT
5.1. Case Study 1: Maximization of Gasoline and (RXN(7,2)) and HTFT (RXN(9,2)) have the same flow
Diesel Fuel Production. The objective of the first case study rate, as presented in Table S5. The values of the resulting
is to generate a process route that can promote the integration decision variable Xn are also reported in Table S5.
simultaneous production of gasoline and diesel and minimize Concerning the possible mass and energy integrations, from
the TCOM. For this case study, as initial values, the feedstock Table S5, is possible to observe that if the objective is to
flow rate going to the pyrolysis section was initially set to 250 increase the production of biofuels and minimize the TCOM,
kg/h; likewise, the feedstock flow rate going to the gasification then steam and electricity generation in the cogeneration plant
section was set to 250 kg/h. These initial values are given to should be considered. The capital cost, energy cost, operating
guarantee that both processes are considered in the labor cost, raw material costs, waste streams cost, TCOM, and
optimization solving. Concerning the feedstock composition, productivity of the proposed route with and without the
as the starting point, both feedstock streams contain 50 wt % integration will be presented and compared with the base case
spruce and 50 wt % pine. Additionally, the initial values for the processes in section 6.
integer variables that define the mass and energy integrations 5.2. Case Study 2: Maximization of Gasoline Fuel
were set as 1, so the optimization can consider initially all Production. In this case study, the production of gasoline is
I https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 5. Global optimal solution CA1-GLTHT for case study 1.

Figure 6. Global optimal solution CA2-GLTHT for case study 2.

preferred, and thus, the diesel flow rate was set to be lower Regarding the decision values of the integration variables,
than or equal to 200 kg/h, and the gasoline flow rate was set to steam and electricity generation in the cogeneration plant is
be greater than or equal to 500 kg/h. These values are given considered similarly as in case study 1. Except that in this
according to the biofuel flow rates obtained from the scenario, the electricity generated might be lower, given that
simulations. For instance, the diesel production for most of most of the gases are upgraded to gasoline components and
the base case processes is around 200−343 kg/h, and the there is a low amount of waste gases available for power
gasoline production is around 523−830 kg/h. Therefore, generation.
values between these ranges were given as the maximum diesel The detailed numerical comparison between the CA2-
and minimum gasoline flow rates. On the other hand, the GLTHT optimal process route and base case processes is
initial values for the feedstock flow rate, feedstock composition, presented in section 6.
and integer variables (integrations) were the same given as in 5.3. Case Study 3: Maximization of Diesel Fuel
case study 1. Production. Contrary to the previous case, in this case
For this case study, the MINLP optimization solution was study, the production of diesel fuel is preferred, and thus, the
designated as CA2-GLTHT and is presented in Figure 6. As gasoline flow rate was set to be lower than or equal to 200 kg/
can be observed, this process route is a combined flowsheet of h, and the diesel flow rate was set to be greater than or equal to
Gasification followed by simultaneous high- and low-temper- 300 kg/h. The limits for the biofuels flow rates were given
ature FT reactions and fractional upgrading units, as obtained based on the minimum gasoline and maximum diesel flow rates
for case study 1. Likewise, the same feedstock flow rates, obtained from the base case process simulations and as
compositions and fractions of syngas sent to the HTFT and presented in section 4.4. Moreover, the same initial values for
LTFT, were obtained by the MINLP solution. However, the the feedstock flow rate, feedstock compositions, and integer
fractional upgrading units needed are different from the variables were given as in the previous cases.
optimal solution for case study 1; because in this case study, The MINLP solution proposed, as well as in the other two
the production of gasoline is preferred, and the diesel case studies, a combined flowsheet of gasification followed by
production is not a priority. Therefore, in this case, tail gas simultaneous HTFT and LTFT reactions and fractional
alkylation and naphtha reforming (RXN(10,2)), LPG upgrading units as the global optimal technological route.
oligomerization (RXN(12,2)), and wax hydrocracking units However, in this process route, the values obtained for the
(RXN(13,2)/RXN(14,2)) are required in which most of the fractional variables Yn were different. For instance, only 20% of
components are upgraded to gasoline blends. From Figure 6, it the syngas product should be sent to HTFT and 80% to LTFT,
can be observed that there is no waste gases from the lumped- which is understandable since LTFT produces more long chain
block RXN(10,2)-SEP(8,3) since the gases are being upgraded hydrocarbons to be upgraded to diesel. This process route was
to gasoline components via alkylation and oligomerization. designated as CA3-GLTHT and is presented in Figure 7. As
J https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 7. Global optimal solution CA3-GLTHT for case study 3.

Figure 8. CA1-GLTHT process flowsheet for simulation and evaluation.

Figure 9. CA2-GLTHT process flowsheet for simulation and evaluation.

can be observed, since the production of diesel is preferred, the were the same as in the other two cases, and thus, the gases
fractional upgrading units required are olefin alkylation, and wastewater were sent to the cogeneration plant.
naphtha hydrotreating and reforming (RXN(10,2)), distillate The detailed numerical comparison between the CA3-
hydrogenation (RXN(8,2)), and wax hydrocracking (RXN- GLTHT optimal process route and base case processes is
(13,2),(14,2)), which are the same units as in the first case. presented in section 6.
The main differences between this case study and the previous
cases are the amount of syngas being sent to each 6. OPTIMAL SCENARIO RIGOROUS SIMULATION
corresponding FT reactor, and that the alkyl benzenes AND EVALUATION
produced are mixed with the diesel components instead of The main task of the MINLP optimization algorithm is to
being mixed with all the upgraded streams, which allows synthesize the optimal network flowsheet. Therefore, given the
reducing the cetane number and increases the diesel density. product profile specifications and constraints, the MINLP
On the other hand, the decision values of the integer variables optimization algorithm searches all the possible networks from
K https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 10. CA3-GLTHT process flowsheet for simulation and evaluation.

Table 3. Comparison of the Three Optimization Processes with the Base Case Processes
parameter BC-PHS BC-PCCF BC-GHTUF BC-GLTUF BC-GHTH CA1-GLTHT CA2-GLTHT CA3-GLTHT
upgraded prod. (kg/h) 942 902 818 700 985 1022 1069 1030
total fuel prod. wt % 78 60 70 87 89 90 97 45
gasoline wt % 54 45 64 38 84 69 80 10
diesel wt % 24 15 6 49 5 21 17 35
CCtot($/year) 473,509 205,791 107,409 82,834 123,445 88,412 88,652 95,115
COL($/year) 1,621,120 1,215,840 1,621,120 1,621,120 1,621,120 1,621,120 1,621,120 1,621,120
CRM($/year) 17,653,109 5,712,142 5,977,087 4,680,232 7,989,420 6,934,368 7,313,331 3,955,324
ECtot($/year) 357,069 131,666 1,359,842 152,704 1,401,245 235,434 229,412 122,105
CWT($/year) 6,344,611 6,541,891 4,612,630 4,289,046 4,583,126 4,484,208 4,371,508 4,456,510
TCOM($/year) 34,514,631 18,611,274 19,153,691 15,668,889 21,647,985 18,784,844 19,105,008 14,949,035
TCOM w/int. ($/year) 34,114,491 18,262,395 17,619,701 15,341,811 20,949,776 18,274,260 18,766,123 14,680,510
productivity (GGE/year) 2,742,610 2,133,829 1,590,358 1,903,308 2,202,228 2,416,304 3,586,161 1,800,690
TCOMGGE($/GGE) 12.4 8.5 11.1 8.1 9.5 7.5 5.2 8.1

the superstructure including possible mass and energy were done to the design parameters to achieve the desired
integrations and gives the optimal process flowsheet that conversions and recoveries. For the rigorous simulation of the
meets the objective function, product profiles, and constraints. proposed process routes and cogeneration plant, the selection
Moreover, to interpret the model predictions correctly, it is of the thermodynamic model for each block has been selected
desirable to quantify the prediction uncertainty. Therefore, according to the stream composition. The thermodynamic
from the obtained MINLP optimal network, final rigorous package Soave−Redlich−Kwong equation of state with
design, simulation, and evaluation of the process flowsheet Kabadi−Danner mixing rules was selected due to its
with the Aspen process simulator should be carried out. The recommended application for mixtures containing water and
results from the rigorous simulation allow the consistent hydrocarbons.19 In addition, the electrolyte and non-electro-
numerical comparison between the optimal solutions and base lyte NRTL models with Redlich−Kwong equation of state
cases. Likewise, the GAMS solution and results from the were employed for the separation units. For instance, when the
rigorous simulation can be compared and the prediction of the stream contains ammonia, water, and some gases dissolved that
uncertainty and modelling error can be quantified. Moreover,
were carried over from the Scrubber, the Electrolyte NRTL
the final objective of the rigorous simulation is to finalize the
model was selected because is suited for solutions with
parameter designs for the obtained optimal flowsheet from the
multiple solvents, and dissolved gases, and it is very suitable for
optimization algorithm, which serves as to reconcile the blocks
and stream information for the obtained process flowsheet. any low-to-moderate pressure applications.19
6.1. Process Flowsheet Setups for Simulation and From the simulations, product profiles, product properties,
Evaluation with the Process Simulator. From the MINLP waste streams flow rates, and auxiliary stream flow rates were
solutions, three corresponding process flowsheets were collected. Aspen process economic analyzer V8.8 was used to
designed for the evaluation and simulation of the CA1- calculate energy and equipment costs for the process routes
GLTHT, CA2-GLTHT, and CA3-GLTHT network process with and without the recommended mass and energy
routes, as presented in Figures 8, 9, and 10, respectively. The integrations. The numerical evaluation was used to compare
rigorous simulations were performed in Aspen Plus V8.8 using the new process routes with the initial five base case process
as starting points the setups and designs of the gasification- routes introduced in Part 1 of this study. The data and
based process routes introduced in part 1 of this study and information obtained throughout process setups and rigorous
evaluated by Ibarra-Gonzalez and Rong.14 Then, because of simulations ensure that the evaluation, comparison, and
changes in the composition and flow rates of the FT fractions, analysis of the different process route performance is within
as well as the upgrading units required, minor adjustments the same framework and platform.
L https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

6.2. BtL Process Routes Comparison. The numerical TCOMGGE is presented. The fuel product profiles and
comparison between the new and base case process routes was properties for each process route were calculated from the
performed considering the total upgraded product flow rate, rigorous simulations and are presented in Figure 12 and Table
product profiles, economic evaluation, and annual productivity. 4, respectively.
For the economic evaluation and comparison, the results from
the rigorous simulations, the prices presented in Table S1, the
equations presented in section 4.2 (capital cost (eq 16), energy
costs (eqs 17 and 18), operating labor costs (eq 19), wastes
disposal costs (eq 21), raw material costs (eq 22), and
TCOMGGE (eq 1)) were employed together with the TCOM
calculation depicted in eq 23. For the implementation of these
equations, feedstock flow rates, auxiliary reagent flow rates,
emission streams, capital costs, and energy costs were collected
from the rigorous simulations. On the other hand, component
market prices, average operator salary, and costs of hazardous
and non-hazardous waste components were collected from
literature.15,16
TCOM = 0.280CCtot + 2.73COL + 1.23(ECtot + C WT
+ C RM) (23)
In Table 3, the economic results of all the process routes Figure 12. Product profiles of the BtL process route base cases and
were collected including capital costs, energy costs, operating three optimization cases.
labor costs, waste disposal stream and raw material costs, and
TCOM considering 8000 h of annual operation. Likewise, for Moreover, the prediction of the uncertainty of the MINLP
comparison purposes, the upgraded product flow rate, product model was performed by comparing the results obtained from
profiles, the TCOM of the process routes considering heat and the optimization in GAMS with the results from the rigorous
power integration, the productivity, and the fuel product value simulations, as depicted in Table 5. To calculate the
per GGE (TCOMGGE) are reported. In Figure 11, the uncertainty of the model, the mass outlet flowrate of the
contribution of the different costs (dollars per GGE) in the components produced from the algorithm implementation and
the total mass outlet flowrates of the gasoline and diesel
produced from the rigorous simulation of the process
configuration generated by the optimization were presented.
From the numerical results, it was demonstrated that when
simultaneous production of gasoline and diesel is preferred, the
CA1-GLTHT is the optimal technological route, producing
considerable amounts of both diesel and gasoline at lower
TCOM per GGE. For instance, the CA1-GLTHT could
produce 33% more biofuels (kilograms per hour) than BC-
GLTUF and 20% more than BC-PHS, and it presented a
gasoline equivalent cost of $7.5/GGE, which is $0.6/GGE
cheaper than the biofuels produced from BC-GLTUF. On the
other hand, if gasoline production is desired, the CA2-GLTHT
is the optimal process route since it can increase the gasoline
production by 4% compared to BC-GHTH and the TCOM is
$4.3/GGE cheaper. The detailed stream table for CA2-
GLTHT is presented in Table S6. Finally, if diesel production
is desired, the CA3-GLTHT is the optimal process route since
diesel production can be increased by 5% compared to BC-
GLTUF and the TCOM considering heat and power
integration is 4.4% lower. Moreover, even though BC-
GLTUF presents a higher productivity of total biofuels, if
diesel production is preferred, the CA3-GLTHT is the optimal
route since it only focuses on its production and also improves
the diesel properties. For instance, the cetane number was
reduced and the HHV and density were increased and can be
observed in Table 4. Likewise, as can be observed, the TCOMs
of CA1-GLTHT and CA2-GLTHT are higher compared to
BC-GLTUF, BC-GHTUF, and BC-PCCF. This is due to the
fact that, in the combined alternatives, the upgrading block
configurations allow the upgrading of the majority of the
Figure 11. Cost contribution in the TCOMGGE($/GGE) for the base products into fuels, and therefore, more auxiliary reagents are
cases and three optimization case processes. needed to achieve the conversions, which leads to an increase
M https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Table 5. Product Flow Rate Comparison between GAMS

1.08
diesel

38.6
CA3-GLTHT
and Aspen Plus Simulation Results for the Three

804

54
Optimization Cases
GAMS Aspen Plus difference between
gasoline

46.33

1.04
flow rate flow rate modeling

85
691
case studies fuel (kg/h) (kg/h) approaches %
CA1-GLTHT gasoline 755 705 7.09
diesel 217.3 215 1.07
1.21
diesel

37.7

CA2-GLTHT gasoline 839 848 1.06


CA2-GLTHT

802

59

diesel 199 184.6 7.80


CA3-GLTHT gasoline 100 103 2.91
gasoline

1.09

diesel 385.2 360.5 6.84


43.5
842

101

in the energy and raw material costs. On the other hand, by


1.09
diesel

37.3

upgrading the majority of the products, the productivity


CA1-GLTHT

802

58

increases leading to lower TCOM per GGE.


Concerning the fuel product value per GGE, the CA2-
gasoline

GLTHT is the optimal process route among all the process


0.89
43.22
848

100

routes evaluated in this study. It presented the highest biofuel


production and lowest product value of $5.2/GGE if heat and
power integrations are considered. However, compared to the
38.95

1.06
diesel

commercial gasoline price of $2.54/gallon reported for USA,20


753

77
BC-GHTH

the fuel product values per GGE obtained for all the process
routes explored are higher, which is expected for pioneer plants
during the first several years of operation. For instance, a study
gasoline

44.60

0.87

performed by researchers from the National Renewable Energy


803

88

Laboratory (NREL) stated that for a first of its kind plant, such
Table 4. Advanced Biofuel Properties for the Base Case and Three Optimization Case Processes

as high-temperature and low-temperature pioneer plants, the


1.28
diesel

fuel product values are estimated to be $7.60/GGE and $8.10/


771
34
82
BC-GLTUF

GGE, respectively.21
To further increase the commercial application of these
technologies, other aspects including the raw material
gasoline

44.52

0.96

availability, location, plant capacity, conversion steps for total


787

78

production processes, desired products and relevant by-


products, transportation costs as well as the state of technical
38.14

1.15
diesel

development, overall production costs, and government


72
760
BC-GHTUF

policies need to be considered. For instance, if the processing


plant is considered to be in Denmark where the cost of
gasoline is $6.22/Gallon20 and sufficient raw material can be
gasoline

44.25

0.91

available, then the feasibility increases.


816

104

Furthermore, the differences between the values obtained in


GAMS and Aspen are due to the fact that in the material
0.99

balances where the conversion factors are introduced (eqs 4


diesel

34
13
1091

and 7), the conversions are calculated according to the flow of


BC-PCCF

the limiting reagent.


gasoline

45.35

0.91

7. CONCLUSIONS
811

93

In this work, the softwood BtL superstructure, proposed in


Part 1 of this work, with the technical and economic data and
information prepared for its units and blocks, has been
45.83

1.11
diesel
706.5

42

developed into an MINLP optimization algorithm. The


BC-PHS

defined MINLP problem sets the objective to minimize the


TCOM of BtL fuels under different cases and integration
gasoline

46.83

0.97
738.1

scenarios. The superstructure was coded in GAMS 24.5.6, and


60

the solver BARON 14.4 was employed to determine the


process route that meets the objective function. From the
mathematical modeling of the superstructure and its
avg. octane value
density (kg/m3)
HHV (MJ/kg)
cetane number

implementation, optimal technological routes for the produc-


properties

tion of gasoline and diesel considering different product


profiles and integration possibilities were generated.
GGE

To evaluate the algorithm, three cases with different product


specifications have been studied, which demonstrated that the
N https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

optimization methodology can generate optimal process routes BtL biomass to liquid
for the specified cases. From the process routes evaluated, it CA1-GLTHT combined alternative one
was found that the CA2-GLTHT is the optimal technological CA2-GLTHT combined alternative two
route for the production of advanced biofuels as gasoline and CA3-GLTHT combined alternative three
diesel. This process can increase the production of advanced Caux(RXN(i, j, k)) auxiliary stream costs
biofuels while reducing the TCOM when heat and power CCadj(i, j) processing block-adjusted capital cost
integrations are considered. The process can produce gasoline CCbase(i, j) processing block-base capital cost
and diesel with similar fuel properties to petroleum fuels and CCtot total capital cost
with a fuel product price as of $5.2/GGE. Cfeed feedstocks costs
The results demonstrated that the integrated methodology COL operating labor costs
with the superstructure-based optimization can synthesize convf conversion factor
optimal total BtL biofuel production processes compared to CRM raw materials cost
the previous base case processes. CWT waste disposal costs


*
ASSOCIATED CONTENT
sı Supporting Information
DC

EC(i, j)
hazardous or nonhazardous disposal costs
($/kg)
processing block energy cost
The Supporting Information is available free of charge at ECtot total energy cost
https://pubs.acs.org/doi/10.1021/acs.iecr.0c02903. FT Fischer−Tropsch
Detailed information regarding the softwood BtL GGE gasoline gallon equivalent
superstructure diagram, market prices for raw materials, GHG greenhouse gas
utilities, and intermediate chemicals as well as processing GHSV gas hourly space velocity
block performance parameters, catalyst information, GLTHT gasification followed by simultaneous high-
integer, and fractional variables for a case study, and and low-temperature FT reactions and frac-
detailed stream table of the optimal process route (PDF) tional upgrading units


HHV high heating value
HTFT high-temperature Fischer−Tropsch
AUTHOR INFORMATION LHSV liquid hourly space velocity
Corresponding Author LPG liquified petroleum gas
Ben-Guang Rong − Department of Chemical Engineering, LTFT low-temperature Fischer−Tropsch
Biotechnology and Environmental Technology, University of mbase base mass inlet flow rate
Southern Denmark, Odense M DK-5230, Denmark; mfeedstock feedstock mass flow rate
Department of Chemistry and Bioscience, Aalborg University, min(i, j, k) actual mass inlet flow rate
Esbjerg DK-6700, Denmark; orcid.org/0000-0001-8730- MINLP mixed Integer Nonlinear Programming
5058; Phone: +45 9940 7668; Email: bro@bio.aau.dk min(RXN(i, j, k)) reaction block total inlet flow
min(SEP(i, j, k)) separation block inlet flow rate
Author
min(WWD(i, j, k)) total waste stream flow rate
Paola Ibarra-Gonzalez − Department of Chemical Engineering, MJ megajoule
Biotechnology and Environmental Technology, University of mnew(RXN(i, j, k)) auxiliary stream flow rate
Southern Denmark, Odense M DK-5230, Denmark mout(RXN(i, j, k)) reaction block outlet flow rate
Complete contact information is available at: mout(SEP(i, j, k)) total separation block outlet flow rate
https://pubs.acs.org/10.1021/acs.iecr.0c02903 moutn(SEP(i, j, k)) separation block outlet stream flow rate
moutPB flow rate coming from a previous block
Notes mrec(SEP(i, j, k)) recycle stream
The authors declare no competing financial interest. NCGs noncondensable gases

■ ACKNOWLEDGMENTS
The authors acknowledge financial support from CONACYT-
Newf(RXN(i, j, k))
NLP
Nnp(i, j)
auxiliary component ratio
nonlinear programming
total number of unit operations included in
The Mexican National Council for Science and Technology each processing block
(grant 326204/439098) and the University of Southern NOL number of operators per shift
Denmark. NREL National Renewable Energy Laboratory


Avg
NOMENCLATURE
average
OL
OY
P
operating labor
operational year
number of processing steps involving handling
BARON branch-and-reduce optimization navigator of solids
BC base case prod product
BC-GHTH gasification-HTFT hydrocarbon production- RXN reaction block
FT hydrocarbon hydroprocessing SEP separation block
BC-GHTUF gasification-HTFT hydrocarbon production- Sepf recovery factor
FT hydrocarbon upgrading-fractionation TCOM total cost of manufacturing
BC-GLTUF gasification-LTFT hydrocarbon production- TCOMGGE total cost of manufacturing per gasoline gallon
FT hydrocarbon upgrading-fractionation equivalent
BC-PCCF pyrolysis-catalytic cracking-fractionation w/int with integration
BC-PHS pyrolysis-hydroprocessing-separation WHSV weight hourly space velocity

O https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Xn integer variable (21) Swanson, R. M.; Satrio, J. A.; Brown, R. C.; Platon, A.; Hsu, D.
fractional variable, fraction of product sent to D. Techno-Economic Analysis of Biofuels Production Based on Gas-
Yn upgrading units ification; NREL, 2010.

■ REFERENCES
(1) Elliott, D. C. Transportation Fuels from Biomass via Fast
Pyrolysis and Hydroprocessing. Wiley Interdiscip. Rev. Energy Environ.
2013, 2, 525−533.
(2) Elliott, D. C.; Schiefelbein, G. F. Liquid Hydrocarbon Fuels from
Biomass. Am. Chem. Soc. Div. Fuel Chem. Prepr. 1989, 34, 1160−1166.
(3) Ward, C. C.; Schwartz, F. G.; Adams, N. G. Composition of
Fischer-Tropsch Diesel Fuel. Ind. Eng. Chem. 1951, 43, 1117−1119.
(4) Lappas, A. A.; Bezergianni, S.; Vasalos, I. A. Production of
Biofuels via Co-Processing in Conventional Refining Processes. Catal.
Today 2009, 145, 55−62.
(5) Baliban, R. C.; Elia, J. A.; Floudas, C. A. Biomass and Natural
Gas to Liquid Transportation Fuels: Process Synthesis, Global
Optimization, and Topology Analysis. Ind. Eng. Chem. Res. 2013,
52, 3381−3406.
(6) Baliban, R. C.; Elia, J. A.; Floudas, C. A. Biomass to Liquid
Transportation Fuels (BTL) Systems: Process Synthesis and Global
Optimization Framework. Energy Environ. Sci. 2013, 6, 267−287.
(7) Santibañ ez-Aguilar, J. E.; Guillen-Gosálbez, G.; Morales-
Rodriguez, R.; Jiménez-Esteller, L.; Castro-Montoya, A. J.; Ponce-
Ortega, J. M. Financial Risk Assessment and Optimal Planning of
Biofuels Supply Chains under Uncertainty. Bioenergy Res. 2016, 9,
1053−1069.
(8) Santibañez-Aguilar, J. E.; González-Campos, J. B.; Ponce-Ortega,
J. M.; Serna-González, M.; El-Halwagi, M. M. Optimal Planning and
Site Selection for Distributed Multiproduct Biorefineries Involving
Economic, Environmental and Social Objectives. J. Cleaner Prod.
2014, 65, 270−294.
(9) Martín, M.; Grossmann, I. E. Process Optimization of FT-Diesel
Production from Lignocellulosic Switchgrass. Ind. Eng. Chem. Res.
2011, 50, 13485−13499.
(10) Baliban, R. C.; Elia, J. A.; Floudas, C. A.; Gurau, B.;
Weingarten, M. B.; Klotz, S. D. Hardwood Biomass to Gasoline,
Diesel, and Jet Fuel: 1. Process Synthesis and Global Optimization of
a Thermochemical Refinery. Energy Fuels 2013, 27, 4302−4324.
(11) Torres-Ortega, C. E.; Gong, J.; You, F.; Rong, B.-G. Optimal
Synthesis of Integrated Process for Co-Production of Biodiesel and
Hydrotreated Vegetable Oil (HVO) Diesel from Hybrid Oil
Feedstocks. Comput. Aided Chem. Eng. 2017, 40, 673−678.
(12) Torres-Ortega, C. E.; Rong, B.-G. Integrated Biofuels Process
Synthesis: Integration between Bioethanol and Biodiesel Processes. In
Process Synthesis and Process Intensification: Methodological Approaches;
Rong, B.-G. Ed.; De Gruyter Graduate, 2017; pp. 241−289.
(13) Caballero, J. A.; Milán-Yañez, D.; Grossmann, I. E. Rigorous
Design of Distillation Columns: Integration of Disjunctive Program-
ming and Process Simulators. Ind. Eng. Chem. Res. 2005, 44, 6760−
6775.
(14) Ibarra-Gonzalez, P.; Rong, B.-G. Systematic Synthesis and
Evaluation of Thermochemical Conversion Processes for Lignocellu-
losic Biofuels Production: Total Process Evaluation and Integration.
Ind. Eng. Chem. Res. 2018, 57, 9925−9942.
(15) Turton, R. Analysis, Synthesis, and Design of Chemical Processes;
2012.
(16) Chemical Plant and System Operators; https://www.bls.gov/
oes/current/oes518091.htm (accessed Sep 30, 2019).
(17) Bradley, S. P. Integer Programming. In Applied Mathematical
Programming; Addison-Wesley Publishing, Ed.; 1977; pp. 272−319.
(18) Sahinidis, N. BARON; https://www.gams.com/latest/docs/S_
BARON.html (accessed Jul 17, 2019).
(19) Aspen Physical Property System: Physical Property Methods and
Models 11.1. Aspen Tech 2001.
(20) Gasoline prices around the world; 13-Jul-2020 https://www.
globalpetrolprices.com/gasoline_prices/ (accessed Jul 19, 2020).

P https://dx.doi.org/10.1021/acs.iecr.0c02903
Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

You might also like